From dhurjati at cs.uiuc.edu  Mon Jun 23 07:14:02 2003
From: dhurjati at cs.uiuc.edu (Dinakar Dhurjati)
Date: Mon Jun 23 07:14:02 2003
Subject: [llvm-commits] CVS: llvm/www/pubs/2003-05-05-LCTES03-CodeSafety.html
Message-ID: <200306231213.HAA27460@tank.cs.uiuc.edu>


Changes in directory llvm/www/pubs:

2003-05-05-LCTES03-CodeSafety.html updated: 1.2 -> 1.3

---
Log message:

Added presentation to downloads


---
Diffs of the changes:

Index: llvm/www/pubs/2003-05-05-LCTES03-CodeSafety.html
diff -u llvm/www/pubs/2003-05-05-LCTES03-CodeSafety.html:1.2 llvm/www/pubs/2003-05-05-LCTES03-CodeSafety.html:1.3
--- llvm/www/pubs/2003-05-05-LCTES03-CodeSafety.html:1.2	Wed May 21 12:15:41 2003
+++ llvm/www/pubs/2003-05-05-LCTES03-CodeSafety.html	Mon Jun 23 07:13:36 2003
@@ -58,6 +58,7 @@
 <ul>
 <li><a href="2003-05-05-LCTES03-CodeSafety.ps">Memory Safety Without Runtime Checks or Garbage Collection</a> (PS)
 <li><a href="2003-05-05-LCTES03-CodeSafety.pdf">Memory Safety Without Runtime Checks or Garbage Collection</a> (PDF)
+<li><a href="2003-06-12-LCTES03-CodeSafety.ppt">Memory Safety Without Runtime Checks or Garbage Collection - Presentation</a> (PPT)
 </ul>
 
 <h2>Bibtex Entry:</h2>
@@ -72,5 +73,9 @@
     URL = {http://llvm.cs.uiuc.edu/pubs/2003-05-05-LCTES03-CodeSafety.html}
   }
 </pre>
+
+<h2> Links: </h2>
+
+<a href="http://safecode.cs.uiuc.edu/">SAFECode project</a> 
 
 </body></html>




From lattner at cs.uiuc.edu  Mon Jun 23 12:16:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Mon Jun 23 12:16:01 2003
Subject: [llvm-commits] CVS: llvm/test/Programs/External/SPEC/CFP2000/179.art/Makefile
Message-ID: <200306231715.MAA13730@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Programs/External/SPEC/CFP2000/179.art:

Makefile added (r1.1)

---
Log message:

Initial checkin of new bm


---
Diffs of the changes:

Index: llvm/test/Programs/External/SPEC/CFP2000/179.art/Makefile
diff -c /dev/null llvm/test/Programs/External/SPEC/CFP2000/179.art/Makefile:1.1
*** /dev/null	Mon Jun 23 12:15:28 2003
--- llvm/test/Programs/External/SPEC/CFP2000/179.art/Makefile	Mon Jun 23 12:15:18 2003
***************
*** 0 ****
--- 1,11 ----
+ LEVEL = ../../../../../..
+ ifdef LARGE_PROBLEM_SIZE
+ RUN_OPTIONS     = -scanfile c756hel.in -trainfile1 a10.img -stride 2 -startx 134 -starty 220 -endx 184 -endy 240 -objects 3
+ else
+ RUN_OPTIONS     = -scanfile c756hel.in -trainfile1 a10.img -stride 2 -startx 134 -starty 220 -endx 139 -endy 225 -objects 1
+ endif
+ #STDIN_FILENAME  = $(RUN_TYPE).in
+ STDOUT_FILENAME = $(RUN_TYPE).out
+ 
+ CPPFLAGS =
+ include ../../Makefile.spec




From lattner at cs.uiuc.edu  Mon Jun 23 12:16:06 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Mon Jun 23 12:16:06 2003
Subject: [llvm-commits] CVS: llvm/test/Programs/External/SPEC/CFP2000/179.art/
Message-ID: <200306231715.MAA13721@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Programs/External/SPEC/CFP2000/179.art:


---
Log message:

Directory /home/vadve/vadve/Research/DynOpt/CVSRepository/llvm/test/Programs/External/SPEC/CFP2000/179.art added to the repository


---
Diffs of the changes:



From lattner at cs.uiuc.edu  Mon Jun 23 12:20:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Mon Jun 23 12:20:01 2003
Subject: [llvm-commits] CVS: llvm/test/Programs/External/SPEC/CFP2000/183.equake/Makefile
Message-ID: <200306231719.MAA14094@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Programs/External/SPEC/CFP2000/183.equake:

Makefile added (r1.1)

---
Log message:

Initial checkin


---
Diffs of the changes:

Index: llvm/test/Programs/External/SPEC/CFP2000/183.equake/Makefile
diff -c /dev/null llvm/test/Programs/External/SPEC/CFP2000/183.equake/Makefile:1.1
*** /dev/null	Mon Jun 23 12:19:55 2003
--- llvm/test/Programs/External/SPEC/CFP2000/183.equake/Makefile	Mon Jun 23 12:19:45 2003
***************
*** 0 ****
--- 1,6 ----
+ LEVEL = ../../../../../..
+ RUN_OPTIONS     = 
+ STDIN_FILENAME  = inp.in
+ STDOUT_FILENAME = inp.out
+ CPPFLAGS =
+ include ../../Makefile.spec




From lattner at cs.uiuc.edu  Mon Jun 23 12:20:03 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Mon Jun 23 12:20:03 2003
Subject: [llvm-commits] CVS: llvm/test/Programs/External/SPEC/CFP2000/183.equake/
Message-ID: <200306231719.MAA14085@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Programs/External/SPEC/CFP2000/183.equake:


---
Log message:

Directory /home/vadve/vadve/Research/DynOpt/CVSRepository/llvm/test/Programs/External/SPEC/CFP2000/183.equake added to the repository


---
Diffs of the changes:



From lattner at cs.uiuc.edu  Mon Jun 23 12:38:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Mon Jun 23 12:38:00 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/ExprTypeConvert.cpp TransformInternals.cpp
Message-ID: <200306231737.MAA15542@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms:

ExprTypeConvert.cpp updated: 1.73 -> 1.74
TransformInternals.cpp updated: 1.35 -> 1.36

---
Log message:

avoid dividing by zero when dealing with zero sized types (like [0 x double])


---
Diffs of the changes:

Index: llvm/lib/Transforms/ExprTypeConvert.cpp
diff -u llvm/lib/Transforms/ExprTypeConvert.cpp:1.73 llvm/lib/Transforms/ExprTypeConvert.cpp:1.74
--- llvm/lib/Transforms/ExprTypeConvert.cpp:1.73	Thu May 29 10:12:27 2003
+++ llvm/lib/Transforms/ExprTypeConvert.cpp	Mon Jun 23 12:36:49 2003
@@ -49,6 +49,7 @@
 
   // Get information about the base datatype being allocated, before & after
   int ReqTypeSize = TD.getTypeSize(Ty);
+  if (ReqTypeSize == 0) return false;
   unsigned OldTypeSize = TD.getTypeSize(MI->getType()->getElementType());
 
   // Must have a scale or offset to analyze it...


Index: llvm/lib/Transforms/TransformInternals.cpp
diff -u llvm/lib/Transforms/TransformInternals.cpp:1.35 llvm/lib/Transforms/TransformInternals.cpp:1.36
--- llvm/lib/Transforms/TransformInternals.cpp:1.35	Sat Jun  7 16:45:42 2003
+++ llvm/lib/Transforms/TransformInternals.cpp	Mon Jun 23 12:36:49 2003
@@ -119,6 +119,7 @@
       if (!ElTy->isSized() || (isa<PointerType>(CompTy) && !Indices.empty()))
         return 0; // Type is unreasonable... escape!
       unsigned ElSize = TD.getTypeSize(ElTy);
+      if (ElSize == 0) return 0;   // Avoid division by zero...
       int64_t ElSizeS = ElSize;
 
       // See if the user is indexing into a different cell of this array...




From lattner at cs.uiuc.edu  Mon Jun 23 13:07:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Mon Jun 23 13:07:00 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/CFrontend/2003-06-23-GCC-fold-infinite-recursion.c
Message-ID: <200306231806.NAA17029@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/CFrontend:

2003-06-23-GCC-fold-infinite-recursion.c added (r1.1)

---
Log message:

New testcase for a GCC bug


---
Diffs of the changes:

Index: llvm/test/Regression/CFrontend/2003-06-23-GCC-fold-infinite-recursion.c
diff -c /dev/null llvm/test/Regression/CFrontend/2003-06-23-GCC-fold-infinite-recursion.c:1.1
*** /dev/null	Mon Jun 23 13:06:31 2003
--- llvm/test/Regression/CFrontend/2003-06-23-GCC-fold-infinite-recursion.c	Mon Jun 23 13:06:21 2003
***************
*** 0 ****
--- 1,4 ----
+ double Test(double A, double B, double C, double D) {
+   return -(A-B) - (C-D);
+ }
+ 




From lattner at cs.uiuc.edu  Mon Jun 23 13:18:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Mon Jun 23 13:18:01 2003
Subject: [llvm-commits] CVS: llvm/test/Programs/External/SPEC/CFP2000/188.ammp/Makefile
Message-ID: <200306231817.NAA17517@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Programs/External/SPEC/CFP2000/188.ammp:

Makefile added (r1.1)

---
Log message:

New testcase


---
Diffs of the changes:

Index: llvm/test/Programs/External/SPEC/CFP2000/188.ammp/Makefile
diff -c /dev/null llvm/test/Programs/External/SPEC/CFP2000/188.ammp/Makefile:1.1
*** /dev/null	Mon Jun 23 13:17:28 2003
--- llvm/test/Programs/External/SPEC/CFP2000/188.ammp/Makefile	Mon Jun 23 13:17:18 2003
***************
*** 0 ****
--- 1,6 ----
+ LEVEL = ../../../../../..
+ RUN_OPTIONS     = 
+ STDIN_FILENAME  = ammp.in
+ STDOUT_FILENAME = ammp.out
+ CPPFLAGS =
+ include ../../Makefile.spec




From lattner at cs.uiuc.edu  Mon Jun 23 13:18:03 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Mon Jun 23 13:18:03 2003
Subject: [llvm-commits] CVS: llvm/test/Programs/External/SPEC/CFP2000/188.ammp/
Message-ID: <200306231817.NAA17507@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Programs/External/SPEC/CFP2000/188.ammp:


---
Log message:

Directory /home/vadve/vadve/Research/DynOpt/CVSRepository/llvm/test/Programs/External/SPEC/CFP2000/188.ammp added to the repository


---
Diffs of the changes:



From criswell at choi.cs.uiuc.edu  Mon Jun 23 13:44:01 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 13:44:01 2003
Subject: [llvm-commits] CVS: llvm/include/Config/
Message-ID: <200306231843.h5NIhnC01755@choi.cs.uiuc.edu>


Changes in directory llvm/include/Config:


---
Log message:

Directory /home/vadve/vadve/Research/DynOpt/CVSRepository/llvm/include/Config added to the repository
--> Using per-directory sticky tag `autoconf'


---
Diffs of the changes:



From criswell at choi.cs.uiuc.edu  Mon Jun 23 13:46:01 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 13:46:01 2003
Subject: [llvm-commits] CVS: llvm/include/Config/alloca.h assert.h dlfcn.h errno.h fcntl.h limits.h link.h malloc.h memory.h stdint.h stdio.h stdlib.h string.h strings.h time.h unistd.h
Message-ID: <200306231845.h5NIjlk01813@choi.cs.uiuc.edu>


Changes in directory llvm/include/Config:

alloca.h added (r1.1.2.1)
assert.h added (r1.1.2.1)
dlfcn.h added (r1.1.2.1)
errno.h added (r1.1.2.1)
fcntl.h added (r1.1.2.1)
limits.h added (r1.1.2.1)
link.h added (r1.1.2.1)
malloc.h added (r1.1.2.1)
memory.h added (r1.1.2.1)
stdint.h added (r1.1.2.1)
stdio.h added (r1.1.2.1)
stdlib.h added (r1.1.2.1)
string.h added (r1.1.2.1)
strings.h added (r1.1.2.1)
time.h added (r1.1.2.1)
unistd.h added (r1.1.2.1)

---
Log message:

Added header files that wrap normal header files.  Source code should include
these header files, which will use information from autoconf to determine
which system headers should be included.


---
Diffs of the changes:



































From criswell at choi.cs.uiuc.edu  Mon Jun 23 13:47:02 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 13:47:02 2003
Subject: [llvm-commits] CVS: llvm/include/Config/sys/mman.h resource.h stat.h time.h types.h wait.h
Message-ID: <200306231846.h5NIkjF01850@choi.cs.uiuc.edu>


Changes in directory llvm/include/Config/sys:

mman.h added (r1.1.2.1)
resource.h added (r1.1.2.1)
stat.h added (r1.1.2.1)
time.h added (r1.1.2.1)
types.h added (r1.1.2.1)
wait.h added (r1.1.2.1)

---
Log message:

Wrapper header files for the system header files.  These include the correct
header files based upon what autoconf has found.


---
Diffs of the changes:















From criswell at choi.cs.uiuc.edu  Mon Jun 23 13:47:04 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 13:47:04 2003
Subject: [llvm-commits] CVS: llvm/include/Config/sys/
Message-ID: <200306231846.h5NIk3U01826@choi.cs.uiuc.edu>


Changes in directory llvm/include/Config/sys:


---
Log message:

Directory /home/vadve/vadve/Research/DynOpt/CVSRepository/llvm/include/Config/sys added to the repository
--> Using per-directory sticky tag `autoconf'


---
Diffs of the changes:



From criswell at choi.cs.uiuc.edu  Mon Jun 23 13:49:01 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 13:49:01 2003
Subject: [llvm-commits] CVS: llvm/include/Config/config.h.in
Message-ID: <200306231848.h5NImAC01869@choi.cs.uiuc.edu>


Changes in directory llvm/include/Config:

config.h.in added (r1.1.2.1)

---
Log message:

Template config.h file for autoconf.


---
Diffs of the changes:





From criswell at choi.cs.uiuc.edu  Mon Jun 23 13:51:01 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 13:51:01 2003
Subject: [llvm-commits] CVS: llvm/include/Support/Annotation.h BitSetVector.h Casting.h CommandLine.h DataTypes.h Signals.h TarjanSCCIterator.h Timer.h Tree.h hash_map hash_set ilist iterator slist
Message-ID: <200306231850.h5NIomc01933@choi.cs.uiuc.edu>


Changes in directory llvm/include/Support:

Annotation.h updated: 1.8 -> 1.8.2.1
BitSetVector.h updated: 1.5 -> 1.5.2.1
Casting.h updated: 1.5 -> 1.5.2.1
CommandLine.h updated: 1.12 -> 1.12.2.1
DataTypes.h updated: 1.8 -> 1.8.2.1
Signals.h updated: 1.2 -> 1.2.2.1
TarjanSCCIterator.h updated: 1.5 -> 1.5.2.1
Timer.h updated: 1.6 -> 1.6.2.1
Tree.h updated: 1.3 -> 1.3.2.1
hash_map updated: 1.6 -> 1.6.2.1
hash_set updated: 1.6 -> 1.6.2.1
ilist updated: 1.6 -> 1.6.2.1
iterator updated: 1.2 -> 1.2.2.1
slist updated: 1.2 -> 1.2.2.1

---
Log message:

Initial checkin of autoconf code.



---
Diffs of the changes:

Index: llvm/include/Support/Annotation.h
diff -u llvm/include/Support/Annotation.h:1.8 llvm/include/Support/Annotation.h:1.8.2.1
--- llvm/include/Support/Annotation.h:1.8	Wed Jun 11 09:01:21 2003
+++ llvm/include/Support/Annotation.h	Mon Jun 23 13:50:38 2003
@@ -16,7 +16,8 @@
 #define SUPPORT_ANNOTATION_H
 
 #include <string>
-#include <assert.h>
+#include "Config/assert.h"
+
 class AnnotationID;
 class Annotation;
 class Annotable;


Index: llvm/include/Support/BitSetVector.h
diff -u llvm/include/Support/BitSetVector.h:1.5 llvm/include/Support/BitSetVector.h:1.5.2.1
--- llvm/include/Support/BitSetVector.h:1.5	Wed Jun 11 09:01:21 2003
+++ llvm/include/Support/BitSetVector.h	Mon Jun 23 13:50:38 2003
@@ -31,7 +31,7 @@
 #include <functional>
 #include <iostream>
 
-#include <assert.h>
+#include "Config/assert.h"
 
 #define WORDSIZE (32U)
 


Index: llvm/include/Support/Casting.h
diff -u llvm/include/Support/Casting.h:1.5 llvm/include/Support/Casting.h:1.5.2.1
--- llvm/include/Support/Casting.h:1.5	Thu May 29 10:07:48 2003
+++ llvm/include/Support/Casting.h	Mon Jun 23 13:50:38 2003
@@ -8,7 +8,7 @@
 #ifndef SUPPORT_CASTING_H
 #define SUPPORT_CASTING_H
 
-#include <assert.h>
+#include "Config/assert.h"
 
 //===----------------------------------------------------------------------===//
 //                          isa<x> Support Templates


Index: llvm/include/Support/CommandLine.h
diff -u llvm/include/Support/CommandLine.h:1.12 llvm/include/Support/CommandLine.h:1.12.2.1
--- llvm/include/Support/CommandLine.h:1.12	Wed Jun 11 09:01:21 2003
+++ llvm/include/Support/CommandLine.h	Mon Jun 23 13:50:38 2003
@@ -19,7 +19,7 @@
 #include <cstdarg>
 #include "boost/type_traits/object_traits.hpp"
 
-#include <assert.h>
+#include "Config/assert.h"
 
 /// cl Namespace - This namespace contains all of the command line option
 /// processing machinery.  It is intentionally a short name to make qualified


Index: llvm/include/Support/DataTypes.h
diff -u llvm/include/Support/DataTypes.h:1.8 llvm/include/Support/DataTypes.h:1.8.2.1
--- llvm/include/Support/DataTypes.h:1.8	Wed Jun 11 14:44:48 2003
+++ llvm/include/Support/DataTypes.h	Mon Jun 23 13:50:38 2003
@@ -19,44 +19,21 @@
 #ifndef LLVM_SUPPORT_DATATYPES_H
 #define LLVM_SUPPORT_DATATYPES_H
 
-#define __STDC_LIMIT_MACROS 1
-#include <inttypes.h>
-
-#ifdef __linux__
-#  include <endian.h>
-#  if BYTE_ORDER == LITTLE_ENDIAN
-#    undef BIG_ENDIAN
-#  else
-#    undef LITTLE_ENDIAN
-#  endif
-#else
-#  if (BSD >= 199103)
-#    include <machine/endian.h>
-#  endif
-#endif
-
-#ifdef __sparc__
-#  include <sys/types.h>
-#  ifdef _LITTLE_ENDIAN
-#    define LITTLE_ENDIAN 1
-#  else
-#    define BIG_ENDIAN 1
-#  endif
-#endif
-
 //
-// Convert the information from the header files into our own local
-// endian macros.  We do this because various strange systems define both
-// BIG_ENDIAN and LITTLE_ENDIAN, and we don't want to conflict with them.
+// Un-define these constants so that we can re-define only the one that we want
+// (which is really to say that we'll define LITTLE_ENDIAN if the machine is
+// little endian, and not define anything otherwise).
 //
-// Don't worry; once we introduce autoconf, this will look a lot nicer.
-// 
-#ifdef LITTLE_ENDIAN
-#define ENDIAN_LITTLE
+#include "Config/config.h"
+
+#define __STDC_LIMIT_MACROS 1
+
+#ifdef HAVE_INTTYPES_H
+#include <inttypes.h>
 #endif
 
-#ifdef BIG_ENDIAN
-#define ENDIAN_BIG
+#ifdef HAVE_SYS_TYPES_H
+#include <sys/types.h>
 #endif
 
 #if (defined(ENDIAN_LITTLE) && defined(ENDIAN_BIG))


Index: llvm/include/Support/Signals.h
diff -u llvm/include/Support/Signals.h:1.2 llvm/include/Support/Signals.h:1.2.2.1
--- llvm/include/Support/Signals.h:1.2	Wed Dec 11 21:43:30 2002
+++ llvm/include/Support/Signals.h	Mon Jun 23 13:50:38 2003
@@ -10,6 +10,11 @@
 
 #include <string>
 
+//
+// This will include the signal handler return type.
+//
+#include "Config/config.h"
+
 // RemoveFileOnSignal - This function registers signal handlers to ensure that
 // if a signal gets delivered that the named file is removed.
 //


Index: llvm/include/Support/TarjanSCCIterator.h
diff -u llvm/include/Support/TarjanSCCIterator.h:1.5 llvm/include/Support/TarjanSCCIterator.h:1.5.2.1
--- llvm/include/Support/TarjanSCCIterator.h:1.5	Wed Jun 11 09:01:21 2003
+++ llvm/include/Support/TarjanSCCIterator.h	Mon Jun 23 13:50:38 2003
@@ -21,7 +21,7 @@
 #include <stack>
 #include <map>
 
-#include <assert.h>
+#include "Config/assert.h"
 
 //--------------------------------------------------------------------------
 // class SCC : A simple representation of an SCC in a generic Graph.


Index: llvm/include/Support/Timer.h
diff -u llvm/include/Support/Timer.h:1.6 llvm/include/Support/Timer.h:1.6.2.1
--- llvm/include/Support/Timer.h:1.6	Wed Jun 11 09:01:21 2003
+++ llvm/include/Support/Timer.h	Mon Jun 23 13:50:38 2003
@@ -29,7 +29,7 @@
 #include <vector>
 #include <iosfwd>
 
-#include <assert.h>
+#include "Config/assert.h"
 
 class TimerGroup;
 


Index: llvm/include/Support/Tree.h
diff -u llvm/include/Support/Tree.h:1.3 llvm/include/Support/Tree.h:1.3.2.1
--- llvm/include/Support/Tree.h:1.3	Wed Jun 11 09:01:21 2003
+++ llvm/include/Support/Tree.h	Mon Jun 23 13:50:38 2003
@@ -10,7 +10,7 @@
 
 #include <vector>
 
-#include <assert.h>
+#include "Config/assert.h"
 
 template<class ConcreteTreeNode, class Payload>
 class Tree {


Index: llvm/include/Support/hash_map
diff -u llvm/include/Support/hash_map:1.6 llvm/include/Support/hash_map:1.6.2.1
--- llvm/include/Support/hash_map:1.6	Fri Jan 31 22:14:28 2003
+++ llvm/include/Support/hash_map	Mon Jun 23 13:50:38 2003
@@ -17,23 +17,25 @@
 //  3.0.4       std      ext/hash_map
 //  3.1      __gnu_cxx   ext/hash_map
 //
-#if __GNUC__ == 3
-#include <ext/hash_map>
 
-#ifndef HASH_NAMESPACE
-#if __GNUC_MINOR__ == 0
-#define HASH_NAMESPACE std
-#else
+#include "Config/config.h"
+
+#ifdef HAVE_GNU_EXT_HASH_MAP
+
+#include <ext/hash_map>
 #define HASH_NAMESPACE __gnu_cxx
-#endif
-#endif
 
 #else
 
-#include <hash_map>
-#ifndef HASH_NAMESPACE
+#ifdef HAVE_STD_EXT_HASH_MAP
+#include <ext/hash_map>
 #define HASH_NAMESPACE std
+
+#else
+#include <hash_map>
+#define HASH_NAMESPACE
 #endif
+
 #endif
 
 using HASH_NAMESPACE::hash_map;


Index: llvm/include/Support/hash_set
diff -u llvm/include/Support/hash_set:1.6 llvm/include/Support/hash_set:1.6.2.1
--- llvm/include/Support/hash_set:1.6	Fri Nov  8 08:07:33 2002
+++ llvm/include/Support/hash_set	Mon Jun 23 13:50:38 2003
@@ -17,23 +17,25 @@
 //  3.0.4       std      ext/hash_set
 //  3.1      __gnu_cxx   ext/hash_set
 //
-#if __GNUC__==3
-#include <ext/hash_set>
 
-#ifndef HASH_NAMESPACE
-#if __GNUC_MINOR__ == 0
-#define HASH_NAMESPACE std
-#else
+#include "Config/config.h"
+
+#ifdef HAVE_GNU_EXT_HASH_SET
+
+#include <ext/hash_set>
 #define HASH_NAMESPACE __gnu_cxx
-#endif
-#endif
 
 #else
 
-#include <hash_set>
-#ifndef HASH_NAMESPACE
+#ifdef HAVE_STD_EXT_HASH_SET
+#include <ext/hash_set>
 #define HASH_NAMESPACE std
+
+#else
+#include <hash_set>
+#define HASH_NAMESPACE
 #endif
+
 #endif
 
 using HASH_NAMESPACE::hash_set;


Index: llvm/include/Support/ilist
diff -u llvm/include/Support/ilist:1.6 llvm/include/Support/ilist:1.6.2.1
--- llvm/include/Support/ilist:1.6	Wed Apr 23 11:17:53 2003
+++ llvm/include/Support/ilist	Mon Jun 23 13:50:38 2003
@@ -31,7 +31,7 @@
 #ifndef INCLUDED_SUPPORT_ILIST
 #define INCLUDED_SUPPORT_ILIST
 
-#include <assert.h>
+#include "Config/assert.h"
 #include <algorithm>
 #include <Support/iterator>
 


Index: llvm/include/Support/iterator
diff -u llvm/include/Support/iterator:1.2 llvm/include/Support/iterator:1.2.2.1
--- llvm/include/Support/iterator:1.2	Wed Jul 24 17:07:13 2002
+++ llvm/include/Support/iterator	Mon Jun 23 13:50:38 2003
@@ -19,9 +19,11 @@
 #ifndef SUPPORT_ITERATOR_H
 #define SUPPORT_ITERATOR_H
 
+#include "Config/config.h"
+
 #include <iterator>
 
-#if __GNUC__ == 3
+#ifdef HAVE_STD_ITERATOR
 
 // Define stupid wrappers around std::iterator...
 template<class Ty, class PtrDiffTy>
@@ -35,6 +37,7 @@
 };
 
 #else
+
 // Just use bidirectional_iterator directly.
 using std::bidirectional_iterator;
 using std::forward_iterator;


Index: llvm/include/Support/slist
diff -u llvm/include/Support/slist:1.2 llvm/include/Support/slist:1.2.2.1
--- llvm/include/Support/slist:1.2	Sun Feb 23 22:40:35 2003
+++ llvm/include/Support/slist	Mon Jun 23 13:50:38 2003
@@ -10,6 +10,8 @@
 #ifndef SUPPORT_SLIST_H
 #define SUPPORT_SLIST_H
 
+#include "Config/config.h"
+
 // Compiler Support Matrix
 //
 // Version   Namespace   Header File
@@ -17,18 +19,19 @@
 //  3.0.4       std      ext/slist
 //  3.1      __gnu_cxx   ext/slist
 //
-#if __GNUC__ == 3
+
+#ifdef HAVE_EXT_SLIST
 #include <ext/slist>
+#else
+#include <slist>
+#endif
 
-#if __GNUC_MINOR__ == 0
+#if HAVE_EXT_SLIST == std
 using std::slist;
-#else
-using __gnu_cxx::slist;
 #endif
 
-#else
-// GCC 2.x
-#include <slist>
+#if HAVE_EXT_SLIST == gnu
+using __gnu_cxx::slist;
 #endif
 
 #endif




From criswell at choi.cs.uiuc.edu  Mon Jun 23 13:53:02 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 13:53:02 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Analysis/AliasSetTracker.h CallGraph.h ConstantsScanner.h DSGraph.h DSGraphTraits.h DSNode.h DSSupport.h DataStructure.h DependenceGraph.h Dominators.h IPModRef.h InstForest.h IntervalIterator.h MemoryDepAnalysis.h PgmDependenceGraph.h
Message-ID: <200306231852.h5NIqRM02108@choi.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Analysis:

AliasSetTracker.h updated: 1.5 -> 1.5.2.1
CallGraph.h updated: 1.27 -> 1.27.2.1
ConstantsScanner.h updated: 1.10 -> 1.10.2.1
DSGraph.h updated: 1.48 -> 1.48.2.1
DSGraphTraits.h updated: 1.14 -> 1.14.2.1
DSNode.h updated: 1.24 -> 1.24.2.1
DSSupport.h updated: 1.16 -> 1.16.2.1
DataStructure.h updated: 1.63 -> 1.63.2.1
DependenceGraph.h updated: 1.4 -> 1.4.2.1
Dominators.h updated: 1.33 -> 1.33.2.1
IPModRef.h updated: 1.10 -> 1.10.2.1
InstForest.h updated: 1.17 -> 1.17.2.1
IntervalIterator.h updated: 1.11 -> 1.11.2.1
MemoryDepAnalysis.h updated: 1.2 -> 1.2.2.1
PgmDependenceGraph.h updated: 1.2 -> 1.2.2.1

---
Log message:

Initial checkin of autoconf code.


---
Diffs of the changes:

Index: llvm/include/llvm/Analysis/AliasSetTracker.h
diff -u llvm/include/llvm/Analysis/AliasSetTracker.h:1.5 llvm/include/llvm/Analysis/AliasSetTracker.h:1.5.2.1
--- llvm/include/llvm/Analysis/AliasSetTracker.h:1.5	Wed Jun 11 09:01:29 2003
+++ llvm/include/llvm/Analysis/AliasSetTracker.h	Mon Jun 23 13:51:46 2003
@@ -10,7 +10,7 @@
 #ifndef LLVM_ANALYSIS_ALIASSETTRACKER_H
 #define LLVM_ANALYSIS_ALIASSETTRACKER_H
 
-#include <assert.h>
+#include "Config/assert.h"
 
 #include "llvm/Support/CallSite.h"
 #include "Support/iterator"


Index: llvm/include/llvm/Analysis/CallGraph.h
diff -u llvm/include/llvm/Analysis/CallGraph.h:1.27 llvm/include/llvm/Analysis/CallGraph.h:1.27.2.1
--- llvm/include/llvm/Analysis/CallGraph.h:1.27	Wed Jun 11 09:01:29 2003
+++ llvm/include/llvm/Analysis/CallGraph.h	Mon Jun 23 13:51:47 2003
@@ -41,7 +41,7 @@
 #ifndef LLVM_ANALYSIS_CALLGRAPH_H
 #define LLVM_ANALYSIS_CALLGRAPH_H
 
-#include <assert.h>
+#include "Config/assert.h"
 
 #include "Support/GraphTraits.h"
 #include "Support/STLExtras.h"


Index: llvm/include/llvm/Analysis/ConstantsScanner.h
diff -u llvm/include/llvm/Analysis/ConstantsScanner.h:1.10 llvm/include/llvm/Analysis/ConstantsScanner.h:1.10.2.1
--- llvm/include/llvm/Analysis/ConstantsScanner.h:1.10	Wed Jun 11 09:01:29 2003
+++ llvm/include/llvm/Analysis/ConstantsScanner.h	Mon Jun 23 13:51:47 2003
@@ -9,7 +9,7 @@
 #ifndef LLVM_ANALYSIS_CONSTANTSSCANNER_H
 #define LLVM_ANALYSIS_CONSTANTSSCANNER_H
 
-#include <assert.h>
+#include "Config/assert.h"
 
 #include "llvm/Support/InstIterator.h"
 #include "llvm/Instruction.h"


Index: llvm/include/llvm/Analysis/DSGraph.h
diff -u llvm/include/llvm/Analysis/DSGraph.h:1.48 llvm/include/llvm/Analysis/DSGraph.h:1.48.2.1
--- llvm/include/llvm/Analysis/DSGraph.h:1.48	Wed Jun 11 09:01:29 2003
+++ llvm/include/llvm/Analysis/DSGraph.h	Mon Jun 23 13:51:47 2003
@@ -7,7 +7,7 @@
 #ifndef LLVM_ANALYSIS_DSGRAPH_H
 #define LLVM_ANALYSIS_DSGRAPH_H
 
-#include <assert.h>
+#include "Config/assert.h"
 
 #include "llvm/Analysis/DSNode.h"
 


Index: llvm/include/llvm/Analysis/DSGraphTraits.h
diff -u llvm/include/llvm/Analysis/DSGraphTraits.h:1.14 llvm/include/llvm/Analysis/DSGraphTraits.h:1.14.2.1
--- llvm/include/llvm/Analysis/DSGraphTraits.h:1.14	Wed Jun 11 09:01:29 2003
+++ llvm/include/llvm/Analysis/DSGraphTraits.h	Mon Jun 23 13:51:47 2003
@@ -9,7 +9,7 @@
 #ifndef LLVM_ANALYSIS_DSGRAPHTRAITS_H
 #define LLVM_ANALYSIS_DSGRAPHTRAITS_H
 
-#include <assert.h>
+#include "Config/assert.h"
 
 #include "llvm/Analysis/DSGraph.h"
 #include "Support/GraphTraits.h"


Index: llvm/include/llvm/Analysis/DSNode.h
diff -u llvm/include/llvm/Analysis/DSNode.h:1.24 llvm/include/llvm/Analysis/DSNode.h:1.24.2.1
--- llvm/include/llvm/Analysis/DSNode.h:1.24	Wed Jun 11 09:01:29 2003
+++ llvm/include/llvm/Analysis/DSNode.h	Mon Jun 23 13:51:47 2003
@@ -7,7 +7,7 @@
 #ifndef LLVM_ANALYSIS_DSNODE_H
 #define LLVM_ANALYSIS_DSNODE_H
 
-#include <assert.h>
+#include "Config/assert.h"
 
 #include "llvm/Analysis/DSSupport.h"
 template<typename BaseType>


Index: llvm/include/llvm/Analysis/DSSupport.h
diff -u llvm/include/llvm/Analysis/DSSupport.h:1.16 llvm/include/llvm/Analysis/DSSupport.h:1.16.2.1
--- llvm/include/llvm/Analysis/DSSupport.h:1.16	Wed Jun 11 09:01:29 2003
+++ llvm/include/llvm/Analysis/DSSupport.h	Mon Jun 23 13:51:47 2003
@@ -7,7 +7,7 @@
 #ifndef LLVM_ANALYSIS_DSSUPPORT_H
 #define LLVM_ANALYSIS_DSSUPPORT_H
 
-#include <assert.h>
+#include "Config/assert.h"
 
 #include <vector>
 #include <functional>


Index: llvm/include/llvm/Analysis/DataStructure.h
diff -u llvm/include/llvm/Analysis/DataStructure.h:1.63 llvm/include/llvm/Analysis/DataStructure.h:1.63.2.1
--- llvm/include/llvm/Analysis/DataStructure.h:1.63	Wed Jun 11 09:01:29 2003
+++ llvm/include/llvm/Analysis/DataStructure.h	Mon Jun 23 13:51:47 2003
@@ -7,7 +7,7 @@
 #ifndef LLVM_ANALYSIS_DATA_STRUCTURE_H
 #define LLVM_ANALYSIS_DATA_STRUCTURE_H
 
-#include <assert.h>
+#include "Config/assert.h"
 
 #include "llvm/Pass.h"
 #include "Support/HashExtras.h"


Index: llvm/include/llvm/Analysis/DependenceGraph.h
diff -u llvm/include/llvm/Analysis/DependenceGraph.h:1.4 llvm/include/llvm/Analysis/DependenceGraph.h:1.4.2.1
--- llvm/include/llvm/Analysis/DependenceGraph.h:1.4	Wed Jun 11 09:01:29 2003
+++ llvm/include/llvm/Analysis/DependenceGraph.h	Mon Jun 23 13:51:47 2003
@@ -23,7 +23,7 @@
 #include <vector>
 #include <utility>
 
-#include <assert.h>
+#include "Config/assert.h"
 
 class Instruction;
 class Function;


Index: llvm/include/llvm/Analysis/Dominators.h
diff -u llvm/include/llvm/Analysis/Dominators.h:1.33 llvm/include/llvm/Analysis/Dominators.h:1.33.2.1
--- llvm/include/llvm/Analysis/Dominators.h:1.33	Wed Jun 11 09:01:29 2003
+++ llvm/include/llvm/Analysis/Dominators.h	Mon Jun 23 13:51:47 2003
@@ -20,7 +20,7 @@
 
 #include "llvm/Pass.h"
 #include <set>
-#include <assert.h>
+#include "Config/assert.h"
 class Instruction;
 
 template <typename GraphType> struct GraphTraits;


Index: llvm/include/llvm/Analysis/IPModRef.h
diff -u llvm/include/llvm/Analysis/IPModRef.h:1.10 llvm/include/llvm/Analysis/IPModRef.h:1.10.2.1
--- llvm/include/llvm/Analysis/IPModRef.h:1.10	Wed Jun 11 09:01:29 2003
+++ llvm/include/llvm/Analysis/IPModRef.h	Mon Jun 23 13:51:47 2003
@@ -39,7 +39,7 @@
 #ifndef LLVM_ANALYSIS_IPMODREF_H
 #define LLVM_ANALYSIS_IPMODREF_H
 
-#include <assert.h>
+#include "Config/assert.h"
 
 #include "llvm/Pass.h"
 #include "Support/BitSetVector.h"


Index: llvm/include/llvm/Analysis/InstForest.h
diff -u llvm/include/llvm/Analysis/InstForest.h:1.17 llvm/include/llvm/Analysis/InstForest.h:1.17.2.1
--- llvm/include/llvm/Analysis/InstForest.h:1.17	Wed Jun 11 09:01:29 2003
+++ llvm/include/llvm/Analysis/InstForest.h	Mon Jun 23 13:51:47 2003
@@ -19,7 +19,7 @@
 #include "llvm/Function.h"
 #include "Support/Tree.h"
 #include <map>
-#include <assert.h>
+#include "Config/assert.h"
 
 template<class Payload> class InstTreeNode;
 template<class Payload> class InstForest;


Index: llvm/include/llvm/Analysis/IntervalIterator.h
diff -u llvm/include/llvm/Analysis/IntervalIterator.h:1.11 llvm/include/llvm/Analysis/IntervalIterator.h:1.11.2.1
--- llvm/include/llvm/Analysis/IntervalIterator.h:1.11	Wed Jun 11 09:01:29 2003
+++ llvm/include/llvm/Analysis/IntervalIterator.h	Mon Jun 23 13:51:47 2003
@@ -33,7 +33,7 @@
 #include <stack>
 #include <set>
 #include <algorithm>
-#include <assert.h>
+#include "Config/assert.h"
 
 // getNodeHeader - Given a source graph node and the source graph, return the 
 // BasicBlock that is the header node.  This is the opposite of


Index: llvm/include/llvm/Analysis/MemoryDepAnalysis.h
diff -u llvm/include/llvm/Analysis/MemoryDepAnalysis.h:1.2 llvm/include/llvm/Analysis/MemoryDepAnalysis.h:1.2.2.1
--- llvm/include/llvm/Analysis/MemoryDepAnalysis.h:1.2	Wed Jun 11 09:01:29 2003
+++ llvm/include/llvm/Analysis/MemoryDepAnalysis.h	Mon Jun 23 13:51:47 2003
@@ -20,7 +20,7 @@
 #include "Support/NonCopyable.h"
 #include "Support/hash_map"
 
-#include <assert.h>
+#include "Config/assert.h"
 
 class Instruction;
 class Function;


Index: llvm/include/llvm/Analysis/PgmDependenceGraph.h
diff -u llvm/include/llvm/Analysis/PgmDependenceGraph.h:1.2 llvm/include/llvm/Analysis/PgmDependenceGraph.h:1.2.2.1
--- llvm/include/llvm/Analysis/PgmDependenceGraph.h:1.2	Wed Jun 11 09:01:29 2003
+++ llvm/include/llvm/Analysis/PgmDependenceGraph.h	Mon Jun 23 13:51:47 2003
@@ -40,7 +40,7 @@
 #include "Support/NonCopyable.h"
 #include <iterator>
 
-#include <assert.h>
+#include "Config/assert.h"
 
 class Instruction;
 class Function;




From criswell at choi.cs.uiuc.edu  Mon Jun 23 13:53:04 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 13:53:04 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/AbstractTypeUser.h Constant.h ConstantHandling.h Constants.h DerivedTypes.h GlobalVariable.h Instruction.h Pass.h PassAnalysisSupport.h PassSupport.h SymbolTable.h User.h iOther.h iPHINode.h iTerminators.h
Message-ID: <200306231852.h5NIqK402072@choi.cs.uiuc.edu>


Changes in directory llvm/include/llvm:

AbstractTypeUser.h updated: 1.6 -> 1.6.2.1
Constant.h updated: 1.7 -> 1.7.2.1
ConstantHandling.h updated: 1.26 -> 1.26.2.1
Constants.h updated: 1.26 -> 1.26.2.1
DerivedTypes.h updated: 1.29 -> 1.29.2.1
GlobalVariable.h updated: 1.20 -> 1.20.2.1
Instruction.h updated: 1.40 -> 1.40.2.1
Pass.h updated: 1.33 -> 1.33.2.1
PassAnalysisSupport.h updated: 1.12 -> 1.12.2.1
PassSupport.h updated: 1.12 -> 1.12.2.1
SymbolTable.h updated: 1.21 -> 1.21.2.1
User.h updated: 1.18 -> 1.18.2.1
iOther.h updated: 1.35 -> 1.35.2.1
iPHINode.h updated: 1.9 -> 1.9.2.1
iTerminators.h updated: 1.27 -> 1.27.2.1

---
Log message:

Initial checkin of autoconf code.


---
Diffs of the changes:

Index: llvm/include/llvm/AbstractTypeUser.h
diff -u llvm/include/llvm/AbstractTypeUser.h:1.6 llvm/include/llvm/AbstractTypeUser.h:1.6.2.1
--- llvm/include/llvm/AbstractTypeUser.h:1.6	Tue Jan 14 15:29:52 2003
+++ llvm/include/llvm/AbstractTypeUser.h	Mon Jun 23 13:51:40 2003
@@ -21,7 +21,7 @@
 #ifndef LLVM_ABSTRACT_TYPE_USER_H
 #define LLVM_ABSTRACT_TYPE_USER_H
 
-#include <assert.h>
+#include "Config/assert.h"
 class Type;
 class DerivedType;
 


Index: llvm/include/llvm/Constant.h
diff -u llvm/include/llvm/Constant.h:1.7 llvm/include/llvm/Constant.h:1.7.2.1
--- llvm/include/llvm/Constant.h:1.7	Wed Jun 11 09:01:26 2003
+++ llvm/include/llvm/Constant.h	Mon Jun 23 13:51:40 2003
@@ -7,7 +7,7 @@
 #ifndef LLVM_CONSTANT_H
 #define LLVM_CONSTANT_H
 
-#include <assert.h>
+#include "Config/assert.h"
 #include "llvm/User.h"
 
 class Constant : public User {


Index: llvm/include/llvm/ConstantHandling.h
diff -u llvm/include/llvm/ConstantHandling.h:1.26 llvm/include/llvm/ConstantHandling.h:1.26.2.1
--- llvm/include/llvm/ConstantHandling.h:1.26	Wed Jun 11 09:01:26 2003
+++ llvm/include/llvm/ConstantHandling.h	Mon Jun 23 13:51:40 2003
@@ -33,7 +33,7 @@
 #ifndef LLVM_CONSTANTHANDLING_H
 #define LLVM_CONSTANTHANDLING_H
 
-#include <assert.h>
+#include "Config/assert.h"
 
 #include "llvm/Constants.h"
 #include "llvm/Type.h"


Index: llvm/include/llvm/Constants.h
diff -u llvm/include/llvm/Constants.h:1.26 llvm/include/llvm/Constants.h:1.26.2.1
--- llvm/include/llvm/Constants.h:1.26	Wed Jun 11 09:01:26 2003
+++ llvm/include/llvm/Constants.h	Mon Jun 23 13:51:40 2003
@@ -8,7 +8,7 @@
 #ifndef LLVM_CONSTANTS_H
 #define LLVM_CONSTANTS_H
 
-#include <assert.h>
+#include "Config/assert.h"
 
 #include "llvm/Constant.h"
 #include "Support/DataTypes.h"


Index: llvm/include/llvm/DerivedTypes.h
diff -u llvm/include/llvm/DerivedTypes.h:1.29 llvm/include/llvm/DerivedTypes.h:1.29.2.1
--- llvm/include/llvm/DerivedTypes.h:1.29	Wed Jun 11 09:01:26 2003
+++ llvm/include/llvm/DerivedTypes.h	Mon Jun 23 13:51:40 2003
@@ -11,7 +11,7 @@
 #ifndef LLVM_DERIVED_TYPES_H
 #define LLVM_DERIVED_TYPES_H
 
-#include <assert.h>
+#include "Config/assert.h"
 
 #include "llvm/Type.h"
 


Index: llvm/include/llvm/GlobalVariable.h
diff -u llvm/include/llvm/GlobalVariable.h:1.20 llvm/include/llvm/GlobalVariable.h:1.20.2.1
--- llvm/include/llvm/GlobalVariable.h:1.20	Wed Jun 11 09:01:26 2003
+++ llvm/include/llvm/GlobalVariable.h	Mon Jun 23 13:51:40 2003
@@ -13,7 +13,7 @@
 #ifndef LLVM_GLOBAL_VARIABLE_H
 #define LLVM_GLOBAL_VARIABLE_H
 
-#include <assert.h>
+#include "Config/assert.h"
 
 #include "llvm/GlobalValue.h"
 class Module;


Index: llvm/include/llvm/Instruction.h
diff -u llvm/include/llvm/Instruction.h:1.40 llvm/include/llvm/Instruction.h:1.40.2.1
--- llvm/include/llvm/Instruction.h:1.40	Wed Jun 11 09:01:26 2003
+++ llvm/include/llvm/Instruction.h	Mon Jun 23 13:51:40 2003
@@ -8,7 +8,7 @@
 #ifndef LLVM_INSTRUCTION_H
 #define LLVM_INSTRUCTION_H
 
-#include <assert.h>
+#include "Config/assert.h"
 
 #include "llvm/User.h"
 template<typename SC> struct ilist_traits;


Index: llvm/include/llvm/Pass.h
diff -u llvm/include/llvm/Pass.h:1.33 llvm/include/llvm/Pass.h:1.33.2.1
--- llvm/include/llvm/Pass.h:1.33	Wed Jun 11 09:01:26 2003
+++ llvm/include/llvm/Pass.h	Mon Jun 23 13:51:40 2003
@@ -22,7 +22,7 @@
 #ifndef LLVM_PASS_H
 #define LLVM_PASS_H
 
-#include <assert.h>
+#include "Config/assert.h"
 
 #include <vector>
 #include <map>


Index: llvm/include/llvm/PassAnalysisSupport.h
diff -u llvm/include/llvm/PassAnalysisSupport.h:1.12 llvm/include/llvm/PassAnalysisSupport.h:1.12.2.1
--- llvm/include/llvm/PassAnalysisSupport.h:1.12	Wed Jun 11 09:01:26 2003
+++ llvm/include/llvm/PassAnalysisSupport.h	Mon Jun 23 13:51:40 2003
@@ -14,7 +14,7 @@
 
 // No need to include Pass.h, we are being included by it!
 
-#include <assert.h>
+#include "Config/assert.h"
 
 //===----------------------------------------------------------------------===//
 // AnalysisUsage - Represent the analysis usage information of a pass.  This


Index: llvm/include/llvm/PassSupport.h
diff -u llvm/include/llvm/PassSupport.h:1.12 llvm/include/llvm/PassSupport.h:1.12.2.1
--- llvm/include/llvm/PassSupport.h:1.12	Wed Jun 11 09:01:26 2003
+++ llvm/include/llvm/PassSupport.h	Mon Jun 23 13:51:40 2003
@@ -14,7 +14,7 @@
 #ifndef LLVM_PASS_SUPPORT_H
 #define LLVM_PASS_SUPPORT_H
 
-#include <assert.h>
+#include "Config/assert.h"
 
 // No need to include Pass.h, we are being included by it!
 


Index: llvm/include/llvm/SymbolTable.h
diff -u llvm/include/llvm/SymbolTable.h:1.21 llvm/include/llvm/SymbolTable.h:1.21.2.1
--- llvm/include/llvm/SymbolTable.h:1.21	Wed Jun 11 09:01:26 2003
+++ llvm/include/llvm/SymbolTable.h	Mon Jun 23 13:51:40 2003
@@ -16,7 +16,7 @@
 #ifndef LLVM_SYMBOL_TABLE_H
 #define LLVM_SYMBOL_TABLE_H
 
-#include <assert.h>
+#include "Config/assert.h"
 
 #include "llvm/Value.h"
 #include <map>


Index: llvm/include/llvm/User.h
diff -u llvm/include/llvm/User.h:1.18 llvm/include/llvm/User.h:1.18.2.1
--- llvm/include/llvm/User.h:1.18	Wed Jun 11 09:01:26 2003
+++ llvm/include/llvm/User.h	Mon Jun 23 13:51:40 2003
@@ -12,7 +12,7 @@
 #ifndef LLVM_USER_H
 #define LLVM_USER_H
 
-#include <assert.h>
+#include "Config/assert.h"
 
 #include "llvm/Value.h"
 


Index: llvm/include/llvm/iOther.h
diff -u llvm/include/llvm/iOther.h:1.35 llvm/include/llvm/iOther.h:1.35.2.1
--- llvm/include/llvm/iOther.h:1.35	Wed Jun 11 09:01:26 2003
+++ llvm/include/llvm/iOther.h	Mon Jun 23 13:51:40 2003
@@ -8,7 +8,7 @@
 #ifndef LLVM_IOTHER_H
 #define LLVM_IOTHER_H
 
-#include <assert.h>
+#include "Config/assert.h"
 
 #include "llvm/InstrTypes.h"
 


Index: llvm/include/llvm/iPHINode.h
diff -u llvm/include/llvm/iPHINode.h:1.9 llvm/include/llvm/iPHINode.h:1.9.2.1
--- llvm/include/llvm/iPHINode.h:1.9	Wed Jun 11 09:01:26 2003
+++ llvm/include/llvm/iPHINode.h	Mon Jun 23 13:51:40 2003
@@ -7,7 +7,7 @@
 #ifndef LLVM_IPHINODE_H
 #define LLVM_IPHINODE_H
 
-#include <assert.h>
+#include "Config/assert.h"
 
 #include "llvm/Instruction.h"
 class BasicBlock;


Index: llvm/include/llvm/iTerminators.h
diff -u llvm/include/llvm/iTerminators.h:1.27 llvm/include/llvm/iTerminators.h:1.27.2.1
--- llvm/include/llvm/iTerminators.h:1.27	Wed Jun 11 09:01:26 2003
+++ llvm/include/llvm/iTerminators.h	Mon Jun 23 13:51:40 2003
@@ -9,7 +9,7 @@
 #ifndef LLVM_ITERMINATORS_H
 #define LLVM_ITERMINATORS_H
 
-#include <assert.h>
+#include "Config/assert.h"
 
 #include "llvm/InstrTypes.h"
 




From criswell at choi.cs.uiuc.edu  Mon Jun 23 13:53:05 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 13:53:05 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Target/MRegisterInfo.h TargetCacheInfo.h TargetInstrInfo.h TargetRegInfo.h TargetSchedInfo.h
Message-ID: <200306231852.h5NIq9M02038@choi.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Target:

MRegisterInfo.h updated: 1.16 -> 1.16.2.1
TargetCacheInfo.h updated: 1.9 -> 1.9.2.1
TargetInstrInfo.h updated: 1.45 -> 1.45.2.1
TargetRegInfo.h updated: 1.38 -> 1.38.2.1
TargetSchedInfo.h updated: 1.17 -> 1.17.2.1

---
Log message:

Initial checkin of autoconf code.


---
Diffs of the changes:

Index: llvm/include/llvm/Target/MRegisterInfo.h
diff -u llvm/include/llvm/Target/MRegisterInfo.h:1.16 llvm/include/llvm/Target/MRegisterInfo.h:1.16.2.1
--- llvm/include/llvm/Target/MRegisterInfo.h:1.16	Sun Jan 12 18:19:44 2003
+++ llvm/include/llvm/Target/MRegisterInfo.h	Mon Jun 23 13:51:59 2003
@@ -10,7 +10,7 @@
 #define LLVM_TARGET_MREGISTERINFO_H
 
 #include "llvm/CodeGen/MachineBasicBlock.h"
-#include <assert.h>
+#include "Config/assert.h"
 
 class Type;
 class MachineFunction;


Index: llvm/include/llvm/Target/TargetCacheInfo.h
diff -u llvm/include/llvm/Target/TargetCacheInfo.h:1.9 llvm/include/llvm/Target/TargetCacheInfo.h:1.9.2.1
--- llvm/include/llvm/Target/TargetCacheInfo.h:1.9	Wed Jun 11 09:01:36 2003
+++ llvm/include/llvm/Target/TargetCacheInfo.h	Mon Jun 23 13:51:59 2003
@@ -8,7 +8,7 @@
 #define LLVM_TARGET_TARGETCACHEINFO_H
 
 #include "Support/DataTypes.h"
-#include <assert.h>
+#include "Config/assert.h"
 
 class TargetMachine;
 


Index: llvm/include/llvm/Target/TargetInstrInfo.h
diff -u llvm/include/llvm/Target/TargetInstrInfo.h:1.45 llvm/include/llvm/Target/TargetInstrInfo.h:1.45.2.1
--- llvm/include/llvm/Target/TargetInstrInfo.h:1.45	Wed Jun 11 09:01:36 2003
+++ llvm/include/llvm/Target/TargetInstrInfo.h	Mon Jun 23 13:51:59 2003
@@ -9,7 +9,7 @@
 
 #include "Support/DataTypes.h"
 #include <vector>
-#include <assert.h>
+#include "Config/assert.h"
 
 class MachineInstr;
 class TargetMachine;


Index: llvm/include/llvm/Target/TargetRegInfo.h
diff -u llvm/include/llvm/Target/TargetRegInfo.h:1.38 llvm/include/llvm/Target/TargetRegInfo.h:1.38.2.1
--- llvm/include/llvm/Target/TargetRegInfo.h:1.38	Wed Jun 11 09:01:36 2003
+++ llvm/include/llvm/Target/TargetRegInfo.h	Mon Jun 23 13:51:59 2003
@@ -10,7 +10,7 @@
 
 #include "Support/hash_map"
 #include <string>
-#include <assert.h>
+#include "Config/assert.h"
 
 class TargetMachine;
 class IGNode;


Index: llvm/include/llvm/Target/TargetSchedInfo.h
diff -u llvm/include/llvm/Target/TargetSchedInfo.h:1.17 llvm/include/llvm/Target/TargetSchedInfo.h:1.17.2.1
--- llvm/include/llvm/Target/TargetSchedInfo.h:1.17	Wed Jun 11 09:01:36 2003
+++ llvm/include/llvm/Target/TargetSchedInfo.h	Mon Jun 23 13:51:59 2003
@@ -10,7 +10,7 @@
 #include "llvm/Target/TargetInstrInfo.h"
 #include "Support/hash_map"
 #include <string>
-#include <assert.h>
+#include "Config/assert.h"
 
 typedef long long cycles_t; 
 static const cycles_t HUGE_LATENCY = ~((long long) 1 << (sizeof(cycles_t)-2));




From criswell at choi.cs.uiuc.edu  Mon Jun 23 13:53:07 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 13:53:07 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Support/CFG.h InstVisitor.h PassNameParser.h
Message-ID: <200306231852.h5NIq7w02022@choi.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Support:

CFG.h updated: 1.11 -> 1.11.2.1
InstVisitor.h updated: 1.17 -> 1.17.2.1
PassNameParser.h updated: 1.6 -> 1.6.2.1

---
Log message:

Initial checkin of autoconf code.


---
Diffs of the changes:

Index: llvm/include/llvm/Support/CFG.h
diff -u llvm/include/llvm/Support/CFG.h:1.11 llvm/include/llvm/Support/CFG.h:1.11.2.1
--- llvm/include/llvm/Support/CFG.h:1.11	Wed Jun 11 09:01:35 2003
+++ llvm/include/llvm/Support/CFG.h	Mon Jun 23 13:51:57 2003
@@ -13,7 +13,7 @@
 #include "llvm/InstrTypes.h"
 #include "Support/iterator"
 
-#include <assert.h>
+#include "Config/assert.h"
 
 //===--------------------------------------------------------------------===//
 // BasicBlock pred_iterator definition


Index: llvm/include/llvm/Support/InstVisitor.h
diff -u llvm/include/llvm/Support/InstVisitor.h:1.17 llvm/include/llvm/Support/InstVisitor.h:1.17.2.1
--- llvm/include/llvm/Support/InstVisitor.h:1.17	Wed Jun 11 09:01:35 2003
+++ llvm/include/llvm/Support/InstVisitor.h	Mon Jun 23 13:51:57 2003
@@ -44,7 +44,7 @@
 #define LLVM_SUPPORT_INSTVISITOR_H
 
 #include "llvm/Instruction.h"
-#include <assert.h>
+#include "Config/assert.h"
 
 class Module;
 


Index: llvm/include/llvm/Support/PassNameParser.h
diff -u llvm/include/llvm/Support/PassNameParser.h:1.6 llvm/include/llvm/Support/PassNameParser.h:1.6.2.1
--- llvm/include/llvm/Support/PassNameParser.h:1.6	Wed Jun 11 09:01:35 2003
+++ llvm/include/llvm/Support/PassNameParser.h	Mon Jun 23 13:51:57 2003
@@ -20,7 +20,7 @@
 #include "llvm/Pass.h"
 #include <algorithm>
 #include <iostream>
-#include <assert.h>
+#include "Config/assert.h"
 
 //===----------------------------------------------------------------------===//
 // PassNameParser class - Make use of the pass registration mechanism to




From criswell at choi.cs.uiuc.edu  Mon Jun 23 13:53:08 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 13:53:08 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Reoptimizer/BinInterface/regmask.h sparcbin.h
Message-ID: <200306231852.h5NIq5w02010@choi.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Reoptimizer/BinInterface:

regmask.h updated: 1.2 -> 1.2.2.1
sparcbin.h updated: 1.3 -> 1.3.2.1

---
Log message:

Initial checkin of autoconf code.


---
Diffs of the changes:

Index: llvm/include/llvm/Reoptimizer/BinInterface/regmask.h
diff -u llvm/include/llvm/Reoptimizer/BinInterface/regmask.h:1.2 llvm/include/llvm/Reoptimizer/BinInterface/regmask.h:1.2.2.1
--- llvm/include/llvm/Reoptimizer/BinInterface/regmask.h:1.2	Wed Jun 11 09:01:34 2003
+++ llvm/include/llvm/Reoptimizer/BinInterface/regmask.h	Mon Jun 23 13:51:55 2003
@@ -11,7 +11,7 @@
 #include "bitmath.h"
 #include <vector>
 #include <iostream>
-#include <assert.h>
+#include "Config/assert.h"
 
 #define VREG_ISMEM(x) (x >= 32)
 #define VREG_ISREG(x) (x <  32)


Index: llvm/include/llvm/Reoptimizer/BinInterface/sparcbin.h
diff -u llvm/include/llvm/Reoptimizer/BinInterface/sparcbin.h:1.3 llvm/include/llvm/Reoptimizer/BinInterface/sparcbin.h:1.3.2.1
--- llvm/include/llvm/Reoptimizer/BinInterface/sparcbin.h:1.3	Wed Jun 11 09:01:34 2003
+++ llvm/include/llvm/Reoptimizer/BinInterface/sparcbin.h	Mon Jun 23 13:51:55 2003
@@ -18,7 +18,7 @@
 #include "sparcpriv.h"
 #include <map>
 #include <vector>
-#include <assert.h>
+#include "Config/assert.h"
 
 using std::vector;
 




From criswell at choi.cs.uiuc.edu  Mon Jun 23 13:53:09 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 13:53:09 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Reoptimizer/TraceCache.h VirtualMem.h
Message-ID: <200306231852.h5NIq2102000@choi.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Reoptimizer:

TraceCache.h updated: 1.10 -> 1.10.2.1
VirtualMem.h updated: 1.7 -> 1.7.2.1

---
Log message:

Initial checkin of autoconf code.


---
Diffs of the changes:

Index: llvm/include/llvm/Reoptimizer/TraceCache.h
diff -u llvm/include/llvm/Reoptimizer/TraceCache.h:1.10 llvm/include/llvm/Reoptimizer/TraceCache.h:1.10.2.1
--- llvm/include/llvm/Reoptimizer/TraceCache.h:1.10	Wed Jun 11 09:01:32 2003
+++ llvm/include/llvm/Reoptimizer/TraceCache.h	Mon Jun 23 13:51:52 2003
@@ -11,7 +11,7 @@
 #include <vector>
 #include <list>
 #include <utility>
-#include <assert.h>
+#include "Config/assert.h"
 
 class VirtualMem;
 class MemoryManager;


Index: llvm/include/llvm/Reoptimizer/VirtualMem.h
diff -u llvm/include/llvm/Reoptimizer/VirtualMem.h:1.7 llvm/include/llvm/Reoptimizer/VirtualMem.h:1.7.2.1
--- llvm/include/llvm/Reoptimizer/VirtualMem.h:1.7	Sat May 31 21:31:24 2003
+++ llvm/include/llvm/Reoptimizer/VirtualMem.h	Mon Jun 23 13:51:52 2003
@@ -8,10 +8,11 @@
 #define LLVM_REOPTIMIZER_TRACECACHE_VIRTUALMEMORY_H
 
 #include "Support/DataTypes.h"
-#include <unistd.h>
-#include <fcntl.h>
-#include <stdio.h>
-#include <assert.h>
+#include "Config/fcntl.h"
+#include "Config/unistd.h"
+
+#include "Config/stdio.h"
+#include "Config/assert.h"
 #include <iostream>
 
 #include <vector>




From criswell at choi.cs.uiuc.edu  Mon Jun 23 13:53:11 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 13:53:11 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/CodeGen/IGNode.h InstrForest.h InstrSelection.h LiveRange.h LiveVariables.h MachineCodeForInstruction.h MachineFrameInfo.h MachineInstr.h MachineInstrAnnot.h SSARegMap.h
Message-ID: <200306231852.h5NIq0101988@choi.cs.uiuc.edu>


Changes in directory llvm/include/llvm/CodeGen:

IGNode.h updated: 1.12 -> 1.12.2.1
InstrForest.h updated: 1.26 -> 1.26.2.1
InstrSelection.h updated: 1.23 -> 1.23.2.1
LiveRange.h updated: 1.19 -> 1.19.2.1
LiveVariables.h updated: 1.5 -> 1.5.2.1
MachineCodeForInstruction.h updated: 1.9 -> 1.9.2.1
MachineFrameInfo.h updated: 1.5 -> 1.5.2.1
MachineInstr.h updated: 1.103 -> 1.103.2.1
MachineInstrAnnot.h updated: 1.9 -> 1.9.2.1
SSARegMap.h updated: 1.4 -> 1.4.2.1

---
Log message:

Initial checkin of autoconf code.


---
Diffs of the changes:

Index: llvm/include/llvm/CodeGen/IGNode.h
diff -u llvm/include/llvm/CodeGen/IGNode.h:1.12 llvm/include/llvm/CodeGen/IGNode.h:1.12.2.1
--- llvm/include/llvm/CodeGen/IGNode.h:1.12	Wed Jun 11 09:01:30 2003
+++ llvm/include/llvm/CodeGen/IGNode.h	Mon Jun 23 13:51:50 2003
@@ -26,7 +26,7 @@
 #define IG_NODE_H
 
 #include "llvm/CodeGen/LiveRange.h"
-#include <assert.h>
+#include "Config/assert.h"
 class RegClass;
 
 //----------------------------------------------------------------------------


Index: llvm/include/llvm/CodeGen/InstrForest.h
diff -u llvm/include/llvm/CodeGen/InstrForest.h:1.26 llvm/include/llvm/CodeGen/InstrForest.h:1.26.2.1
--- llvm/include/llvm/CodeGen/InstrForest.h:1.26	Wed Jun 11 09:01:30 2003
+++ llvm/include/llvm/CodeGen/InstrForest.h	Mon Jun 23 13:51:50 2003
@@ -21,7 +21,7 @@
 
 #include "llvm/Instruction.h"
 #include "Support/HashExtras.h"
-#include <assert.h>
+#include "Config/assert.h"
 
 class Constant;
 class BasicBlock;


Index: llvm/include/llvm/CodeGen/InstrSelection.h
diff -u llvm/include/llvm/CodeGen/InstrSelection.h:1.23 llvm/include/llvm/CodeGen/InstrSelection.h:1.23.2.1
--- llvm/include/llvm/CodeGen/InstrSelection.h:1.23	Wed Jun 11 09:01:30 2003
+++ llvm/include/llvm/CodeGen/InstrSelection.h	Mon Jun 23 13:51:50 2003
@@ -8,7 +8,7 @@
 #define LLVM_CODEGEN_INSTR_SELECTION_H
 
 #include "llvm/Instruction.h"
-#include <assert.h>
+#include "Config/assert.h"
 class Function;
 class InstrForest;
 class MachineInstr;


Index: llvm/include/llvm/CodeGen/LiveRange.h
diff -u llvm/include/llvm/CodeGen/LiveRange.h:1.19 llvm/include/llvm/CodeGen/LiveRange.h:1.19.2.1
--- llvm/include/llvm/CodeGen/LiveRange.h:1.19	Wed Jun 11 09:01:30 2003
+++ llvm/include/llvm/CodeGen/LiveRange.h	Mon Jun 23 13:51:50 2003
@@ -14,7 +14,7 @@
 #include "llvm/CodeGen/ValueSet.h"
 #include "llvm/Value.h"
 
-#include <assert.h>
+#include "Config/assert.h"
 
 class RegClass;
 class IGNode;


Index: llvm/include/llvm/CodeGen/LiveVariables.h
diff -u llvm/include/llvm/CodeGen/LiveVariables.h:1.5 llvm/include/llvm/CodeGen/LiveVariables.h:1.5.2.1
--- llvm/include/llvm/CodeGen/LiveVariables.h:1.5	Wed Jun 11 09:01:30 2003
+++ llvm/include/llvm/CodeGen/LiveVariables.h	Mon Jun 23 13:51:50 2003
@@ -24,7 +24,7 @@
 
 #include "llvm/CodeGen/MachineFunctionPass.h"
 #include <map>
-#include <assert.h>
+#include "Config/assert.h"
 
 class MRegisterInfo;
 


Index: llvm/include/llvm/CodeGen/MachineCodeForInstruction.h
diff -u llvm/include/llvm/CodeGen/MachineCodeForInstruction.h:1.9 llvm/include/llvm/CodeGen/MachineCodeForInstruction.h:1.9.2.1
--- llvm/include/llvm/CodeGen/MachineCodeForInstruction.h:1.9	Wed Jun 11 09:01:30 2003
+++ llvm/include/llvm/CodeGen/MachineCodeForInstruction.h	Mon Jun 23 13:51:50 2003
@@ -20,7 +20,7 @@
 
 #include "Support/Annotation.h"
 #include <vector>
-#include <assert.h>
+#include "Config/assert.h"
 
 class MachineInstr;
 class Instruction;


Index: llvm/include/llvm/CodeGen/MachineFrameInfo.h
diff -u llvm/include/llvm/CodeGen/MachineFrameInfo.h:1.5 llvm/include/llvm/CodeGen/MachineFrameInfo.h:1.5.2.1
--- llvm/include/llvm/CodeGen/MachineFrameInfo.h:1.5	Wed Jun 11 09:01:30 2003
+++ llvm/include/llvm/CodeGen/MachineFrameInfo.h	Mon Jun 23 13:51:50 2003
@@ -35,7 +35,7 @@
 class TargetRegisterClass;
 class MachineFunction;
 #include <vector>
-#include <assert.h>
+#include "Config/assert.h"
 
 class MachineFrameInfo {
 


Index: llvm/include/llvm/CodeGen/MachineInstr.h
diff -u llvm/include/llvm/CodeGen/MachineInstr.h:1.103 llvm/include/llvm/CodeGen/MachineInstr.h:1.103.2.1
--- llvm/include/llvm/CodeGen/MachineInstr.h:1.103	Wed Jun 11 09:01:30 2003
+++ llvm/include/llvm/CodeGen/MachineInstr.h	Mon Jun 23 13:51:50 2003
@@ -13,7 +13,7 @@
 #include "Support/Annotation.h"
 #include "Support/iterator"
 #include <set>
-#include <assert.h>
+#include "Config/assert.h"
 
 class Value;
 class Function;


Index: llvm/include/llvm/CodeGen/MachineInstrAnnot.h
diff -u llvm/include/llvm/CodeGen/MachineInstrAnnot.h:1.9 llvm/include/llvm/CodeGen/MachineInstrAnnot.h:1.9.2.1
--- llvm/include/llvm/CodeGen/MachineInstrAnnot.h:1.9	Wed Jun 11 09:01:31 2003
+++ llvm/include/llvm/CodeGen/MachineInstrAnnot.h	Mon Jun 23 13:51:50 2003
@@ -10,7 +10,7 @@
 #include "llvm/CodeGen/MachineInstr.h"
 #include "llvm/Target/TargetRegInfo.h"
 
-#include <assert.h>
+#include "Config/assert.h"
 
 class Value;
 class TmpInstruction;


Index: llvm/include/llvm/CodeGen/SSARegMap.h
diff -u llvm/include/llvm/CodeGen/SSARegMap.h:1.4 llvm/include/llvm/CodeGen/SSARegMap.h:1.4.2.1
--- llvm/include/llvm/CodeGen/SSARegMap.h:1.4	Wed Jun 11 09:01:31 2003
+++ llvm/include/llvm/CodeGen/SSARegMap.h	Mon Jun 23 13:51:50 2003
@@ -12,7 +12,7 @@
 
 #include "llvm/Target/MRegisterInfo.h"
 
-#include <assert.h>
+#include "Config/assert.h"
 
 class TargetRegisterClass;
 




From criswell at choi.cs.uiuc.edu  Mon Jun 23 13:55:00 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 13:55:00 2003
Subject: [llvm-commits] CVS: llvm/mkinstalldirs ltmain.sh install-sh aclocal.m4 config.guess config.sub configure.ac
Message-ID: <200306231854.h5NIsXq02144@choi.cs.uiuc.edu>


Changes in directory llvm:

mkinstalldirs added (r1.1.2.1)
ltmain.sh added (r1.1.2.1)
install-sh added (r1.1.2.1)
aclocal.m4 added (r1.1.2.1)
config.guess added (r1.1.2.1)
config.sub added (r1.1.2.1)
configure.ac added (r1.1.2.1)

---
Log message:

Initial autoconf checkin.


---
Diffs of the changes:

















From criswell at choi.cs.uiuc.edu  Mon Jun 23 14:00:01 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 14:00:01 2003
Subject: [llvm-commits] CVS: llvm/Makefile.config.in
Message-ID: <200306231859.h5NIxYM02186@choi.cs.uiuc.edu>


Changes in directory llvm:

Makefile.config.in added (r1.1.2.1)

---
Log message:

Initial checkin of autoconf.


---
Diffs of the changes:





From criswell at choi.cs.uiuc.edu  Mon Jun 23 14:03:02 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 14:03:02 2003
Subject: [llvm-commits] CVS: llvm/utils/Burg/Makefile
Message-ID: <200306231902.h5NJ2kn02232@choi.cs.uiuc.edu>


Changes in directory llvm/utils/Burg:

Makefile updated: 1.13 -> 1.13.2.1

---
Log message:

Made clean target clean up extra .c files.


---
Diffs of the changes:

Index: llvm/utils/Burg/Makefile
diff -u llvm/utils/Burg/Makefile:1.13 llvm/utils/Burg/Makefile:1.13.2.1
--- llvm/utils/Burg/Makefile:1.13	Wed Jun 11 08:57:48 2003
+++ llvm/utils/Burg/Makefile	Mon Jun 23 14:02:36 2003
@@ -1,4 +1,4 @@
-# $Id: Makefile,v 1.13 2003/06/11 13:57:48 criswell Exp $
+# $Id: Makefile,v 1.13.2.1 2003/06/23 19:02:36 criswell Exp $
 LEVEL = ../..
 TOOLNAME = burg
 EXTRASOURCES = gram.tab.c
@@ -26,3 +26,4 @@
 	$(TOOLEXENAME_G) -I      sample.gr -o tmp && cmp tmp sample.c
 	$(TOOLEXENAME_G) -I -O0 <sample.gr   >tmp && cmp tmp sample.c
 	$(TOOLEXENAME_G) -I -=  <sample.gr   >tmp && cmp tmp sample.c
+	$(RM) -f tmp sample.c




From criswell at choi.cs.uiuc.edu  Mon Jun 23 14:05:01 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 14:05:01 2003
Subject: [llvm-commits] CVS: llvm/tools/bugpoint/Makefile SystemUtils.cpp
Message-ID: <200306231904.h5NJ4xn02280@choi.cs.uiuc.edu>


Changes in directory llvm/tools/bugpoint:

Makefile updated: 1.3 -> 1.3.2.1
SystemUtils.cpp updated: 1.3 -> 1.3.2.1

---
Log message:

Initial autoconf checkin.


---
Diffs of the changes:

Index: llvm/tools/bugpoint/Makefile
diff -u llvm/tools/bugpoint/Makefile:1.3 llvm/tools/bugpoint/Makefile:1.3.2.1
--- llvm/tools/bugpoint/Makefile:1.3	Tue Jan 14 15:30:30 2003
+++ llvm/tools/bugpoint/Makefile	Mon Jun 23 14:04:49 2003
@@ -8,6 +8,4 @@
 USEDLIBS = ipo scalaropts analysis $(OPTLIBS) $(ANALIBS) \
            transformutils asmparser bcreader bcwriter vmcore support
 
-TOOLLINKOPTS = -ldl
-
 include $(LEVEL)/Makefile.common


Index: llvm/tools/bugpoint/SystemUtils.cpp
diff -u llvm/tools/bugpoint/SystemUtils.cpp:1.3 llvm/tools/bugpoint/SystemUtils.cpp:1.3.2.1
--- llvm/tools/bugpoint/SystemUtils.cpp:1.3	Wed Jan 29 12:15:34 2003
+++ llvm/tools/bugpoint/SystemUtils.cpp	Mon Jun 23 14:04:49 2003
@@ -10,12 +10,12 @@
 #include <fstream>
 #include <iostream>
 #include <cstdlib>
-#include <alloca.h>
+#include "Config/alloca.h"
 #include <sys/types.h>
 #include <sys/stat.h>
-#include <fcntl.h>
+#include "Config/fcntl.h"
 #include <sys/wait.h>
-#include <unistd.h>
+#include "Config/unistd.h"
 #include <errno.h>
 
 /// removeFile - Delete the specified file




From criswell at choi.cs.uiuc.edu  Mon Jun 23 14:05:03 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 14:05:03 2003
Subject: [llvm-commits] CVS: llvm/tools/analyze/Makefile
Message-ID: <200306231904.h5NJ4uh02268@choi.cs.uiuc.edu>


Changes in directory llvm/tools/analyze:

Makefile updated: 1.18 -> 1.18.2.1

---
Log message:

Initial autoconf checkin.


---
Diffs of the changes:

Index: llvm/tools/analyze/Makefile
diff -u llvm/tools/analyze/Makefile:1.18 llvm/tools/analyze/Makefile:1.18.2.1
--- llvm/tools/analyze/Makefile:1.18	Mon Mar 31 11:30:35 2003
+++ llvm/tools/analyze/Makefile	Mon Jun 23 14:04:46 2003
@@ -2,7 +2,6 @@
 TOOLNAME = analyze
 USEDLIBS = asmparser bcreader scalaropts.a transforms.a analysis ipa \
            datastructure target.a transformutils.a scalaropts.a vmcore support
-TOOLLINKOPTS = -ldl
 
 include $(LEVEL)/Makefile.common
 




From criswell at choi.cs.uiuc.edu  Mon Jun 23 14:06:02 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 14:06:02 2003
Subject: [llvm-commits] CVS: llvm/tools/opt/Makefile
Message-ID: <200306231905.h5NJ55T02326@choi.cs.uiuc.edu>


Changes in directory llvm/tools/opt:

Makefile updated: 1.39 -> 1.39.2.1

---
Log message:

Initial autoconf checkin.


---
Diffs of the changes:

Index: llvm/tools/opt/Makefile
diff -u llvm/tools/opt/Makefile:1.39 llvm/tools/opt/Makefile:1.39.2.1
--- llvm/tools/opt/Makefile:1.39	Wed Apr 16 17:55:55 2003
+++ llvm/tools/opt/Makefile	Mon Jun 23 14:04:55 2003
@@ -6,6 +6,4 @@
            ipo ipa.a datastructure transforms target.a analysis \
            transformutils vmcore support
 
-TOOLLINKOPTS = -ldl
-
 include $(LEVEL)/Makefile.common




From criswell at choi.cs.uiuc.edu  Mon Jun 23 14:06:05 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 14:06:05 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/ExecutionEngine.cpp Makefile
Message-ID: <200306231905.h5NJ51D02300@choi.cs.uiuc.edu>


Changes in directory llvm/tools/lli:

ExecutionEngine.cpp updated: 1.12 -> 1.12.2.1
Makefile updated: 1.24 -> 1.24.2.1

---
Log message:

Initial autoconf checkin.


---
Diffs of the changes:

Index: llvm/tools/lli/ExecutionEngine.cpp
diff -u llvm/tools/lli/ExecutionEngine.cpp:1.12 llvm/tools/lli/ExecutionEngine.cpp:1.12.2.1
--- llvm/tools/lli/ExecutionEngine.cpp:1.12	Wed May 14 12:53:49 2003
+++ llvm/tools/lli/ExecutionEngine.cpp	Mon Jun 23 14:04:51 2003
@@ -12,7 +12,7 @@
 #include "llvm/Module.h"
 #include "llvm/Target/TargetData.h"
 #include "Support/Statistic.h"
-#include <dlfcn.h>
+#include "Config/dlfcn.h"
 
 Statistic<> NumInitBytes("lli", "Number of bytes of global vars initialized");
 


Index: llvm/tools/lli/Makefile
diff -u llvm/tools/lli/Makefile:1.24 llvm/tools/lli/Makefile:1.24.2.1
--- llvm/tools/lli/Makefile:1.24	Tue May 27 16:42:05 2003
+++ llvm/tools/lli/Makefile	Mon Jun 23 14:04:51 2003
@@ -22,9 +22,4 @@
 USEDLIBS = lli-interpreter $(JITLIBS) bcreader vmcore scalaropts \
            analysis.a support.a target.a $(ARCHLIBS)
 
-# Have gcc tell the linker to export symbols from the program so that
-# dynamically loaded modules can be linked against them.
-#
-TOOLLINKOPTS = -ldl
-
 include $(LEVEL)/Makefile.common




From criswell at choi.cs.uiuc.edu  Mon Jun 23 14:06:07 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 14:06:07 2003
Subject: [llvm-commits] CVS: llvm/tools/llc/Makefile
Message-ID: <200306231905.h5NJ50002289@choi.cs.uiuc.edu>


Changes in directory llvm/tools/llc:

Makefile updated: 1.35 -> 1.35.2.1

---
Log message:

Initial autoconf checkin.


---
Diffs of the changes:

Index: llvm/tools/llc/Makefile
diff -u llvm/tools/llc/Makefile:1.35 llvm/tools/llc/Makefile:1.35.2.1
--- llvm/tools/llc/Makefile:1.35	Tue May 27 16:23:02 2003
+++ llvm/tools/llc/Makefile	Mon Jun 23 14:04:50 2003
@@ -18,7 +18,6 @@
 	   bcwriter \
            vmcore \
            support
-TOOLLINKOPTS = -ldl
 
 include $(LEVEL)/Makefile.common
 




From criswell at choi.cs.uiuc.edu  Mon Jun 23 14:06:09 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 14:06:09 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/Interpreter/ExternalFunctions.cpp
Message-ID: <200306231905.h5NJ52P02308@choi.cs.uiuc.edu>


Changes in directory llvm/tools/lli/Interpreter:

ExternalFunctions.cpp updated: 1.52 -> 1.52.2.1

---
Log message:

Initial autoconf checkin.


---
Diffs of the changes:

Index: llvm/tools/lli/Interpreter/ExternalFunctions.cpp
diff -u llvm/tools/lli/Interpreter/ExternalFunctions.cpp:1.52 llvm/tools/lli/Interpreter/ExternalFunctions.cpp:1.52.2.1
--- llvm/tools/lli/Interpreter/ExternalFunctions.cpp:1.52	Wed May 14 09:21:30 2003
+++ llvm/tools/lli/Interpreter/ExternalFunctions.cpp	Mon Jun 23 14:04:52 2003
@@ -17,10 +17,10 @@
 #include "llvm/SymbolTable.h"
 #include "llvm/Target/TargetData.h"
 #include <map>
-#include <dlfcn.h>
-#include <link.h>
+#include "Config/dlfcn.h"
+#include "Config/link.h"
 #include <math.h>
-#include <stdio.h>
+#include "Config/stdio.h"
 using std::vector;
 
 typedef GenericValue (*ExFunc)(FunctionType *, const vector<GenericValue> &);




From criswell at choi.cs.uiuc.edu  Mon Jun 23 14:06:12 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 14:06:12 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/JIT/Emitter.cpp Intercept.cpp
Message-ID: <200306231905.h5NJ54202318@choi.cs.uiuc.edu>


Changes in directory llvm/tools/lli/JIT:

Emitter.cpp updated: 1.14 -> 1.14.2.1
Intercept.cpp updated: 1.4 -> 1.4.2.1

---
Log message:

Initial autoconf checkin.


---
Diffs of the changes:

Index: llvm/tools/lli/JIT/Emitter.cpp
diff -u llvm/tools/lli/JIT/Emitter.cpp:1.14 llvm/tools/lli/JIT/Emitter.cpp:1.14.2.1
--- llvm/tools/lli/JIT/Emitter.cpp:1.14	Sun Jun  8 01:43:57 2003
+++ llvm/tools/lli/JIT/Emitter.cpp	Mon Jun 23 14:04:54 2003
@@ -6,6 +6,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "VM.h"
+#include "Config/sys/mman.h"
 #include "llvm/CodeGen/MachineCodeEmitter.h"
 #include "llvm/CodeGen/MachineFunction.h"
 #include "llvm/CodeGen/MachineConstantPool.h"
@@ -79,7 +80,7 @@
   static unsigned long Counter = 0;
   pa = mmap((void*)(0x140000000UL+Counter), pageSize*NumPages,
             PROT_READ|PROT_WRITE|PROT_EXEC,
-            MAP_PRIVATE|MAP_ANON|MAP_FIXED, -1, 0); /* fd = -1 */
+            MAP_PRIVATE|MAP_ANONYMOUS|MAP_FIXED, -1, 0); /* fd = -1 */
   Counter += pageSize*NumPages;
 #else
   std::cerr << "This architecture is not supported by the JIT\n";


Index: llvm/tools/lli/JIT/Intercept.cpp
diff -u llvm/tools/lli/JIT/Intercept.cpp:1.4 llvm/tools/lli/JIT/Intercept.cpp:1.4.2.1
--- llvm/tools/lli/JIT/Intercept.cpp:1.4	Tue Jun  3 20:57:22 2003
+++ llvm/tools/lli/JIT/Intercept.cpp	Mon Jun 23 14:04:54 2003
@@ -9,7 +9,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "VM.h"
-#include <dlfcn.h>    // dlsym access
+#include "Config/dlfcn.h"    // dlsym access
 #include <iostream>
 
 // AtExitList - List of functions registered with the at_exit function




From criswell at choi.cs.uiuc.edu  Mon Jun 23 14:08:08 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 14:08:08 2003
Subject: [llvm-commits] CVS: llvm/lib/Makefile
Message-ID: <200306231907.h5NJ7dH02488@choi.cs.uiuc.edu>


Changes in directory llvm/lib:

Makefile updated: 1.11 -> 1.11.2.1

---
Log message:

Initial autoconf checkin.


---
Diffs of the changes:

Index: llvm/lib/Makefile
diff -u llvm/lib/Makefile:1.11 llvm/lib/Makefile:1.11.2.1
--- llvm/lib/Makefile:1.11	Tue Sep 17 22:25:55 2002
+++ llvm/lib/Makefile	Mon Jun 23 14:06:59 2003
@@ -1,6 +1,8 @@
 LEVEL = ..
 
-PARALLEL_DIRS = VMCore Analysis Transforms AsmParser Bytecode Support CodeGen Target CWriter Reoptimizer
+PARALLEL_DIRS = VMCore Analysis Transforms AsmParser Bytecode Support CodeGen Target CWriter
+
+OPTIONAL_DIRS = Reoptimizer
 
 include $(LEVEL)/Makefile.common
 




From criswell at choi.cs.uiuc.edu  Mon Jun 23 14:08:15 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 14:08:15 2003
Subject: [llvm-commits] CVS: llvm/lib/VMCore/PassManagerT.h
Message-ID: <200306231907.h5NJ7Pt02482@choi.cs.uiuc.edu>


Changes in directory llvm/lib/VMCore:

PassManagerT.h updated: 1.38 -> 1.38.2.1

---
Log message:

Modified code so that it compiles under GCC 3.3, but should otherwise function
the same.


---
Diffs of the changes:

Index: llvm/lib/VMCore/PassManagerT.h
diff -u llvm/lib/VMCore/PassManagerT.h:1.38 llvm/lib/VMCore/PassManagerT.h:1.38.2.1
--- llvm/lib/VMCore/PassManagerT.h:1.38	Thu Apr 24 15:07:38 2003
+++ llvm/lib/VMCore/PassManagerT.h	Mon Jun 23 14:06:45 2003
@@ -130,8 +130,14 @@
   typedef typename Traits::BatcherClass BatcherClass;
   typedef typename Traits::ParentClass   ParentClass;
 
-  friend typename Traits::PassClass;
-  friend typename Traits::SubPassClass;  
+  //
+  // JTC:
+  //	GCC 3.3 does not permit friend declarations without the class keyword.
+  //	Hence, I changed the following two lines to allow the code to compiler
+  //	under the new compiler.
+  //
+  friend class PassManagerTraits<UnitType>::PassClass;
+  friend class PassManagerTraits<UnitType>::SubPassClass;  
   friend class Traits;
   friend class ImmutablePass;
 




From criswell at choi.cs.uiuc.edu  Mon Jun 23 14:08:19 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 14:08:19 2003
Subject: [llvm-commits] CVS: llvm/lib/VMCore/Pass.cpp
Message-ID: <200306231907.h5NJ7Lq02476@choi.cs.uiuc.edu>


Changes in directory llvm/lib/VMCore:

Pass.cpp updated: 1.41 -> 1.41.2.1

---
Log message:

Initial autoconf checkin.


---
Diffs of the changes:

Index: llvm/lib/VMCore/Pass.cpp
diff -u llvm/lib/VMCore/Pass.cpp:1.41 llvm/lib/VMCore/Pass.cpp:1.41.2.1
--- llvm/lib/VMCore/Pass.cpp:1.41	Fri Mar 21 15:41:02 2003
+++ llvm/lib/VMCore/Pass.cpp	Mon Jun 23 14:07:11 2003
@@ -11,10 +11,10 @@
 #include "llvm/Module.h"
 #include "Support/STLExtras.h"
 #include "Support/TypeInfo.h"
-#include <stdio.h>
-#include <sys/resource.h>
-#include <sys/time.h>
-#include <sys/unistd.h>
+#include "Config/stdio.h"
+#include "Config/sys/resource.h"
+#include "Config/sys/time.h"
+#include "Config/unistd.h"
 #include <set>
 
 // IncludeFile - Stub function used to help linking out.




From criswell at choi.cs.uiuc.edu  Mon Jun 23 14:08:21 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 14:08:21 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Instrumentation/ProfilePaths/EdgeCode.cpp ProfilePaths.cpp
Message-ID: <200306231907.h5NJ7K002468@choi.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Instrumentation/ProfilePaths:

EdgeCode.cpp updated: 1.21 -> 1.21.2.1
ProfilePaths.cpp updated: 1.28 -> 1.28.2.1

---
Log message:

Initial autoconf checkin.


---
Diffs of the changes:

Index: llvm/lib/Transforms/Instrumentation/ProfilePaths/EdgeCode.cpp
diff -u llvm/lib/Transforms/Instrumentation/ProfilePaths/EdgeCode.cpp:1.21 llvm/lib/Transforms/Instrumentation/ProfilePaths/EdgeCode.cpp:1.21.2.1
--- llvm/lib/Transforms/Instrumentation/ProfilePaths/EdgeCode.cpp:1.21	Thu Jun  5 01:02:27 2003
+++ llvm/lib/Transforms/Instrumentation/ProfilePaths/EdgeCode.cpp	Mon Jun 23 14:07:10 2003
@@ -16,7 +16,7 @@
 #include "llvm/iOperators.h"
 #include "llvm/iPHINode.h"
 #include "llvm/Module.h"
-#include <stdio.h>
+#include "Config/stdio.h"
 
 #define INSERT_LOAD_COUNT
 #define INSERT_STORE


Index: llvm/lib/Transforms/Instrumentation/ProfilePaths/ProfilePaths.cpp
diff -u llvm/lib/Transforms/Instrumentation/ProfilePaths/ProfilePaths.cpp:1.28 llvm/lib/Transforms/Instrumentation/ProfilePaths/ProfilePaths.cpp:1.28.2.1
--- llvm/lib/Transforms/Instrumentation/ProfilePaths/ProfilePaths.cpp:1.28	Thu Jun  5 01:02:46 2003
+++ llvm/lib/Transforms/Instrumentation/ProfilePaths/ProfilePaths.cpp	Mon Jun 23 14:07:10 2003
@@ -34,7 +34,7 @@
 #include "llvm/Module.h"
 #include "Graph.h"
 #include <fstream>
-#include <stdio.h>
+#include "Config/stdio.h"
 using std::vector;
 
 struct ProfilePaths : public FunctionPass {




From criswell at choi.cs.uiuc.edu  Mon Jun 23 14:08:24 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 14:08:24 2003
Subject: [llvm-commits] CVS: llvm/lib/Target/X86/X86CodeEmitter.cpp
Message-ID: <200306231907.h5NJ7JL02458@choi.cs.uiuc.edu>


Changes in directory llvm/lib/Target/X86:

X86CodeEmitter.cpp updated: 1.28 -> 1.28.2.1

---
Log message:

Initial autoconf checkin.


---
Diffs of the changes:

Index: llvm/lib/Target/X86/X86CodeEmitter.cpp
diff -u llvm/lib/Target/X86/X86CodeEmitter.cpp:1.28 llvm/lib/Target/X86/X86CodeEmitter.cpp:1.28.2.1
--- llvm/lib/Target/X86/X86CodeEmitter.cpp:1.28	Fri Jun  6 13:25:33 2003
+++ llvm/lib/Target/X86/X86CodeEmitter.cpp	Mon Jun 23 14:07:08 2003
@@ -13,7 +13,7 @@
 #include "llvm/CodeGen/MachineInstr.h"
 #include "llvm/Value.h"
 #include "Support/Statistic.h"
-#include <alloca.h>
+#include "Config/alloca.h"
 
 namespace {
   Statistic<>




From criswell at choi.cs.uiuc.edu  Mon Jun 23 14:08:27 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 14:08:27 2003
Subject: [llvm-commits] CVS: llvm/lib/Target/Sparc/Makefile SparcInstrInfo.cpp SparcInternals.h SparcOptInfo.cpp
Message-ID: <200306231907.h5NJ7Ih02450@choi.cs.uiuc.edu>


Changes in directory llvm/lib/Target/Sparc:

Makefile updated: 1.23 -> 1.23.2.1
SparcInstrInfo.cpp updated: 1.47 -> 1.47.2.1
SparcInternals.h updated: 1.90 -> 1.90.2.1
SparcOptInfo.cpp updated: 1.8 -> 1.8.2.1

---
Log message:

Initial autoconf checkin.


---
Diffs of the changes:

Index: llvm/lib/Target/Sparc/Makefile
diff -u llvm/lib/Target/Sparc/Makefile:1.23 llvm/lib/Target/Sparc/Makefile:1.23.2.1
--- llvm/lib/Target/Sparc/Makefile:1.23	Wed Jun 11 08:49:11 2003
+++ llvm/lib/Target/Sparc/Makefile	Mon Jun 23 14:07:07 2003
@@ -24,10 +24,10 @@
 	$(CompileP) $< -o $@
 
 Debug/Sparc.burg.in1 : Sparc.burg.in Debug/.dir
-	$(CXX) -E -I$(LEVEL)/include $(DEBUG_FLAG) -x c++ $< | sed '/^# /d' | sed 's/Ydefine/#define/' > $@
+	$(CXX) -E -I$(LEVEL)/include $(DEBUG_FLAG) -x c++ $< | ${SED} '/^# /d' | ${SED} 's/Ydefine/#define/' > $@
 
 Debug/Sparc.burm : Debug/Sparc.burg.in1
-	$(CXX) -E -I$(LEVEL)/include $(DEBUG_FLAG) -x c++ $< | sed '/^# /d' | sed 's/Xinclude/#include/g' | sed 's/Xdefine/#define/g' > $@
+	$(CXX) -E -I$(LEVEL)/include $(DEBUG_FLAG) -x c++ $< | ${SED} '/^# /d' | ${SED} 's/Xinclude/#include/g' | ${SED} 's/Xdefine/#define/g' > $@
 
 $(BUILD_OBJ_DIR)/Depend/Sparc.burm.d: $(BUILD_OBJ_DIR)/Depend/.dir
 	touch $@
@@ -35,8 +35,8 @@
 SparcV9CodeEmitter.cpp: SparcV9CodeEmitter.inc
 
 SparcV9CodeEmitter.inc: SparcV9.td SparcV9_F2.td SparcV9_F3.td SparcV9_F4.td SparcV9_Reg.td $(TBLGEN)
-	@echo "TableGen-erating $@"
+	@${ECHO} "TableGen-erating $@"
 	cpp -P SparcV9.td | $(TBLGEN) -gen-emitter -o $@
 
 clean::
-	rm -f SparcV9CodeEmitter.inc
+	${RM} -f SparcV9CodeEmitter.inc


Index: llvm/lib/Target/Sparc/SparcInstrInfo.cpp
diff -u llvm/lib/Target/Sparc/SparcInstrInfo.cpp:1.47 llvm/lib/Target/Sparc/SparcInstrInfo.cpp:1.47.2.1
--- llvm/lib/Target/Sparc/SparcInstrInfo.cpp:1.47	Fri Jun  6 04:52:23 2003
+++ llvm/lib/Target/Sparc/SparcInstrInfo.cpp	Mon Jun 23 14:07:07 2003
@@ -13,7 +13,7 @@
 #include "llvm/Function.h"
 #include "llvm/Constants.h"
 #include "llvm/DerivedTypes.h"
-#include <stdlib.h>
+#include "Config/stdlib.h"
 
 static const uint32_t MAXLO   = (1 << 10) - 1; // set bits set by %lo(*)
 static const uint32_t MAXSIMM = (1 << 12) - 1; // set bits in simm13 field of OR


Index: llvm/lib/Target/Sparc/SparcInternals.h
diff -u llvm/lib/Target/Sparc/SparcInternals.h:1.90 llvm/lib/Target/Sparc/SparcInternals.h:1.90.2.1
--- llvm/lib/Target/Sparc/SparcInternals.h:1.90	Mon Jun 16 10:31:09 2003
+++ llvm/lib/Target/Sparc/SparcInternals.h	Mon Jun 23 14:07:07 2003
@@ -17,7 +17,7 @@
 #include "llvm/Target/TargetOptInfo.h"
 #include "llvm/Type.h"
 #include "SparcRegClassInfo.h"
-#include <sys/types.h>
+#include "Config/sys/types.h"
 
 class LiveRange;
 class UltraSparc;


Index: llvm/lib/Target/Sparc/SparcOptInfo.cpp
diff -u llvm/lib/Target/Sparc/SparcOptInfo.cpp:1.8 llvm/lib/Target/Sparc/SparcOptInfo.cpp:1.8.2.1
--- llvm/lib/Target/Sparc/SparcOptInfo.cpp:1.8	Tue May 27 17:39:29 2003
+++ llvm/lib/Target/Sparc/SparcOptInfo.cpp	Mon Jun 23 14:07:07 2003
@@ -7,7 +7,7 @@
 #include "SparcInternals.h"
 #include "llvm/Target/TargetRegInfo.h"
 #include "llvm/CodeGen/MachineInstr.h"
-#include <stdlib.h>
+#include "Config/stdlib.h"
 
 //----------------------------------------------------------------------------
 // Function: IsUselessCopy




From criswell at choi.cs.uiuc.edu  Mon Jun 23 14:08:30 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 14:08:30 2003
Subject: [llvm-commits] CVS: llvm/lib/Support/PluginLoader.cpp Signals.cpp Timer.cpp
Message-ID: <200306231907.h5NJ7Gp02435@choi.cs.uiuc.edu>


Changes in directory llvm/lib/Support:

PluginLoader.cpp updated: 1.2 -> 1.2.2.1
Signals.cpp updated: 1.5 -> 1.5.2.1
Timer.cpp updated: 1.18 -> 1.18.2.1

---
Log message:

Initial autoconf checkin.


---
Diffs of the changes:

Index: llvm/lib/Support/PluginLoader.cpp
diff -u llvm/lib/Support/PluginLoader.cpp:1.2 llvm/lib/Support/PluginLoader.cpp:1.2.2.1
--- llvm/lib/Support/PluginLoader.cpp:1.2	Thu Jul 25 01:17:46 2002
+++ llvm/lib/Support/PluginLoader.cpp	Mon Jun 23 14:07:06 2003
@@ -11,8 +11,8 @@
 //===----------------------------------------------------------------------===//
 
 #include "Support/CommandLine.h"
-#include <dlfcn.h>
-#include <link.h>
+#include "Config/dlfcn.h"
+#include "Config/link.h"
 #include <iostream>
 
 namespace {


Index: llvm/lib/Support/Signals.cpp
diff -u llvm/lib/Support/Signals.cpp:1.5 llvm/lib/Support/Signals.cpp:1.5.2.1
--- llvm/lib/Support/Signals.cpp:1.5	Tue May 27 11:25:04 2003
+++ llvm/lib/Support/Signals.cpp	Mon Jun 23 14:07:06 2003
@@ -32,7 +32,7 @@
 
 
 // SignalHandler - The signal handler that runs...
-static void SignalHandler(int Sig) {
+static RETSIGTYPE SignalHandler(int Sig) {
   while (!FilesToRemove.empty()) {
     std::remove(FilesToRemove.back().c_str());
     FilesToRemove.pop_back();


Index: llvm/lib/Support/Timer.cpp
diff -u llvm/lib/Support/Timer.cpp:1.18 llvm/lib/Support/Timer.cpp:1.18.2.1
--- llvm/lib/Support/Timer.cpp:1.18	Fri Jun  6 17:13:01 2003
+++ llvm/lib/Support/Timer.cpp	Mon Jun 23 14:07:06 2003
@@ -4,14 +4,15 @@
 //
 //===----------------------------------------------------------------------===//
 
+#include "Config/malloc.h"
+
 #include "Support/Timer.h"
 #include "Support/CommandLine.h"
-#include <sys/resource.h>
-#include <sys/time.h>
-#include <sys/unistd.h>
-#include <unistd.h>
-#include <malloc.h>
-#include <stdio.h>
+
+#include "Config/sys/resource.h"
+#include "Config/sys/time.h"
+#include "Config/unistd.h"
+#include "Config/stdio.h"
 #include <iostream>
 #include <algorithm>
 #include <functional>
@@ -20,10 +21,12 @@
 std::string LibSupportInfoOutputFilename;
 
 namespace {
+#ifdef HAVE_MALLINFO
   cl::opt<bool>
   TrackSpace("track-memory", cl::desc("Enable -time-passes memory "
                                       "tracking (this may be slow)"),
              cl::Hidden);
+#endif
 
   cl::opt<std::string, true>
   InfoOutputFilename("info-output-file",
@@ -75,12 +78,16 @@
 }
 
 static long getMemUsage() {
+#ifdef HAVE_MALLINFO
   if (TrackSpace) {
     struct mallinfo MI = mallinfo();
     return MI.uordblks/*+MI.hblkhd*/;
   } else {
     return 0;
   }
+#else
+    return 0;
+#endif
 }
 
 struct TimeRecord {




From criswell at choi.cs.uiuc.edu  Mon Jun 23 14:08:33 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 14:08:33 2003
Subject: [llvm-commits] CVS: llvm/lib/CodeGen/InstrSelection/InstrForest.cpp
Message-ID: <200306231907.h5NJ7Fi02423@choi.cs.uiuc.edu>


Changes in directory llvm/lib/CodeGen/InstrSelection:

InstrForest.cpp updated: 1.40 -> 1.40.2.1

---
Log message:

Initial autoconf checkin.


---
Diffs of the changes:

Index: llvm/lib/CodeGen/InstrSelection/InstrForest.cpp
diff -u llvm/lib/CodeGen/InstrSelection/InstrForest.cpp:1.40 llvm/lib/CodeGen/InstrSelection/InstrForest.cpp:1.40.2.1
--- llvm/lib/CodeGen/InstrSelection/InstrForest.cpp:1.40	Sat Aug 24 16:02:09 2002
+++ llvm/lib/CodeGen/InstrSelection/InstrForest.cpp	Mon Jun 23 14:07:05 2003
@@ -21,7 +21,7 @@
 #include "llvm/Type.h"
 #include "llvm/CodeGen/MachineInstr.h"
 #include "Support/STLExtras.h"
-#include <alloca.h>
+#include "Config/alloca.h"
 using std::cerr;
 using std::vector;
 




From criswell at choi.cs.uiuc.edu  Mon Jun 23 14:08:37 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 14:08:37 2003
Subject: [llvm-commits] CVS: llvm/lib/CodeGen/MachineFunction.cpp
Message-ID: <200306231907.h5NJ7EP02414@choi.cs.uiuc.edu>


Changes in directory llvm/lib/CodeGen:

MachineFunction.cpp updated: 1.39 -> 1.39.2.1

---
Log message:

Initial autoconf checkin.


---
Diffs of the changes:

Index: llvm/lib/CodeGen/MachineFunction.cpp
diff -u llvm/lib/CodeGen/MachineFunction.cpp:1.39 llvm/lib/CodeGen/MachineFunction.cpp:1.39.2.1
--- llvm/lib/CodeGen/MachineFunction.cpp:1.39	Wed Apr 23 11:36:09 2003
+++ llvm/lib/CodeGen/MachineFunction.cpp	Mon Jun 23 14:07:04 2003
@@ -19,7 +19,7 @@
 #include "llvm/Function.h"
 #include "llvm/iOther.h"
 #include "llvm/Pass.h"
-#include <limits.h>
+#include "Config/limits.h"
 
 const int INVALID_FRAME_OFFSET = INT_MAX; // std::numeric_limits<int>::max();
 




From criswell at choi.cs.uiuc.edu  Mon Jun 23 14:08:40 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 14:08:40 2003
Subject: [llvm-commits] CVS: llvm/lib/Bytecode/Writer/Writer.cpp
Message-ID: <200306231907.h5NJ7D202405@choi.cs.uiuc.edu>


Changes in directory llvm/lib/Bytecode/Writer:

Writer.cpp updated: 1.34 -> 1.34.2.1

---
Log message:

Initial autoconf checkin.


---
Diffs of the changes:

Index: llvm/lib/Bytecode/Writer/Writer.cpp
diff -u llvm/lib/Bytecode/Writer/Writer.cpp:1.34 llvm/lib/Bytecode/Writer/Writer.cpp:1.34.2.1
--- llvm/lib/Bytecode/Writer/Writer.cpp:1.34	Thu May 22 13:35:38 2003
+++ llvm/lib/Bytecode/Writer/Writer.cpp	Mon Jun 23 14:07:03 2003
@@ -26,7 +26,7 @@
 #include "llvm/DerivedTypes.h"
 #include "Support/STLExtras.h"
 #include "Support/Statistic.h"
-#include <string.h>
+#include "Config/string.h"
 #include <algorithm>
 
 static RegisterPass<WriteBytecodePass> X("emitbytecode", "Bytecode Writer");




From criswell at choi.cs.uiuc.edu  Mon Jun 23 14:08:43 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 14:08:43 2003
Subject: [llvm-commits] CVS: llvm/lib/Bytecode/Reader/ReadArchive.cpp Reader.cpp
Message-ID: <200306231907.h5NJ7CH02396@choi.cs.uiuc.edu>


Changes in directory llvm/lib/Bytecode/Reader:

ReadArchive.cpp updated: 1.3 -> 1.3.2.1
Reader.cpp updated: 1.56 -> 1.56.2.1

---
Log message:

Initial autoconf checkin.


---
Diffs of the changes:

Index: llvm/lib/Bytecode/Reader/ReadArchive.cpp
diff -u llvm/lib/Bytecode/Reader/ReadArchive.cpp:1.3 llvm/lib/Bytecode/Reader/ReadArchive.cpp:1.3.2.1
--- llvm/lib/Bytecode/Reader/ReadArchive.cpp:1.3	Tue Apr 22 21:59:05 2003
+++ llvm/lib/Bytecode/Reader/ReadArchive.cpp	Mon Jun 23 14:07:02 2003
@@ -11,9 +11,9 @@
 
 #include "llvm/Bytecode/Reader.h"
 #include "llvm/Module.h"
-#include <sys/stat.h>
-#include <sys/mman.h>
-#include <fcntl.h>
+#include "Config/sys/stat.h"
+#include "Config/sys/mman.h"
+#include "Config/fcntl.h"
 
 namespace {
   struct ar_hdr {


Index: llvm/lib/Bytecode/Reader/Reader.cpp
diff -u llvm/lib/Bytecode/Reader/Reader.cpp:1.56 llvm/lib/Bytecode/Reader/Reader.cpp:1.56.2.1
--- llvm/lib/Bytecode/Reader/Reader.cpp:1.56	Thu May 22 13:26:48 2003
+++ llvm/lib/Bytecode/Reader/Reader.cpp	Mon Jun 23 14:07:02 2003
@@ -11,17 +11,17 @@
 //===----------------------------------------------------------------------===//
 
 #include "ReaderInternals.h"
+#include "Config/sys/mman.h"
 #include "llvm/Bytecode/Reader.h"
 #include "llvm/Bytecode/Format.h"
 #include "llvm/Module.h"
 #include "llvm/Constants.h"
 #include "llvm/iPHINode.h"
 #include "llvm/iOther.h"
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <sys/mman.h>
-#include <fcntl.h>
-#include <unistd.h>
+#include "Config/sys/types.h"
+#include "Config/sys/stat.h"
+#include "Config/fcntl.h"
+#include "Config/unistd.h"
 #include <algorithm>
 
 bool BytecodeParser::getTypeSlot(const Type *Ty, unsigned &Slot) {




From criswell at choi.cs.uiuc.edu  Mon Jun 23 14:08:46 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 14:08:46 2003
Subject: [llvm-commits] CVS: llvm/lib/AsmParser/Lexer.l
Message-ID: <200306231907.h5NJ7BD02386@choi.cs.uiuc.edu>


Changes in directory llvm/lib/AsmParser:

Lexer.l updated: 1.34 -> 1.34.2.1

---
Log message:

Initial autoconf checkin.


---
Diffs of the changes:

Index: llvm/lib/AsmParser/Lexer.l
diff -u llvm/lib/AsmParser/Lexer.l:1.34 llvm/lib/AsmParser/Lexer.l:1.34.2.1
--- llvm/lib/AsmParser/Lexer.l:1.34	Wed May  7 21:44:01 2003
+++ llvm/lib/AsmParser/Lexer.l	Mon Jun 23 14:07:01 2003
@@ -23,7 +23,7 @@
 #include <list>
 #include "llvmAsmParser.h"
 #include <ctype.h>
-#include <stdlib.h>
+#include "Config/stdlib.h"
 
 #define RET_TOK(type, Enum, sym) \
   llvmAsmlval.type = Instruction::Enum; return sym




From criswell at choi.cs.uiuc.edu  Mon Jun 23 14:12:02 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Mon Jun 23 14:12:02 2003
Subject: [llvm-commits] CVS: llvm/Makefile.common
Message-ID: <200306231911.h5NJBq502544@choi.cs.uiuc.edu>


Changes in directory llvm:

Makefile.common updated: 1.88 -> 1.88.2.1

---
Log message:

Initial autoconf checkin.
Uses a modified version of libtool for building files.


---
Diffs of the changes:

Index: llvm/Makefile.common
diff -u llvm/Makefile.common:1.88 llvm/Makefile.common:1.88.2.1
--- llvm/Makefile.common:1.88	Wed Jun 11 08:55:26 2003
+++ llvm/Makefile.common	Mon Jun 23 14:11:41 2003
@@ -117,8 +117,13 @@
 # gross and should be autoconfiscated (automake actually), but should hopefully
 # work on Linux and solaris (SunOS).
 #
-UNAME := $(shell uname)
-include $(LLVM_SRC_ROOT)/Makefile.$(UNAME)
+include $(LLVM_SRC_ROOT)/Makefile.$(OS)
+
+###########################################################################
+# Default Targets:
+#	The following targets are the standard top level targets for
+#	building.
+###########################################################################
 
 ifdef SHARED_LIBRARY
 # if SHARED_LIBRARY is specified, the default is to build the dynamic lib
@@ -134,6 +139,12 @@
 # Default rule for test.  It ensures everything has a test rule
 test::
 
+###########################################################################
+# Miscellaneous paths and commands:
+#	This section defines various configuration macros, such as where
+#	to find burg, tblgen, and libtool.
+###########################################################################
+
 #--------------------------------------------------------------------
 # Variables derived from configuration options... 
 #--------------------------------------------------------------------
@@ -157,8 +168,21 @@
   endif
 endif
 
+#
+# Enable this for profiling support with 'gprof'
+# This automatically enables optimized builds.
+#
+ifdef ENABLE_PROFILING
+  PROFILE = -pg
+endif
+
+#
+# Suffixes for library compilation rules
+#
+.SUFFIXES: .so
+
 ###########################################################################
-# Library Locations
+# Library Locations:
 #	These variables describe various library locations:
 #
 #	DEST* = Location of where libraries that are built will be placed.
@@ -208,36 +232,45 @@
 PROJTOOLPROFILE := $(BUILD_OBJ_ROOT)/tools/Profile
 PROJTOOLCURRENT := $(BUILD_OBJ_ROOT)/tools/$(CONFIGURATION)
 
+#
+# Libtool is found in the current directory.
+#
+ifdef VERBOSE
+LIBTOOL=$(LLVM_SRC_ROOT)/libtool
+else
+LIBTOOL=$(LLVM_SRC_ROOT)/libtool --silent
+endif
+
+#
 # Verbosity levels
+#
 ifndef VERBOSE
 VERB := @
 endif
 
-#---------------------------------------------------------
-# Compilation options...
-#---------------------------------------------------------
-
 ###########################################################################
-# Special tools used while building the LLVM tree.  Burg is built as part of the
-# utils directory.
+# Miscellaneous paths and commands (part deux):
+#	This section defines various configuration macros, such as where
+#	to find burg, tblgen, and libtool.
 ###########################################################################
+
+#--------------------------------------------------------------------------
+# Special tools used while building the LLVM tree.  Burg is built as part
+# of the utils directory.
+#--------------------------------------------------------------------------
 BURG    := $(LLVMTOOLCURRENT)/burg
 RunBurg := $(BURG) $(BURG_OPTS)
 
 TBLGEN  := $(LLVMTOOLCURRENT)/tblgen
 
-# Enable this for profiling support with 'gprof'
-# This automatically enables optimized builds.
-ifdef ENABLE_PROFILING
-  PROFILE = -pg
-endif
 
 ###########################################################################
 # Compile Time Flags
 ###########################################################################
 
 #
-# Include both the project headers and the LLVM headers for compilation
+# Include both the project headers and the LLVM headers for compilation and
+# dependency computation.
 #
 CPPFLAGS += -I$(BUILD_SRC_ROOT)/include -I$(LLVM_SRC_ROOT)/include
 
@@ -254,47 +287,71 @@
 CompileCommonOpts := -Wall -W  -Wwrite-strings -Wno-unused -I$(LEVEL)/include
 CompileOptimizeOpts := -O3 -DNDEBUG -finline-functions -fshort-enums
 
+#
+# Compile commands with libtool.
+#
+Compile  := $(LIBTOOL) --mode=compile $(CXX) -c $(CPPFLAGS) $(CXXFLAGS) $(CompileCommonOpts)
+CompileC  := $(LIBTOOL) --mode=compile $(CC) -c $(CPPFLAGS) $(CCFLAGS) $(CompileCommonOpts)
+
+#
+# Add the LLVM specific "-only-static" option so that we only compile .o files
+# once when not building a shared library.
+#
+# For shared libraries, we will end up building twice, but that doesn't happen
+# very often, so we'll let it go.
+#
+ifndef SHARED_LIBRARY
+Compile  := $(Compile)  -only-static
+CompileC := $(CompileC) -only-static
+endif
+
 # Compile a cpp file, don't link...
-Compile  := $(CXX) -c $(CPPFLAGS) $(CXXFLAGS) $(CompileCommonOpts)
 CompileG := $(Compile) -g  -D_DEBUG
 CompileO := $(Compile) $(CompileOptimizeOpts) -felide-constructors -fomit-frame-pointer
 CompileP := $(Compile) $(CompileOptimizeOpts) -felide-constructors $(PROFILE)
 
 # Compile a c file, don't link...
-CompileC  := $(CC) -c $(CPPFLAGS) $(CCFLAGS) $(CompileCommonOpts)
 CompileCG := $(CompileC) -g  -D_DEBUG
 CompileCO := $(CompileC) $(CompileOptimizeOpts) -fomit-frame-pointer
 CompileCP := $(CompileC) $(CompileOptimizeOpts) $(PROFILE)
 
+###########################################################################
+# Link Time Options
+###########################################################################
 
+#
 # Link final executable
-
+#	(Note that we always link with the C++ compiler).
+#
 ifdef ENABLE_PURIFY # To enable purify, build with 'gmake ENABLE_PURIFY=1'
-Link     := $(PURIFY) $(CXX) -static
+Link     := $(PURIFY) $(LIBTOOL) --mode=link $(CXX) -static
 else
-Link     := $(CXX)
+Link     := $(LIBTOOL) --mode=link $(CXX)
 endif
 
-ifdef PROJ_COMPILE
-# include both projlib source and llvmlib source
+# link both projlib and llvmlib libraries
 LinkG    := $(Link) -g -L$(PROJLIBDEBUGSOURCE)  -L$(LLVMLIBDEBUGSOURCE) $(STRIP)
 LinkO    := $(Link) -O3 -L$(PROJLIBRELEASESOURCE) -L$(LLVMLIBRELEASESOURCE)
 LinkP    := $(Link) -O3 -L$(PROJLIBPROFILESOURCE) -L$(LLVMLIBPROFILESOURCE) $(PROFILE)
-else
-LinkG    := $(Link) -g  -L$(LLVMLIBDEBUGSOURCE) $(STRIP)
-LinkO    := $(Link) -O3 -L$(LLVMLIBRELEASESOURCE)
-LinkP    := $(Link) -O3 -L$(LLVMLIBPROFILESOURCE) $(PROFILE)
-endif
-
-
 
 # Create one .o file from a bunch of .o files...
-Relink = ${LD} -r
+Relink = ${LIBTOOL} --mode=link $(CXX)
 
 # MakeSO - Create a .so file from a .o files...
-MakeSO   := $(CXX) $(MakeSharedObjectOption)
-MakeSOO  := $(MakeSO) -O3
-MakeSOP  := $(MakeSOO) $(PROFILE)
+#MakeSO   := $(LIBTOOL) --mode=link $(CXX) $(MakeSharedObjectOption)
+#MakeSOO  := $(MakeSO) -O3
+#MakeSOP  := $(MakeSOO) $(PROFILE)
+
+#
+# Configure where the item being compiled should go.
+#
+ifdef LIBRARYNAME
+Link := $(Link) -rpath $(DESTLIBCURRENT)
+endif
+
+ifdef TOOLNAME
+Link := $(Link) -rpath $(DESTTOOLCURRENT)
+endif
 
 # Create dependancy file from CPP file, send to stdout.
 Depend   := $(CXX) -MM -I$(LEVEL)/include $(CPPFLAGS) 
@@ -318,6 +375,10 @@
 ObjectsP := $(addprefix $(BUILD_OBJ_DIR)/Profile/,$(Objs))
 ObjectsG := $(addprefix $(BUILD_OBJ_DIR)/Debug/,$(Objs))
 
+LObjs := $(sort $(patsubst Debug/%.lo, %.lo, $(addsuffix .lo,$(notdir $(basename $(Source))))))
+LObjectsO := $(addprefix $(BUILD_OBJ_DIR)/Release/,$(LObjs))
+LObjectsP := $(addprefix $(BUILD_OBJ_DIR)/Profile/,$(LObjs))
+LObjectsG := $(addprefix $(BUILD_OBJ_DIR)/Debug/,$(LObjs))
 
 #---------------------------------------------------------
 # Handle the DIRS and PARALLEL_DIRS options
@@ -341,6 +402,9 @@
 	$(VERB) cd $(@D); $(MAKE) $(subst $(@D)/.make,,$@)
 endif
 
+###########################################################################
+# Library Build Rules:
+#
 #---------------------------------------------------------
 # Handle the LIBRARYNAME option - used when building libs...
 #---------------------------------------------------------
@@ -357,6 +421,7 @@
 #  it's built as a .o file, then all of the constituent .o files in it will be
 #  linked into tools (for example gccas) even if they only use one of the parts
 #  of it.  For this reason, sometimes it's useful to use libraries as .a files.
+###########################################################################
 
 ifdef LIBRARYNAME
 
@@ -373,6 +438,11 @@
 LIBNAME_OBJP := $(DESTLIBPROFILE)/$(LIBRARYNAME).o
 LIBNAME_OBJG := $(DESTLIBDEBUG)/$(LIBRARYNAME).o
 
+#--------------------------------------------------------------------
+# Library Targets
+#	Modify the top level targets to build the desired libraries.
+#--------------------------------------------------------------------
+
 # dynamic target builds a shared object version of the library...
 dynamic:: $(DESTLIBCURRENT)/lib$(LIBRARYNAME).so
 
@@ -386,33 +456,49 @@
 all:: $(DESTLIBCURRENT)/lib$(LIBRARYNAME).a
 endif
 
-$(LIBNAME_O): $(ObjectsO) $(LibSubDirs) $(DESTLIBRELEASE)/.dir
+#--------------------------------------------------------------------
+# Rules for building libraries
+#--------------------------------------------------------------------
+
+#
+# Rules for building dynamically linked libraries.
+#
+$(LIBNAME_O): $(LObjectsO) $(LibSubDirs) $(DESTLIBRELEASE)/.dir
 	@echo ======= Linking $(LIBRARYNAME) release library =======
-	$(VERB) $(MakeSOO) -o $@ $(ObjectsO) $(LibSubDirs) $(LibLinkOpts)
+	$(VERB) $(Link) -o $*.la $(LObjectsO) $(LibSubDirs) $(LibLinkOpts); \
+	$(VERB) $(LIBTOOL) --mode=install $(INSTALL) $*.la $(DESTLIBCURRENT);
 
-$(LIBNAME_P): $(ObjectsP) $(LibSubDirs) $(DESTLIBPROFILE)/.dir
+$(LIBNAME_P): $(LObjectsP) $(LibSubDirs) $(DESTLIBPROFILE)/.dir
 	@echo ======= Linking $(LIBRARYNAME) profile library =======
-	$(VERB) $(MakeSOP) -o $@ $(ObjectsP) $(LibSubDirs) $(LibLinkOpts)
+	$(VERB) $(Link) -o $*.la $(LObjectsP) $(LibSubDirs) $(LibLinkOpts); \
+	$(VERB) $(LIBTOOL) --mode=install $(INSTALL) $*.la $(DESTLIBCURRENT);
 
-$(LIBNAME_G): $(ObjectsG) $(LibSubDirs) $(DESTLIBDEBUG)/.dir
+$(LIBNAME_G): $(LObjectsG) $(LibSubDirs) $(DESTLIBDEBUG)/.dir
 	@echo ======= Linking $(LIBRARYNAME) debug library =======
-	$(VERB) $(MakeSO) -g -o $@ $(ObjectsG) $(LibSubDirs) $(LibLinkOpts)
+	$(VERB) $(Link) -o $*.la $(LObjectsG) $(LibSubDirs) $(LibLinkOpts); \
+	$(VERB) $(LIBTOOL) --mode=install $(INSTALL) $*.la $(DESTLIBCURRENT);
 
-$(LIBNAME_AO): $(ObjectsO) $(LibSubDirs) $(DESTLIBRELEASE)/.dir
+#
+# Rules for building static archive libraries.
+#
+$(LIBNAME_AO): $(LObjectsO) $(LibSubDirs) $(DESTLIBRELEASE)/.dir
 	@echo ======= Linking $(LIBRARYNAME) release library =======
-	@rm -f $@
-	$(VERB) $(AR) $@ $(ObjectsO) $(LibSubDirs)
+	@$(RM) -f $@
+	$(VERB) $(LinkO) -o $@ $(LObjectsO) $(LibSubDirs) -static
 
-$(LIBNAME_AP): $(ObjectsP) $(LibSubDirs) $(DESTLIBPROFILE)/.dir
+$(LIBNAME_AP): $(LObjectsP) $(LibSubDirs) $(DESTLIBPROFILE)/.dir
 	@echo ======= Linking $(LIBRARYNAME) profile library =======
-	@rm -f $@
-	$(VERB) $(AR) $@ $(ObjectsP) $(LibSubDirs)
+	@$(RM) -f $@
+	$(VERB) $(LinkP) -o $@ $(LObjectsP) $(LibSubDirs) -static
 
 $(LIBNAME_AG): $(ObjectsG) $(LibSubDirs) $(DESTLIBDEBUG)/.dir
 	@echo ======= Linking $(LIBRARYNAME) debug library =======
-	@rm -f $@
-	$(VERB) $(AR) $@ $(ObjectsG) $(LibSubDirs)
+	@$(RM) -f $@
+	$(VERB) $(LinkG) -o $@ $(LObjectsG) $(LibSubDirs) -static
 
+#
+# Rules for building .o libraries.
+#
 $(LIBNAME_OBJO): $(ObjectsO) $(LibSubDirs) $(DESTLIBRELEASE)/.dir
 	@echo "Linking $@"
 	$(VERB) $(Relink) -o $@ $(ObjectsO) $(LibSubDirs)
@@ -431,11 +517,16 @@
 # Create a TAGS database for emacs
 #------------------------------------------------------------------------
 
+ifdef ETAGS
 ifeq ($(LEVEL), .)
 tags:
-	etags -l c++ `find include lib tools -name '*.cpp' -o -name '*.h'`
+	$(ETAGS) -l c++ `find include lib tools -name '*.cpp' -o -name '*.h'`
 all:: tags
 endif
+else
+tags:
+	${ECHO} "Cannot build $@: The program etags is not installed"
+endif
 
 #------------------------------------------------------------------------
 # Handle the TOOLNAME option - used when building tool executables...
@@ -477,8 +568,14 @@
 USED_LIB_PATHS_G := $(addprefix $(DESTLIBDEBUG)/, $(STATICUSEDLIBS))
 USED_LIB_PATHS_O := $(addprefix $(DESTLIBRELEASE)/, $(STATICUSEDLIBS))
 USED_LIB_PATHS_P := $(addprefix $(DESTLIBPROFILE)/, $(STATICUSEDLIBS))
-LINK_OPTS        := $(TOOLLINKOPTS) $(PLATFORMLINKOPTS)
+#LINK_OPTS        := $(TOOLLINKOPTS) $(PLATFORMLINKOPTS)
 
+#
+# Libtool link options:
+#	Ensure that all binaries have their symbols exported so that they can
+#	by dlsym'ed.
+#
+LINK_OPTS := -export-dynamic $(TOOLLINKOPTS)
 
 
 
@@ -491,19 +588,19 @@
 
 all::   $(TOOLEXENAMES)
 clean::
-	$(VERB) rm -f $(TOOLEXENAMES)
+	$(VERB) $(RM) -f $(TOOLEXENAMES)
 
 $(TOOLEXENAME_G): $(ObjectsG) $(USED_LIB_PATHS_G) $(DESTTOOLDEBUG)/.dir
 	@echo ======= Linking $(TOOLNAME) debug executable $(STRIP_WARN_MSG)=======
-	$(VERB) $(LinkG) -o $@ $(ObjectsG) $(LIB_OPTS_G) $(LINK_OPTS)
+	$(VERB) $(LinkG) -o $@ $(ObjectsG) $(LIB_OPTS_G) $(LINK_OPTS) $(LIBS)
 
 $(TOOLEXENAME_O): $(ObjectsO) $(USED_LIB_PATHS_O) $(DESTTOOLRELEASE)/.dir
 	@echo ======= Linking $(TOOLNAME) release executable =======
-	$(VERB) $(LinkO) -o $@ $(ObjectsO) $(LIB_OPTS_O) $(LINK_OPTS)
+	$(VERB) $(LinkO) -o $@ $(ObjectsO) $(LIB_OPTS_O) $(LINK_OPTS) $(LIBS)
 
 $(TOOLEXENAME_P): $(ObjectsP) $(USED_LIB_PATHS_P) $(DESTTOOLPROFILE)/.dir
 	@echo ======= Linking $(TOOLNAME) profile executable =======
-	$(VERB) $(LinkP) -o $@ $(ObjectsP) $(LIB_OPTS_P) $(LINK_OPTS)
+	$(VERB) $(LinkP) -o $@ $(ObjectsP) $(LIB_OPTS_P) $(LINK_OPTS) $(LIBS)
 
 endif
 
@@ -536,6 +633,29 @@
 $(BUILD_OBJ_DIR)/Debug/%.o: $(SourceDir)%.c $(BUILD_OBJ_DIR)/Debug/.dir 
 	$(VERB) $(CompileCG) $< -o $@
 
+# Create .lo files in the ObjectFiles directory from the .cpp and .c files...
+$(BUILD_OBlJ_DIR)/Release/%.lo: $(SourceDir)%.cpp $(BUILD_OBJ_DIR)/Release/.dir
+	@echo "Compiling $<"
+	$(VERB) $(CompileO) $< -o $@
+
+$(BUILD_OBJ_DIR)/Release/%.lo: $(SourceDir)%.c $(BUILD_OBJ_DIR)/Release/.dir
+	$(VERB) $(CompileCO) $< -o $@
+
+$(BUILD_OBJ_DIR)/Profile/%.lo: $(SourceDir)%.cpp $(BUILD_OBJ_DIR)/Profile/.dir
+	@echo "Compiling $<"
+	$(VERB) $(CompileP) $< -o $@
+
+$(BUILD_OBJ_DIR)/Profile/%.lo: $(SourceDir)%.c $(BUILD_OBJ_DIR)/Profile/.dir
+	@echo "Compiling $<"
+	$(VERB) $(CompileCP) $< -o $@
+
+$(BUILD_OBJ_DIR)/Debug/%.lo: $(SourceDir)%.cpp $(BUILD_OBJ_DIR)/Debug/.dir
+	@echo "Compiling $<"
+	$(VERB) $(CompileG) $< -o $@
+
+$(BUILD_OBJ_DIR)/Debug/%.lo: $(SourceDir)%.c $(BUILD_OBJ_DIR)/Debug/.dir 
+	$(VERB) $(CompileCG) $< -o $@
+
 #
 # Rules for building lex/yacc files
 #
@@ -554,10 +674,10 @@
 # FIXME.  (f.e. char Buffer[10000]; )
 #
 %.cpp: %.l
-	$(FLEX) -t $< | sed '/^find_rule/d' | \
-                     sed 's/void yyunput/inline void yyunput/' | \
-          sed 's/void \*yy_flex_realloc/inline void *yy_flex_realloc/' | \
-          sed 's/#define YY_BUF_SIZE 16384/#define YY_BUF_SIZE (16384*64)/' > $@
+	$(FLEX) -t $< | $(SED) '/^find_rule/d' | \
+                     $(SED) 's/void yyunput/inline void yyunput/' | \
+          $(SED) 's/void \*yy_flex_realloc/inline void *yy_flex_realloc/' | \
+          $(SED) 's/#define YY_BUF_SIZE 16384/#define YY_BUF_SIZE (16384*64)/' > $@
 
 # Rule for building the bison parsers...
 %.c: %.y     # Cancel built-in rules for yacc
@@ -565,23 +685,28 @@
 %.cpp %.h : %.y
 	@echo Bison\'ing $<...
 	$(VERB) $(BISON) -v -d -p $(<:%Parser.y=%) $*.y
-	$(VERB) mv -f $*.tab.c $*.cpp
-	$(VERB) mv -f $*.tab.h $*.h
+	$(VERB) ${MV} -f $*.tab.c $*.cpp
+	$(VERB) ${MV} -f $*.tab.h $*.h
 
 # To create the directories...
 %/.dir:
-	$(VERB) mkdir -p $*
-	@date > $@
+	$(VERB) ${MKDIR} $*
+	@$(DATE) > $@
 
 # To create postscript files from dot files...
+ifdef DOT
 %.ps: %.dot
-	dot -Tps < $< > $@
+	${DOT} -Tps < $< > $@
+else
+%.ps: %.dot
+	${ECHO} "Cannot build $@: The program dot is not installed"
+endif
 
 # 'make clean' nukes the tree
 clean::
-	$(VERB) rm -rf $(BUILD_OBJ_DIR)/Debug $(BUILD_OBJ_DIR)/Release $(BUILD_OBJ_DIR)/Profile $(BUILD_OBJ_DIR)/Depend
-	$(VERB) rm -f core core.[0-9][0-9]* *.o *.d *.so *~ *.flc
-	$(VERB) rm -f $(LEX_OUTPUT) $(YACC_OUTPUT)
+	$(VERB) $(RM) -rf $(BUILD_OBJ_DIR)/Debug $(BUILD_OBJ_DIR)/Release $(BUILD_OBJ_DIR)/Profile $(BUILD_OBJ_DIR)/Depend
+	$(VERB) $(RM) -f core core.[0-9][0-9]* *.o *.d *.so *~ *.flc
+	$(VERB) $(RM) -f $(LEX_OUTPUT) $(YACC_OUTPUT)
 
 # If dependencies were generated for the file that included this file,
 # include the dependancies now...
@@ -592,12 +717,12 @@
 # Create dependencies for the *.cpp files...
 #$(SourceDepend): \x
 $(BUILD_OBJ_DIR)/Depend/%.d: $(SourceDir)%.cpp $(BUILD_OBJ_DIR)/Depend/.dir
-	$(VERB) $(Depend) $< | sed 's|$*\.o *|$(BUILD_OBJ_DIR)/Release/& $(BUILD_OBJ_DIR)/Profile/& $(BUILD_OBJ_DIR)/Debug/& $(BUILD_OBJ_DIR)/Depend/$(@F)|g' > $@
+	$(VERB) $(Depend) $< | $(SED) 's|$*\.o *|$(BUILD_OBJ_DIR)/Release/& $(BUILD_OBJ_DIR)/Profile/& $(BUILD_OBJ_DIR)/Debug/& $(BUILD_OBJ_DIR)/Depend/$(@F)|g' > $@
 
 # Create dependencies for the *.c files...
 #$(SourceDepend): \x
 $(BUILD_OBJ_DIR)/Depend/%.d: $(SourceDir)%.c $(BUILD_OBJ_DIR)/Depend/.dir
-	$(VERB) $(DependC) $< | sed 's|$*\.o *|Release/& Profile/& Debug/& Depend/$(@F)|g' > $@
+	$(VERB) $(DependC) $< | $(SED) 's|$*\.o *|Release/& Profile/& Debug/& Depend/$(@F)|g' > $@
 
 ifneq ($(SourceDepend),)
 -include $(SourceDepend)




From lattner at cs.uiuc.edu  Mon Jun 23 14:17:02 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Mon Jun 23 14:17:02 2003
Subject: [llvm-commits] CVS: llvm/lib/VMCore/PassManagerT.h
Message-ID: <200306231916.OAA20598@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/VMCore:

PassManagerT.h updated: 1.38 -> 1.39

---
Log message:

Add support for GCC 3.3


---
Diffs of the changes:

Index: llvm/lib/VMCore/PassManagerT.h
diff -u llvm/lib/VMCore/PassManagerT.h:1.38 llvm/lib/VMCore/PassManagerT.h:1.39
--- llvm/lib/VMCore/PassManagerT.h:1.38	Thu Apr 24 15:07:38 2003
+++ llvm/lib/VMCore/PassManagerT.h	Mon Jun 23 14:16:20 2003
@@ -130,8 +130,8 @@
   typedef typename Traits::BatcherClass BatcherClass;
   typedef typename Traits::ParentClass   ParentClass;
 
-  friend typename Traits::PassClass;
-  friend typename Traits::SubPassClass;  
+  friend class PassManagerTraits<UnitType>::PassClass;
+  friend class PassManagerTraits<UnitType>::SubPassClass;  
   friend class Traits;
   friend class ImmutablePass;
 




From gaeke at cs.uiuc.edu  Mon Jun 23 14:43:01 2003
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Mon Jun 23 14:43:01 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/Interpreter/ExternalFunctions.cpp
Message-ID: <200306231942.OAA21346@neo.cs.uiuc.edu>


Changes in directory llvm/tools/lli/Interpreter:

ExternalFunctions.cpp updated: 1.53 -> 1.54

---
Log message:

Include <cmath> instead of <math.h>
Remove isnan; it's too unportable to handle cleanly at this point.


---
Diffs of the changes:

Index: llvm/tools/lli/Interpreter/ExternalFunctions.cpp
diff -u llvm/tools/lli/Interpreter/ExternalFunctions.cpp:1.53 llvm/tools/lli/Interpreter/ExternalFunctions.cpp:1.54
--- llvm/tools/lli/Interpreter/ExternalFunctions.cpp:1.53	Tue Jun 17 14:59:17 2003
+++ llvm/tools/lli/Interpreter/ExternalFunctions.cpp	Mon Jun 23 14:41:55 2003
@@ -19,7 +19,7 @@
 #include <map>
 #include <dlfcn.h>
 #include <link.h>
-#include <math.h>
+#include <cmath>
 #include <stdio.h>
 using std::vector;
 
@@ -217,14 +217,6 @@
   return GV;
 }
 
-// int isnan(double value);
-GenericValue lle_X_isnan(FunctionType *F, const vector<GenericValue> &Args) {
-  assert(Args.size() == 1);
-  GenericValue GV;
-  GV.IntVal = std::isnan(Args[0].DoubleVal);
-  return GV;
-}
-
 // double floor(double)
 GenericValue lle_X_floor(FunctionType *M, const vector<GenericValue> &Args) {
   assert(Args.size() == 1);
@@ -744,7 +736,6 @@
   FuncNames["lle_X_pow"]          = lle_X_pow;
   FuncNames["lle_X_exp"]          = lle_X_exp;
   FuncNames["lle_X_log"]          = lle_X_log;
-  FuncNames["lle_X_isnan"]        = lle_X_isnan;
   FuncNames["lle_X_floor"]        = lle_X_floor;
   FuncNames["lle_X_srand"]        = lle_X_srand;
   FuncNames["lle_X_drand48"]      = lle_X_drand48;




From gaeke at cs.uiuc.edu  Mon Jun 23 15:00:03 2003
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Mon Jun 23 15:00:03 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/CBackend/Makefile
Message-ID: <200306231959.OAA08571@zion.cs.uiuc.edu>


Changes in directory llvm/test/Regression/CBackend:

Makefile updated: 1.7 -> 1.8

---
Log message:

Allow .llx testcases (i.e., those that contain explicit RUN: stmts.)


---
Diffs of the changes:

Index: llvm/test/Regression/CBackend/Makefile
diff -u llvm/test/Regression/CBackend/Makefile:1.7 llvm/test/Regression/CBackend/Makefile:1.8
--- llvm/test/Regression/CBackend/Makefile:1.7	Mon May 12 10:33:52 2003
+++ llvm/test/Regression/CBackend/Makefile	Mon Jun 23 14:58:50 2003
@@ -10,9 +10,9 @@
 .PRECIOUS: Output/%.c
 
 TESTS := $(wildcard *.ll)
+FTESTS := $(wildcard *.llx)   # Freeform tests
 
-all:: $(addprefix Output/, $(TESTS:%.ll=%.to))
-
+all:: $(addprefix Output/, $(TESTS:%.ll=%.to)) $(addprefix Output/, $(FTESTS:%.llx=%.llx.out))
 
 Output/%.to: Output/%.c
 	$(CC) -c -Werror $< -o $@ || \
@@ -21,3 +21,7 @@
 Output/%.c: %.ll Output/.dir $(LAS) $(LDIS)
 	$(LAS) < $< | $(LDIS) -c > $@ || \
 		(rm -f $@; $(FAILURE) $@ )
+
+Output/%.llx.out: %.llx Output/.dir $(LAS) $(LDIS)
+	-$(TESTRUNR) $<
+




From gaeke at cs.uiuc.edu  Mon Jun 23 15:00:06 2003
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Mon Jun 23 15:00:06 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/CBackend/2003-06-23-PromotedExprs.llx
Message-ID: <200306231959.OAA08564@zion.cs.uiuc.edu>


Changes in directory llvm/test/Regression/CBackend:

2003-06-23-PromotedExprs.llx added (r1.1)

---
Log message:

New test case for C Writer regression found in 256.bzip2


---
Diffs of the changes:

Index: llvm/test/Regression/CBackend/2003-06-23-PromotedExprs.llx
diff -c /dev/null llvm/test/Regression/CBackend/2003-06-23-PromotedExprs.llx:1.1
*** /dev/null	Mon Jun 23 14:59:27 2003
--- llvm/test/Regression/CBackend/2003-06-23-PromotedExprs.llx	Mon Jun 23 14:59:17 2003
***************
*** 0 ****
--- 1,16 ----
+ 
+ ; RUN: as < %s | dis -c > Output/%s.cbe.c
+ ; RUN: gcc -B/usr/bin/ Output/%s.cbe.c -o Output/%s.cbe
+ ; RUN: Output/%s.cbe
+ 
+ bool %doTest(ubyte %x) {
+ 	%dec.0 = add ubyte %x, 255
+     %tmp.1001 = cast ubyte %dec.0 to bool
+     ret bool %tmp.1001
+ }
+ 
+ int %main () {
+     %result = call bool %doTest(ubyte 1)
+     %p = cast bool %result to int
+     ret int %p
+ }




From gaeke at cs.uiuc.edu  Mon Jun 23 15:02:03 2003
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Mon Jun 23 15:02:03 2003
Subject: [llvm-commits] CVS: llvm/lib/CWriter/Writer.cpp
Message-ID: <200306232001.PAA08587@zion.cs.uiuc.edu>


Changes in directory llvm/lib/CWriter:

Writer.cpp updated: 1.97 -> 1.98

---
Log message:

Fix 2003-06-23-PromotedExprs.llx -- if we are adding two bytes we better
explicitly cast the result to be a byte, or C will gleefully promote it
to int.


---
Diffs of the changes:

Index: llvm/lib/CWriter/Writer.cpp
diff -u llvm/lib/CWriter/Writer.cpp:1.97 llvm/lib/CWriter/Writer.cpp:1.98
--- llvm/lib/CWriter/Writer.cpp:1.97	Tue Jun 17 18:55:31 2003
+++ llvm/lib/CWriter/Writer.cpp	Mon Jun 23 15:00:51 2003
@@ -1029,6 +1029,16 @@
 void CWriter::visitBinaryOperator(Instruction &I) {
   // binary instructions, shift instructions, setCond instructions.
   assert(!isa<PointerType>(I.getType()));
+
+  // We must cast the results of binary operations which might be promoted.
+  bool needsCast = false;
+  if ((I.getType() == Type::UByteTy) || (I.getType() == Type::SByteTy)
+      || (I.getType() == Type::UShortTy) || (I.getType() == Type::ShortTy)) {
+    needsCast = true;
+    Out << "((";
+    printType(Out, I.getType(), "", false, false);
+    Out << ")(";
+  }
       
   writeOperand(I.getOperand(0));
 
@@ -1053,6 +1063,10 @@
   }
 
   writeOperand(I.getOperand(1));
+
+  if (needsCast) {
+    Out << "))";
+  }
 }
 
 void CWriter::visitCastInst(CastInst &I) {




From lattner at cs.uiuc.edu  Mon Jun 23 16:49:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Mon Jun 23 16:49:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/InstCombine/cast.ll
Message-ID: <200306232148.QAA25554@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/InstCombine:

cast.ll updated: 1.6 -> 1.7

---
Log message:

Test cases for when casts to bool can be eliminated


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/InstCombine/cast.ll
diff -u llvm/test/Regression/Transforms/InstCombine/cast.ll:1.6 llvm/test/Regression/Transforms/InstCombine/cast.ll:1.7
--- llvm/test/Regression/Transforms/InstCombine/cast.ll:1.6	Mon Sep 23 18:39:17 2002
+++ llvm/test/Regression/Transforms/InstCombine/cast.ll	Mon Jun 23 16:48:26 2003
@@ -68,3 +68,16 @@
 	%c2 = cast uint %c1 to short
 	ret short %c2
 }
+
+bool %test11(ubyte %A, ubyte %B) {
+        %C = sub ubyte %A, %B
+        %D = cast ubyte %C to bool    ; == setne A, B
+        ret bool %D
+}
+
+bool %test12(ubyte %A) {
+        %B = add ubyte %A, 255
+        %C = cast ubyte %B to bool    ; === A != 1
+        ret bool %C
+}
+




From lattner at cs.uiuc.edu  Mon Jun 23 16:50:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Mon Jun 23 16:50:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/InstCombine/cast.ll
Message-ID: <200306232149.QAA25637@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/InstCombine:

cast.ll updated: 1.7 -> 1.8

---
Log message:

Yet another testcase


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/InstCombine/cast.ll
diff -u llvm/test/Regression/Transforms/InstCombine/cast.ll:1.7 llvm/test/Regression/Transforms/InstCombine/cast.ll:1.8
--- llvm/test/Regression/Transforms/InstCombine/cast.ll:1.7	Mon Jun 23 16:48:26 2003
+++ llvm/test/Regression/Transforms/InstCombine/cast.ll	Mon Jun 23 16:49:25 2003
@@ -81,3 +81,9 @@
         ret bool %C
 }
 
+bool %test13(ubyte %A, ubyte %b) {
+        %B = add ubyte %A, %b
+        %C = cast ubyte %B to bool    ; === A != 1
+        ret bool %C
+}
+




From lattner at cs.uiuc.edu  Mon Jun 23 16:54:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Mon Jun 23 16:54:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/InstCombine/cast.ll
Message-ID: <200306232153.QAA29864@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/InstCombine:

cast.ll updated: 1.8 -> 1.9

---
Log message:

Fix comment


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/InstCombine/cast.ll
diff -u llvm/test/Regression/Transforms/InstCombine/cast.ll:1.8 llvm/test/Regression/Transforms/InstCombine/cast.ll:1.9
--- llvm/test/Regression/Transforms/InstCombine/cast.ll:1.8	Mon Jun 23 16:49:25 2003
+++ llvm/test/Regression/Transforms/InstCombine/cast.ll	Mon Jun 23 16:52:59 2003
@@ -83,7 +83,7 @@
 
 bool %test13(ubyte %A, ubyte %b) {
         %B = add ubyte %A, %b
-        %C = cast ubyte %B to bool    ; === A != 1
+        %C = cast ubyte %B to bool    ; === A != -b
         ret bool %C
 }
 




From lattner at cs.uiuc.edu  Mon Jun 23 17:01:12 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Mon Jun 23 17:01:12 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Scalar/InstructionCombining.cpp
Message-ID: <200306232200.RAA00349@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Scalar:

InstructionCombining.cpp updated: 1.91 -> 1.92

---
Log message:

Implement new transforms:
  Replace (cast (sub A, B) to bool) -> (setne A, B)
  Replace (cast (add A, B) to bool) -> (setne A, -B)



---
Diffs of the changes:

Index: llvm/lib/Transforms/Scalar/InstructionCombining.cpp
diff -u llvm/lib/Transforms/Scalar/InstructionCombining.cpp:1.91 llvm/lib/Transforms/Scalar/InstructionCombining.cpp:1.92
--- llvm/lib/Transforms/Scalar/InstructionCombining.cpp:1.91	Sat Jun 21 18:12:02 2003
+++ llvm/lib/Transforms/Scalar/InstructionCombining.cpp	Mon Jun 23 16:59:52 2003
@@ -898,15 +898,17 @@
 // CastInst simplification
 //
 Instruction *InstCombiner::visitCastInst(CastInst &CI) {
+  Value *Src = CI.getOperand(0);
+
   // If the user is casting a value to the same type, eliminate this cast
   // instruction...
-  if (CI.getType() == CI.getOperand(0)->getType())
-    return ReplaceInstUsesWith(CI, CI.getOperand(0));
+  if (CI.getType() == Src->getType())
+    return ReplaceInstUsesWith(CI, Src);
 
   // If casting the result of another cast instruction, try to eliminate this
   // one!
   //
-  if (CastInst *CSrc = dyn_cast<CastInst>(CI.getOperand(0))) {
+  if (CastInst *CSrc = dyn_cast<CastInst>(Src)) {
     if (isEliminableCastOfCast(CI, CSrc)) {
       // This instruction now refers directly to the cast's src operand.  This
       // has a good chance of making CSrc dead.
@@ -933,7 +935,7 @@
   // If casting the result of a getelementptr instruction with no offset, turn
   // this into a cast of the original pointer!
   //
-  if (GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(CI.getOperand(0))) {
+  if (GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(Src)) {
     bool AllZeroOperands = true;
     for (unsigned i = 1, e = GEP->getNumOperands(); i != e; ++i)
       if (!isa<Constant>(GEP->getOperand(i)) ||
@@ -944,6 +946,33 @@
     if (AllZeroOperands) {
       CI.setOperand(0, GEP->getOperand(0));
       return &CI;
+    }
+  }
+
+  // If this is a cast to bool (which is effectively a "!=0" test), then we can
+  // perform a few optimizations...
+  //
+  if (CI.getType() == Type::BoolTy) {
+    if (BinaryOperator *BO = dyn_cast<BinaryOperator>(Src)) {
+      Value *Op0 = BO->getOperand(0), *Op1 = BO->getOperand(1);
+
+      // Replace (cast (sub A, B) to bool) with (setne A, B)
+      if (BO->getOpcode() == Instruction::Sub)
+        return new SetCondInst(Instruction::SetNE, Op0, Op1);
+
+      // Replace (cast (add A, B) to bool) with (setne A, -B) if B is
+      // efficiently invertible, or if the add has just this one use.
+      if (BO->getOpcode() == Instruction::Add)
+        if (Value *NegVal = dyn_castNegVal(Op1))
+          return new SetCondInst(Instruction::SetNE, Op0, NegVal);
+        else if (Value *NegVal = dyn_castNegVal(Op0))
+          return new SetCondInst(Instruction::SetNE, NegVal, Op1);
+        else if (BO->use_size() == 1) {
+          Instruction *Neg = BinaryOperator::createNeg(Op1, BO->getName());
+          BO->setName("");
+          InsertNewInstBefore(Neg, CI);
+          return new SetCondInst(Instruction::SetNE, Op0, Neg);
+        }
     }
   }
 




From lattner at cs.uiuc.edu  Mon Jun 23 17:18:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Mon Jun 23 17:18:01 2003
Subject: [llvm-commits] CVS: llvm/test/Programs/External/SPEC/CFP2000/Makefile
Message-ID: <200306232217.RAA03845@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Programs/External/SPEC/CFP2000:

Makefile updated: 1.1 -> 1.2

---
Log message:

Add support for disabled directories


---
Diffs of the changes:

Index: llvm/test/Programs/External/SPEC/CFP2000/Makefile
diff -u llvm/test/Programs/External/SPEC/CFP2000/Makefile:1.1 llvm/test/Programs/External/SPEC/CFP2000/Makefile:1.2
--- llvm/test/Programs/External/SPEC/CFP2000/Makefile:1.1	Sun Jun  1 21:45:11 2003
+++ llvm/test/Programs/External/SPEC/CFP2000/Makefile	Mon Jun 23 17:17:25 2003
@@ -1,3 +1,3 @@
 LEVEL = ../../../../..
-DIRS := $(sort $(filter-out CVS/, $(wildcard */)))
+PARALLEL_DIRS := $(filter-out %-disabled/, $(sort $(filter-out CVS/, $(wildcard */))))
 include ${LEVEL}/test/Makefile.tests




From lattner at cs.uiuc.edu  Mon Jun 23 17:37:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Mon Jun 23 17:37:01 2003
Subject: [llvm-commits] CVS: llvm/test/Programs/TEST.nightly.Makefile
Message-ID: <200306232236.RAA06300@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Programs:

TEST.nightly.Makefile updated: 1.15 -> 1.16

---
Log message:

Prevent crafty from stopping the nightly tester!


---
Diffs of the changes:

Index: llvm/test/Programs/TEST.nightly.Makefile
diff -u llvm/test/Programs/TEST.nightly.Makefile:1.15 llvm/test/Programs/TEST.nightly.Makefile:1.16
--- llvm/test/Programs/TEST.nightly.Makefile:1.15	Sat Jun 21 21:38:29 2003
+++ llvm/test/Programs/TEST.nightly.Makefile	Mon Jun 23 17:36:09 2003
@@ -21,8 +21,9 @@
 # Compilation tests
 $(PROGRAMS_TO_TEST:%=Output/%.nightly.compile.report.txt): \
 Output/%.nightly.compile.report.txt: Output/%.llvm.bc $(LGCCAS)
-	@echo '$(LGCCAS) Output/$*.linked.rll -o /dev/null $(TIMEOPT) > $@ 2>&1'
-	@if ($(LGCCAS) Output/$*.linked.rll -o /dev/null $(TIMEOPT) > $@ 2>&1)\
+	@echo > $@
+	@echo '$(LGCCAS) Output/$*.linked.rll -o/dev/null $(TIMEOPT) >> $@ 2>&1'
+	@if ($(LGCCAS) Output/$*.linked.rll -o /dev/null $(TIMEOPT) >> $@ 2>&1)\
 	;then \
 	  echo "TEST-PASS: compile $(RELDIR)/$*" >> $@;\
 	  echo -n "TEST-RESULT-compile: " >> $@;\
@@ -42,14 +43,16 @@
 # NAT tests
 $(PROGRAMS_TO_TEST:%=Output/%.nightly.nat.report.txt): \
 Output/%.nightly.nat.report.txt: Output/%.out-nat
+	@echo > $@
 	echo -n "TEST-RESULT-nat-time: " >> $@
 	-grep "^real" Output/$*.out-nat.time >> $@
 
 # LLC tests
 $(PROGRAMS_TO_TEST:%=Output/%.nightly.llc.report.txt): \
 Output/%.nightly.llc.report.txt: Output/%.llvm.bc $(LLC)
+	@echo > $@
 	@echo 'time -p $(LLC) -f $(TIMEOPT) $< -o /dev/null) > $@ 2>&1'
-	@if (time -p $(LLC) -f $(TIMEOPT) $< -o /dev/null) > $@ 2>&1; then \
+	@if (time -p $(LLC) -f $(TIMEOPT) $< -o /dev/null) >> $@ 2>&1; then \
 	  echo "TEST-PASS: llc $(RELDIR)/$*" >> $@;\
 	  echo -n "TEST-RESULT-llc: " >> $@;\
 	  grep "Total Execution Time" $@.info >> $@;\
@@ -65,7 +68,8 @@
 # CBE tests
 $(PROGRAMS_TO_TEST:%=Output/%.nightly.cbe.report.txt): \
 Output/%.nightly.cbe.report.txt: Output/%.llvm.bc Output/%.exe-cbe $(LDIS)
-	-head -n 100 Output/$*.exe-cbe > $@
+	@echo > $@
+	-head -n 100 Output/$*.exe-cbe >> $@
 	@if test -f Output/$*.exe-cbe; then \
 	  echo "TEST-PASS: cbe $(RELDIR)/$*" >> $@;\
           echo "TEST-RESULT-cbe: YES" >> $@;\
@@ -79,7 +83,8 @@
 # JIT tests
 $(PROGRAMS_TO_TEST:%=Output/%.nightly.jit.report.txt): \
 Output/%.nightly.jit.report.txt: Output/%.llvm.bc Output/%.exe-jit $(LLI)
-	-head -n 100 Output/$*.exe-jit > $@
+	@echo > $@
+	-head -n 100 Output/$*.exe-jit >> $@
 	@if test -f Output/$*.exe-jit; then \
           echo "TEST-PASS: jit $(RELDIR)/$*" >> $@;\
 	  echo -n "TEST-RESULT-jit-time: " >> $@;\




From jstanley at cs.uiuc.edu  Mon Jun 23 17:45:01 2003
From: jstanley at cs.uiuc.edu (Joel Stanley)
Date: Mon Jun 23 17:45:01 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Reoptimizer/BinInterface/sparc9.h
Message-ID: <200306232244.RAA02526@trinity.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Reoptimizer/BinInterface:

sparc9.h updated: 1.14 -> 1.15

---
Log message:



---
Diffs of the changes:

Index: llvm/include/llvm/Reoptimizer/BinInterface/sparc9.h
diff -u llvm/include/llvm/Reoptimizer/BinInterface/sparc9.h:1.14 llvm/include/llvm/Reoptimizer/BinInterface/sparc9.h:1.15
--- llvm/include/llvm/Reoptimizer/BinInterface/sparc9.h:1.14	Sat May 31 17:16:52 2003
+++ llvm/include/llvm/Reoptimizer/BinInterface/sparc9.h	Mon Jun 23 17:44:25 2003
@@ -430,12 +430,111 @@
    MK_FLD(INSTR_I, 1)             |            \
    MK_FLD(INSTR_OP3, OP3_ADD))
 
-#define MK_INSTR_BRANCH_RS1(cond)               \
+#define MK_INSTR_BRANCH_RS1(cond)              \
    (MK_FLD(INSTR_OP, OP_BRANCH)   |            \
     MK_FLD(INSTR_BPR, BPR)        |            \
     MK_FLD(INSTR_BPR_COND, cond))
 
 #define MK_INSTR_BRANCH_NO_RS1(cond)           \
    (MK_FLD(INSTR_OP, OP_BRANCH)   |            \
-    MK_FLD(INSTR_INT_BR, Bicc)        |            \
+    MK_FLD(INSTR_INT_BR, Bicc)    |            \
     MK_FLD(INSTR_INTBR_COND, cond))
+
+// Add a register to a register
+
+#define MK_ADD_R_R(dreg, sreg1, sreg2)         \
+   (MK_FLD(INSTR_OP, OP_2)        |            \
+    MK_FLD(INSTR_RD, dreg)        |            \
+    MK_FLD(INSTR_OP3, OP3_ADD)    |            \
+    MK_FLD(INSTR_RS1, sreg1)      |            \
+    MK_FLD(INSTR_I, 0)            |            \
+    MK_FLD(INSTR_RS2, sreg2))
+
+// Construct immediate-valued logical operation
+
+#define MK_LOGIC_IMM(op3, dreg, sreg, imm)  \
+ (MK_FLD(INSTR_OP,     OP_2) |              \
+  MK_FLD(INSTR_RD,     dreg) |              \
+  MK_FLD(INSTR_RS1,    sreg) |              \
+  MK_FLD(INSTR_SIMM13, imm)  |              \
+  MK_FLD(INSTR_I,      1)    |              \
+  MK_FLD(INSTR_OP3,    op3))
+
+// Construct reg-to-reg logical operation
+
+#define MK_LOGIC(op3, dreg, sreg1, sreg2) \
+ (MK_FLD(INSTR_OP,  OP_2)  |              \
+  MK_FLD(INSTR_RD,  dreg)  |              \
+  MK_FLD(INSTR_RS1, sreg1) |              \
+  MK_FLD(INSTR_RS2, sreg2) |              \
+  MK_FLD(INSTR_I,   0)     |              \
+  MK_FLD(INSTR_OP3, op3))
+
+// Construct sethi instruction
+
+#define MK_SETHI(dreg, imm)                \
+  (MK_FLD(INSTR_OP, OP_BRANCH)  |          \
+   MK_FLD(INSTR_RD, dreg)       |          \
+   MK_FLD(INSTR_OP2, OP2_SETHI) |          \
+   MK_FLD(INSTR_IMM22, imm))
+
+// Construct S{LL,RL,RA}X shift instruction. 
+
+#define MK_SHIFTX(op3, dreg, sreg, shcnt)   \
+  (MK_FLD(INSTR_OP, OP_2)  |                \
+   MK_FLD(INSTR_RD, dreg)  |                \
+   MK_FLD(INSTR_OP3, op3)  |                \
+   MK_FLD(INSTR_RS1, sreg) |                \
+   MK_FLD(INSTR_I, 1)      |                \
+   MK_FLD(INSTR_X, 1)      |                \
+   MK_FLD(INSTR_SHCNT64, shcnt))
+
+#define MK_STORE_IMM(srcreg, basereg, offset)  \
+  (MK_FLD(INSTR_OP, OP_3)         |            \
+   MK_FLD(INSTR_RD, srcreg)       |            \
+   MK_FLD(INSTR_OP3, OP3_STX)     |            \
+   MK_FLD(INSTR_RS1, basereg)     |            \
+   MK_FLD(INSTR_I, 1)             |            \
+   MK_FLD(INSTR_SIMM13, offset))
+
+#define MK_LOAD_IMM(destreg, basereg, offset)  \
+  (MK_FLD(INSTR_OP, OP_3)         |            \
+   MK_FLD(INSTR_RD, destreg)      |            \
+   MK_FLD(INSTR_OP3, OP3_LDX)     |            \
+   MK_FLD(INSTR_RS1, basereg)     |            \
+   MK_FLD(INSTR_I, 1)             |            \
+   MK_FLD(INSTR_SIMM13, offset))
+
+// Construct save instruction
+
+#define MK_SAVE_IMM(dreg, sreg, imm)        \
+  (MK_FLD(INSTR_OP, OP_2)         |         \
+   MK_FLD(INSTR_RD, dreg)         |         \
+   MK_FLD(INSTR_OP3, OP3_SAVE)    |         \
+   MK_FLD(INSTR_RS1, sreg)        |         \
+   MK_FLD(INSTR_I, 1)             |         \
+   MK_FLD(INSTR_SIMM13, imm))
+
+#define MK_SAVE_REG(dreg, sreg1, sreg2)     \
+  (MK_FLD(INSTR_OP, OP_2)         |         \
+   MK_FLD(INSTR_RD, dreg)         |         \
+   MK_FLD(INSTR_OP3, OP3_SAVE)    |         \
+   MK_FLD(INSTR_RS1, sreg1)       |         \
+   MK_FLD(INSTR_I, 0)             |         \
+   MK_FLD(INSTR_RS2, sreg2))
+
+#define MK_RESTORE_IMM(dreg, sreg, imm)     \
+  (MK_FLD(INSTR_OP, OP_2)         |         \
+   MK_FLD(INSTR_RD, dreg)         |         \
+   MK_FLD(INSTR_OP3, OP3_RESTORE) |         \
+   MK_FLD(INSTR_RS1, sreg)        |         \
+   MK_FLD(INSTR_I, 1)             |         \
+   MK_FLD(INSTR_SIMM13, imm))
+
+#define MK_JMPL_INDIRECT(dreg, sreg, imm)   \
+  (MK_FLD(INSTR_OP, OP_2)         |         \
+   MK_FLD(INSTR_RD, dreg)         |         \
+   MK_FLD(INSTR_OP3, OP3_JMPL)    |         \
+   MK_FLD(INSTR_RS1, sreg)        |         \
+   MK_FLD(INSTR_I, 1)             |         \
+   MK_FLD(INSTR_SIMM13, imm))




From jstanley at cs.uiuc.edu  Mon Jun 23 18:32:00 2003
From: jstanley at cs.uiuc.edu (Joel Stanley)
Date: Mon Jun 23 18:32:00 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Reoptimizer/BinInterface/bitmath.h
Message-ID: <200306232331.SAA03187@trinity.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Reoptimizer/BinInterface:

bitmath.h updated: 1.9 -> 1.10

---
Log message:



---
Diffs of the changes:

Index: llvm/include/llvm/Reoptimizer/BinInterface/bitmath.h
diff -u llvm/include/llvm/Reoptimizer/BinInterface/bitmath.h:1.9 llvm/include/llvm/Reoptimizer/BinInterface/bitmath.h:1.10
--- llvm/include/llvm/Reoptimizer/BinInterface/bitmath.h:1.9	Sat May 31 17:16:41 2003
+++ llvm/include/llvm/Reoptimizer/BinInterface/bitmath.h	Mon Jun 23 18:31:47 2003
@@ -16,6 +16,8 @@
 #ifndef __BITMATH_H__
 #define __BITMATH_H__
 
+#include "Support/DataTypes.h"
+
 //*********************************
 //	Misc functions
 //*********************************
@@ -29,7 +31,7 @@
 //*********************************
 #define FLD_UPPER(FLD_DEF) (1 ? FLD_DEF)
 #define FLD_LOWER(FLD_DEF) (0 ? FLD_DEF)
-#define MASKBELOW(V)       ((1 << V) - 1)
+#define MASKBELOW(V)       ((((uint64_t) 1) << V) - 1) // 64-bit '1' constant here to avoid overflow warnings
 #define MASKEQBELOW(V)	   ((1 << V) | MASKBELOW(V)) //masks off everything ABOVE
 #define RD_FLD(x, FLD)     ((x & MASKEQBELOW(FLD_UPPER(FLD)))  >> FLD_LOWER(FLD))
 #define MK_FLD(FLD, val)   ((val << FLD_LOWER(FLD)) & MASKEQBELOW(FLD_UPPER(FLD)))
@@ -129,7 +131,9 @@
    return ((0xFFFFAA50 >> w) & 3)+n;   // this is a LUT
 }
 
-
+static unsigned LOW10(unsigned value)    { return value & 0x000003ff;       }
+static unsigned HIGH22(unsigned value)   { return value >> 10;              }
+static unsigned HIGHWORD(uint64_t value) { return (unsigned) (value >> 32); }
+static unsigned LOWWORD(uint64_t value)  { return (unsigned) value;         }
 
 #endif
-




From jstanley at cs.uiuc.edu  Mon Jun 23 21:47:01 2003
From: jstanley at cs.uiuc.edu (Joel Stanley)
Date: Mon Jun 23 21:47:01 2003
Subject: [llvm-commits] CVS: llvm/test/Libraries/libinstr/tracelib.c
Message-ID: <200306240246.VAA20238@trinity.cs.uiuc.edu>


Changes in directory llvm/test/Libraries/libinstr:

tracelib.c updated: 1.3 -> 1.4

---
Log message:

Prevented inclusion of (non-existent) stdint.h on Sparc.


---
Diffs of the changes:

Index: llvm/test/Libraries/libinstr/tracelib.c
diff -u llvm/test/Libraries/libinstr/tracelib.c:1.3 llvm/test/Libraries/libinstr/tracelib.c:1.4
--- llvm/test/Libraries/libinstr/tracelib.c:1.3	Tue May 27 16:43:14 2003
+++ llvm/test/Libraries/libinstr/tracelib.c	Mon Jun 23 21:46:47 2003
@@ -10,7 +10,9 @@
 #include <stdlib.h>
 #include <stdio.h>
 #include <string.h>
+#ifndef sun
 #include <stdint.h>
+#endif
 
 /*===---------------------------------------------------------------------=====
  * HASH FUNCTIONS




From jstanley at cs.uiuc.edu  Tue Jun 24 10:30:02 2003
From: jstanley at cs.uiuc.edu (Joel Stanley)
Date: Tue Jun 24 10:30:02 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Reoptimizer/inst/Makefile Test1.c
Message-ID: <200306241529.KAA00266@trinity.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Reoptimizer/inst:

Makefile updated: 1.5 -> 1.5.2.1
Test1.c updated: 1.4.2.1 -> 1.4.2.2

---
Log message:



---
Diffs of the changes:

Index: llvm/test/Regression/Reoptimizer/inst/Makefile
diff -u llvm/test/Regression/Reoptimizer/inst/Makefile:1.5 llvm/test/Regression/Reoptimizer/inst/Makefile:1.5.2.1
--- llvm/test/Regression/Reoptimizer/inst/Makefile:1.5	Thu May 22 22:25:43 2003
+++ llvm/test/Regression/Reoptimizer/inst/Makefile	Tue Jun 24 10:29:11 2003
@@ -1,4 +1,5 @@
-LEVEL = /localhome/$(USER)/Research/llvm
+LEVEL = ../../../../
+#LEVEL = /localhome/$(USER)/Research/llvm
 
 CC = /usr/dcs/software/evaluation/bin/gcc
 CXX = /usr/dcs/software/evaluation/bin/g++
@@ -10,8 +11,8 @@
 	   $(LEVEL)/lib/Debug/libtracecache.a   \
 	   $(LEVEL)/lib/Debug/libbininterface.a
 
-
-LIBINSTRBC = /home/vadve/lattner/cvs/gcc_install_sparc/llvm/lib/libinstr.bc
+CRTEND = $(LLVM_LIB_SEARCH_PATH)/crtend.o
+LIBINSTRBC = $(LEVEL)/test/Libraries/Output/libinstr.bc
 
 Test1: Test1.o $(PERFOBJS)
         # NB: The only reason we don't include the PAPI library in the list of exclusions
@@ -19,8 +20,8 @@
         #     tables, and thus will never try to instrument the PAPI functions, which are
         #     dynamically linked.
 
-	mkexcl $(PERFOBJS) libinstr.o > mkexcl.cpp
-	$(CXX) -o Test1 $^ mkexcl.cpp -L$(PAPIDIR) -lpapi -lelf
+	mkexcl $(PERFOBJS) libinstr.o crtend.o > mkexcl.cpp
+	$(CXX) -o Test1 $^ mkexcl.cpp crtend.o -L$(PAPIDIR) -lpapi -lelf
 
 Test1.o: Test1.opt.c
 	$(CC) -c Test1.opt.c -o Test1.o
@@ -28,12 +29,16 @@
 Test1.opt.c: Test1.opt.bc
 	dis -c < Test1.opt.bc > Test1.opt.c
 
-Test1.opt.bc: Test1.bc libinstr.o $(LEVEL)/lib/Debug/libperf.so
+Test1.opt.bc: Test1.bc libinstr.o crtend.o $(LEVEL)/lib/Debug/libperf.so
 	opt -load $(LEVEL)/lib/Debug/libperf.so -pp1 -emitfuncs -deadtypeelim < Test1.bc > Test1.opt.o
 	gccld -disable-internalize $(LIBINSTRBC) Test1.opt.o -o Test1.opt
 
 Test1.bc: Test1.c
 	$(LLVMGCC) -I$(LOCAL_CVSROOT)/llvm/lib/Reoptimizer/Inst/rtl -c Test1.c -o Test1.bc
+
+crtend.o: $(CRTEND)
+	dis -c < $^ > $(@:.o=.c)
+	$(CC) -c $(@:.o=.c) 2>/dev/null
 
 libinstr.o: $(LIBINSTRBC)
 	dis -c < $^ > $(@:.o=.c)


Index: llvm/test/Regression/Reoptimizer/inst/Test1.c
diff -u llvm/test/Regression/Reoptimizer/inst/Test1.c:1.4.2.1 llvm/test/Regression/Reoptimizer/inst/Test1.c:1.4.2.2
--- llvm/test/Regression/Reoptimizer/inst/Test1.c:1.4.2.1	Fri May 30 10:23:50 2003
+++ llvm/test/Regression/Reoptimizer/inst/Test1.c	Tue Jun 24 10:29:11 2003
@@ -22,10 +22,6 @@
     printf("Just about to call fibs() for the first time...\n");
     fibs();
 
-    registerIntervalInst(42, pp_L1_cache_miss_start,
-                         pp_L1_cache_miss_end, sizeof(int),
-                         &l1cachemiss);
-
     printf("Just about to call fibs() for the second time...\n");
     fibs();
 
@@ -55,6 +51,10 @@
 
     printf("after for loop...\n");
     fflush(stdout);
+
+    registerIntervalInst(42, pp_L1_cache_miss_start,
+                         pp_L1_cache_miss_end, sizeof(int),
+                         &l1cachemiss);
 
     pp_declareIntervalEnd(42, &tmp);
 




From jstanley at cs.uiuc.edu  Tue Jun 24 10:34:01 2003
From: jstanley at cs.uiuc.edu (Joel Stanley)
Date: Tue Jun 24 10:34:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Reoptimizer/inst/Makefile Test1.c
Message-ID: <200306241533.KAA00287@trinity.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Reoptimizer/inst:

Makefile updated: 1.5 -> 1.6
Test1.c updated: 1.4 -> 1.5

---
Log message:

Merge from branch.


---
Diffs of the changes:

Index: llvm/test/Regression/Reoptimizer/inst/Makefile
diff -u llvm/test/Regression/Reoptimizer/inst/Makefile:1.5 llvm/test/Regression/Reoptimizer/inst/Makefile:1.6
--- llvm/test/Regression/Reoptimizer/inst/Makefile:1.5	Thu May 22 22:25:43 2003
+++ llvm/test/Regression/Reoptimizer/inst/Makefile	Tue Jun 24 10:33:31 2003
@@ -1,4 +1,5 @@
-LEVEL = /localhome/$(USER)/Research/llvm
+LEVEL = ../../../../
+#LEVEL = /localhome/$(USER)/Research/llvm
 
 CC = /usr/dcs/software/evaluation/bin/gcc
 CXX = /usr/dcs/software/evaluation/bin/g++
@@ -10,8 +11,8 @@
 	   $(LEVEL)/lib/Debug/libtracecache.a   \
 	   $(LEVEL)/lib/Debug/libbininterface.a
 
-
-LIBINSTRBC = /home/vadve/lattner/cvs/gcc_install_sparc/llvm/lib/libinstr.bc
+CRTEND = $(LLVM_LIB_SEARCH_PATH)/crtend.o
+LIBINSTRBC = $(LEVEL)/test/Libraries/Output/libinstr.bc
 
 Test1: Test1.o $(PERFOBJS)
         # NB: The only reason we don't include the PAPI library in the list of exclusions
@@ -19,8 +20,8 @@
         #     tables, and thus will never try to instrument the PAPI functions, which are
         #     dynamically linked.
 
-	mkexcl $(PERFOBJS) libinstr.o > mkexcl.cpp
-	$(CXX) -o Test1 $^ mkexcl.cpp -L$(PAPIDIR) -lpapi -lelf
+	mkexcl $(PERFOBJS) libinstr.o crtend.o > mkexcl.cpp
+	$(CXX) -o Test1 $^ mkexcl.cpp crtend.o -L$(PAPIDIR) -lpapi -lelf
 
 Test1.o: Test1.opt.c
 	$(CC) -c Test1.opt.c -o Test1.o
@@ -28,12 +29,16 @@
 Test1.opt.c: Test1.opt.bc
 	dis -c < Test1.opt.bc > Test1.opt.c
 
-Test1.opt.bc: Test1.bc libinstr.o $(LEVEL)/lib/Debug/libperf.so
+Test1.opt.bc: Test1.bc libinstr.o crtend.o $(LEVEL)/lib/Debug/libperf.so
 	opt -load $(LEVEL)/lib/Debug/libperf.so -pp1 -emitfuncs -deadtypeelim < Test1.bc > Test1.opt.o
 	gccld -disable-internalize $(LIBINSTRBC) Test1.opt.o -o Test1.opt
 
 Test1.bc: Test1.c
 	$(LLVMGCC) -I$(LOCAL_CVSROOT)/llvm/lib/Reoptimizer/Inst/rtl -c Test1.c -o Test1.bc
+
+crtend.o: $(CRTEND)
+	dis -c < $^ > $(@:.o=.c)
+	$(CC) -c $(@:.o=.c) 2>/dev/null
 
 libinstr.o: $(LIBINSTRBC)
 	dis -c < $^ > $(@:.o=.c)


Index: llvm/test/Regression/Reoptimizer/inst/Test1.c
diff -u llvm/test/Regression/Reoptimizer/inst/Test1.c:1.4 llvm/test/Regression/Reoptimizer/inst/Test1.c:1.5
--- llvm/test/Regression/Reoptimizer/inst/Test1.c:1.4	Thu May 22 22:25:43 2003
+++ llvm/test/Regression/Reoptimizer/inst/Test1.c	Tue Jun 24 10:33:31 2003
@@ -3,41 +3,30 @@
 // The following is actually the semantic result of processing a metric
 // declaration (need more sigfun support)
 
-double elapsedTime = 0;
-double elapsedTime2 = 0;
+double elapsedTime = 3.14;
 int l1cachemiss = 0;
 
 #include <stdio.h>
 
-void phase2();
-
 int fib(int n);
 void fibs();
 
-void someOtherFunction()
-{
-    int x;
-    pp_L1_cache_miss_start(&x);
-    printf("this statement resides within a region!\n");
-    pp_L1_cache_miss_end(&l1cachemiss, &x);
-    printf("someOtherFunction complete: retval is %d\n", l1cachemiss);
-}
-
 int main(int argc, char** argv) 
 {
     phase2();
 
-    printf("Just about to call fibs()...\n");
+    registerIntervalInst(42, pp_elapsed_time_start,
+                         pp_elapsed_time_end, sizeof(double),
+                         &elapsedTime);
+
+    printf("Just about to call fibs() for the first time...\n");
     fibs();
 
-    printf("Just about to call someOtherFunction()...\n");
-    someOtherFunction();
+    printf("Just about to call fibs() for the second time...\n");
+    fibs();
 
     printf("Leaving main...\n");
 
-    pp_regionPair(pp_elapsed_time_start, pp_elapsed_time_end);
-    pp_regionPair(pp_L1_cache_miss_start, pp_L1_cache_miss_end);
-
     return 0;
 }
 
@@ -49,39 +38,12 @@
         return fib(n-2) + fib(n-1);
 }
 
-void functionToBeInlined()
-{
-    double y;
-    pp_elapsed_time_start(&y);
-    printf("This function should be inlined...");
-    pp_elapsed_time_end(&elapsedTime2, &y);
-}
-
 void fibs() 
 {
     int i;
+    double tmp;
 
-    printf("Inside fibs now, calling functionToBeInlined...\n");
-    
-    functionToBeInlined();
-
-    ////////////////
-    // The following is the "scope" that the sampling applies to, represented by
-    // sigfuns. We need better sigfun support.
-    //
-    // That is, the below should look like:
-    //
-    // pp_region_start(elapsedTime);
-    // <for loop>
-    // pp_region_end(elaspedTime);
-    //
-    // or something similar. We should not be putting the actual invoked functions below,
-    // although the post-phase1 transformed code should be "similar" (calls removed and
-    // volatile loads put in their place, of course!)
-
-    long x;
-    double y;
-    pp_elapsed_time_start(&y);
+    pp_declareIntervalStart(&tmp, 42);
 
     for(i = 10; i > 0; i--) {
         printf("fib(%d) = %d\n", i, fib(i));
@@ -90,9 +52,14 @@
     printf("after for loop...\n");
     fflush(stdout);
 
-    pp_elapsed_time_end(&elapsedTime, &y);
+    registerIntervalInst(42, pp_L1_cache_miss_start,
+                         pp_L1_cache_miss_end, sizeof(int),
+                         &l1cachemiss);
+
+    pp_declareIntervalEnd(42, &tmp);
 
-    printf("ElapsedTime (metric variable) has address %lx, value %f\n", &elapsedTime, elapsedTime);
+    printf("value of variable elapsedTime: %f\n", elapsedTime);
+    printf("value of variable l1cachemiss: %d\n", l1cachemiss);
     
     //
     ////////////////




From jstanley at cs.uiuc.edu  Tue Jun 24 10:36:00 2003
From: jstanley at cs.uiuc.edu (Joel Stanley)
Date: Tue Jun 24 10:36:00 2003
Subject: [llvm-commits] CVS: llvm/lib/Reoptimizer/Inst/lib/design.txt
Message-ID: <200306241535.KAA00318@trinity.cs.uiuc.edu>


Changes in directory llvm/lib/Reoptimizer/Inst/lib:

design.txt updated: 1.15 -> 1.15.2.1

---
Log message:




---
Diffs of the changes:

Index: llvm/lib/Reoptimizer/Inst/lib/design.txt
diff -u llvm/lib/Reoptimizer/Inst/lib/design.txt:1.15 llvm/lib/Reoptimizer/Inst/lib/design.txt:1.15.2.1
--- llvm/lib/Reoptimizer/Inst/lib/design.txt:1.15	Sun May 18 12:45:26 2003
+++ llvm/lib/Reoptimizer/Inst/lib/design.txt	Tue Jun 24 10:35:06 2003
@@ -880,91 +880,70 @@
 
 {{{ MILESTONES
 
+  - Experiments
+  - The Paper
+  - The Thesis
 
 }}}
 
 {{{ TODO
 
-    - Move statically-sized spill regions so that they are internal to SparcInstManip.
-      (do not need variable-sized spill region except for phase5 invocations)
+    In priority order:
 
-    - Start table-of-stacks implementation for phase4 authorship of phase 5 slots.
-      - Placed on hold temporary because of "alloca-finding" approach. However, see the 
-        following e-mail for the current state of things:
-
-        {{{ E-mail regarding alloca-finding and table-of-stacks approach
-Okay, this is starting to seem intractable. I have another problem that
-I don't think can be resolved without resorting to a custom-stack
-mechanism that will incur prohibitive overhead.
-
-Everything is working for start-region instrumentation sites.  For
-end-region instrumentation sites, however, there's a problem. In order
-to write the slot for end sites, I have to know (or know how to compute)
-the address of the return value of the corresponding start site. I had
-originally thought that I would just store this in the GBT, or
-"something", but I clearly didn't think through the problem well enough.
-
-There are only two ways I can think of that this can occur:
-
-(a) Write the effective address of the return value of the start inst
-func, so that it gets passed to the end inst func.
-
-or
-
-(b) Somehow encode the stack offset to the return value from the start
-inst, where the offset is from the %sp *at the end-region site*
-
-Both of these have problems.
-
-First, I don't think (b) can work at all, given that there may be
-alloca's present in the original application that would change the %sp,
-and thus the offset value that we'd need, and we can't determine the
-exact allocas that are executed statically.
-
-For (a), the effective address isn't known until runtime. We can store
-this address in some global table where the phase 4 invocation for the
-end site can find it, but it is not sufficient to have a single scalar
-address here -- we must have a stack, due to potential recursive
-invocations. I think that this is clear, please let me know if I'm not
-making sense. :) 
-
-Hence, we'd need to maintain a stack of effective addresses, which was
-pushed during the execution of phase 5 for the start site, and then read
-and popped during the execution of phase 5 for the end site.  We're
-already really bloated with how many instructions we've got going on for
-all of the spills, etc, and I'm concerned about the effect that this
-stack manipulation will have on our overhead, as we talked about before.
-
-The way I see it, we only have two options if we're to make forward
-progress and not obliterate our chances of having lower overhead
-numbers. Hopefully we have some better choices. In the interests of
-short-term forward progress, I'm going to go with #1 for now.
-
-#1 - Make another common-case assumption that there will be no allocas
-between start and end sites, on *any* control path. If this is the case,
-then we know that the stack pointer will not have been manipulated (I
-think) between the start and end sites, and so the %sp offsets to the
-requisite data will be unchanged since when the phase 5 step occurred
-for the start site.  
-
-#2 - Just implement our fall-back solution that everything seems to be
-pointing to. I'm not sure exactly what other logistic nightmares might
-be entailed in this, though, because I've only a sketch of the idea.
-
-I wanted to point out, also, that the so-called "fall back" approach we
-discussed previous also involves manipulation of a stack at runtime
-(push/pop actions still have to occur at runtime), so perhaps the stack
-of effective addresses is less prohibitive than I thought, if only in
-the sense that we cannot avoid it. :(
-        }}}
-
-    - Write phase 5 stuff for end-region sites -- will assume that not allocas lie between
-    the start and end sites, which is not particularly a fair assumption.
-
-    - Optimizations:
-        - No need to save registers (other than those clobbered) in phase 3 slot, since phase 3
-          is invoked at the start of the function. Must still spill/restore shared, though.
-        - No need to save registers (other than those clobbered) in general.
+    1) New implementation for instrumenting at function-level granularity
+    2) Apache through LLVM, experiments
+    3) Writing, writing, writing: ICS version 2 paper, do
+       a) outline
+       b) intro
+       c) language section
+       d) compiler section
+
+    The three top-level items above are more-or-less interchangable. However, the
+    experiments will not be able to be completed unless the "instrumentation @
+    function-level granularity" implementation is done, and there should be an emphasis on
+    getting Apache through LLVM in the short-term because Chris is leaving for Norway in
+    early July.
+
+    However, *all* of the writing (including experimentation sections) for ICS v2 must be
+    done by the end of June, so that Adve can approve it, make the desired corrections,
+    etc., so that I can get started on thesis authorship and submission.  Realistically,
+    the timeframe should look something like this:
+
+    Week of 6/10 (5 days): Implementation, Apache w/ bug reports
+    Week of 6/16 (6 days): Implementation, Small example, continue Apache & do tests
+
+    -- At this point, all experiments should be more-or-less completed --
+
+    Week of 6/23 (6 days): Write, write, write. 3 days nonstop for content, 3
+                           days of Adve making corrections, etc.
+
+    Monday, 6/30 is D-Day...
+
+    Schedule revision 13 Jun 2003: Apache is in stasis waiting to hear back from
+    Chris. Some effort might be expended to see if any more work can be done on compiling
+    Apache even though the build process currently fails altogether.  Implementation for
+    instrumentation at function-level granularity is in stasis until I hear back from
+    Vikram.  This leaves three immediate options open until either of the two are
+    resolved, in which case forward progress on either the implementation or Apache should
+    be made.
+
+       Option 1) POV-Ray through LLVM.
+       Option 2) Writing
+       Option 3) "Small example"
+
+    Option 3 really shouldn't be undertaken until we know if obtaining things like I/O
+    elapsed time (via function-level) is possible, or until I can talk with Vikram in our
+    next meeting about what the heck this nebulous example should look like...this leaves
+    only options 1 and 2 above as viable.
+
+    POV-Ray should be an easy thing to start, and would be good both as a fall-back if
+    Apache isn't possible as well as a useful additional example if the latter is
+    possible.  This would also give V & C some time to respond to the pending queries.
+
+    - Optimizations: - No need to save registers (other than those clobbered) in
+    phase 3 slot, since phase 3 is invoked at the start of the function. Must
+    still spill/restore shared, though.  - No need to save registers (other than
+    those clobbered) in general.
 
 }}}
 
@@ -1460,5 +1439,226 @@
 Also, Chris remarked that any novel page-management mechanisms, etc., (for the
 code that is jumped to in newly-allocated pages) that I devise should perhaps be
 integrated into the LLVM JIT if they are suitable.
+
+}}}
+
+{{{ Experiments
+
+We must devise experiments that demonstrate the novel aspects of the work.  We
+are currently planning on using Apache and/or POV-Ray, and demonstrating how a
+"deep" performance analysis can be encoded using performance primitives. A
+"deep" performance analysis is one which essentially (using Hollingsworth's
+terminology from his PhD thesis) gets at the "what, where, and when" aspects of
+performance bottlenecks.  However, instead of doing this at the "arbitrary
+program during execution" level, as the W^3 search model does, we will encode
+these performance aspects at the application level itself.
+
+Here's what Vikram suggested for a good start:
+
+  I just thought that the examples of performance issues he explores in his
+  automatic search would give you (a) some insights into what performance issues
+  make sense to consider, and (b) some ideas about how to do a systematic
+  diagnosis, albeit at the application level.
+  
+  But isn't detecting the "what, why, and when" of performance bottlenecks
+  pretty closely related to the goal of performance diagnosis?  We'd be looking
+  for bottlenecks too, except that we can use application domain information, we
+  can look for bottlenecks at the algorithmic level instead of the general
+  system level, and we can record it permanently in the program as a first class
+  feature of the program.
+  
+  Anyway, about your second question: here's a way to do what I suggesting:
+  
+  -- think about 2-3 key performance issues with Apache (or POV-Ray) that you'd
+  want to diagnose e.g., cache misses, TLB misses, thread overhead (estimating
+  that could be interesting), I/O delays
+  
+  -- if those issues make sense with a small sort example, try to diagnose those
+  issues in the small example first.  e.g., I think cache misses, TLB misses,
+  and I/O delays would all be issues if you were sorting a huge file of some
+  kind.
+  
+  This is purely to give you a small, well-understood code to try out before
+  going to the big ones where it may be difficult to know, when one diagnosis
+  attempt fails, whether it failed because you misunderstood the performance
+  issues or because the guess was wrong or both.
+
+TLB misses aren't an option, because we cannot get at them with PAPI. Cache
+misses are available, so that'd certainly be a good place to start.  As for I/O
+delays, I have no idea how we'd measure this, either.  The following is the
+comprehensive list of the low-level metrics that are exposed to us via PAPI.  On
+our own, we can support simple things like elapsed time, load average, etc.  I'm
+not altogether clear how we'd determine how much time (for a particular region)
+was spent doing I/O-bound activities...
+
+Number of hardware counters: 2
+Name: PAPI_L1_ICM 	Description: Level 1 instruction cache misses
+Name: PAPI_L2_TCM 	Description: Level 2 cache misses
+Name: PAPI_CA_SNP 	Description: Requests for a snoop
+Name: PAPI_CA_INV 	Description: Requests for cache line invalidation
+Name: PAPI_L1_LDM 	Description: Level 1 load misses
+Name: PAPI_L1_STM 	Description: Level 1 store misses
+Name: PAPI_BR_MSP 	Description: Conditional branch instructions mispredicted
+Name: PAPI_TOT_IIS 	Description: Instructions issued
+Name: PAPI_TOT_INS 	Description: Instructions completed
+Name: PAPI_LD_INS 	Description: Load instructions
+Name: PAPI_SR_INS 	Description: Store instructions
+Name: PAPI_TOT_CYC 	Description: Total cycles
+Name: PAPI_IPS    	Description: Instructions per second
+Name: PAPI_L1_DCR 	Description: Level 1 data cache reads
+Name: PAPI_L1_DCW 	Description: Level 1 data cache writes
+Name: PAPI_L1_ICH 	Description: Level 1 instruction cache hits
+Name: PAPI_L2_ICH 	Description: Level 2 instruction cache hits
+Name: PAPI_L1_ICA 	Description: Level 1 instruction cache accesses
+Name: PAPI_L2_TCH 	Description: Level 2 total cache hits
+Name: PAPI_L2_TCA 	Description: Level 2 total cache accesses
+
+So, in the short-term, we have two outstanding problems. First, what kinds of
+metric would we want to apply to Apache/POV-Ray/Simple example? Second, of those
+metrics, which can we actually realize with the current system? Third, we should
+create the simple program (such as sorting large amounts of data from a file,
+etc.) so that it can use these metrics in such a way that "models" or
+"anticipates" the way they will be used in the bigger applications.
+
+This is an outstanding issue, and I don't really know where to go with it yet.
+
+More notes about this as of 4 Jun 2003:
+
+One way to obtain metric values that are based on the elapsed times of
+particular functions is to somehow register instrumentation for those particular
+functions, and for a particular region -- Vikram argues that we have the ability
+to do this dynamically and we don't need any markers or phase-1 actions because
+we're operating at function-level granularity. 
+
+Here is a sample scenario: We have defined an interval I over some scoped region
+of code.  During phase 1 and phase 2, no instrumentation is registered for this
+interval.  Later on, we construct a metric that is qualified by a list of
+functions that (for example) are to have their runtimes measured and added to
+some running total.  Let's call this the "measure_functions_start" and
+"measure_functions_end" metric, and have it yield a value of type double which
+is the aggregate runtime of the list of functions when they get executed within
+interval I.  The metric registration function will have to have some way
+(varargs?) of denoting what the functions are: perhaps it can simply pass in an
+array of function names together with a size.
+
+Example: 
+
+pp_registerIntervalInst(intervalID, measure_functions_start,
+                        measure_functions_end, &retVal, sizeof(double), 
+                        func1, func2, func3, ...);
+
+However, what do "measure_functions_start" have to do with anything? More than
+likely, what we need to do is specify a particular metric to apply to a
+paritcular function, such that the value will be sampled each time that function
+gets executed.  Then, since there can be multiple invocations (and hence,
+multiple samples) for this selected function within I, we will have to have some
+default or user-specified way of aggregating the data :(.  This is gross.  In
+other words, we should probably simplify the above to something like:
+
+pp_registerIntervalInst(intervalID, some_metric_start, some_metric_end, 
+                        &retVal, sizeof(double), HOW_TO_AGGREGATE, func1);
+
+Where func1 is the function to be instrumented and HOW_TO_AGGREGATE is some
+value that specifies one from a couple of ways of combining the data.  For now,
+HOW_TO_AGGREGATE will not exist and will implicitly sum all return
+values...hence, if the some_metric_{start,end} function ptrs above were to
+elapsed time, at the end of the interval I with interval id intervalID, retVal
+would contain the combined elapsed time of all time spent in function func1.
+
+Enabling this measurement at the start of the interval, disabling at the end of
+the interval; clearly, the start-instrumentation site will need to transform
+func1 (to compute the metric value) when crossed, and the end-insturmentation
+site must remove such instrumentation when crossed. This doesn't follow the
+normal model of what occurs at the instrumentation sites, in terms of just
+ripping down a list of functions...or does it? Perhaps one of the functions in
+the list is just this function that does the transformation on the target
+function...in this case, the process would look something like:
+
+1. Register the transformation function (for the start point) -- call this
+xform_start -- as a regular instrumentation function.  The runtime call to do
+the registration will build the appropriate data structures which will encode
+what metric to associated with the target function, aggregation method, return
+value, etc.
+
+2. Register the transformation function (for the end point) -- call this
+xform_end -- as a regular instrumentation function.
+
+3. When xform_start is invoked as a regular instrumentation function, it will
+instrument the target function with the selected instrumentation.
+
+This is the hardest step to conceptualize and realize.  The problem is that,
+without any placeholders from phase 1, it's not clear that we can instrument the
+target function easily.  Clearly, our instrumentation points are at the start
+and end points of the function (entry and exit).  But this is not really true.
+Our instrumentation points are really at the entry and *all* function exits.
+
+The important question is, what if we have all of the exit points at our
+disposal? Would that change anything?
+
+It would. The entry point together with all exit points would form a set of
+instrumentation points. For each of these instrumentation points, we could
+over-write a branch to a new slot that would call the desired instrumentation
+function, restore the replaced instruction, and return to the instrumentation
+point to continue execution.  This would potentially work. One major problem
+that comes to mind is that for system calls (such as read()), the body of the
+function is highly likely to be out of short-jmp range to the
+tracecache-allocated slot.  The only way around this would be to create a heap
+region and copy the target function into it, etc.  We don't have any code to do
+this yet, so again there is no code reuse or (much) leveraging of existing
+functionality.  Additionally, work will have to be done to make the data
+structure that maps the address of a function to its extents (which come from
+the ELf information).
+
+--
+
+The other alternative is to make sure a "wrapper" function (i.e., for the
+function read()):
+
+int read_wrapper(args for read) {
+  start_inst();
+  int rc = read(args for read);
+  end_inst();
+  return rc;
+}
+
+But this isn't an option because we cannot locate the calls to read() to replace
+them with the wrapper.  We could do the following:
+
+3a. Copy the entire target function to a heap region, and instrument it to our
+heart's content.  However, finding exit points may not be easy without a CFG,
+etc.
+
+3b. Replace the body of the real target function with a call to the modified
+duplicate of the target function, returning whatever the modified duplicate
+returns.
+
+This works, I think, but is incredibly cumbersome and, contrary to what was
+previously discussed, we do *not* possess all of the mechanisms.
+
+--
+
+Notes on POV-Ray experiment 23 Jun 2003
+
+We propose using a user-defined metric, flops/display pixel to compute the
+computational cost per pixel.
+
+1. Compute flops/dpixel and report using pp mechanisms.
+  1a. flops can be obtained via PAPI, IIRC.
+
+2. Create a moving average of flops/dpixel.
+
+3. When the flops/dpixel exceeds mavg flops/dpixel by some multiplicative
+threshold (e.g.), a performance assertion is violated.  When the PA is violated,
+we report other metrics (perhaps ranked in some manner) that had been being
+recorded but not reported.  Initial suggestions for the other metrics to measure
+_over the same region_ (say, the main routine for a single ray-trace) are:
+
+    1) MAverage L1 icache misses vs. this-ray L1 icache misses
+    2) MAverage L2 cache misses vs. this-ray L2 cache misses
+    3) MAverage load count vs. this-ray load count
+    4) MAverage store count vs. this-ray store count
+
+Immediate problem: I don't think we can monitor more than 2 of these values
+concurrently due to hardware limitations on the SPARC...
 
 }}}




From jstanley at cs.uiuc.edu  Tue Jun 24 10:41:01 2003
From: jstanley at cs.uiuc.edu (Joel Stanley)
Date: Tue Jun 24 10:41:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Reoptimizer/Inst/rtl/pprtl.cpp pprtl.h
Message-ID: <200306241540.KAA01771@trinity.cs.uiuc.edu>


Changes in directory llvm/lib/Reoptimizer/Inst/rtl:

pprtl.cpp updated: 1.2 -> 1.3
pprtl.h updated: 1.3 -> 1.4

---
Log message:

Merge from branch jrsdev.



---
Diffs of the changes:

Index: llvm/lib/Reoptimizer/Inst/rtl/pprtl.cpp
diff -u llvm/lib/Reoptimizer/Inst/rtl/pprtl.cpp:1.2 llvm/lib/Reoptimizer/Inst/rtl/pprtl.cpp:1.3
--- llvm/lib/Reoptimizer/Inst/rtl/pprtl.cpp:1.2	Thu May 22 22:26:13 2003
+++ llvm/lib/Reoptimizer/Inst/rtl/pprtl.cpp	Tue Jun 24 10:39:54 2003
@@ -57,7 +57,7 @@
     printf("pp_L1_cache_miss_start invoked: retVal addr is 0x%lx\n", (unsigned long) retVal);
     start_papi_sample(events, 1);
     *retVal = PAPI_L1_ICM;
-    printf("pp_L1_cache_miss_start returning!\n");
+    printf("pp_L1_cache_miss_start returning value %x!\n", *retVal);
 }
 
 void pp_L1_cache_miss_end(int* retVal, int* start)


Index: llvm/lib/Reoptimizer/Inst/rtl/pprtl.h
diff -u llvm/lib/Reoptimizer/Inst/rtl/pprtl.h:1.3 llvm/lib/Reoptimizer/Inst/rtl/pprtl.h:1.4
--- llvm/lib/Reoptimizer/Inst/rtl/pprtl.h:1.3	Thu May 22 22:26:13 2003
+++ llvm/lib/Reoptimizer/Inst/rtl/pprtl.h	Tue Jun 24 10:39:54 2003
@@ -16,11 +16,22 @@
     void     pp_L1_cache_miss_end(int* retVal, int* start);
     unsigned pp_counterPrim(void);
 
-    // significant functions
+    // significant functions and rtl functions
+
+    void phase2();
+    void registerIntervalInst(unsigned siteID,
+                              void* startFunc,
+                              void* endFunc,
+                              unsigned paramSize,
+                              void* retVal);
 
     void     pp_regionPair(void*, void*);
     void     pp_metricSpec(void*);
 
+    // Interval declaration sigfuns: the placeholder variable is to connect the
+    // two sigfuns via uses.
+    void     pp_declareIntervalStart(double* placeHolder, unsigned intervalID);
+    void     pp_declareIntervalEnd(unsigned intervalID, double* placeHolder);
 
 #ifdef _BUILDING_RUNTIME_LIBRARY_
 }




From jstanley at cs.uiuc.edu  Tue Jun 24 10:41:05 2003
From: jstanley at cs.uiuc.edu (Joel Stanley)
Date: Tue Jun 24 10:41:05 2003
Subject: [llvm-commits] CVS: llvm/lib/Reoptimizer/Inst/lib/Phase1/Phase1.cpp PrimInfo.cpp PrimInfo.h
Message-ID: <200306241540.KAA01765@trinity.cs.uiuc.edu>


Changes in directory llvm/lib/Reoptimizer/Inst/lib/Phase1:

Phase1.cpp updated: 1.26 -> 1.27
PrimInfo.cpp updated: 1.15 -> 1.16
PrimInfo.h updated: 1.10 -> 1.11

---
Log message:

Merge from branch jrsdev.



---
Diffs of the changes:

Index: llvm/lib/Reoptimizer/Inst/lib/Phase1/Phase1.cpp
diff -u llvm/lib/Reoptimizer/Inst/lib/Phase1/Phase1.cpp:1.26 llvm/lib/Reoptimizer/Inst/lib/Phase1/Phase1.cpp:1.27
--- llvm/lib/Reoptimizer/Inst/lib/Phase1/Phase1.cpp:1.26	Thu May 22 22:26:13 2003
+++ llvm/lib/Reoptimizer/Inst/lib/Phase1/Phase1.cpp	Tue Jun 24 10:39:53 2003
@@ -12,7 +12,7 @@
 #include <set>
 #include <algorithm>
 #include <map>
-#include <typeinfo>
+#include <string>
 
 #include "llvm/Pass.h"
 #include "llvm/Module.h"
@@ -78,6 +78,12 @@
     Module*          m_module;             // Current module
     vector<PrimInfo> m_primInfos;          // Deferred documentation structures
     TargetData       m_targetData;         // For obtaining target-specific info
+
+    static const unsigned NUM_START_OPDS = 3;
+    static const unsigned NUM_END_OPDS = 3;
+    static const unsigned NUM_STOPD_USES = 2;
+    static const unsigned ENDCALL_ID_OPIDX = 1;
+    static const unsigned POINTCALL_ID_OPIDX = 1;
 };
 
 //////////////// Phase1 implementation ////////////////
@@ -112,7 +118,7 @@
                                      m_module);
 
     // Invoke PrimInfo's static initializer
-    PrimInfo::buildStructType(&m);
+    PrimInfo::buildStructTypes(&m);
 
     //
     ////////////////
@@ -181,12 +187,17 @@
 
     // Populate with builtins
 
-    // TODO: Replace the hard-coded mechanism for the point metrics with a
-    // sigfun registration mechanism like that which exists for region metrics
-    // (pp_regionPair()).
+    // The new approach: pp_declareInterval{Start,End} are the only sigfuns used to
+    // declare an interval, instead of the interval being implicit by the given locations
+    // of the calls to (registered) region sigfuns.  For now, we will still use the old
+    // code for processing the contents of 'rpairs', but the only element of rpairs will
+    // be the pair of interval-declaration sigfuns.  TODO: Clean up this approach, and do
+    // the same for point-site declarations.
+
+    Function* intervalStartDeclFunc = m_module->getNamedFunction("pp_declareIntervalStart");
+    Function* intervalEndDeclFunc = m_module->getNamedFunction("pp_declareIntervalEnd");
 
-    // Find user-declared region-pairs and add them to the list
-    findRegionPairs(rpairs);
+    rpairs.push_back(std::make_pair(intervalStartDeclFunc, intervalEndDeclFunc));
 
     ////////////////
     // Build list of point metric spec sigfuns, finding ppMetricSpec will yield
@@ -218,26 +229,10 @@
     for(unsigned i = 0, e = rpairs.size(); i < e; ++i)
         transformSites(rpairs[i]);
 
-    // Build the interval PrimInfo instance -> GBT index map. This maps the address of a
-    // particular interval PrimInfo instance to the index in the GBT that corresponds to
-    // the record that represents the start-interval global-volatile and related
-    // information.
-
-    std::map<PrimInfo*, unsigned> gbtIdxMap;
-    int idx = 0;
-    for(vector<PrimInfo>::iterator i = m_primInfos.begin(), e = m_primInfos.end(); i != e; ++i){
-        if(i->getType() == PrimInfo::REGION) {
-            gbtIdxMap[&*i] = idx;
-            idx += 2;
-        }
-        else
-            idx++;
-    }
-
     // Build the global bookkeeping table (GBT) contents
     vector<Constant*> gbtElems;
     for(vector<PrimInfo>::iterator i = m_primInfos.begin(), e = m_primInfos.end(); i != e; ++i)
-        i->buildStructInstances(gbtElems, gbtIdxMap, &m_targetData);
+        i->buildStructInstances(gbtElems);
 
     // Make the GBT itself, and the corresponding global variable.
     ArrayType* gbtType = ArrayType::get(PrimInfo::getStructType(), gbtElems.size());
@@ -295,19 +290,19 @@
 
         CallInst* startCall = dyn_cast<CallInst>(*u);
         assert(startCall && "Use of a registered pp sigfun not in a call");
-        
+
         // Find call to endpoint: The first operand of the start function is used in only
         // two places: the call to the start function, and the call to the end
         // function. Find the call to the end function.
 
-        assert(startCall->getNumOperands() == 2 &&
-               "Start-region call insts should only have 2 operands");
+        assert(startCall->getNumOperands() == NUM_START_OPDS &&
+               "Start-region call inst has unexpected number of operands");
 
         Value* stOpd = startCall->getOperand(1);
 
         assert(isa<AllocaInst>(stOpd) && "Unexpected first operand of start call");
-        assert(stOpd->use_size() == 2 &&
-               "Expect only two uses of start-region operand");
+        assert(stOpd->use_size() == NUM_STOPD_USES &&
+               "Unexpected number of uses of of start-region (\"connector\") operand");
         
         CallInst* endCall;
         if((endCall = dyn_cast<CallInst>(stOpd->use_back()))) {
@@ -338,11 +333,21 @@
 
     return cast<GlobalVariable>(ci->getOperand(1));
 }
-    
+
+static unsigned extractUnsignedValue(CallInst* ci, unsigned nth) 
+{
+    // extract the literal value of the nth actual of ci, which must be of type
+    // unsigned.
+
+    if(ConstantUInt* ui = dyn_cast<ConstantUInt>(ci->getOperand(nth)))
+        return ui->getValue();
+
+    assert(0 && "Failed to find unsigned value as call actual");
+    return 0; // squash warning
+}
+
 void Phase1::transformSite(CallInst* startCall, CallInst* endCall) 
 {
-    GlobalVariable* metricVar = GetLHSGlobal(endCall);
-    
     // Insert placeholder code
     PlaceholderPair startPair, endPair;
     generatePlaceholder(startCall, startPair);
@@ -350,8 +355,9 @@
 
     // Create info about this site for later documentation
 
-    m_primInfos.push_back(PrimInfo(PrimInfo::REGION, startPair.first, endPair.first,
-                                   metricVar, m_module, 
+    unsigned siteID = extractUnsignedValue(endCall, ENDCALL_ID_OPIDX);
+    m_primInfos.push_back(PrimInfo(PrimInfo::REGION, siteID,
+                                   startPair.first, endPair.first, m_module,
                                    startCall->getCalledFunction(),
                                    endCall->getCalledFunction()));
 
@@ -379,8 +385,6 @@
 
 void Phase1::transformSite(CallInst* pointCall)
 {
-    GlobalVariable* metricVar = GetLHSGlobal(pointCall);
-
     // Insert placeholder code
     PlaceholderPair php;
     generatePlaceholder(pointCall, php);
@@ -391,7 +395,8 @@
     PrimInfo::PrimType type = pointCall->getCalledFunction() == m_counterPrimFunc ?
         PrimInfo::COUNTER : PrimInfo::POINT;
 
-    m_primInfos.push_back(PrimInfo(type, php.first, metricVar, m_module,
+    unsigned siteID = extractUnsignedValue(pointCall, POINTCALL_ID_OPIDX);
+    m_primInfos.push_back(PrimInfo(type, siteID, php.first, m_module,
                                    pointCall->getCalledFunction()));
 
     // Remove the call


Index: llvm/lib/Reoptimizer/Inst/lib/Phase1/PrimInfo.cpp
diff -u llvm/lib/Reoptimizer/Inst/lib/Phase1/PrimInfo.cpp:1.15 llvm/lib/Reoptimizer/Inst/lib/Phase1/PrimInfo.cpp:1.16
--- llvm/lib/Reoptimizer/Inst/lib/Phase1/PrimInfo.cpp:1.15	Tue May 13 14:43:03 2003
+++ llvm/lib/Reoptimizer/Inst/lib/Phase1/PrimInfo.cpp	Tue Jun 24 10:39:53 2003
@@ -2,10 +2,8 @@
 // programmer: Joel Stanley
 //       date: Tue Jan 28 14:07:05 CST 2003
 //     fileid: PrimInfo.cpp
-//    purpose: Captures information about an instance of a performance primitive.  The
-//    information is intended to be carried between phases by reconstructing PrimInfo
-//    contents that have been placed in global static data structures by Phase1's
-//    invocation of PrimInfo::Document.
+//    purpose: Captures information about an instance of a performance primitive "marker"
+//             in the code.
 
 #include "llvm/Module.h"
 #include "llvm/Type.h"
@@ -15,7 +13,6 @@
 #include "llvm/GlobalVariable.h"
 #include "llvm/Constant.h"
 #include "Support/VectorExtras.h"
-#include "llvm/Target/TargetData.h"
 
 #include "PrimInfo.h"
 #include "Intraphase.h"
@@ -31,16 +28,16 @@
 StructType* PrimInfo::sm_structType = 0;
 
 PrimInfo::PrimInfo(PrimType type,
+                   unsigned siteID,
                    GlobalVariable* startGlob,
                    GlobalVariable* endGlob,
-                   GlobalVariable* metricVar,
                    Module* module,
                    Function* startFunc,
                    Function* endFunc):
     m_type(type),
+    m_siteID(siteID),
     m_globVol(startGlob),
     m_globVolEnd(endGlob),
-    m_metricVar(metricVar),
     m_module(module),
     m_startFunc(startFunc),
     m_endFunc(endFunc)
@@ -48,126 +45,75 @@
 }
 
 PrimInfo::PrimInfo(PrimType type,
+                   unsigned siteID,
                    GlobalVariable* globVol,
-                   GlobalVariable* metricVar,
                    Module* module,
                    Function* calledFunc):
     m_type(type),
+    m_siteID(siteID),
     m_globVol(globVol),
-    m_metricVar(metricVar),
     m_module(module),
     m_startFunc(calledFunc)
 {
     assert(type != POINT || calledFunc && "point type --> non-null called function");
 }
 
-void PrimInfo::buildStructType(Module* module)
+void PrimInfo::buildStructTypes(Module* module)
 {
-    // The structure deposited into the LLVM bytecode essentially looks like:
-    // struct PrimInfo
+    static bool init = false;
+
+    assert(!init && "Expect only one invocation of this function");
+    init = true;
+
+    // {{{ struct PrimInfo
     // {
+    //   // Site ID for this site (one value identifies one entire interval as well)
+    //   unsigned siteID;
+    //   
     //   // Type of GBT entry type (i.e., member of enum GBTEntryType)
     //   unsigned gbtType;
     //
     //   // Address of global variable corresponding to this primitive.
     //   unsigned short* loadVar;
     //
-    //   // Index in GBT of struct of corresponding GBT_INTERVAL_START entry
-    //   // (NB: Only valid if gbtType == GBT_INTERVAL_END)
-    //   unsigned gbtStartIdx;
-    //
-    //   // Size, in bytes, of the parameter to the instrumentation function that must be
-    //   // invoked at the instrumentation that corresponds to this structure instance.
-    //   unsigned paramSize;
-    //
-    //   // Pointer to memory for instFunc's return value
-    //   // This is initalized to null for start-interval sites, and filled in at runtime
-    //   // by a pointer to heap-allocated memory of the appropriate size. For all other
-    //   // site types, this is initialized to the metric variable address.
-    //   void* retVal;
-    //   
-    //   // Pointer to instrumentation function
-    //   void* instFunc;
-    //
-    // Other stuff will need to go here...(TODO)
     // };
-
-    static bool init = false;
-
-    assert(!init && "Expect only one invocation of this function");
-    init = true;
+    // }}}
 
     PointerType* uspt = PointerType::get(Type::UShortTy);
-    PointerType* vpt = PointerType::get(Type::VoidTy);
+
     sm_structType =
         StructType::get(make_vector<const Type*>(
-          Type::UIntTy, uspt, Type::UIntTy, Type::UIntTy, vpt, vpt, 0));
-    module->addTypeName("PrimInfo", sm_structType);
-}
-
+          Type::UIntTy, Type::UIntTy, uspt, 0));
 
-// For start sites
-static ConstantStruct* makeConstStruct(StructType* st,
-                                       unsigned gbtType,
-                                       GlobalVariable* loadVar,
-                                       unsigned paramSize,
-                                       Function* instFunc)
-{
-    std::vector<Constant*> init;
-    init.push_back(ConstantUInt::get(Type::UIntTy, gbtType));
-    init.push_back(ConstantPointerRef::get(loadVar));
-    init.push_back(ConstantUInt::get(Type::UIntTy, 0));
-    init.push_back(ConstantUInt::get(Type::UIntTy, paramSize));
-    init.push_back(Constant::getNullValue(PointerType::get(Type::VoidTy)));
-    init.push_back(ConstantExpr::getCast(ConstantPointerRef::get(instFunc),
-                                         PointerType::get(Type::VoidTy)));
-    return ConstantStruct::get(st, init);
+    module->addTypeName("PrimInfo", sm_structType);
 }
 
-// For non-start sites
 static ConstantStruct* makeConstStruct(StructType* st,
+                                       unsigned siteID,
                                        unsigned gbtType,
-                                       GlobalVariable* loadVar,
-                                       unsigned startLinkIdx,
-                                       unsigned paramSize,
-                                       GlobalVariable* metricVar,
-                                       Function* instFunc)
+                                       GlobalVariable* loadVar)
 {
     std::vector<Constant*> init;
-    init.push_back(ConstantUInt::get(Type::UIntTy, gbtType));
-    init.push_back(ConstantPointerRef::get(loadVar));
-    init.push_back(ConstantUInt::get(Type::UIntTy, startLinkIdx));
-    init.push_back(ConstantUInt::get(Type::UIntTy, paramSize));
-    init.push_back(ConstantExpr::getCast(ConstantPointerRef::get(metricVar),
-                                         PointerType::get(Type::VoidTy)));
-    init.push_back(ConstantExpr::getCast(ConstantPointerRef::get(instFunc),
-                                         PointerType::get(Type::VoidTy)));
+    init.push_back(ConstantUInt::get(Type::UIntTy, siteID));  // siteID
+    init.push_back(ConstantUInt::get(Type::UIntTy, gbtType)); // gbtType
+    init.push_back(ConstantPointerRef::get(loadVar));         // loadVar
     return ConstantStruct::get(st, init);
 }
 
-void PrimInfo::buildStructInstances(std::vector<Constant*>& gbtElems,
-                                    std::map<PrimInfo*, unsigned>& gbtIdxMap,
-                                    const TargetData* targetData)
+void PrimInfo::buildStructInstances(std::vector<Constant*>& gbtElems)
 {
     StructType* st = getStructType();
 
     switch(m_type) {
         case REGION: {
-            
-            std::map<PrimInfo*, unsigned>::iterator i = gbtIdxMap.find(this);
-            assert(i != gbtIdxMap.end() && "No entry in map for this PrimInfo instance");
-            
+
             // Create the struct for the start of the region
-            assert(m_startFunc->asize() == 1 && "Unexpected # args for start func");
-            unsigned psize = targetData->getTypeSize(m_startFunc->afront().getType());
-            gbtElems.push_back(makeConstStruct(st, GBT_INTERVAL_START, m_globVol,
-                                               psize, m_startFunc));
+            assert(m_startFunc->asize() == 2 && "Unexpected # args for start func");
+            gbtElems.push_back(makeConstStruct(st, m_siteID, GBT_INTERVAL_START, m_globVol));
 
             // Create the struct for the end of the region
             assert(m_endFunc->asize() == 2 && "Unexpected # args for end func");
-            psize = targetData->getTypeSize(m_endFunc->afront().getType());
-            gbtElems.push_back(makeConstStruct(st, GBT_INTERVAL_END, m_globVolEnd,
-                                               i->second, psize, m_metricVar, m_endFunc));
+            gbtElems.push_back(makeConstStruct(st, m_siteID, GBT_INTERVAL_END, m_globVolEnd));
 
             break;
         }


Index: llvm/lib/Reoptimizer/Inst/lib/Phase1/PrimInfo.h
diff -u llvm/lib/Reoptimizer/Inst/lib/Phase1/PrimInfo.h:1.10 llvm/lib/Reoptimizer/Inst/lib/Phase1/PrimInfo.h:1.11
--- llvm/lib/Reoptimizer/Inst/lib/Phase1/PrimInfo.h:1.10	Mon May 12 21:00:24 2003
+++ llvm/lib/Reoptimizer/Inst/lib/Phase1/PrimInfo.h	Tue Jun 24 10:39:54 2003
@@ -13,8 +13,6 @@
 
 #include <map>
 
-class TargetData;
-
 namespace pp
 {
 
@@ -26,7 +24,6 @@
     PrimInfo(Module* m):
         m_globVol(0),
         m_globVolEnd(0),
-        m_metricVar(0),
         m_module(m),
         m_startFunc(0),
         m_endFunc(0)
@@ -35,16 +32,16 @@
 
     // Phase 1 construction -- point/counter computation site
     PrimInfo(PrimType type,
+             unsigned siteID,
              GlobalVariable* globVol,
-             GlobalVariable* metricVar,
              Module* m,
              Function* calledFunc = 0);    
 
     // Phase 1 construction -- region computation site
     PrimInfo(PrimType type,
+             unsigned siteID,
              GlobalVariable* startGlob,
              GlobalVariable* endGlob,
-             GlobalVariable* metricVar,
              Module* m,
              Function* startFunc,
              Function* endFunc);
@@ -57,15 +54,11 @@
     const Function*       getStartFunc() const  { return m_startFunc;  }
     Function*             getEndFunc()          { return m_endFunc;    }
     const Function*       getEndFunc() const    { return m_endFunc;    }
-    GlobalVariable*       getMetricVar()        { return m_metricVar;  }
-    const GlobalVariable* getMetricVar() const  { return m_metricVar;  }
     PrimType              getType() const       { return m_type;       }
 
-    void buildStructInstances(std::vector<Constant*>& gbtElems,
-                              std::map<PrimInfo*, unsigned>& gbtIdxMap,
-                              const TargetData* targetData);
+    void buildStructInstances(std::vector<Constant*>& gbtElems);
 
-    static void           buildStructType(Module* module);
+    static void           buildStructTypes(Module* module);
     static StructType*    getStructType() { return sm_structType; }
     
   private:
@@ -74,9 +67,9 @@
     // supplied configuration parameters.
 
     PrimType           m_type;
+    unsigned           m_siteID;
     GlobalVariable*    m_globVol;
     GlobalVariable*    m_globVolEnd;
-    GlobalVariable*    m_metricVar;
     Module*            m_module;
     Function*          m_startFunc;
     Function*          m_endFunc;




From jstanley at cs.uiuc.edu  Tue Jun 24 10:41:08 2003
From: jstanley at cs.uiuc.edu (Joel Stanley)
Date: Tue Jun 24 10:41:08 2003
Subject: [llvm-commits] CVS: llvm/lib/Reoptimizer/Inst/lib/PhaseInfo.h Phases.cpp SparcInstManip.cpp design.txt
Message-ID: <200306241540.KAA01748@trinity.cs.uiuc.edu>


Changes in directory llvm/lib/Reoptimizer/Inst/lib:

PhaseInfo.h updated: 1.9 -> 1.10
Phases.cpp updated: 1.35 -> 1.36
SparcInstManip.cpp updated: 1.16 -> 1.17
design.txt updated: 1.15 -> 1.16

---
Log message:

Merge from branch jrsdev.



---
Diffs of the changes:

Index: llvm/lib/Reoptimizer/Inst/lib/PhaseInfo.h
diff -u llvm/lib/Reoptimizer/Inst/lib/PhaseInfo.h:1.9 llvm/lib/Reoptimizer/Inst/lib/PhaseInfo.h:1.10
--- llvm/lib/Reoptimizer/Inst/lib/PhaseInfo.h:1.9	Sun May 18 15:33:46 2003
+++ llvm/lib/Reoptimizer/Inst/lib/PhaseInfo.h	Tue Jun 24 10:39:52 2003
@@ -27,12 +27,9 @@
 typedef std::pair<uint64_t, uint64_t> AddressRange;
 
 typedef struct GBTElem {
-    unsigned gbtType;
-    unsigned short* loadVar;
-    unsigned gbtStartIdx;
-    unsigned paramSize;
-    void*    retVal;
-    void*    instFunc;
+    unsigned      siteID;
+    unsigned      gbtType;
+    unsigned      short* loadVar;
 };
 
 class Phase3Info 


Index: llvm/lib/Reoptimizer/Inst/lib/Phases.cpp
diff -u llvm/lib/Reoptimizer/Inst/lib/Phases.cpp:1.35 llvm/lib/Reoptimizer/Inst/lib/Phases.cpp:1.36
--- llvm/lib/Reoptimizer/Inst/lib/Phases.cpp:1.35	Thu May 22 22:26:12 2003
+++ llvm/lib/Reoptimizer/Inst/lib/Phases.cpp	Tue Jun 24 10:39:52 2003
@@ -45,10 +45,13 @@
 //
 //      2. Change the branch to the phase 4 slot to branch to a (new) phase 5 slot. See
 //      appropriate InstManip instance for detailed information about phase 5 slot
-//      contents.
+//      contents.  If there is no registered instrumentation for the phase 5 slot, leave a
+//      nop in place of the branch to reduce runtime overhead.
 //
 //      3. Deallocate the slot that originated this invocation of phase4().
 //
+// PHASE 5: Phase 5 isn't like the other phases; rather, it simply invokes all
+//          registered instrumentation functions for a particular site.
 
 #include <algorithm>
 #include <iomanip>
@@ -56,6 +59,7 @@
 #include <set>
 #include <stdlib.h>
 #include <vector>
+#include <list>
 
 #include "llvm/Reoptimizer/Inst/ElfReader.h"
 #include "llvm/Reoptimizer/MemoryManager.h"
@@ -146,11 +150,173 @@
     uint64_t    m_tag;         // Entry to look for in the GBT
 };
 
+// InstFunctionInfo is the class used to represent information (e.g., address of return
+// value, pointer-to-instrumentation function) about particular registered instrumentation
+// functions.  In the case of end-interval functions, the link 'm_pStart' is filled in to
+// refer to the InstFunctionInfo instance that contains information about the
+// corresponding start-interval function.
+
+class InstFunctionInfo 
+{
+  public:
+    InstFunctionInfo(void* pRetVal,
+                     void* pInstFunc,
+                     InstFunctionInfo* pStart = 0):
+        m_pRetVal(pRetVal),
+        m_pInstFunc(pInstFunc),
+        m_pStart(pStart),
+        m_invoked(false)
+    {
+    }
+
+    InstFunctionInfo():
+        m_pRetVal(0),
+        m_pInstFunc(0),
+        m_pStart(0),
+        m_invoked(false)
+    {
+    }
+
+    void invoke();
+
+  protected:
+    void*             m_pRetVal;
+    void*             m_pInstFunc;
+    InstFunctionInfo* m_pStart;   // Info about start-interval function
+    bool              m_invoked;  // Has this function been invoked yet?
+};
+
+// InstSiteInfo instances contain information about the state of particular
+// instrumentation sites.  More specifically, it holds the instrumentation
+// status (e.g. whether or not the site has been handled by phase 4 yet) of the
+// sites as well as the list of (pointers to) the InstFunctionInfo instances
+// registered with a particular site.
+
+class InstSiteInfo 
+{
+  public:
+    InstSiteInfo():
+        m_branchInstalled(false),
+        m_instrumented(false)
+    {
+    }
+
+    // For start-interval sites only -- allocates memory for the return value of
+    // the instrumentation function (size of allocated memory is retValBytes).
+    // Returns a pointer to the InstFunctionInfo* that corresponds to the
+    // instrumentation function, or null if the function has already been
+    // registered.
+
+    InstFunctionInfo* push_back(unsigned retValBytes,
+                                void* func);
+
+    // For end-interval sites (w/ optional link to corresponding start); other
+    // site types should use this routine as well.  The provided retVal ptr is
+    // used as the return-value parameter of the instrumentation function.
+
+    void push_back(void* retVal,
+                   void* func,
+                   InstFunctionInfo* startInfo = 0);
+
+    void invokeFunctions();
+
+  protected:
+    bool                      m_branchInstalled;     // Installed branch to slot yet?
+    bool                      m_instrumented;        // Has phase 4 instrumented site?
+    uint64_t                  m_brInstallAddr;       // Address to install branch inst
+    unsigned                  m_branchInst;          // The branch inst to install
+    std::set<void*>           m_registeredFuncs;     // Set of func-ptrs registered here
+    vector<InstFunctionInfo*> m_instFuncInfos;       // Info for all registered funcs
+
+    friend void Phase4::transform();
+
+    void installBranch();
+
+    bool isRegistered(void* func) 
+    {
+        return m_registeredFuncs.find(func) != m_registeredFuncs.end();
+    }
+
+    void setInstrumented() { m_instrumented = true; }
+
+    void setBranchInst(uint64_t addr, unsigned branchInst)
+    {
+        m_brInstallAddr = addr;
+        m_branchInst = branchInst;
+    }
+};
+
+// InstInfo is the class that holds data about the instrumentation that gets
+// bound to instrumentation sites and intervals at runtime.  There should only
+// be on instance of this class (i.e., it is a singleton class).  The
+// implementation hides an STL map that maps the unique identifier associated
+// with an instrumentation interval/site to a pair of InstSiteInfo instances,
+// which contains the information about the instrumentations registered for the
+// given interval or point site.  In the case of intervals, the first element of
+// the pair is the InstSiteInfo instance that contains data about the start
+// site, whereas the second element of the pair contains data about the end
+// site.  For point sites, only the first element of the pair contains valid
+// data.
+
+class InstInfo 
+{
+  public:
+    typedef std::pair<InstSiteInfo, InstSiteInfo> SiteInfoPair;
+
+    static InstInfo* instance() 
+    {
+        if(!m_pInstance)
+            m_pInstance = new InstInfo;
+        return m_pInstance;
+    }
+
+    static InstSiteInfo* findSiteInfo(unsigned siteID, unsigned gbtType) 
+    {
+        SiteInfoPair* sip = instance()->findSiteInfo(siteID);
+        InstSiteInfo* siteInfo;
+
+        switch(gbtType) {
+            case pp::GBT_INTERVAL_START: siteInfo = &sip->first; break;
+            case pp::GBT_INTERVAL_END:   siteInfo = &sip->second; break;
+            default: assert(0 && "Unhandled gbtType encountered"); break;
+        }
+
+        return siteInfo;
+    }
+
+    SiteInfoPair* findSiteInfo(unsigned siteID)
+    {
+        SiteInfoMap::iterator i = m_siteInfoMap.find(siteID);
+        if(i == m_siteInfoMap.end())
+            return &m_siteInfoMap[siteID];
+        return &i->second;
+    }
+
+    void setVM(VirtualMem* vm) { m_pVM = vm;   }
+    VirtualMem* getVM() const  { return m_pVM; }
+
+  protected:
+    typedef std::map<unsigned, SiteInfoPair> SiteInfoMap;
+
+    InstInfo(): m_pVM(0)
+    {
+    }
+    
+    SiteInfoMap m_siteInfoMap;
+    VirtualMem* m_pVM;
+
+  private:
+    static InstInfo* m_pInstance;
+};
+
+InstInfo* InstInfo::m_pInstance = 0;
+
 //////////////// Phase 2 implementation ////////////////
 
 extern "C" void phase2() 
 {
     TraceCache* pTC = new TraceCache();
+    InstInfo::instance()->setVM(pTC->getVM());
     Phase2 ph(pTC, new SparcInstManip(pTC));
     ph.transform();
 }
@@ -192,7 +358,7 @@
         if(m_excludeSet.find(i->first) == m_excludeSet.end()) {
             // Function is not in exclude set, so go ahead and transform it
 
-            DEBUG_MSG(1, "Transforming function " << i->first
+            DEBUG_MSG(4, "Transforming function " << i->first
                       << "[" << HEX(i->second.first)
                       << ", " << HEX(i->second.second) << "]...\n");
             
@@ -217,17 +383,30 @@
     ::doFlush(slotBase, slotBase + im->getInstWidth() * snippet.size());
 }
 
+static uint64_t makeNewSlot(uint64_t srcAddr,
+                            unsigned slotSize,
+                            unsigned& branchInst,
+                            TraceCache* tc,
+                            InstManip* im)
+{
+    // Return a branch instruction to the new slot via branchInst.
+    uint64_t slotBase = tc->getMemMgr()->getMemory(slotSize);
+    assert(slotBase && "Unable to obtain memory from MemoryManager instance");
+
+    branchInst = im->getBranchAlways(slotBase, srcAddr);
+    return slotBase;
+}
+
 static uint64_t replaceInstWithBrToSlot(uint64_t srcAddr,
                                         unsigned slotSize,
                                         TraceCache* tc,
                                         InstManip* im) 
 {
-    // Obtain a new slot of the given size
-    uint64_t slotBase = tc->getMemMgr()->getMemory(slotSize);
-    assert(slotBase && "Unable to obtain memory from MemoryManager instance");
+    unsigned branchInst;
+    uint64_t slotBase = makeNewSlot(srcAddr, slotSize, branchInst, tc, im);
 
     // Replace instruction at srcAddr with branch to start of new slot
-    tc->getVM()->writeInstToVM(srcAddr, im->getBranchAlways(slotBase, srcAddr));
+    tc->getVM()->writeInstToVM(srcAddr, branchInst);
     ::doFlush(srcAddr, srcAddr + im->getInstWidth());
 
     return slotBase;
@@ -363,8 +542,8 @@
         vector<unsigned> snippet;
         m_pIM->buildSlot(p4info, snippet);
 
-        DEBUG_MSG(3, "phase4 slot instructions:\n");
-#if VERBOSE > 2
+        DEBUG_MSG(4, "phase4 slot instructions:\n");
+#if VERBOSE > 3
         dumpSnippet(snippet, m_pIM);
 #endif
 
@@ -430,7 +609,7 @@
 
     for(unsigned i = 0; i < ppGBTSize; ++i) {
         ostr << "[pp] ppGBT[" << i << "]: " << ppGBT[i].gbtType << ", "
-             << ppGBT[i].loadVar << ", " << ppGBT[i].gbtStartIdx << endl;
+             << ppGBT[i].loadVar << endl;
     }
 }
 
@@ -460,29 +639,44 @@
                && "Unexpected number of instructions in candidate");
 
         // Write NOPs over the original instructions that were associated with the elected
-        // candidate.  No need to no-op over the candidate load instruction itself since
-        // we're about to write over it with a branch to the phase 5 slot.
+        // candidate.
 
         VirtualMem* vm = m_pTC->getVM();
-        for(vector<std::pair<uint64_t, unsigned> >::const_iterator i = cand.getInsts().begin() + 1,
+        for(vector<std::pair<uint64_t, unsigned> >::const_iterator i = cand.getInsts().begin(),
                 e = cand.getInsts().end(); i != e; ++i)
             vm->writeInstToVM(i->first, m_pIM->getNOP());
 
-        // Obtain memory (& rewrite branch) to the phase 5 jump slot.
+        // Obtain new slot, the phase 5 jump slot.
         
         unsigned slotSize = m_pIM->getSlotSize(this);
         uint64_t repAddr = cand.front().first;
-        uint64_t slotBase = replaceInstWithBrToSlot(repAddr, slotSize, m_pTC, m_pIM);
+        unsigned branchInst;
+        uint64_t slotBase = makeNewSlot(repAddr, slotSize, branchInst, m_pTC, m_pIM);
 
         vector<unsigned> snippet;
         m_pIM->buildSlot(gbte, slotBase, repAddr, m_pPhase4Info->getRange(), snippet);
 
-        DEBUG_MSG(3, "phase 5 slot contents:\n");
-#if VERBOSE > 2
+        DEBUG_MSG(4, "phase 5 slot contents:\n");
+#if VERBOSE > 3
         dumpSnippet(snippet, m_pIM);
 #endif
 
         copySnippetToSlot(snippet, slotBase, m_pTC->getVM(), m_pIM);
+
+        // Grab the information about this particular site.
+
+        assert(gbte->gbtType == pp::GBT_INTERVAL_START ||
+               gbte->gbtType == pp::GBT_INTERVAL_END &&
+               "Unhandled gbtType encountered (must implement)");
+        
+        InstSiteInfo* siteInfo = InstInfo::findSiteInfo(gbte->siteID, gbte->gbtType);
+
+        // Take steps to install the branch; note that the InstSiteInfo instance
+        // knows whether or not to actually write the branch instruction, etc.
+
+        siteInfo->setInstrumented();
+        siteInfo->setBranchInst(repAddr, branchInst);
+        siteInfo->installBranch();
     }
     else {
         DEBUG_MSG(1, "does not match\n");
@@ -495,14 +689,6 @@
                           m_pPhase4Info->getCandidate().front().second);
     }
 
-#if 0
-    // (TEMP) For now, restore the candidate load to its original position for debugging
-    // purposes.
-
-    m_pPhase4Info->getTraceCache()->getVM()->writeInstToVM(m_pPhase4Info->getCandidate().front().first,
-                                                           m_pPhase4Info->getCandidate().front().second);
-#endif
-
     DEBUG_MSG(1, "================ End Phase 4 ================\n");
 }
 
@@ -510,27 +696,131 @@
 
 void phase5(GBTElem* gbte)
 {
-    switch(gbte->gbtType){
-        case pp::GBT_INTERVAL_START: {
-            DEBUG_MSG(1, "--- phase 5 start site invocation ---\n");
-            DEBUG_MSG(2, "retVal address is " << HEX(gbte->retVal) << endl);
-
-            void (*instFunc)(void*) = (void (*)(void*)) gbte->instFunc;
-            instFunc(gbte->retVal);
-            break;
-        }
-        case pp::GBT_INTERVAL_END: {
-            DEBUG_MSG(1, "--- phase 5 end site invocation ---\n");
-            DEBUG_MSG(2, "start parameter is at gbt index " << gbte->gbtStartIdx << endl);
-            DEBUG_MSG(2, "start parameter addr is "
-                      << HEX(ppGBT[gbte->gbtStartIdx].retVal) << endl);
+    DEBUG_MSG(1, "================ Begin Phase 5 ================\n");    
+    InstSiteInfo* siteInfo = InstInfo::findSiteInfo(gbte->siteID, gbte->gbtType);
+    siteInfo->invokeFunctions();
+    DEBUG_MSG(1, "================ End Phase 5 ================\n");
+}
+
+//////////////// InstSiteInfo implementation ////////////////
+
+InstFunctionInfo* InstSiteInfo::push_back(unsigned retValBytes,
+                                          void* func)
+{
+    DEBUG_MSG(3, "Inside InstSiteInfo::push_back, registering func w/ address "
+              << HEX(func) << ", new retVal of size " << retValBytes << endl);
+
+    InstFunctionInfo* fi = 0;
+
+    if(!isRegistered(func)) {
+        DEBUG_MSG(3, "not yet registered, registering...\n");
+
+        void* retVal = static_cast<void*>(new char[retValBytes]);
+        m_registeredFuncs.insert(func);
+        m_instFuncInfos.push_back(fi = new InstFunctionInfo(retVal, func));
+        installBranch();
+    }
+    else
+        DEBUG_MSG(3, "WARNING: Attempt to register instrumentation at site that was already instrumented\n");
+
+    return fi;
+}
 
-            void (*instFunc)(void*, void*) = (void (*)(void*, void*)) gbte->instFunc;
-            instFunc(gbte->retVal, ppGBT[gbte->gbtStartIdx].retVal);
+void InstSiteInfo::push_back(void* retVal,
+                             void* func,
+                             InstFunctionInfo* startInfo)
+{
+    DEBUG_MSG(3, "Inside InstSiteInfo::push_back, registering func w/ address "
+              << HEX(func) << ", retVal addr " << HEX(retVal) << endl);
+
+    if(!isRegistered(func)) {
+        DEBUG_MSG(3, "not yet registered, registering...\n");
+
+        m_registeredFuncs.insert(func);
+        m_instFuncInfos.push_back(new InstFunctionInfo(retVal, func, startInfo));
+        installBranch();
+    }
+    else
+        DEBUG_MSG(3, "WARNING: Attempt to register instrumentation at site that was already instrumented\n");
+}
+
+void InstSiteInfo::invokeFunctions()
+{
+    assert(m_branchInstalled && m_instrumented && "Invoking functions not permitted");
+
+    for(int i = 0, e = m_instFuncInfos.size(); i < e; ++i) {
+        DEBUG_MSG(3, "Invoking function " << i << endl);
+        m_instFuncInfos[i]->invoke();
+    }
+}
+
+void InstSiteInfo::installBranch()
+{
+    if(!m_branchInstalled && m_instrumented && m_instFuncInfos.size() > 0) {
+        DEBUG_MSG(3, "(InstSiteInfo::installBranch) Installing branch...\n");
+        VirtualMem* vm = InstInfo::instance()->getVM();
+        vm->writeInstToVM(m_brInstallAddr, m_branchInst);
+        m_branchInstalled = true;
+    }
+    else
+        DEBUG_MSG(3, "(InstSiteInfo::installBranch) Not installing branch, it's not needed\n");
+}
+
+//////////////// InstFunctionInfo implementation ////////////////
 
-            break;
+void InstFunctionInfo::invoke()
+{
+    // In the case of start-interval functions, the boolean m_invoked is set to true upon
+    // completion of the invocation.  This same boolean (i.e, the one associated with the
+    // start-interval InstFunctionInfo instance) is cleared by the corresponding
+    // end-interval function invocation.  This is to ensure that no end-interval
+    // instrumentation function is ever invoked unless its corresponding start-interval
+    // instrumentation function has been invoked (which implies that the start-interval
+    // site has been handled properly and all of the registration mechanisms).
+
+    DEBUG_MSG(3, "(InstFunctionInfo::invoke) retVal address is: " << HEX(m_pRetVal) << endl);
+    if(m_pStart) {
+        DEBUG_MSG(3, "End-interval inst func detected\n");
+        void (*instFunc)(void*, void*) = (void (*)(void*, void*)) m_pInstFunc;
+
+        if(m_pStart->m_invoked) {
+            DEBUG_MSG(4, "Corresponding start function has been invoked, so invoking this inst func\n");
+            instFunc(m_pRetVal, m_pStart->m_pRetVal);
+            m_pStart->m_invoked = false;
         }
+        else
+            DEBUG_MSG(4, "Corresponding start func has not been invoked, so not invoking\n");
+    }
+    else {
+        DEBUG_MSG(3, "Start-interval or non-end inst func detected\n");
+        void (*instFunc)(void*) = (void (*)(void*)) m_pInstFunc;
+        instFunc(m_pRetVal);
+        m_invoked = true;
     }
+        
+    DEBUG_MSG(3, "(InstFunctionInfo::invoke) instrumentation function returned\n");
+}
+
+//////////////// register{Interval,Point}Inst implementation ////////////////
 
-    DEBUG_MSG(1, "--- phase 5 invocation completed ---\n" << std::flush);
+extern "C" void registerIntervalInst(unsigned siteID,
+                                     void* startFunc,
+                                     void* endFunc,
+                                     unsigned paramSize,
+                                     void* retVal)
+{
+    InstInfo::SiteInfoPair* sip = InstInfo::instance()->findSiteInfo(siteID);
+    assert(sip && "Unable to obtain SiteInfoPair");
+    
+    // Handle start function
+    DEBUG_MSG(3, "registerIntervalInst: Registering start function...\n");
+    InstFunctionInfo* fi = sip->first.push_back(paramSize, startFunc);
+
+    if(fi) {
+        // Handle end function
+        DEBUG_MSG(3, "registerIntervalInst: Registering end function...\n");
+        sip->second.push_back(retVal, endFunc, fi);
+    }
+    else
+        DEBUG_MSG(3, "WARNING: Register-start-function returned 0, which implies redundant registration");
 }


Index: llvm/lib/Reoptimizer/Inst/lib/SparcInstManip.cpp
diff -u llvm/lib/Reoptimizer/Inst/lib/SparcInstManip.cpp:1.16 llvm/lib/Reoptimizer/Inst/lib/SparcInstManip.cpp:1.17
--- llvm/lib/Reoptimizer/Inst/lib/SparcInstManip.cpp:1.16	Wed May 28 08:52:41 2003
+++ llvm/lib/Reoptimizer/Inst/lib/SparcInstManip.cpp	Tue Jun 24 10:39:53 2003
@@ -242,14 +242,6 @@
 
     DEBUG_MSG(2, "buildPhase5HeapSlot completed\n");
 
-    // If we're dealing with a start-interval instrumentation function, heap-allocate
-    // its parameter memory
-
-    if(gbte->gbtType == pp::GBT_INTERVAL_START) {
-        assert(!gbte->retVal && "Expected null retVal value");
-        gbte->retVal = static_cast<void*>(new char[gbte->paramSize]);
-    }
-
     ////////////////
     // Construct the phase 5 jump slot
 
@@ -363,7 +355,7 @@
                         unsigned sizeBytes,
                         int protBits)
 {
-    DEBUG_MSG(3, "Setting access bits on heap slot page(s)" << endl);
+    DEBUG_MSG(4, "Setting access bits on heap slot page(s)" << endl);
 
     int rc = mprotect(pageBase, sizeBytes, protBits);
     if(rc < 0) {
@@ -416,10 +408,10 @@
 
     setPageBits(heapSlot, numHeapBytes, PROT_READ | PROT_WRITE | PROT_EXEC);
 
-#if VERBOSE > 2
-    DEBUG_MSG(3, "Dumping contents of heap-slot memory...\n");
+#if VERBOSE > 3
+    DEBUG_MSG(4, "Dumping contents of heap-slot memory...\n");
     dumpHeapSlot(heapSlot, getPhase5HeapSize(), this);
-    DEBUG_MSG(3, "Done with heap region construction, moving on to jump slot\n");
+    DEBUG_MSG(4, "Done with heap region construction, moving on to jump slot\n");
 #endif
 
     return heapSlot;
@@ -462,7 +454,7 @@
         return addr + getInstWidth();
     }
     else {
-        DEBUG_MSG(2, "WARNING: Non-save instruction at function entry\n");
+        DEBUG_MSG(4, "WARNING: Non-save instruction at function entry\n");
         return addr;
     }
     
@@ -569,7 +561,7 @@
                                                m_logicalToActualReg[useForIndirect],
                                                offset));
 
-    DEBUG_MSG(3, "JMPL instruction word is " << HEX(m_pCurrSnippet->back()) << endl);
+    DEBUG_MSG(4, "JMPL instruction word is " << HEX(m_pCurrSnippet->back()) << endl);
     m_pCurrSnippet->push_back(getNOP());
 
     assert(m_pCurrSnippet->size() - initSize == GEN_JMPL_SIZE &&


Index: llvm/lib/Reoptimizer/Inst/lib/design.txt
diff -u llvm/lib/Reoptimizer/Inst/lib/design.txt:1.15 llvm/lib/Reoptimizer/Inst/lib/design.txt:1.16
--- llvm/lib/Reoptimizer/Inst/lib/design.txt:1.15	Sun May 18 12:45:26 2003
+++ llvm/lib/Reoptimizer/Inst/lib/design.txt	Tue Jun 24 10:39:53 2003
@@ -880,91 +880,70 @@
 
 {{{ MILESTONES
 
+  - Experiments
+  - The Paper
+  - The Thesis
 
 }}}
 
 {{{ TODO
 
-    - Move statically-sized spill regions so that they are internal to SparcInstManip.
-      (do not need variable-sized spill region except for phase5 invocations)
+    In priority order:
 
-    - Start table-of-stacks implementation for phase4 authorship of phase 5 slots.
-      - Placed on hold temporary because of "alloca-finding" approach. However, see the 
-        following e-mail for the current state of things:
-
-        {{{ E-mail regarding alloca-finding and table-of-stacks approach
-Okay, this is starting to seem intractable. I have another problem that
-I don't think can be resolved without resorting to a custom-stack
-mechanism that will incur prohibitive overhead.
-
-Everything is working for start-region instrumentation sites.  For
-end-region instrumentation sites, however, there's a problem. In order
-to write the slot for end sites, I have to know (or know how to compute)
-the address of the return value of the corresponding start site. I had
-originally thought that I would just store this in the GBT, or
-"something", but I clearly didn't think through the problem well enough.
-
-There are only two ways I can think of that this can occur:
-
-(a) Write the effective address of the return value of the start inst
-func, so that it gets passed to the end inst func.
-
-or
-
-(b) Somehow encode the stack offset to the return value from the start
-inst, where the offset is from the %sp *at the end-region site*
-
-Both of these have problems.
-
-First, I don't think (b) can work at all, given that there may be
-alloca's present in the original application that would change the %sp,
-and thus the offset value that we'd need, and we can't determine the
-exact allocas that are executed statically.
-
-For (a), the effective address isn't known until runtime. We can store
-this address in some global table where the phase 4 invocation for the
-end site can find it, but it is not sufficient to have a single scalar
-address here -- we must have a stack, due to potential recursive
-invocations. I think that this is clear, please let me know if I'm not
-making sense. :) 
-
-Hence, we'd need to maintain a stack of effective addresses, which was
-pushed during the execution of phase 5 for the start site, and then read
-and popped during the execution of phase 5 for the end site.  We're
-already really bloated with how many instructions we've got going on for
-all of the spills, etc, and I'm concerned about the effect that this
-stack manipulation will have on our overhead, as we talked about before.
-
-The way I see it, we only have two options if we're to make forward
-progress and not obliterate our chances of having lower overhead
-numbers. Hopefully we have some better choices. In the interests of
-short-term forward progress, I'm going to go with #1 for now.
-
-#1 - Make another common-case assumption that there will be no allocas
-between start and end sites, on *any* control path. If this is the case,
-then we know that the stack pointer will not have been manipulated (I
-think) between the start and end sites, and so the %sp offsets to the
-requisite data will be unchanged since when the phase 5 step occurred
-for the start site.  
-
-#2 - Just implement our fall-back solution that everything seems to be
-pointing to. I'm not sure exactly what other logistic nightmares might
-be entailed in this, though, because I've only a sketch of the idea.
-
-I wanted to point out, also, that the so-called "fall back" approach we
-discussed previous also involves manipulation of a stack at runtime
-(push/pop actions still have to occur at runtime), so perhaps the stack
-of effective addresses is less prohibitive than I thought, if only in
-the sense that we cannot avoid it. :(
-        }}}
-
-    - Write phase 5 stuff for end-region sites -- will assume that not allocas lie between
-    the start and end sites, which is not particularly a fair assumption.
-
-    - Optimizations:
-        - No need to save registers (other than those clobbered) in phase 3 slot, since phase 3
-          is invoked at the start of the function. Must still spill/restore shared, though.
-        - No need to save registers (other than those clobbered) in general.
+    1) New implementation for instrumenting at function-level granularity
+    2) Apache through LLVM, experiments
+    3) Writing, writing, writing: ICS version 2 paper, do
+       a) outline
+       b) intro
+       c) language section
+       d) compiler section
+
+    The three top-level items above are more-or-less interchangable. However, the
+    experiments will not be able to be completed unless the "instrumentation @
+    function-level granularity" implementation is done, and there should be an emphasis on
+    getting Apache through LLVM in the short-term because Chris is leaving for Norway in
+    early July.
+
+    However, *all* of the writing (including experimentation sections) for ICS v2 must be
+    done by the end of June, so that Adve can approve it, make the desired corrections,
+    etc., so that I can get started on thesis authorship and submission.  Realistically,
+    the timeframe should look something like this:
+
+    Week of 6/10 (5 days): Implementation, Apache w/ bug reports
+    Week of 6/16 (6 days): Implementation, Small example, continue Apache & do tests
+
+    -- At this point, all experiments should be more-or-less completed --
+
+    Week of 6/23 (6 days): Write, write, write. 3 days nonstop for content, 3
+                           days of Adve making corrections, etc.
+
+    Monday, 6/30 is D-Day...
+
+    Schedule revision 13 Jun 2003: Apache is in stasis waiting to hear back from
+    Chris. Some effort might be expended to see if any more work can be done on compiling
+    Apache even though the build process currently fails altogether.  Implementation for
+    instrumentation at function-level granularity is in stasis until I hear back from
+    Vikram.  This leaves three immediate options open until either of the two are
+    resolved, in which case forward progress on either the implementation or Apache should
+    be made.
+
+       Option 1) POV-Ray through LLVM.
+       Option 2) Writing
+       Option 3) "Small example"
+
+    Option 3 really shouldn't be undertaken until we know if obtaining things like I/O
+    elapsed time (via function-level) is possible, or until I can talk with Vikram in our
+    next meeting about what the heck this nebulous example should look like...this leaves
+    only options 1 and 2 above as viable.
+
+    POV-Ray should be an easy thing to start, and would be good both as a fall-back if
+    Apache isn't possible as well as a useful additional example if the latter is
+    possible.  This would also give V & C some time to respond to the pending queries.
+
+    - Optimizations: - No need to save registers (other than those clobbered) in
+    phase 3 slot, since phase 3 is invoked at the start of the function. Must
+    still spill/restore shared, though.  - No need to save registers (other than
+    those clobbered) in general.
 
 }}}
 
@@ -1460,5 +1439,226 @@
 Also, Chris remarked that any novel page-management mechanisms, etc., (for the
 code that is jumped to in newly-allocated pages) that I devise should perhaps be
 integrated into the LLVM JIT if they are suitable.
+
+}}}
+
+{{{ Experiments
+
+We must devise experiments that demonstrate the novel aspects of the work.  We
+are currently planning on using Apache and/or POV-Ray, and demonstrating how a
+"deep" performance analysis can be encoded using performance primitives. A
+"deep" performance analysis is one which essentially (using Hollingsworth's
+terminology from his PhD thesis) gets at the "what, where, and when" aspects of
+performance bottlenecks.  However, instead of doing this at the "arbitrary
+program during execution" level, as the W^3 search model does, we will encode
+these performance aspects at the application level itself.
+
+Here's what Vikram suggested for a good start:
+
+  I just thought that the examples of performance issues he explores in his
+  automatic search would give you (a) some insights into what performance issues
+  make sense to consider, and (b) some ideas about how to do a systematic
+  diagnosis, albeit at the application level.
+  
+  But isn't detecting the "what, why, and when" of performance bottlenecks
+  pretty closely related to the goal of performance diagnosis?  We'd be looking
+  for bottlenecks too, except that we can use application domain information, we
+  can look for bottlenecks at the algorithmic level instead of the general
+  system level, and we can record it permanently in the program as a first class
+  feature of the program.
+  
+  Anyway, about your second question: here's a way to do what I suggesting:
+  
+  -- think about 2-3 key performance issues with Apache (or POV-Ray) that you'd
+  want to diagnose e.g., cache misses, TLB misses, thread overhead (estimating
+  that could be interesting), I/O delays
+  
+  -- if those issues make sense with a small sort example, try to diagnose those
+  issues in the small example first.  e.g., I think cache misses, TLB misses,
+  and I/O delays would all be issues if you were sorting a huge file of some
+  kind.
+  
+  This is purely to give you a small, well-understood code to try out before
+  going to the big ones where it may be difficult to know, when one diagnosis
+  attempt fails, whether it failed because you misunderstood the performance
+  issues or because the guess was wrong or both.
+
+TLB misses aren't an option, because we cannot get at them with PAPI. Cache
+misses are available, so that'd certainly be a good place to start.  As for I/O
+delays, I have no idea how we'd measure this, either.  The following is the
+comprehensive list of the low-level metrics that are exposed to us via PAPI.  On
+our own, we can support simple things like elapsed time, load average, etc.  I'm
+not altogether clear how we'd determine how much time (for a particular region)
+was spent doing I/O-bound activities...
+
+Number of hardware counters: 2
+Name: PAPI_L1_ICM 	Description: Level 1 instruction cache misses
+Name: PAPI_L2_TCM 	Description: Level 2 cache misses
+Name: PAPI_CA_SNP 	Description: Requests for a snoop
+Name: PAPI_CA_INV 	Description: Requests for cache line invalidation
+Name: PAPI_L1_LDM 	Description: Level 1 load misses
+Name: PAPI_L1_STM 	Description: Level 1 store misses
+Name: PAPI_BR_MSP 	Description: Conditional branch instructions mispredicted
+Name: PAPI_TOT_IIS 	Description: Instructions issued
+Name: PAPI_TOT_INS 	Description: Instructions completed
+Name: PAPI_LD_INS 	Description: Load instructions
+Name: PAPI_SR_INS 	Description: Store instructions
+Name: PAPI_TOT_CYC 	Description: Total cycles
+Name: PAPI_IPS    	Description: Instructions per second
+Name: PAPI_L1_DCR 	Description: Level 1 data cache reads
+Name: PAPI_L1_DCW 	Description: Level 1 data cache writes
+Name: PAPI_L1_ICH 	Description: Level 1 instruction cache hits
+Name: PAPI_L2_ICH 	Description: Level 2 instruction cache hits
+Name: PAPI_L1_ICA 	Description: Level 1 instruction cache accesses
+Name: PAPI_L2_TCH 	Description: Level 2 total cache hits
+Name: PAPI_L2_TCA 	Description: Level 2 total cache accesses
+
+So, in the short-term, we have two outstanding problems. First, what kinds of
+metric would we want to apply to Apache/POV-Ray/Simple example? Second, of those
+metrics, which can we actually realize with the current system? Third, we should
+create the simple program (such as sorting large amounts of data from a file,
+etc.) so that it can use these metrics in such a way that "models" or
+"anticipates" the way they will be used in the bigger applications.
+
+This is an outstanding issue, and I don't really know where to go with it yet.
+
+More notes about this as of 4 Jun 2003:
+
+One way to obtain metric values that are based on the elapsed times of
+particular functions is to somehow register instrumentation for those particular
+functions, and for a particular region -- Vikram argues that we have the ability
+to do this dynamically and we don't need any markers or phase-1 actions because
+we're operating at function-level granularity. 
+
+Here is a sample scenario: We have defined an interval I over some scoped region
+of code.  During phase 1 and phase 2, no instrumentation is registered for this
+interval.  Later on, we construct a metric that is qualified by a list of
+functions that (for example) are to have their runtimes measured and added to
+some running total.  Let's call this the "measure_functions_start" and
+"measure_functions_end" metric, and have it yield a value of type double which
+is the aggregate runtime of the list of functions when they get executed within
+interval I.  The metric registration function will have to have some way
+(varargs?) of denoting what the functions are: perhaps it can simply pass in an
+array of function names together with a size.
+
+Example: 
+
+pp_registerIntervalInst(intervalID, measure_functions_start,
+                        measure_functions_end, &retVal, sizeof(double), 
+                        func1, func2, func3, ...);
+
+However, what do "measure_functions_start" have to do with anything? More than
+likely, what we need to do is specify a particular metric to apply to a
+paritcular function, such that the value will be sampled each time that function
+gets executed.  Then, since there can be multiple invocations (and hence,
+multiple samples) for this selected function within I, we will have to have some
+default or user-specified way of aggregating the data :(.  This is gross.  In
+other words, we should probably simplify the above to something like:
+
+pp_registerIntervalInst(intervalID, some_metric_start, some_metric_end, 
+                        &retVal, sizeof(double), HOW_TO_AGGREGATE, func1);
+
+Where func1 is the function to be instrumented and HOW_TO_AGGREGATE is some
+value that specifies one from a couple of ways of combining the data.  For now,
+HOW_TO_AGGREGATE will not exist and will implicitly sum all return
+values...hence, if the some_metric_{start,end} function ptrs above were to
+elapsed time, at the end of the interval I with interval id intervalID, retVal
+would contain the combined elapsed time of all time spent in function func1.
+
+Enabling this measurement at the start of the interval, disabling at the end of
+the interval; clearly, the start-instrumentation site will need to transform
+func1 (to compute the metric value) when crossed, and the end-insturmentation
+site must remove such instrumentation when crossed. This doesn't follow the
+normal model of what occurs at the instrumentation sites, in terms of just
+ripping down a list of functions...or does it? Perhaps one of the functions in
+the list is just this function that does the transformation on the target
+function...in this case, the process would look something like:
+
+1. Register the transformation function (for the start point) -- call this
+xform_start -- as a regular instrumentation function.  The runtime call to do
+the registration will build the appropriate data structures which will encode
+what metric to associated with the target function, aggregation method, return
+value, etc.
+
+2. Register the transformation function (for the end point) -- call this
+xform_end -- as a regular instrumentation function.
+
+3. When xform_start is invoked as a regular instrumentation function, it will
+instrument the target function with the selected instrumentation.
+
+This is the hardest step to conceptualize and realize.  The problem is that,
+without any placeholders from phase 1, it's not clear that we can instrument the
+target function easily.  Clearly, our instrumentation points are at the start
+and end points of the function (entry and exit).  But this is not really true.
+Our instrumentation points are really at the entry and *all* function exits.
+
+The important question is, what if we have all of the exit points at our
+disposal? Would that change anything?
+
+It would. The entry point together with all exit points would form a set of
+instrumentation points. For each of these instrumentation points, we could
+over-write a branch to a new slot that would call the desired instrumentation
+function, restore the replaced instruction, and return to the instrumentation
+point to continue execution.  This would potentially work. One major problem
+that comes to mind is that for system calls (such as read()), the body of the
+function is highly likely to be out of short-jmp range to the
+tracecache-allocated slot.  The only way around this would be to create a heap
+region and copy the target function into it, etc.  We don't have any code to do
+this yet, so again there is no code reuse or (much) leveraging of existing
+functionality.  Additionally, work will have to be done to make the data
+structure that maps the address of a function to its extents (which come from
+the ELf information).
+
+--
+
+The other alternative is to make sure a "wrapper" function (i.e., for the
+function read()):
+
+int read_wrapper(args for read) {
+  start_inst();
+  int rc = read(args for read);
+  end_inst();
+  return rc;
+}
+
+But this isn't an option because we cannot locate the calls to read() to replace
+them with the wrapper.  We could do the following:
+
+3a. Copy the entire target function to a heap region, and instrument it to our
+heart's content.  However, finding exit points may not be easy without a CFG,
+etc.
+
+3b. Replace the body of the real target function with a call to the modified
+duplicate of the target function, returning whatever the modified duplicate
+returns.
+
+This works, I think, but is incredibly cumbersome and, contrary to what was
+previously discussed, we do *not* possess all of the mechanisms.
+
+--
+
+Notes on POV-Ray experiment 23 Jun 2003
+
+We propose using a user-defined metric, flops/display pixel to compute the
+computational cost per pixel.
+
+1. Compute flops/dpixel and report using pp mechanisms.
+  1a. flops can be obtained via PAPI, IIRC.
+
+2. Create a moving average of flops/dpixel.
+
+3. When the flops/dpixel exceeds mavg flops/dpixel by some multiplicative
+threshold (e.g.), a performance assertion is violated.  When the PA is violated,
+we report other metrics (perhaps ranked in some manner) that had been being
+recorded but not reported.  Initial suggestions for the other metrics to measure
+_over the same region_ (say, the main routine for a single ray-trace) are:
+
+    1) MAverage L1 icache misses vs. this-ray L1 icache misses
+    2) MAverage L2 cache misses vs. this-ray L2 cache misses
+    3) MAverage load count vs. this-ray load count
+    4) MAverage store count vs. this-ray store count
+
+Immediate problem: I don't think we can monitor more than 2 of these values
+concurrently due to hardware limitations on the SPARC...
 
 }}}




From criswell at choi.cs.uiuc.edu  Tue Jun 24 11:01:01 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Tue Jun 24 11:01:01 2003
Subject: [llvm-commits] CVS: llvm/Makefile.common
Message-ID: <200306241600.h5OG0C011480@choi.cs.uiuc.edu>


Changes in directory llvm:

Makefile.common updated: 1.88.2.1 -> 1.88.2.2

---
Log message:

Added some code from the GNU Make manual that prevents dependencies from
being generated if the user is attempting to do a make clean.


---
Diffs of the changes:

Index: llvm/Makefile.common
diff -u llvm/Makefile.common:1.88.2.1 llvm/Makefile.common:1.88.2.2
--- llvm/Makefile.common:1.88.2.1	Mon Jun 23 14:11:41 2003
+++ llvm/Makefile.common	Tue Jun 24 11:00:01 2003
@@ -708,6 +708,12 @@
 	$(VERB) $(RM) -f core core.[0-9][0-9]* *.o *.d *.so *~ *.flc
 	$(VERB) $(RM) -f $(LEX_OUTPUT) $(YACC_OUTPUT)
 
+###########################################################################
+# C/C++ Dependencies
+#	Define variables and rules that generate header file dependencies
+#	from C/C++ source files.
+###########################################################################
+
 # If dependencies were generated for the file that included this file,
 # include the dependancies now...
 #
@@ -724,6 +730,12 @@
 $(BUILD_OBJ_DIR)/Depend/%.d: $(SourceDir)%.c $(BUILD_OBJ_DIR)/Depend/.dir
 	$(VERB) $(DependC) $< | $(SED) 's|$*\.o *|Release/& Profile/& Debug/& Depend/$(@F)|g' > $@
 
+#
+# Include dependencies generated from C/C++ source files, but not if we
+# are cleaning (this example taken from the GNU Make Manual).
+#
+ifneq ($(MAKECMDGOALS),clean)
 ifneq ($(SourceDepend),)
 -include $(SourceDepend)
+endif
 endif




From lattner at cs.uiuc.edu  Tue Jun 24 12:32:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Jun 24 12:32:01 2003
Subject: [llvm-commits] CVS: llvm/test/Programs/TEST.nightly.Makefile
Message-ID: <200306241731.MAA09127@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Programs:

TEST.nightly.Makefile updated: 1.16 -> 1.17

---
Log message:

Really fix crafty in nightly tester


---
Diffs of the changes:

Index: llvm/test/Programs/TEST.nightly.Makefile
diff -u llvm/test/Programs/TEST.nightly.Makefile:1.16 llvm/test/Programs/TEST.nightly.Makefile:1.17
--- llvm/test/Programs/TEST.nightly.Makefile:1.16	Mon Jun 23 17:36:09 2003
+++ llvm/test/Programs/TEST.nightly.Makefile	Tue Jun 24 12:31:24 2003
@@ -104,7 +104,7 @@
 # Overall tests: just run subordinate tests
 $(PROGRAMS_TO_TEST:%=Output/%.$(TEST).report.txt): \
 Output/%.$(TEST).report.txt: $(addprefix Output/%.nightly., $(REPORTS_SUFFIX))
-	cat $(addprefix Output/$*.nightly., $(REPORTS_SUFFIX)) > $@
+	-cat $(addprefix Output/$*.nightly., $(REPORTS_SUFFIX)) > $@
 
 
 




From criswell at choi.cs.uiuc.edu  Tue Jun 24 13:27:01 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Tue Jun 24 13:27:01 2003
Subject: [llvm-commits] CVS: llvm/utils/TableGen/Makefile
Message-ID: <200306241826.h5OIQW206583@choi.cs.uiuc.edu>


Changes in directory llvm/utils/TableGen:

Makefile updated: 1.4 -> 1.4.2.1

---
Log message:

Added a dependency that forces make to generate the parser header file
before generating header files dependencies.



---
Diffs of the changes:

Index: llvm/utils/TableGen/Makefile
diff -u llvm/utils/TableGen/Makefile:1.4 llvm/utils/TableGen/Makefile:1.4.2.1
--- llvm/utils/TableGen/Makefile:1.4	Tue Dec  3 00:00:11 2002
+++ llvm/utils/TableGen/Makefile	Tue Jun 24 13:26:22 2003
@@ -6,6 +6,13 @@
 
 include $(LEVEL)/Makefile.common
 
+#
+# Make the source file depend on the header file.  In this way, dependencies
+# (which depend on the source file) won't get generated until bison is done
+# generating the C source and header files for the parser.
+#
+FileLexer.cpp: FileParser.h
+
 clean::
 	-rm -f FileParser.cpp FileParser.h FileLexer.cpp CommandLine.cpp
 	-rm -f FileParser.output




From criswell at choi.cs.uiuc.edu  Tue Jun 24 13:28:03 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Tue Jun 24 13:28:03 2003
Subject: [llvm-commits] CVS: llvm/configure
Message-ID: <200306241827.h5OIREK06604@choi.cs.uiuc.edu>


Changes in directory llvm:

configure added (r1.1.2.1)

---
Log message:

Initial checkin of autoconf.


---
Diffs of the changes:





From criswell at choi.cs.uiuc.edu  Tue Jun 24 13:48:02 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Tue Jun 24 13:48:02 2003
Subject: [llvm-commits] CVS: llvm/lib/AsmParser/Makefile
Message-ID: <200306241847.h5OIlOB23481@choi.cs.uiuc.edu>


Changes in directory llvm/lib/AsmParser:

Makefile updated: 1.2 -> 1.2.2.1

---
Log message:

Added a rule that prevents make from generating header file dependencies
until after the parser file (*.y) is processed by bison.


---
Diffs of the changes:

Index: llvm/lib/AsmParser/Makefile
diff -u llvm/lib/AsmParser/Makefile:1.2 llvm/lib/AsmParser/Makefile:1.2.2.1
--- llvm/lib/AsmParser/Makefile:1.2	Mon Apr  8 15:17:06 2002
+++ llvm/lib/AsmParser/Makefile	Tue Jun 24 13:47:14 2003
@@ -5,3 +5,9 @@
 
 include $(LEVEL)/Makefile.common
 
+#
+# Make the source code file for the lexer depend upon the header file generated
+# by the Bison parser.  This prevents the generation of dependencies from
+# being performed until after the header file has been created.
+#
+Lexer.cpp: llvmAsmParser.h




From lattner at cs.uiuc.edu  Tue Jun 24 13:50:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Jun 24 13:50:01 2003
Subject: [llvm-commits] CVS: llvm/test/Programs/External/SPEC/Sandbox.sh
Message-ID: <200306241849.NAA10723@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Programs/External/SPEC:

Sandbox.sh updated: 1.7 -> 1.8

---
Log message:

Silence a silly warning


---
Diffs of the changes:

Index: llvm/test/Programs/External/SPEC/Sandbox.sh
diff -u llvm/test/Programs/External/SPEC/Sandbox.sh:1.7 llvm/test/Programs/External/SPEC/Sandbox.sh:1.8
--- llvm/test/Programs/External/SPEC/Sandbox.sh:1.7	Fri Jun 20 15:42:46 2003
+++ llvm/test/Programs/External/SPEC/Sandbox.sh	Tue Jun 24 13:49:38 2003
@@ -22,7 +22,7 @@
 cp -r $INPUT_DIR/* .
 
 # If there is an "all" input directory, take it as well...
-cp -r $INPUT_DIR/../../all/input/* . 2>&1 > /dev/null
+cp -r $INPUT_DIR/../../all/input/* . > /dev/null 2>&1 
 
 # Run the program now...
 echo Running: $*




From lattner at cs.uiuc.edu  Tue Jun 24 13:51:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Jun 24 13:51:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/Makefile
Message-ID: <200306241850.NAA10737@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms:

Makefile updated: 1.23 -> 1.24

---
Log message:

Run the tail dup tests


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/Makefile
diff -u llvm/test/Regression/Transforms/Makefile:1.23 llvm/test/Regression/Transforms/Makefile:1.24
--- llvm/test/Regression/Transforms/Makefile:1.23	Tue May 27 11:10:03 2003
+++ llvm/test/Regression/Transforms/Makefile	Tue Jun 24 13:49:51 2003
@@ -23,6 +23,7 @@
 	Reassociate \
 	SCCP \
 	ScalarRepl \
-	SimplifyCFG
+	SimplifyCFG \
+	TailDup
 
 include $(LEVEL)/Makefile.common




From lattner at cs.uiuc.edu  Tue Jun 24 14:18:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Jun 24 14:18:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/TailDup/2003-06-24-Simpleloop.ll
Message-ID: <200306241917.OAA11383@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/TailDup:

2003-06-24-Simpleloop.ll added (r1.1)

---
Log message:

Testcase reduced by bugpoint


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/TailDup/2003-06-24-Simpleloop.ll
diff -c /dev/null llvm/test/Regression/Transforms/TailDup/2003-06-24-Simpleloop.ll:1.1
*** /dev/null	Tue Jun 24 14:17:47 2003
--- llvm/test/Regression/Transforms/TailDup/2003-06-24-Simpleloop.ll	Tue Jun 24 14:17:37 2003
***************
*** 0 ****
--- 1,18 ----
+ ; RUN: as < %s | opt -tailduplicate -disable-output
+ 
+ void %motion_result7() {
+ entry:
+         br label %endif
+ 
+ endif:
+         %i.1 = phi int [ %inc, %no_exit ], [ 0, %entry ]
+         %inc = add int %i.1, 1
+         br bool false, label %no_exit, label %UnifiedExitNode
+ 
+ no_exit:
+         br bool false, label %UnifiedExitNode, label %endif
+ 
+ UnifiedExitNode:
+         ret void
+ }
+ 




From jstanley at cs.uiuc.edu  Tue Jun 24 14:48:03 2003
From: jstanley at cs.uiuc.edu (Joel Stanley)
Date: Tue Jun 24 14:48:03 2003
Subject: [llvm-commits] CVS: llvm/lib/Reoptimizer/BinInterface/sparcbin.cpp
Message-ID: <200306241947.OAA17591@trinity.cs.uiuc.edu>


Changes in directory llvm/lib/Reoptimizer/BinInterface:

sparcbin.cpp updated: 1.7 -> 1.8

---
Log message:

Fix to remove warning.



---
Diffs of the changes:

Index: llvm/lib/Reoptimizer/BinInterface/sparcbin.cpp
diff -u llvm/lib/Reoptimizer/BinInterface/sparcbin.cpp:1.7 llvm/lib/Reoptimizer/BinInterface/sparcbin.cpp:1.8
--- llvm/lib/Reoptimizer/BinInterface/sparcbin.cpp:1.7	Wed Jun  4 04:40:54 2003
+++ llvm/lib/Reoptimizer/BinInterface/sparcbin.cpp	Tue Jun 24 14:47:23 2003
@@ -274,7 +274,7 @@
       if (instr->flags & IF_BRINTERNAL)
 	printf("{dest:s%02d} ", instr->ibranch.dest);
       else
-	printf("{dest:0x%x} ", (intptr_t)instr->ebranch.dest);
+	printf("{dest:0x%lx} ", (intptr_t)instr->ebranch.dest);
       
       sparc_print_pseudo(instr->instr, 
 			 instr->alu.genrs1, 




From jstanley at cs.uiuc.edu  Tue Jun 24 14:49:01 2003
From: jstanley at cs.uiuc.edu (Joel Stanley)
Date: Tue Jun 24 14:49:01 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Reoptimizer/BinInterface/bitmath.h
Message-ID: <200306241948.OAA17815@trinity.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Reoptimizer/BinInterface:

bitmath.h updated: 1.10 -> 1.11

---
Log message:

Fix to remove warning.



---
Diffs of the changes:

Index: llvm/include/llvm/Reoptimizer/BinInterface/bitmath.h
diff -u llvm/include/llvm/Reoptimizer/BinInterface/bitmath.h:1.10 llvm/include/llvm/Reoptimizer/BinInterface/bitmath.h:1.11
--- llvm/include/llvm/Reoptimizer/BinInterface/bitmath.h:1.10	Mon Jun 23 18:31:47 2003
+++ llvm/include/llvm/Reoptimizer/BinInterface/bitmath.h	Tue Jun 24 14:48:43 2003
@@ -33,7 +33,7 @@
 #define FLD_LOWER(FLD_DEF) (0 ? FLD_DEF)
 #define MASKBELOW(V)       ((((uint64_t) 1) << V) - 1) // 64-bit '1' constant here to avoid overflow warnings
 #define MASKEQBELOW(V)	   ((1 << V) | MASKBELOW(V)) //masks off everything ABOVE
-#define RD_FLD(x, FLD)     ((x & MASKEQBELOW(FLD_UPPER(FLD)))  >> FLD_LOWER(FLD))
+#define RD_FLD(x, FLD)     (unsigned)(((x & MASKEQBELOW(FLD_UPPER(FLD)))  >> FLD_LOWER(FLD)))
 #define MK_FLD(FLD, val)   ((val << FLD_LOWER(FLD)) & MASKEQBELOW(FLD_UPPER(FLD)))
 #define MASK_FLD(FLD)      (MASKEQBELOW(FLD_UPPER(FLD)) & ~MASKBELOW(FLD_LOWER(FLD)))
 #define RM_FLD(FLD, val)   (val & ~(MASK_FLD(FLD)))




From lattner at cs.uiuc.edu  Tue Jun 24 14:49:04 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Jun 24 14:49:04 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Scalar/TailDuplication.cpp
Message-ID: <200306241948.OAA17231@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Scalar:

TailDuplication.cpp updated: 1.3 -> 1.4

---
Log message:

Fix bug: TailDup/2003-06-24-Simpleloop.ll


---
Diffs of the changes:

Index: llvm/lib/Transforms/Scalar/TailDuplication.cpp
diff -u llvm/lib/Transforms/Scalar/TailDuplication.cpp:1.3 llvm/lib/Transforms/Scalar/TailDuplication.cpp:1.4
--- llvm/lib/Transforms/Scalar/TailDuplication.cpp:1.3	Sun Jun 22 15:46:00 2003
+++ llvm/lib/Transforms/Scalar/TailDuplication.cpp	Tue Jun 24 14:48:06 2003
@@ -203,7 +203,8 @@
   for (Value::use_iterator I = OrigInst->use_begin(), E = OrigInst->use_end();
        I != E; ++I) {
     Instruction *In = cast<Instruction>(*I);
-    if (In->getParent() != OrigBlock)  // Don't modify uses in the orig block!
+    if (In->getParent() != OrigBlock ||  // Don't modify uses in the orig block!
+        isa<PHINode>(In))
       Users.push_back(In);
   }
 




From lattner at cs.uiuc.edu  Tue Jun 24 15:30:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Jun 24 15:30:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/SCCP/2003-06-24-OverdefinedPHIValue.ll
Message-ID: <200306242029.PAA17770@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/SCCP:

2003-06-24-OverdefinedPHIValue.ll added (r1.1)

---
Log message:

New testcase


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/SCCP/2003-06-24-OverdefinedPHIValue.ll
diff -c /dev/null llvm/test/Regression/Transforms/SCCP/2003-06-24-OverdefinedPHIValue.ll:1.1
*** /dev/null	Tue Jun 24 15:29:46 2003
--- llvm/test/Regression/Transforms/SCCP/2003-06-24-OverdefinedPHIValue.ll	Tue Jun 24 15:29:36 2003
***************
*** 0 ****
--- 1,36 ----
+ ; RUN: if as < %s | opt -sccp -simplifycfg | dis | grep then:
+ ; RUN: then exit 1
+ ; RUN: else exit 0
+ ; RUN: fi
+ 
+ void %cprop_test11(int* %data.1) {
+ entry:		; No predecessors!
+ 	%tmp.1 = load int* %data.1		; <int> [#uses=3]
+ 	%tmp.41 = setgt int %tmp.1, 1		; <bool> [#uses=1]
+ 	br bool %tmp.41, label %no_exit, label %loopexit
+ 
+ no_exit:		; preds = %entry, %then, %endif
+ 	%j.0 = phi int [ %j.0, %endif ], [ %i.0, %then ], [ 1, %entry ]		; <int> [#uses=3]
+ 	%i.0 = phi int [ %inc, %endif ], [ %inc1, %then ], [ 1, %entry ]		; <int> [#uses=4]
+ 	%tmp.8.not = cast int %j.0 to bool		; <bool> [#uses=1]
+ 	br bool %tmp.8.not, label %endif, label %then
+ 
+ then:		; preds = %no_exit
+ 	%inc1 = add int %i.0, 1		; <int> [#uses=3]
+ 	%tmp.42 = setlt int %inc1, %tmp.1		; <bool> [#uses=1]
+ 	br bool %tmp.42, label %no_exit, label %loopexit
+ 
+ endif:		; preds = %no_exit
+ 	%inc = add int %i.0, 1		; <int> [#uses=3]
+ 	%tmp.4 = setlt int %inc, %tmp.1		; <bool> [#uses=1]
+ 	br bool %tmp.4, label %no_exit, label %loopexit
+ 
+ loopexit:		; preds = %entry, %endif, %then
+ 	%j.1 = phi int [ 1, %entry ], [ %j.0, %endif ], [ %i.0, %then ]		; <int> [#uses=1]
+ 	%i.1 = phi int [ 1, %entry ], [ %inc, %endif ], [ %inc1, %then ]		; <int> [#uses=1]
+ 	%tmp.17 = getelementptr int* %data.1, long 1		; <int*> [#uses=1]
+ 	store int %j.1, int* %tmp.17
+ 	%tmp.23 = getelementptr int* %data.1, long 2		; <int*> [#uses=1]
+ 	store int %i.1, int* %tmp.23
+ 	ret void
+ }




From lattner at cs.uiuc.edu  Tue Jun 24 15:31:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Jun 24 15:31:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Scalar/SCCP.cpp
Message-ID: <200306242030.PAA17784@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Scalar:

SCCP.cpp updated: 1.69 -> 1.70

---
Log message:

Fix bug: SCCP/2003-06-24-OverdefinedPHIValue.ll


---
Diffs of the changes:

Index: llvm/lib/Transforms/Scalar/SCCP.cpp
diff -u llvm/lib/Transforms/Scalar/SCCP.cpp:1.69 llvm/lib/Transforms/Scalar/SCCP.cpp:1.70
--- llvm/lib/Transforms/Scalar/SCCP.cpp:1.69	Tue May 20 16:01:16 2003
+++ llvm/lib/Transforms/Scalar/SCCP.cpp	Tue Jun 24 15:29:52 2003
@@ -420,12 +420,13 @@
   for (unsigned i = 0, e = PN.getNumIncomingValues(); i != e; ++i) {
     InstVal &IV = getValueState(PN.getIncomingValue(i));
     if (IV.isUndefined()) continue;  // Doesn't influence PHI node.
-    if (IV.isOverdefined()) {   // PHI node becomes overdefined!
-      markOverdefined(&PN);
-      return;
-    }
     
     if (isEdgeFeasible(PN.getIncomingBlock(i), PN.getParent())) {
+      if (IV.isOverdefined()) {   // PHI node becomes overdefined!
+        markOverdefined(&PN);
+        return;
+      }
+
       if (OperandVal == 0) {   // Grab the first value...
         OperandVal = IV.getConstant();
       } else {                // Another value is being merged in!




From criswell at choi.cs.uiuc.edu  Tue Jun 24 15:37:01 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Tue Jun 24 15:37:01 2003
Subject: [llvm-commits] CVS: llvm/Makefile.common
Message-ID: <200306242036.h5OKaFt05770@choi.cs.uiuc.edu>


Changes in directory llvm:

Makefile.common updated: 1.88.2.2 -> 1.88.2.3

---
Log message:

Added comments about the new environment variables (LLVM_SRC_DIR and friends).
Removed the use of Makefile.Linux and Makefile.SunOS.
Fixed the errors regarding invalid options to libtool for building static
libraries.


---
Diffs of the changes:

Index: llvm/Makefile.common
diff -u llvm/Makefile.common:1.88.2.2 llvm/Makefile.common:1.88.2.3
--- llvm/Makefile.common:1.88.2.2	Tue Jun 24 11:00:01 2003
+++ llvm/Makefile.common	Tue Jun 24 15:36:04 2003
@@ -30,15 +30,24 @@
 #    are in, if they are not in the current directory.  This should include a
 #    trailing / character.
 #
-# 6. PROJ_COMPILE - If set to 1, then this makefile can also be used to
-#    compile other projects using llvm. Note if this option is set then the
-#    following *must* hold
-#    PROJLEVEL should be set to the top of the source directory for the
-#               project files 
-#    LEVEL should be set to the top of LLVM source tree
-#    LLVM_LIB_DIR should be set to the top of the LLVM build tree
+# 6. LLVM_SRC_ROOT - If specified, points to the top of the LLVM source tree.
 #
-#    
+# 7. LLVM_OBJ_ROOT - If specified, points to the top directory where LLVM
+#    object files are placed.
+#
+# 8. BUILD_SRC_DIR - The directory which contains the current set of Makefiles
+#    and usually the source code too (unless SourceDir is set).
+#
+# 9. BUILD_SRC_ROOT - The root directory of the source code being compiled.
+#
+# 10. BUILD_OBJ_DIR - The directory where object code should be placed.
+#
+# 11. BUILD_OBJ_ROOT - The root directory for where object code should be
+#     placed.
+#
+# For building,
+# 	LLVM, LLVM_SRC_ROOT = BUILD_SRC_ROOT, and
+# 	LLVM_OBJ_ROOT = BUILD_OBJ_ROOT.
 #===-----------------------------------------------------------------------====
 
 # Configuration file to set paths specific to local installation of LLVM
@@ -113,12 +122,6 @@
 LLVM_OBJ_ROOT = $(BUILD_OBJ_ROOT)
 endif
 
-# Figure out how to do platform specific stuff on this platform.  This is really
-# gross and should be autoconfiscated (automake actually), but should hopefully
-# work on Linux and solaris (SunOS).
-#
-include $(LLVM_SRC_ROOT)/Makefile.$(OS)
-
 ###########################################################################
 # Default Targets:
 #	The following targets are the standard top level targets for
@@ -345,7 +348,7 @@
 #
 # Configure where the item being compiled should go.
 #
-ifdef LIBRARYNAME
+ifdef SHARED_LIBRARY
 Link := $(Link) -rpath $(DESTLIBCURRENT)
 endif
 
@@ -484,17 +487,17 @@
 $(LIBNAME_AO): $(LObjectsO) $(LibSubDirs) $(DESTLIBRELEASE)/.dir
 	@echo ======= Linking $(LIBRARYNAME) release library =======
 	@$(RM) -f $@
-	$(VERB) $(LinkO) -o $@ $(LObjectsO) $(LibSubDirs) -static
+	$(VERB) $(Link) -03 -o $@ $(LObjectsO) $(LibSubDirs) -static
 
 $(LIBNAME_AP): $(LObjectsP) $(LibSubDirs) $(DESTLIBPROFILE)/.dir
 	@echo ======= Linking $(LIBRARYNAME) profile library =======
 	@$(RM) -f $@
-	$(VERB) $(LinkP) -o $@ $(LObjectsP) $(LibSubDirs) -static
+	$(VERB) $(Link) -03 $(PROFILE) -o $@ $(LObjectsP) $(LibSubDirs) -static
 
 $(LIBNAME_AG): $(ObjectsG) $(LibSubDirs) $(DESTLIBDEBUG)/.dir
 	@echo ======= Linking $(LIBRARYNAME) debug library =======
 	@$(RM) -f $@
-	$(VERB) $(LinkG) -o $@ $(LObjectsG) $(LibSubDirs) -static
+	$(VERB) $(Link) -g $(STRIP) -o $@ $(LObjectsG) $(LibSubDirs) -static
 
 #
 # Rules for building .o libraries.
@@ -708,6 +711,12 @@
 	$(VERB) $(RM) -f core core.[0-9][0-9]* *.o *.d *.so *~ *.flc
 	$(VERB) $(RM) -f $(LEX_OUTPUT) $(YACC_OUTPUT)
 
+distclean:: clean
+	$(VERB) $(RM) -rf $(LEVEL)/Makefile.config \
+	                  $(LEVEL)/include/Config/config.h \
+	                  $(LEVEL)/autom4te.cache \
+	                  $(LEVEL)/config.log
+
 ###########################################################################
 # C/C++ Dependencies
 #	Define variables and rules that generate header file dependencies
@@ -719,6 +728,13 @@
 #
 SourceBaseNames := $(basename $(notdir $(filter-out Debug/%, $(Source))))
 SourceDepend := $(SourceBaseNames:%=$(BUILD_OBJ_DIR)/Depend/%.d)
+
+#
+# Depend target:
+#	This allows a user to manually ask for an update in the dependencies
+#
+depend: $(SourceDepend)
+
 
 # Create dependencies for the *.cpp files...
 #$(SourceDepend): \x




From criswell at choi.cs.uiuc.edu  Tue Jun 24 16:20:03 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Tue Jun 24 16:20:03 2003
Subject: [llvm-commits] CVS: llvm/Makefile.Linux Makefile.SunOS Makefile.config.in configure configure.ac
Message-ID: <200306242119.h5OLJPS12224@choi.cs.uiuc.edu>


Changes in directory llvm:

Makefile.Linux updated: 1.10 -> 1.10.2.1
Makefile.SunOS updated: 1.9 -> 1.9.2.1
Makefile.config.in updated: 1.1.2.1 -> 1.1.2.2
configure updated: 1.1.2.1 -> 1.1.2.2
configure.ac updated: 1.1.2.1 -> 1.1.2.2

---
Log message:

Moved functionality from Makefile.<platform> to the autoconf scripts.



---
Diffs of the changes:

Index: llvm/Makefile.Linux
diff -u llvm/Makefile.Linux:1.10 llvm/Makefile.Linux:1.10.2.1
--- llvm/Makefile.Linux:1.10	Sun May 25 11:52:41 2003
+++ llvm/Makefile.Linux	Tue Jun 24 16:19:14 2003
@@ -8,16 +8,6 @@
 # Set the architecture so that x86 platform dependant code is compmiled
 ARCH := x86
 
-# MakeSharedObjectOption - This option is passed to the linker by
-# Makefile.common when compiling a shared object.
-#
-MakeSharedObjectOption := -shared
-
-# PLATFORMLINKOPTS - On Linux, we must specify this linker option so that all
-# symbols are exported by tools for runtime use by dlsym.
-#
-PLATFORMLINKOPTS := -Wl,--export-dynamic
-
 # PLATFORMSTRIPOPTS - Command to pass to the link phase to cause symbols to be 
 # stripped from the resultant binary.
 PLATFORMSTRIPOPTS := -Wl,-x


Index: llvm/Makefile.SunOS
diff -u llvm/Makefile.SunOS:1.9 llvm/Makefile.SunOS:1.9.2.1
--- llvm/Makefile.SunOS:1.9	Sun May 25 11:52:41 2003
+++ llvm/Makefile.SunOS	Tue Jun 24 16:19:14 2003
@@ -8,15 +8,6 @@
 # Set the architecture so that Sparc platform dependant code is compmiled
 ARCH := Sparc
 
-# MakeSharedObjectOption - This option is passed to the linker by
-# Makefile.common when compiling a shared object.
-#
-MakeSharedObjectOption := -G
-
-# PLATFORMLINKOPTS - On sun we need to link to libmalloc.so to get a definition
-# of mallinfo.
-PLATFORMLINKOPTS := -lmalloc
-
 # PLATFORMSTRIPOPTS - Command to pass to the link phase to cause symbols to be 
 # stripped from the resultant binary.
 #


Index: llvm/Makefile.config.in
diff -u llvm/Makefile.config.in:1.1.2.1 llvm/Makefile.config.in:1.1.2.2
--- llvm/Makefile.config.in:1.1.2.1	Mon Jun 23 13:59:24 2003
+++ llvm/Makefile.config.in	Tue Jun 24 16:19:14 2003
@@ -63,7 +63,7 @@
 #
 # Determine the target for which LLVM should generate code.
 #
-#LLVMGCCARCH := @target@/3.4-llvm
+LLVMGCCARCH := @target@/3.4-llvm
 
 # Path to directory where object files should be stored during a build.
 # Set OBJ_ROOT to "." if you do not want to use a separate place for
@@ -75,7 +75,7 @@
 # Path to location for LLVM front-end this should only be specified here if you
 # want to override the value set in Makefile.$(uname)
 #
-#LLVMGCCDIR := @LLVMGCCDIR@
+LLVMGCCDIR := @LLVMGCCDIR@
 
 # When this setting is set to true, programs in the llvm/test/Programs hierarchy
 # are not recompiled from source code.  Instead, the bytecode for the file is
@@ -146,4 +146,14 @@
 #
 #ENABLE_PURIFY = 1
 @ENABLE_PURIFY@
+
+#
+# Enable JIT for this platform
+#
+ at JIT@
+
+#
+# Disable LLC diffs for testing.
+#
+ at DISABLE_LLC_DIFFS@
 


Index: llvm/configure
diff -u llvm/configure:1.1.2.1 llvm/configure:1.1.2.2
--- llvm/configure:1.1.2.1	Tue Jun 24 13:27:03 2003
+++ llvm/configure	Tue Jun 24 16:19:14 2003
@@ -1009,13 +1009,19 @@
   --enable-spec           Compile SPEC benchmarks (default is NO)
   --enable-precompiled_bytecode
                           Use pre-compiled bytecode (default is NO)
+  --enable-jit            Enable Just In Time Compiling (default is NO)
 
 Optional Packages:
   --with-PACKAGE[=ARG]    use PACKAGE [ARG=yes]
   --without-PACKAGE       do not use PACKAGE (same as --with-PACKAGE=no)
   --with-gnu-ld           assume the C compiler uses GNU ld default=no
   --with-pic              try to use only PIC/non-PIC objects default=use both
-
+  --with-spec             Location of SPEC benchmarks
+  --with-llvmgccdir       Location of LLVM GCC front-end
+  --with-bcrepos          Location of Bytecode Repository
+  --with-papi             Location of PAPI
+  --with-objroot          Location where object files should be placed
+                          (default is .)
 
 Some influential environment variables:
   CXX         C++ compiler command
@@ -1578,14 +1584,30 @@
     NONENONEs,x,x, &&
   program_prefix=${target_alias}-
 
-echo $build
 case $build in
 	*i*86*)  OS=Linux
 
+	         DISABLE_LLC_DIFFS=DISABLE_LLC_DIFFS:=1
+
+	         LLVMGCCDIR=/home/vadve/lattner/local/x86/llvm-gcc/
+
 	         ;;
+
 	*sparc*) OS=SunOS
 
+	         LLVMGCCDIR=/home/vadve/lattner/local/sparc/llvm-gcc/
+
 	         ;;
+
+	*)       OS=Unknown
+
+	         ;;
+esac
+
+case $target in
+	*sparc*solaris*)  target=sparcv9-sun-solaris2.8
+
+	                  ;;
 esac
 
 
@@ -4274,7 +4296,7 @@
 case $host in
 *-*-irix6*)
   # Find out which ABI we are using.
-  echo '#line 4277 "configure"' > conftest.$ac_ext
+  echo '#line 4299 "configure"' > conftest.$ac_ext
   if { (eval echo "$as_me:$LINENO: \"$ac_compile\"") >&5
   (eval $ac_compile) 2>&5
   ac_status=$?
@@ -4824,7 +4846,7 @@
 save_CFLAGS="$CFLAGS"
 CFLAGS="$CFLAGS -o out/conftest2.$ac_objext"
 compiler_c_o=no
-if { (eval echo configure:4827: \"$ac_compile\") 1>&5; (eval $ac_compile) 2>out/conftest.err; } && test -s out/conftest2.$ac_objext; then
+if { (eval echo configure:4849: \"$ac_compile\") 1>&5; (eval $ac_compile) 2>out/conftest.err; } && test -s out/conftest2.$ac_objext; then
   # The compiler can only warn and ignore the option if not recognized
   # So say no if there are warnings
   if test -s out/conftest.err; then
@@ -6655,7 +6677,7 @@
     lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2
   lt_status=$lt_dlunknown
   cat > conftest.$ac_ext <<EOF
-#line 6658 "configure"
+#line 6680 "configure"
 #include "confdefs.h"
 
 #if HAVE_DLFCN_H
@@ -6753,7 +6775,7 @@
     lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2
   lt_status=$lt_dlunknown
   cat > conftest.$ac_ext <<EOF
-#line 6756 "configure"
+#line 6778 "configure"
 #include "confdefs.h"
 
 #if HAVE_DLFCN_H
@@ -11343,6 +11365,78 @@
    { (exit 1); exit 1; }; }
 fi
 
+echo "$as_me:$LINENO: checking for mprotect" >&5
+echo $ECHO_N "checking for mprotect... $ECHO_C" >&6
+if test "${ac_cv_func_mprotect+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  cat >conftest.$ac_ext <<_ACEOF
+#line $LINENO "configure"
+#include "confdefs.h"
+/* System header to define __stub macros and hopefully few prototypes,
+    which can conflict with char mprotect (); below.  */
+#include <assert.h>
+/* Override any gcc2 internal prototype to avoid an error.  */
+#ifdef __cplusplus
+extern "C"
+#endif
+/* We use char because int might match the return type of a gcc2
+   builtin and then its argument prototype would still apply.  */
+char mprotect ();
+char (*f) ();
+
+#ifdef F77_DUMMY_MAIN
+#  ifdef __cplusplus
+     extern "C"
+#  endif
+   int F77_DUMMY_MAIN() { return 1; }
+#endif
+int
+main ()
+{
+/* The GNU C library defines this for functions which it implements
+    to always fail with ENOSYS.  Some functions are actually named
+    something starting with __ and the normal name is an alias.  */
+#if defined (__stub_mprotect) || defined (__stub___mprotect)
+choke me
+#else
+f = mprotect;
+#endif
+
+  ;
+  return 0;
+}
+_ACEOF
+rm -f conftest.$ac_objext conftest$ac_exeext
+if { (eval echo "$as_me:$LINENO: \"$ac_link\"") >&5
+  (eval $ac_link) 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); } &&
+         { ac_try='test -s conftest$ac_exeext'
+  { (eval echo "$as_me:$LINENO: \"$ac_try\"") >&5
+  (eval $ac_try) 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; }; then
+  ac_cv_func_mprotect=yes
+else
+  echo "$as_me: failed program was:" >&5
+cat conftest.$ac_ext >&5
+ac_cv_func_mprotect=no
+fi
+rm -f conftest.$ac_objext conftest$ac_exeext conftest.$ac_ext
+fi
+echo "$as_me:$LINENO: result: $ac_cv_func_mprotect" >&5
+echo "${ECHO_T}$ac_cv_func_mprotect" >&6
+if test $ac_cv_func_mprotect = yes; then
+  :
+else
+  { { echo "$as_me:$LINENO: error: Function mprotect() required but not found" >&5
+echo "$as_me: error: Function mprotect() required but not found" >&2;}
+   { (exit 1); exit 1; }; }
+fi
+
 
 # Check whether --enable-purify or --disable-purify was given.
 if test "${enable_purify+set}" = set; then
@@ -11398,6 +11492,20 @@
   UPB=
 
 fi;
+case $build in
+	*i*86*)  # Check whether --enable-jit or --disable-jit was given.
+if test "${enable_jit+set}" = set; then
+  enableval="$enable_jit"
+  JIT=TARGET_HAS_JIT=1
+
+else
+  JIT=
+
+fi;
+	         ;;
+	*)
+	         ;;
+esac
 
 
 # Check whether --with-spec or --without-spec was given.
@@ -12018,6 +12126,8 @@
 s, at target_vendor@,$target_vendor,;t t
 s, at target_os@,$target_os,;t t
 s, at OS@,$OS,;t t
+s, at DISABLE_LLC_DIFFS@,$DISABLE_LLC_DIFFS,;t t
+s, at LLVMGCCDIR@,$LLVMGCCDIR,;t t
 s, at CXX@,$CXX,;t t
 s, at CXXFLAGS@,$CXXFLAGS,;t t
 s, at LDFLAGS@,$LDFLAGS,;t t
@@ -12059,8 +12169,8 @@
 s, at ENABLE_OPTIMIZED@,$ENABLE_OPTIMIZED,;t t
 s, at USE_SPEC@,$USE_SPEC,;t t
 s, at UPB@,$UPB,;t t
+s, at JIT@,$JIT,;t t
 s, at SPEC_ROOT@,$SPEC_ROOT,;t t
-s, at LLVMGCCDIR@,$LLVMGCCDIR,;t t
 s, at BCR@,$BCR,;t t
 s, at PAPIDIR@,$PAPIDIR,;t t
 s, at OBJROOT@,$OBJROOT,;t t


Index: llvm/configure.ac
diff -u llvm/configure.ac:1.1.2.1 llvm/configure.ac:1.1.2.2
--- llvm/configure.ac:1.1.2.1	Mon Jun 23 13:54:23 2003
+++ llvm/configure.ac	Tue Jun 24 16:19:14 2003
@@ -33,17 +33,32 @@
 
 dnl
 dnl Now, for some of our own magic:
-dnl We will use the build machine information to determine which master
-dnl Makefile dnl to load.  It will either be Linux or SparcV9
+dnl We will use the build machine information to set some variables.
 dnl
-echo $build
 case $build in
 	*i*86*)  AC_SUBST(OS,[Linux])
+	         AC_SUBST(DISABLE_LLC_DIFFS,[[DISABLE_LLC_DIFFS:=1]])
+	         AC_SUBST(LLVMGCCDIR,[/home/vadve/lattner/local/x86/llvm-gcc/])
 	         ;;
+
 	*sparc*) AC_SUBST(OS,[SunOS])
+	         AC_SUBST(LLVMGCCDIR,[/home/vadve/lattner/local/sparc/llvm-gcc/])
+	         ;;
+
+	*)       AC_SUBST(OS,[Unknown])
 	         ;;
 esac
 
+dnl
+dnl If we are on a Solaris machine, pretend that it is V9, since that is all
+dnl that we support at the moment, and autoconf will only tell us we're a
+dnl sparc.
+dnl
+case $target in
+	*sparc*solaris*)  AC_SUBST(target,[[sparcv9-sun-solaris2.8]])
+	                  ;;
+esac
+
 dnl **************************************************************************
 dnl * Check for programs.
 dnl **************************************************************************
@@ -190,6 +205,7 @@
 dnl As the code is made more portable (i.e. less reliant on these functions,
 dnl these checks should go away.
 AC_CHECK_FUNC(mmap,,AC_MSG_ERROR([Function mmap() required but not found]))
+AC_CHECK_FUNC(mprotect,,AC_MSG_ERROR([Function mprotect() required but not found]))
 
 dnl **************************************************************************
 dnl * Enable various compile-time options
@@ -200,14 +216,20 @@
 AC_ARG_ENABLE(optimized,AC_HELP_STRING([--enable-optimized],[Compile with optimizations enabled (default is NO)]), AC_SUBST(ENABLE_OPTIMIZED,[[ENABLE_OPTIMIZED=1]]), AC_SUBST(ENABLE_OPTIMIZED,[[]]))
 AC_ARG_ENABLE(spec,AC_HELP_STRING([--enable-spec],[Compile SPEC benchmarks (default is NO)]), AC_SUBST(USE_SPEC,[[USE_SPEC=1]]), AC_SUBST(USE_SPEC,[[]]))
 AC_ARG_ENABLE(precompiled_bytecode,AC_HELP_STRING([--enable-precompiled_bytecode],[Use pre-compiled bytecode (default is NO)]), AC_SUBST(UPB,[[USE_PRECOMPILED_BYTECODE=1]]), AC_SUBST(UPB,[[]]))
+case $build in
+	*i*86*)  AC_ARG_ENABLE(jit,AC_HELP_STRING([--enable-jit],[Enable Just In Time Compiling (default is NO)]), AC_SUBST(JIT,[[TARGET_HAS_JIT=1]]), AC_SUBST(JIT,[[]]))
+	         ;;
+	*)
+	         ;;
+esac
 
 dnl **************************************************************************
 dnl * Set the location of various third-party software packages
 dnl **************************************************************************
-AC_ARG_WITH(spec,,AC_SUBST(SPEC_ROOT,[$withval]),AC_SUBST(SPEC_ROOT,[/home/vadve/shared/benchmarks/speccpu2000/benchspec]))
-AC_ARG_WITH(llvmgccdir,,AC_SUBST(LLVMGCCDIR,[$withval]))
-AC_ARG_WITH(bcrepos,,AC_SUBST(BCR,[$withval]),AC_SUBST(BCR,[/home/vadve/lattner/LLVMPrograms]))
-AC_ARG_WITH(papi,,AC_SUBST(PAPIDIR,[$withval]),AC_SUBST(PAPIDIR,[/home/vadve/shared/papi-2.3.4.1]))
-AC_ARG_WITH(objroot,,AC_SUBST(OBJROOT,[$withval]),AC_SUBST(OBJROOT,[.]))
+AC_ARG_WITH(spec,AC_HELP_STRING([--with-spec],[Location of SPEC benchmarks]),AC_SUBST(SPEC_ROOT,[$withval]),AC_SUBST(SPEC_ROOT,[/home/vadve/shared/benchmarks/speccpu2000/benchspec]))
+AC_ARG_WITH(llvmgccdir,AC_HELP_STRING([--with-llvmgccdir],[Location of LLVM GCC front-end]),AC_SUBST(LLVMGCCDIR,[$withval]))
+AC_ARG_WITH(bcrepos,AC_HELP_STRING([--with-bcrepos],[Location of Bytecode Repository]),AC_SUBST(BCR,[$withval]),AC_SUBST(BCR,[/home/vadve/lattner/LLVMPrograms]))
+AC_ARG_WITH(papi,AC_HELP_STRING([--with-papi],[Location of PAPI]),AC_SUBST(PAPIDIR,[$withval]),AC_SUBST(PAPIDIR,[/home/vadve/shared/papi-2.3.4.1]))
+AC_ARG_WITH(objroot,AC_HELP_STRING([--with-objroot],[Location where object files should be placed (default is .)]),AC_SUBST(OBJROOT,[$withval]),AC_SUBST(OBJROOT,[.]))
 
 AC_OUTPUT(Makefile.config)




From lattner at cs.uiuc.edu  Tue Jun 24 16:22:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Jun 24 16:22:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/ADCE/basictest1.ll basictest2.ll
Message-ID: <200306242121.QAA19269@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/ADCE:

basictest1.ll updated: 1.2 -> 1.3
basictest2.ll updated: 1.2 -> 1.3

---
Log message:

Modernize testcases


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/ADCE/basictest1.ll
diff -u llvm/test/Regression/Transforms/ADCE/basictest1.ll:1.2 llvm/test/Regression/Transforms/ADCE/basictest1.ll:1.3
--- llvm/test/Regression/Transforms/ADCE/basictest1.ll:1.2	Thu Aug 22 11:39:04 2002
+++ llvm/test/Regression/Transforms/ADCE/basictest1.ll	Tue Jun 24 16:21:31 2003
@@ -31,7 +31,7 @@
 	br bool %cond266, label %bb3, label %bb2
 
 bb2:					;[#uses=1]
-	%cast273 = getelementptr [17 x sbyte]* %.LC12, uint 0, uint 0		; <sbyte*> [#uses=0]
+	%cast273 = getelementptr [17 x sbyte]* %.LC12, long 0, long 0		; <sbyte*> [#uses=0]
 	br label %bb3
 
 bb3:					;[#uses=2]
@@ -39,19 +39,19 @@
 	br bool %cond267, label %bb5, label %bb4
 
 bb4:					;[#uses=2]
-	%reg111 = getelementptr [20 x %FILE]* %__iob, uint 0, uint 1, ubyte 3		; <ubyte*> [#uses=1]
-	%cast274 = getelementptr [34 x sbyte]* %.LC9, uint 0, uint 0		; <sbyte*> [#uses=0]
+	%reg111 = getelementptr [20 x %FILE]* %__iob, long 0, long 1, ubyte 3		; <ubyte*> [#uses=1]
+	%cast274 = getelementptr [34 x sbyte]* %.LC9, long 0, long 0		; <sbyte*> [#uses=0]
 	%cast282 = cast ubyte* %reg111 to %FILE*		; <%FILE*> [#uses=0]
 	call void %exit( int 1 )
 	br label %UnifiedExitNode
 
 bb5:					;[#uses=1]
-	%reg107-idxcast1 = cast int %fd to uint		; <uint> [#uses=2]
-	%reg107-idxcast2 = cast int %fd to uint		; <uint> [#uses=1]
-	%reg1311 = getelementptr [3 x %spec_fd_t]* %spec_fd, uint 0, uint %reg107-idxcast2		; <%spec_fd_t*> [#uses=1]
-	%idx1 = getelementptr [3 x %spec_fd_t]* %spec_fd, uint 0, uint %reg107-idxcast1, ubyte 2		; <int> [#uses=3]
+	%reg107-idxcast1 = cast int %fd to long		; <long> [#uses=2]
+	%reg107-idxcast2 = cast int %fd to long		; <long> [#uses=1]
+	%reg1311 = getelementptr [3 x %spec_fd_t]* %spec_fd, long 0, long %reg107-idxcast2		; <%spec_fd_t*> [#uses=1]
+	%idx1 = getelementptr [3 x %spec_fd_t]* %spec_fd, long 0, long %reg107-idxcast1, ubyte 2		; <int> [#uses=3]
 	%reg1321 = load int* %idx1
-	%idx2 = getelementptr %spec_fd_t* %reg1311, uint 0, ubyte 1		; <int> [#uses=1]
+	%idx2 = getelementptr %spec_fd_t* %reg1311, long 0, ubyte 1		; <int> [#uses=1]
 	%reg1331 = load int* %idx2
 	%cond270 = setlt int %reg1321, %reg1331		; <bool> [#uses=1]
 	br bool %cond270, label %bb9, label %bb6
@@ -62,29 +62,29 @@
 	br bool %cond271, label %bb8, label %bb7
 
 bb7:					;[#uses=1]
-	%cast277 = getelementptr [4 x sbyte]* %.LC10, uint 0, uint 0		; <sbyte*> [#uses=0]
+	%cast277 = getelementptr [4 x sbyte]* %.LC10, long 0, long 0		; <sbyte*> [#uses=0]
 	br label %bb8
 
 bb8:					;[#uses=3]
 	br label %UnifiedExitNode
 
 bb9:					;[#uses=1]
-	%reg107-idxcast3 = cast int %fd to uint		; <uint> [#uses=1]
-	%idx3 = getelementptr [3 x %spec_fd_t]* %spec_fd, uint 0, uint %reg107-idxcast3, ubyte 3		; <ubyte*> [#uses=1]
+	%reg107-idxcast3 = cast int %fd to long		; <long> [#uses=1]
+	%idx3 = getelementptr [3 x %spec_fd_t]* %spec_fd, long 0, long %reg107-idxcast3, ubyte 3		; <ubyte*> [#uses=1]
 	%reg1601 = load ubyte** %idx3
-	%reg132-idxcast1 = cast int %reg1321 to uint		; <uint> [#uses=1]
-	%idx4 = getelementptr ubyte* %reg1601, uint %reg132-idxcast1		; <ubyte> [#uses=2]
+	%reg132-idxcast1 = cast int %reg1321 to long		; <long> [#uses=1]
+	%idx4 = getelementptr ubyte* %reg1601, long %reg132-idxcast1		; <ubyte> [#uses=2]
 	%reg1621 = load ubyte* %idx4
-	%cast108 = cast ubyte %reg1621 to uint		; <uint> [#uses=0]
+	%cast108 = cast ubyte %reg1621 to long		; <long> [#uses=0]
 	%reg157 = add int %reg1321, 1		; <int> [#uses=1]
-	%idx5 = getelementptr [3 x %spec_fd_t]* %spec_fd, uint 0, uint %reg107-idxcast1, ubyte 2
+	%idx5 = getelementptr [3 x %spec_fd_t]* %spec_fd, long 0, long %reg107-idxcast1, ubyte 2
 	store int %reg157, int* %idx5
 	%reg163 = load int* %dbglvl		; <int> [#uses=1]
 	%cond272 = setle int %reg163, 4		; <bool> [#uses=1]
 	br bool %cond272, label %bb11, label %bb10
 
 bb10:					;[#uses=1]
-	%cast279 = getelementptr [4 x sbyte]* %.LC11, uint 0, uint 0		; <sbyte*> [#uses=0]
+	%cast279 = getelementptr [4 x sbyte]* %.LC11, long 0, long 0		; <sbyte*> [#uses=0]
 	br label %bb11
 
 bb11:					;[#uses=3]


Index: llvm/test/Regression/Transforms/ADCE/basictest2.ll
diff -u llvm/test/Regression/Transforms/ADCE/basictest2.ll:1.2 llvm/test/Regression/Transforms/ADCE/basictest2.ll:1.3
--- llvm/test/Regression/Transforms/ADCE/basictest2.ll:1.2	Thu Aug 22 11:39:04 2002
+++ llvm/test/Regression/Transforms/ADCE/basictest2.ll	Tue Jun 24 16:21:31 2003
@@ -31,7 +31,7 @@
 	br bool %cond266, label %bb3, label %bb2
 
 bb2:					;[#uses=1]
-	%cast273 = getelementptr [17 x sbyte]* %.LC12, uint 0, uint 0		; <sbyte*> [#uses=0]
+	%cast273 = getelementptr [17 x sbyte]* %.LC12, long 0, long 0		; <sbyte*> [#uses=0]
 	br label %bb3
 
 bb3:					;[#uses=2]
@@ -40,19 +40,19 @@
 	br label %bb5
 
 bb4:					;[#uses=2]
-	%reg111 = getelementptr [20 x %FILE]* %__iob, uint 0, uint 1, ubyte 3		; <ubyte*> [#uses=1]
-	%cast274 = getelementptr [34 x sbyte]* %.LC9, uint 0, uint 0		; <sbyte*> [#uses=0]
+	%reg111 = getelementptr [20 x %FILE]* %__iob, long 0, long 1, ubyte 3		; <ubyte*> [#uses=1]
+	%cast274 = getelementptr [34 x sbyte]* %.LC9, long 0, long 0		; <sbyte*> [#uses=0]
 	%cast282 = cast ubyte* %reg111 to %FILE*		; <%FILE*> [#uses=0]
 	call void %exit( int 1 )
 	br label %UnifiedExitNode
 
 bb5:					;[#uses=1]
-	%reg107-idxcast1 = cast int %fd to uint		; <uint> [#uses=2]
-	%reg107-idxcast2 = cast int %fd to uint		; <uint> [#uses=1]
-	%reg1311 = getelementptr [3 x %spec_fd_t]* %spec_fd, uint 0, uint %reg107-idxcast2		; <%spec_fd_t*> [#uses=1]
-	%idx1 = getelementptr [3 x %spec_fd_t]* %spec_fd, uint 0, uint %reg107-idxcast1, ubyte 2		; <int> [#uses=3]
+	%reg107-idxcast1 = cast int %fd to long		; <long> [#uses=2]
+	%reg107-idxcast2 = cast int %fd to long		; <long> [#uses=1]
+	%reg1311 = getelementptr [3 x %spec_fd_t]* %spec_fd, long 0, long %reg107-idxcast2		; <%spec_fd_t*> [#uses=1]
+	%idx1 = getelementptr [3 x %spec_fd_t]* %spec_fd, long 0, long %reg107-idxcast1, ubyte 2		; <int> [#uses=3]
 	%reg1321 = load int* %idx1
-	%idx2 = getelementptr %spec_fd_t* %reg1311, uint 0, ubyte 1		; <int> [#uses=1]
+	%idx2 = getelementptr %spec_fd_t* %reg1311, long 0, ubyte 1		; <int> [#uses=1]
 	%reg1331 = load int* %idx2
 	%cond270 = setlt int %reg1321, %reg1331		; <bool> [#uses=1]
 	br bool %cond270, label %bb9, label %bb6
@@ -63,29 +63,29 @@
 	br bool %cond271, label %bb8, label %bb7
 
 bb7:					;[#uses=1]
-	%cast277 = getelementptr [4 x sbyte]* %.LC10, uint 0, uint 0		; <sbyte*> [#uses=0]
+	%cast277 = getelementptr [4 x sbyte]* %.LC10, long 0, long 0		; <sbyte*> [#uses=0]
 	br label %bb8
 
 bb8:					;[#uses=3]
 	br label %UnifiedExitNode
 
 bb9:					;[#uses=1]
-	%reg107-idxcast3 = cast int %fd to uint		; <uint> [#uses=1]
-	%idx3 = getelementptr [3 x %spec_fd_t]* %spec_fd, uint 0, uint %reg107-idxcast3, ubyte 3		; <ubyte*> [#uses=1]
+	%reg107-idxcast3 = cast int %fd to long		; <uint> [#uses=1]
+	%idx3 = getelementptr [3 x %spec_fd_t]* %spec_fd, long 0, long %reg107-idxcast3, ubyte 3		; <ubyte*> [#uses=1]
 	%reg1601 = load ubyte** %idx3
-	%reg132-idxcast1 = cast int %reg1321 to uint		; <uint> [#uses=1]
-	%idx4 = getelementptr ubyte* %reg1601, uint %reg132-idxcast1		; <ubyte> [#uses=2]
+	%reg132-idxcast1 = cast int %reg1321 to long		; <long> [#uses=1]
+	%idx4 = getelementptr ubyte* %reg1601, long %reg132-idxcast1		; <ubyte> [#uses=2]
 	%reg1621 = load ubyte* %idx4
-	%cast108 = cast ubyte %reg1621 to uint		; <uint> [#uses=0]
+	%cast108 = cast ubyte %reg1621 to long		; <long> [#uses=0]
 	%reg157 = add int %reg1321, 1		; <int> [#uses=1]
-	%idx5 = getelementptr [3 x %spec_fd_t]* %spec_fd, uint 0, uint %reg107-idxcast1, ubyte 2
+	%idx5 = getelementptr [3 x %spec_fd_t]* %spec_fd, long 0, long %reg107-idxcast1, ubyte 2
 	store int %reg157, int* %idx5
 	%reg163 = load int* %dbglvl		; <int> [#uses=1]
 	%cond272 = setle int %reg163, 4		; <bool> [#uses=1]
 	br bool %cond272, label %bb11, label %bb10
 
 bb10:					;[#uses=1]
-	%cast279 = getelementptr [4 x sbyte]* %.LC11, uint 0, uint 0		; <sbyte*> [#uses=0]
+	%cast279 = getelementptr [4 x sbyte]* %.LC11, long 0, long 0		; <sbyte*> [#uses=0]
 	br label %bb11
 
 bb11:					;[#uses=3]




From lattner at cs.uiuc.edu  Tue Jun 24 16:23:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Jun 24 16:23:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/ADCE/2003-06-24-BasicFunctionality.ll
Message-ID: <200306242122.QAA19419@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/ADCE:

2003-06-24-BasicFunctionality.ll added (r1.1)

---
Log message:

Make sure adce is basically working!


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/ADCE/2003-06-24-BasicFunctionality.ll
diff -c /dev/null llvm/test/Regression/Transforms/ADCE/2003-06-24-BasicFunctionality.ll:1.1
*** /dev/null	Tue Jun 24 16:22:19 2003
--- llvm/test/Regression/Transforms/ADCE/2003-06-24-BasicFunctionality.ll	Tue Jun 24 16:22:09 2003
***************
*** 0 ****
--- 1,43 ----
+ ; RUN: if as < %s | opt -adce -simplifycfg | dis | grep then:
+ ; RUN: then exit 1
+ ; RUN: else exit 0
+ ; RUN: fi
+ 
+ void %dead_test8(int* %data.1, int %idx.1) {
+ entry:		; No predecessors!
+ 	%tmp.1 = load int* %data.1		; <int> [#uses=2]
+ 	%tmp.41 = setgt int %tmp.1, 0		; <bool> [#uses=1]
+ 	br bool %tmp.41, label %no_exit.preheader, label %return
+ 
+ no_exit.preheader:		; preds = %entry
+ 	%tmp.11 = getelementptr int* %data.1, long 1		; <int*> [#uses=1]
+ 	%tmp.22-idxcast = cast int %idx.1 to long		; <long> [#uses=1]
+ 	%tmp.28 = getelementptr int* %data.1, long %tmp.22-idxcast		; <int*> [#uses=1]
+ 	br label %no_exit
+ 
+ no_exit:		; preds = %no_exit.preheader, %endif
+ 	%k.1 = phi int [ %k.0, %endif ], [ 0, %no_exit.preheader ]		; <int> [#uses=3]
+ 	%i.0 = phi int [ %inc.1, %endif ], [ 0, %no_exit.preheader ]		; <int> [#uses=1]
+ 	%tmp.12 = load int* %tmp.11		; <int> [#uses=1]
+ 	%tmp.14 = sub int 0, %tmp.12		; <int> [#uses=1]
+ 	%tmp.161 = setne int %k.1, %tmp.14		; <bool> [#uses=1]
+ 	br bool %tmp.161, label %then, label %else
+ 
+ then:		; preds = %no_exit
+ 	%inc.0 = add int %k.1, 1		; <int> [#uses=1]
+ 	br label %endif
+ 
+ else:		; preds = %no_exit
+ 	%dec = add int %k.1, -1		; <int> [#uses=1]
+ 	br label %endif
+ 
+ endif:		; preds = %else, %then
+ 	%k.0 = phi int [ %dec, %else ], [ %inc.0, %then ]		; <int> [#uses=1]
+ 	store int 2, int* %tmp.28
+ 	%inc.1 = add int %i.0, 1		; <int> [#uses=2]
+ 	%tmp.4 = setlt int %inc.1, %tmp.1		; <bool> [#uses=1]
+ 	br bool %tmp.4, label %no_exit, label %return
+ 
+ return:		; preds = %entry, %endif
+ 	ret void
+ }




From lattner at cs.uiuc.edu  Tue Jun 24 16:25:02 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Jun 24 16:25:02 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/ADCE/2002-05-28-Crash.ll
Message-ID: <200306242124.QAA19450@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/ADCE:

2002-05-28-Crash.ll updated: 1.2 -> 1.3

---
Log message:

modernize testcase


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/ADCE/2002-05-28-Crash.ll
diff -u llvm/test/Regression/Transforms/ADCE/2002-05-28-Crash.ll:1.2 llvm/test/Regression/Transforms/ADCE/2002-05-28-Crash.ll:1.3
--- llvm/test/Regression/Transforms/ADCE/2002-05-28-Crash.ll:1.2	Tue May 28 16:29:28 2002
+++ llvm/test/Regression/Transforms/ADCE/2002-05-28-Crash.ll	Tue Jun 24 16:23:50 2003
@@ -26,7 +26,8 @@
 	%reg129 = add int %cast109, -1		; <int> [#uses=1]
 	%reg114-idxcast = cast ulong %reg114 to uint		; <uint> [#uses=1]
 	%reg114-idxcast-offset = add uint %reg114-idxcast, 1073741823		; <uint> [#uses=1]
-	%reg124 = getelementptr uint* %set, uint %reg114-idxcast-offset		; <uint*> [#uses=1]
+	%reg114-idxcast-offset = cast uint %reg114-idxcast-offset to long
+	%reg124 = getelementptr uint* %set, long %reg114-idxcast-offset		; <uint*> [#uses=1]
 	%reg125 = load uint* %reg124		; <uint> [#uses=1]
 	%cond232 = setne uint %reg125, 0		; <bool> [#uses=1]
 	br bool %cond232, label %bb3, label %bb2
@@ -38,7 +39,8 @@
 	%add1-indvar = add int %cann-indvar, 1		; <int> [#uses=1]
 	%reg130-idxcast = cast int %reg130 to uint		; <uint> [#uses=1]
 	%reg130-idxcast-offset = add uint %reg130-idxcast, 1073741823		; <uint> [#uses=1]
-	%reg118 = getelementptr uint* %set, uint %reg130-idxcast-offset		; <uint*> [#uses=1]
+	%reg130-idxcast-offset = cast uint %reg130-idxcast-offset to long
+	%reg118 = getelementptr uint* %set, long %reg130-idxcast-offset		; <uint*> [#uses=1]
 	%reg119 = load uint* %reg118		; <uint> [#uses=1]
 	%cond233 = seteq uint %reg119, 0		; <bool> [#uses=1]
 	br bool %cond233, label %bb2, label %bb3




From lattner at cs.uiuc.edu  Tue Jun 24 16:50:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Jun 24 16:50:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Scalar/ADCE.cpp
Message-ID: <200306242149.QAA26055@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Scalar:

ADCE.cpp updated: 1.54 -> 1.55

---
Log message:

Do not mark ALL terminators live if any instruciton in the block is live.  We only
want to mark it live if it is an unconditional branch.  This fixes bug:
ADCE/2002-05-28-Crash.ll and makes this pass _much_ more useful.



---
Diffs of the changes:

Index: llvm/lib/Transforms/Scalar/ADCE.cpp
diff -u llvm/lib/Transforms/Scalar/ADCE.cpp:1.54 llvm/lib/Transforms/Scalar/ADCE.cpp:1.55
--- llvm/lib/Transforms/Scalar/ADCE.cpp:1.54	Mon Jun 16 07:10:45 2003
+++ llvm/lib/Transforms/Scalar/ADCE.cpp	Tue Jun 24 16:49:45 2003
@@ -105,8 +105,11 @@
              bind_obj(this, &ADCE::markTerminatorLive));
   }
   
-  // If this basic block is live, then the terminator must be as well!
-  markTerminatorLive(BB);
+  // If this basic block is live, and it ends in an unconditional branch, then
+  // the branch is alive as well...
+  if (BranchInst *BI = dyn_cast<BranchInst>(BB->getTerminator()))
+    if (BI->isUnconditional())
+      markTerminatorLive(BB);
 }
 
 // dropReferencesOfDeadInstructionsInLiveBlock - Loop over all of the
@@ -248,6 +251,7 @@
       NewEntry->getInstList().push_back(new BranchInst(&Func->front()));
       Func->getBasicBlockList().push_front(NewEntry);
       AliveBlocks.insert(NewEntry);    // This block is always alive!
+      LiveSet.insert(NewEntry->getTerminator());  // The branch is live
     }
     
     // Loop over all of the alive blocks in the function.  If any successor
@@ -260,6 +264,23 @@
         BasicBlock *BB = I;
         TerminatorInst *TI = BB->getTerminator();
       
+        // If the terminator instruction is alive, but the block it is contained
+        // in IS alive, this means that this terminator is a conditional branch
+        // on a condition that doesn't matter.  Make it an unconditional branch
+        // to ONE of the successors.  This has the side effect of dropping a use
+        // of the conditional value, which may also be dead.
+        if (!LiveSet.count(TI)) {
+          assert(TI->getNumSuccessors() > 1 && "Not a conditional?");
+          BranchInst *NB = new BranchInst(TI->getSuccessor(0), TI);
+
+          // Remove entries from PHI nodes to avoid confusing ourself later...
+          for (unsigned i = 1, e = TI->getNumSuccessors(); i != e; ++i)
+              TI->getSuccessor(i)->removePredecessor(BB);
+
+          BB->getInstList().erase(TI);
+          TI = NB;
+        }
+
         // Loop over all of the successors, looking for ones that are not alive.
         // We cannot save the number of successors in the terminator instruction
         // here because we may remove them if we don't have a postdominator...




From lattner at cs.uiuc.edu  Tue Jun 24 16:52:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Jun 24 16:52:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/ADCE/2002-05-23-ZeroArgPHITest.ll
Message-ID: <200306242151.QAA26438@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/ADCE:

2002-05-23-ZeroArgPHITest.ll updated: 1.2 -> 1.3

---
Log message:

Modernize testcase


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/ADCE/2002-05-23-ZeroArgPHITest.ll
diff -u llvm/test/Regression/Transforms/ADCE/2002-05-23-ZeroArgPHITest.ll:1.2 llvm/test/Regression/Transforms/ADCE/2002-05-23-ZeroArgPHITest.ll:1.3
--- llvm/test/Regression/Transforms/ADCE/2002-05-23-ZeroArgPHITest.ll:1.2	Thu Aug 22 11:39:04 2002
+++ llvm/test/Regression/Transforms/ADCE/2002-05-23-ZeroArgPHITest.ll	Tue Jun 24 16:51:03 2003
@@ -23,7 +23,7 @@
 
 bb2:					;[#uses=3]
 	%reg109 = phi %node_t* [ %reg110, %bb2 ], [ %reg107, %bb1 ]		; <%node_t*> [#uses=1]
-	%reg212 = getelementptr %node_t* %reg109, uint 0, ubyte 1		; <%node_t**> [#uses=1]
+	%reg212 = getelementptr %node_t* %reg109, long 0, ubyte 1		; <%node_t**> [#uses=1]
 	%reg110 = load %node_t** %reg212		; <%node_t*> [#uses=2]
 	%cond213 = setne %node_t* %reg110, null		; <bool> [#uses=1]
 	br bool %cond213, label %bb2, label %bb3




From criswell at choi.cs.uiuc.edu  Tue Jun 24 17:11:03 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Tue Jun 24 17:11:03 2003
Subject: [llvm-commits] CVS: llvm/configure configure.ac
Message-ID: <200306242210.h5OMAZg11734@choi.cs.uiuc.edu>


Changes in directory llvm:

configure updated: 1.1.2.2 -> 1.1.2.3
configure.ac updated: 1.1.2.2 -> 1.1.2.3

---
Log message:

Added options for purify.


---
Diffs of the changes:

Index: llvm/configure
diff -u llvm/configure:1.1.2.2 llvm/configure:1.1.2.3
--- llvm/configure:1.1.2.2	Tue Jun 24 16:19:14 2003
+++ llvm/configure	Tue Jun 24 17:10:24 2003
@@ -1022,6 +1022,7 @@
   --with-papi             Location of PAPI
   --with-objroot          Location where object files should be placed
                           (default is .)
+  --with-purify           Location of purify program
 
 Some influential environment variables:
   CXX         C++ compiler command
@@ -4296,7 +4297,7 @@
 case $host in
 *-*-irix6*)
   # Find out which ABI we are using.
-  echo '#line 4299 "configure"' > conftest.$ac_ext
+  echo '#line 4300 "configure"' > conftest.$ac_ext
   if { (eval echo "$as_me:$LINENO: \"$ac_compile\"") >&5
   (eval $ac_compile) 2>&5
   ac_status=$?
@@ -4846,7 +4847,7 @@
 save_CFLAGS="$CFLAGS"
 CFLAGS="$CFLAGS -o out/conftest2.$ac_objext"
 compiler_c_o=no
-if { (eval echo configure:4849: \"$ac_compile\") 1>&5; (eval $ac_compile) 2>out/conftest.err; } && test -s out/conftest2.$ac_objext; then
+if { (eval echo configure:4850: \"$ac_compile\") 1>&5; (eval $ac_compile) 2>out/conftest.err; } && test -s out/conftest2.$ac_objext; then
   # The compiler can only warn and ignore the option if not recognized
   # So say no if there are warnings
   if test -s out/conftest.err; then
@@ -6677,7 +6678,7 @@
     lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2
   lt_status=$lt_dlunknown
   cat > conftest.$ac_ext <<EOF
-#line 6680 "configure"
+#line 6681 "configure"
 #include "confdefs.h"
 
 #if HAVE_DLFCN_H
@@ -6775,7 +6776,7 @@
     lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2
   lt_status=$lt_dlunknown
   cat > conftest.$ac_ext <<EOF
-#line 6778 "configure"
+#line 6779 "configure"
 #include "confdefs.h"
 
 #if HAVE_DLFCN_H
@@ -11552,6 +11553,13 @@
 
 else
   OBJROOT=.
+
+fi;
+
+# Check whether --with-purify or --without-purify was given.
+if test "${with_purify+set}" = set; then
+  withval="$with_purify"
+  PURIFY=$withval
 
 fi;
 


Index: llvm/configure.ac
diff -u llvm/configure.ac:1.1.2.2 llvm/configure.ac:1.1.2.3
--- llvm/configure.ac:1.1.2.2	Tue Jun 24 16:19:14 2003
+++ llvm/configure.ac	Tue Jun 24 17:10:24 2003
@@ -231,5 +231,6 @@
 AC_ARG_WITH(bcrepos,AC_HELP_STRING([--with-bcrepos],[Location of Bytecode Repository]),AC_SUBST(BCR,[$withval]),AC_SUBST(BCR,[/home/vadve/lattner/LLVMPrograms]))
 AC_ARG_WITH(papi,AC_HELP_STRING([--with-papi],[Location of PAPI]),AC_SUBST(PAPIDIR,[$withval]),AC_SUBST(PAPIDIR,[/home/vadve/shared/papi-2.3.4.1]))
 AC_ARG_WITH(objroot,AC_HELP_STRING([--with-objroot],[Location where object files should be placed (default is .)]),AC_SUBST(OBJROOT,[$withval]),AC_SUBST(OBJROOT,[.]))
+AC_ARG_WITH(purify,AC_HELP_STRING([--with-purify],[Location of purify program]),AC_SUBST(PURIFY,[$withval]))
 
 AC_OUTPUT(Makefile.config)




From lattner at cs.uiuc.edu  Tue Jun 24 17:21:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Jun 24 17:21:01 2003
Subject: [llvm-commits] CVS: llvm/lib/VMCore/Value.cpp
Message-ID: <200306242220.RAA28586@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/VMCore:

Value.cpp updated: 1.31 -> 1.32

---
Log message:

Make assertion more descriptive


---
Diffs of the changes:

Index: llvm/lib/VMCore/Value.cpp
diff -u llvm/lib/VMCore/Value.cpp:1.31 llvm/lib/VMCore/Value.cpp:1.32
--- llvm/lib/VMCore/Value.cpp:1.31	Thu Feb 13 13:46:22 2003
+++ llvm/lib/VMCore/Value.cpp	Tue Jun 24 17:20:19 2003
@@ -40,7 +40,7 @@
                 << **I << "\n";
   }
 #endif
-  assert(Uses.begin() == Uses.end());
+  assert(Uses.begin() == Uses.end() &&"Uses remain when a value is destroyed!");
 
   // There should be no uses of this object anymore, remove it.
   LeakDetector::removeGarbageObject(this);




From lattner at cs.uiuc.edu  Tue Jun 24 17:59:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Jun 24 17:59:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/ADCE/2003-06-24-BadSuccessor.ll
Message-ID: <200306242258.RAA30578@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/ADCE:

2003-06-24-BadSuccessor.ll added (r1.1)

---
Log message:

New testcase


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/ADCE/2003-06-24-BadSuccessor.ll
diff -c /dev/null llvm/test/Regression/Transforms/ADCE/2003-06-24-BadSuccessor.ll:1.1
*** /dev/null	Tue Jun 24 17:58:16 2003
--- llvm/test/Regression/Transforms/ADCE/2003-06-24-BadSuccessor.ll	Tue Jun 24 17:58:05 2003
***************
*** 0 ****
--- 1,94 ----
+ ; RUN: as < %s | opt -adce -disable-output
+ 
+ target endian = little
+ target pointersize = 32
+ 	%struct..CppObjTypeDesc = type { uint, ushort, ushort }
+ 	%struct..TypeToken = type { uint, ushort, ushort }
+ 
+ implementation   ; Functions:
+ 
+ uint %C_ReFaxToDb() {
+ entry:		; No predecessors!
+ 	br bool false, label %endif.0, label %then.0
+ 
+ then.0:		; preds = %entry
+ 	ret uint 0
+ 
+ endif.0:		; preds = %entry
+ 	br bool false, label %then.11, label %then.4
+ 
+ then.4:		; preds = %endif.0
+ 	ret uint 0
+ 
+ then.11:		; preds = %endif.0
+ 	br bool false, label %loopentry.0, label %else.2
+ 
+ loopentry.0:		; preds = %then.11, %endif.14, %loopentry.1
+ 	br bool false, label %endif.14, label %loopexit.0
+ 
+ endif.14:		; preds = %loopentry.0
+ 	br bool false, label %loopentry.1, label %loopentry.0
+ 
+ loopentry.1:		; preds = %endif.14, %then.53, %then.53, %then.53, %then.53, %then.53
+ 	%SubArrays.10 = phi uint* [ %SubArrays.8, %then.53 ] , [ null, %endif.14 ]		; <uint*> [#uses=3]
+ 	br bool false, label %no_exit.1, label %loopentry.0
+ 
+ no_exit.1:		; preds = %loopentry.1
+ 	switch uint 0, label %label.17 [
+ 		 uint 2, label %label.11
+ 		 uint 19, label %label.10
+ 	]
+ 
+ label.10:		; preds = %no_exit.1
+ 	br bool false, label %then.43, label %endif.43
+ 
+ then.43:		; preds = %label.10
+ 	br bool false, label %then.44, label %endif.44
+ 
+ then.44:		; preds = %then.43
+ 	br bool false, label %shortcirc_next.4, label %endif.45
+ 
+ shortcirc_next.4:		; preds = %then.44
+ 	br bool false, label %no_exit.2, label %loopexit.2
+ 
+ no_exit.2:		; preds = %shortcirc_next.4
+ 	%tmp.897 = getelementptr uint* %SubArrays.10, long 0		; <uint*> [#uses=1]
+ 	%tmp.899 = load uint* %tmp.897		; <uint> [#uses=1]
+ 	store uint %tmp.899, uint* null
+ 	ret uint 0
+ 
+ loopexit.2:		; preds = %shortcirc_next.4
+ 	ret uint 0
+ 
+ endif.45:		; preds = %then.44
+ 	ret uint 0
+ 
+ endif.44:		; preds = %then.43
+ 	ret uint 0
+ 
+ endif.43:		; preds = %label.10
+ 	ret uint 0
+ 
+ label.11:		; preds = %no_exit.1
+ 	ret uint 0
+ 
+ label.17:		; preds = %no_exit.1, %no_exit.1, %no_exit.1, %no_exit.1, %no_exit.1, %no_exit.1
+ 	br bool false, label %then.53, label %shortcirc_next.7
+ 
+ shortcirc_next.7:		; preds = %label.17
+ 	br bool false, label %then.53, label %shortcirc_next.8
+ 
+ shortcirc_next.8:		; preds = %shortcirc_next.7
+ 	ret uint 0
+ 
+ then.53:		; preds = %shortcirc_next.7, %label.17
+ 	%SubArrays.8 = phi uint* [ %SubArrays.10, %shortcirc_next.7 ], [ %SubArrays.10, %label.17 ]		; <uint*> [#uses=5]
+ 	%tmp.1023 = load uint* null		; <uint> [#uses=1]
+ 	switch uint %tmp.1023, label %loopentry.1 []
+ 
+ loopexit.0:		; preds = %loopentry.0
+ 	ret uint 0
+ 
+ else.2:		; preds = %then.11
+ 	ret uint 0
+ }




From lattner at cs.uiuc.edu  Tue Jun 24 18:03:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Jun 24 18:03:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Scalar/ADCE.cpp
Message-ID: <200306242302.SAA31033@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Scalar:

ADCE.cpp updated: 1.55 -> 1.56

---
Log message:

Fix bug: ADCE/2003-06-24-BadSuccessor.ll



---
Diffs of the changes:

Index: llvm/lib/Transforms/Scalar/ADCE.cpp
diff -u llvm/lib/Transforms/Scalar/ADCE.cpp:1.55 llvm/lib/Transforms/Scalar/ADCE.cpp:1.56
--- llvm/lib/Transforms/Scalar/ADCE.cpp:1.55	Tue Jun 24 16:49:45 2003
+++ llvm/lib/Transforms/Scalar/ADCE.cpp	Tue Jun 24 18:02:45 2003
@@ -73,6 +73,8 @@
   // instructions that are dead according to LiveSet.
   bool dropReferencesOfDeadInstructionsInLiveBlock(BasicBlock *BB);
 
+  TerminatorInst *convertToUnconditionalBranch(TerminatorInst *TI);
+
   inline void markInstructionLive(Instruction *I) {
     if (LiveSet.count(I)) return;
     DEBUG(std::cerr << "Insn Live: " << I);
@@ -139,6 +141,24 @@
 }
 
 
+/// convertToUnconditionalBranch - Transform this conditional terminator
+/// instruction into an unconditional branch because we don't care which of the
+/// successors it goes to.  This eliminate a use of the condition as well.
+///
+TerminatorInst *ADCE::convertToUnconditionalBranch(TerminatorInst *TI) {
+  BranchInst *NB = new BranchInst(TI->getSuccessor(0), TI);
+  BasicBlock *BB = TI->getParent();
+
+  // Remove entries from PHI nodes to avoid confusing ourself later...
+  for (unsigned i = 1, e = TI->getNumSuccessors(); i != e; ++i)
+    TI->getSuccessor(i)->removePredecessor(BB);
+  
+  // Delete the old branch itself...
+  BB->getInstList().erase(TI);
+  return NB;
+}
+
+
 // doADCE() - Run the Aggressive Dead Code Elimination algorithm, returning
 // true if the function was modified.
 //
@@ -234,13 +254,25 @@
   // new entry node...
   //
   if (AliveBlocks.size() == Func->size()) {  // No dead blocks?
-    for (Function::iterator I = Func->begin(), E = Func->end(); I != E; ++I)
+    for (Function::iterator I = Func->begin(), E = Func->end(); I != E; ++I) {
       // Loop over all of the instructions in the function, telling dead
       // instructions to drop their references.  This is so that the next sweep
       // over the program can safely delete dead instructions without other dead
       // instructions still refering to them.
       //
       dropReferencesOfDeadInstructionsInLiveBlock(I);
+
+      // Check to make sure the terminator instruction is live.  If it isn't,
+      // this means that the condition that it branches on (we know it is not an
+      // unconditional branch), is not needed to make the decision of where to
+      // go to, because all outgoing edges go to the same place.  We must remove
+      // the use of the condition (because it's probably dead), so we convert
+      // the terminator to a conditional branch.
+      //
+      TerminatorInst *TI = I->getTerminator();
+      if (!LiveSet.count(TI))
+        convertToUnconditionalBranch(TI);
+    }
     
   } else {                                   // If there are some blocks dead...
     // If the entry node is dead, insert a new entry node to eliminate the entry
@@ -269,17 +301,8 @@
         // on a condition that doesn't matter.  Make it an unconditional branch
         // to ONE of the successors.  This has the side effect of dropping a use
         // of the conditional value, which may also be dead.
-        if (!LiveSet.count(TI)) {
-          assert(TI->getNumSuccessors() > 1 && "Not a conditional?");
-          BranchInst *NB = new BranchInst(TI->getSuccessor(0), TI);
-
-          // Remove entries from PHI nodes to avoid confusing ourself later...
-          for (unsigned i = 1, e = TI->getNumSuccessors(); i != e; ++i)
-              TI->getSuccessor(i)->removePredecessor(BB);
-
-          BB->getInstList().erase(TI);
-          TI = NB;
-        }
+        if (!LiveSet.count(TI))
+          TI = convertToUnconditionalBranch(TI);
 
         // Loop over all of the successors, looking for ones that are not alive.
         // We cannot save the number of successors in the terminator instruction




From lattner at cs.uiuc.edu  Tue Jun 24 18:39:12 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Jun 24 18:39:12 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/BugPoint/Makefile
Message-ID: <200306242338.SAA04465@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/BugPoint:

Makefile updated: 1.2 -> 1.3

---
Log message:

Don't print message, remove temp files


---
Diffs of the changes:

Index: llvm/test/Regression/BugPoint/Makefile
diff -u llvm/test/Regression/BugPoint/Makefile:1.2 llvm/test/Regression/BugPoint/Makefile:1.3
--- llvm/test/Regression/BugPoint/Makefile:1.2	Mon Dec 23 17:47:39 2002
+++ llvm/test/Regression/BugPoint/Makefile	Tue Jun 24 18:38:09 2003
@@ -12,8 +12,8 @@
 
 Output/%.ll.out: %.ll Output/.dir $(LAS) $(LBUGPOINT)
 	-$(TESTRUNR) $<
-	-# Remove temporary files generated...
-	@rm -f bugpoint-$<-*
+	-@# Remove temporary files generated...
+	- at rm -f bugpoint-*
 
 clean::
 	rm -f *.bc




From lattner at cs.uiuc.edu  Tue Jun 24 18:41:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Jun 24 18:41:00 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/BugPoint/remove_arguments_test.ll
Message-ID: <200306242340.SAA04586@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/BugPoint:

remove_arguments_test.ll added (r1.1)

---
Log message:

New testcase


---
Diffs of the changes:

Index: llvm/test/Regression/BugPoint/remove_arguments_test.ll
diff -c /dev/null llvm/test/Regression/BugPoint/remove_arguments_test.ll:1.1
*** /dev/null	Tue Jun 24 18:40:37 2003
--- llvm/test/Regression/BugPoint/remove_arguments_test.ll	Tue Jun 24 18:40:27 2003
***************
*** 0 ****
--- 1,9 ----
+ ; RUN: bugpoint %s  -bugpoint-crashcalls
+ 
+ ; Test to make sure that arguments are removed from the function if they are unnecessary.
+ 
+ declare int %test2()
+ int %test(int %A, int %B, float %C) {
+ 	call int %test2()
+ 	ret int %0
+ }




From gaeke at cs.uiuc.edu  Tue Jun 24 22:06:00 2003
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Tue Jun 24 22:06:00 2003
Subject: [llvm-commits] CVS: llvm/lib/CWriter/Writer.cpp
Message-ID: <200306250305.WAA02888@neo.cs.uiuc.edu>


Changes in directory llvm/lib/CWriter:

Writer.cpp updated: 1.98 -> 1.99

---
Log message:

It seems likely that floats would need a cast too, because they are
ordinarily promoted to doubles.


---
Diffs of the changes:

Index: llvm/lib/CWriter/Writer.cpp
diff -u llvm/lib/CWriter/Writer.cpp:1.98 llvm/lib/CWriter/Writer.cpp:1.99
--- llvm/lib/CWriter/Writer.cpp:1.98	Mon Jun 23 15:00:51 2003
+++ llvm/lib/CWriter/Writer.cpp	Tue Jun 24 22:05:33 2003
@@ -1033,7 +1033,8 @@
   // We must cast the results of binary operations which might be promoted.
   bool needsCast = false;
   if ((I.getType() == Type::UByteTy) || (I.getType() == Type::SByteTy)
-      || (I.getType() == Type::UShortTy) || (I.getType() == Type::ShortTy)) {
+      || (I.getType() == Type::UShortTy) || (I.getType() == Type::ShortTy)
+      || (I.getType() == Type::FloatTy)) {
     needsCast = true;
     Out << "((";
     printType(Out, I.getType(), "", false, false);




From lattner at cs.uiuc.edu  Tue Jun 24 23:13:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Jun 24 23:13:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/IPO/DeadArgumentElimination.cpp
Message-ID: <200306250412.XAA05108@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/IPO:

DeadArgumentElimination.cpp updated: 1.2 -> 1.3

---
Log message:

Add argument to DAE to allow operation on non-internal functions


---
Diffs of the changes:

Index: llvm/lib/Transforms/IPO/DeadArgumentElimination.cpp
diff -u llvm/lib/Transforms/IPO/DeadArgumentElimination.cpp:1.2 llvm/lib/Transforms/IPO/DeadArgumentElimination.cpp:1.3
--- llvm/lib/Transforms/IPO/DeadArgumentElimination.cpp:1.2	Wed Jun 18 11:25:51 2003
+++ llvm/lib/Transforms/IPO/DeadArgumentElimination.cpp	Tue Jun 24 23:12:49 2003
@@ -25,15 +25,27 @@
   Statistic<> NumArgumentsEliminated("deadargelim", "Number of args removed");
 
   struct DAE : public Pass {
+    DAE(bool DFEF = false) : DeleteFromExternalFunctions(DFEF) {}
     bool run(Module &M);
+
+  private:
+    bool DeleteFromExternalFunctions;
+    bool FunctionArgumentsIntrinsicallyAlive(const Function &F);
+    void RemoveDeadArgumentsFromFunction(Function *F,
+                                         std::set<Argument*> &DeadArguments);
   };
   RegisterOpt<DAE> X("deadargelim", "Dead Argument Elimination");
 }
 
-// createDeadArgEliminationPass - This pass removes arguments from functions
-// which are not used by the body of the function.
-//
-Pass *createDeadArgEliminationPass() { return new DAE(); }
+/// createDeadArgEliminationPass - This pass removes arguments from functions
+/// which are not used by the body of the function.  If
+/// DeleteFromExternalFunctions is true, the pass will modify functions that
+/// have external linkage, which is not usually safe (this is used by bugpoint
+/// to reduce testcases).
+///
+Pass *createDeadArgEliminationPass(bool DeleteFromExternalFunctions) {
+  return new DAE(DeleteFromExternalFunctions);
+}
 
 
 // FunctionArgumentsIntrinsicallyAlive - Return true if the arguments of the
@@ -42,8 +54,8 @@
 // We consider arguments of non-internal functions to be intrinsically alive as
 // well as arguments to functions which have their "address taken".
 //
-static bool FunctionArgumentsIntrinsicallyAlive(const Function &F) {
-  if (!F.hasInternalLinkage()) return true;
+bool DAE::FunctionArgumentsIntrinsicallyAlive(const Function &F) {
+  if (!F.hasInternalLinkage() && !DeleteFromExternalFunctions) return true;
 
   for (Value::use_const_iterator I = F.use_begin(), E = F.use_end(); I!=E; ++I){
     // If this use is anything other than a call site, the function is alive.
@@ -151,8 +163,8 @@
 // specified by the DeadArguments list.  Transform the function and all of the
 // callees of the function to not have these arguments.
 //
-static void RemoveDeadArgumentsFromFunction(Function *F,
-                                            std::set<Argument*> &DeadArguments){
+void DAE::RemoveDeadArgumentsFromFunction(Function *F,
+                                          std::set<Argument*> &DeadArguments){
   // Start by computing a new prototype for the function, which is the same as
   // the old function, but has fewer arguments.
   const FunctionType *FTy = F->getFunctionType();
@@ -166,7 +178,7 @@
                                          FTy->isVarArg());
   
   // Create the new function body and insert it into the module...
-  Function *NF = new Function(NFTy, Function::InternalLinkage, F->getName());
+  Function *NF = new Function(NFTy, F->getLinkage(), F->getName());
   F->getParent()->getFunctionList().insert(F, NF);
 
   // Loop over all of the callers of the function, transforming the call sites




From lattner at cs.uiuc.edu  Tue Jun 24 23:13:06 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Jun 24 23:13:06 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Transforms/IPO.h
Message-ID: <200306250412.XAA05096@apoc.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Transforms:

IPO.h updated: 1.17 -> 1.18

---
Log message:

Add argument to DAE to allow operation on non-internal functions


---
Diffs of the changes:

Index: llvm/include/llvm/Transforms/IPO.h
diff -u llvm/include/llvm/Transforms/IPO.h:1.17 llvm/include/llvm/Transforms/IPO.h:1.18
--- llvm/include/llvm/Transforms/IPO.h:1.17	Mon Jun 16 07:16:52 2003
+++ llvm/include/llvm/Transforms/IPO.h	Tue Jun 24 23:12:29 2003
@@ -70,9 +70,12 @@
 
 //===----------------------------------------------------------------------===//
 // createDeadArgEliminationPass - This pass removes arguments from functions
-// which are not used by the body of the function.
+// which are not used by the body of the function.  If
+// DeleteFromExternalFunctions is true, the pass will modify functions that have
+// external linkage, which is not usually safe (this is used by bugpoint to
+// reduce testcases).
 //
-Pass *createDeadArgEliminationPass();
+Pass *createDeadArgEliminationPass(bool DeleteFromExternalFunctions=false);
 
 
 //===----------------------------------------------------------------------===//




From lattner at cs.uiuc.edu  Tue Jun 24 23:15:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Jun 24 23:15:01 2003
Subject: [llvm-commits] CVS: llvm/tools/bugpoint/ExtractFunction.cpp
Message-ID: <200306250414.XAA05134@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/bugpoint:

ExtractFunction.cpp updated: 1.9 -> 1.10

---
Log message:

Run dead arg elimination, and tell it that it's ok to hack up non-internal functions


---
Diffs of the changes:

Index: llvm/tools/bugpoint/ExtractFunction.cpp
diff -u llvm/tools/bugpoint/ExtractFunction.cpp:1.9 llvm/tools/bugpoint/ExtractFunction.cpp:1.10
--- llvm/tools/bugpoint/ExtractFunction.cpp:1.9	Wed May 21 15:38:59 2003
+++ llvm/tools/bugpoint/ExtractFunction.cpp	Tue Jun 24 23:13:36 2003
@@ -97,6 +97,7 @@
   CleanupPasses.add(createFunctionResolvingPass());
   CleanupPasses.add(createGlobalDCEPass());
   CleanupPasses.add(createDeadTypeEliminationPass());
+  CleanupPasses.add(createDeadArgEliminationPass(true));
   CleanupPasses.add(createVerifierPass());
   CleanupPasses.run(*M);
   return M;




From lattner at cs.uiuc.edu  Tue Jun 24 23:15:05 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Jun 24 23:15:05 2003
Subject: [llvm-commits] CVS: llvm/tools/bugpoint/CrashDebugger.cpp
Message-ID: <200306250414.XAA05129@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/bugpoint:

CrashDebugger.cpp updated: 1.10 -> 1.11

---
Log message:

Try to run cleanups even if nothing was modified in the preview passes


---
Diffs of the changes:

Index: llvm/tools/bugpoint/CrashDebugger.cpp
diff -u llvm/tools/bugpoint/CrashDebugger.cpp:1.10 llvm/tools/bugpoint/CrashDebugger.cpp:1.11
--- llvm/tools/bugpoint/CrashDebugger.cpp:1.10	Sun Jun  1 23:54:29 2003
+++ llvm/tools/bugpoint/CrashDebugger.cpp	Tue Jun 24 23:13:52 2003
@@ -360,20 +360,18 @@
   } while (Simplification);
 
   // Try to clean up the testcase by running funcresolve and globaldce...
-  if (AnyReduction) {
-    std::cout << "\n*** Attempting to perform final cleanups: ";
-    Module *M = performFinalCleanups();
-    std::swap(Program, M);
+  std::cout << "\n*** Attempting to perform final cleanups: ";
+  Module *M = performFinalCleanups();
+  std::swap(Program, M);
             
-    // Find out if the pass still crashes on the cleaned up program...
-    if (runPasses(PassesToRun)) {
-      // Yup, it does, keep the reduced version...
-      delete M;
-      AnyReduction = true;
-    } else {
-      delete Program;   // Otherwise, restore the original module...
-      Program = M;
-    }
+  // Find out if the pass still crashes on the cleaned up program...
+  if (runPasses(PassesToRun)) {
+    // Yup, it does, keep the reduced version...
+    delete M;
+    AnyReduction = true;
+  } else {
+    delete Program;   // Otherwise, restore the original module...
+    Program = M;
   }
 
   if (AnyReduction)




From jstanley at cs.uiuc.edu  Tue Jun 24 23:53:01 2003
From: jstanley at cs.uiuc.edu (Joel Stanley)
Date: Tue Jun 24 23:53:01 2003
Subject: [llvm-commits] CVS: llvm/lib/CWriter/Writer.cpp
Message-ID: <200306250452.XAA13272@trinity.cs.uiuc.edu>


Changes in directory llvm/lib/CWriter:

Writer.cpp updated: 1.99 -> 1.100

---
Log message:

- Fixed name mangling conditions to handle 'linkonce' linkage type. In
particular, name mangling for GlobalValues only occurs when the linkage type is
internal or when the name must be mangled to avoid a collision.  See comments in
CWriter::getValueName for more information.

- 'inline' keyword is now emitted for functions with 'linkonce' linkage type.

- Fixed typos.


---
Diffs of the changes:

Index: llvm/lib/CWriter/Writer.cpp
diff -u llvm/lib/CWriter/Writer.cpp:1.99 llvm/lib/CWriter/Writer.cpp:1.100
--- llvm/lib/CWriter/Writer.cpp:1.99	Tue Jun 24 22:05:33 2003
+++ llvm/lib/CWriter/Writer.cpp	Tue Jun 24 23:52:09 2003
@@ -164,14 +164,25 @@
 }
 
 std::string CWriter::getValueName(const Value *V) {
-  if (V->hasName()) {              // Print out the label if it exists...
-    if (isa<GlobalValue>(V) &&     // Do not mangle globals...
-        (cast<GlobalValue>(V)->hasExternalLinkage() &&// Unless it's internal or
-         !MangledGlobals.count(V))) // Unless the name would collide if we don't
-      return makeNameProper(V->getName());
-
+  if (V->hasName()) { // Print out the label if it exists...
+    
+    // Name mangling occurs as follows:
+    // - If V is not a global, mangling always occurs.
+    // - Otherwise, mangling occurs when any of the following are true:
+    //   1) V has internal linkage
+    //   2) V's name would collide if it is not mangled.
+    //
+    
+    if(const GlobalValue* gv = dyn_cast<GlobalValue>(V)) {
+      if(!gv->hasInternalLinkage() && !MangledGlobals.count(gv)) {
+        // No internal linkage, name will not collide -> no mangling.
+        return makeNameProper(gv->getName());
+      }
+    }
+    
+    // Non-global, or global with internal linkage / colliding name -> mangle.
     return "l" + utostr(V->getType()->getUniqueID()) + "_" +
-           makeNameProper(V->getName());      
+      makeNameProper(V->getName());      
   }
 
   int Slot = Table->getValSlot(V);
@@ -220,23 +231,23 @@
   switch (Ty->getPrimitiveID()) {
   case Type::FunctionTyID: {
     const FunctionType *MTy = cast<FunctionType>(Ty);
-    std::stringstream FunctionInards; 
-    FunctionInards << " (" << NameSoFar << ") (";
+    std::stringstream FunctionInnards; 
+    FunctionInnards << " (" << NameSoFar << ") (";
     for (FunctionType::ParamTypes::const_iterator
            I = MTy->getParamTypes().begin(),
            E = MTy->getParamTypes().end(); I != E; ++I) {
       if (I != MTy->getParamTypes().begin())
-        FunctionInards << ", ";
-      printType(FunctionInards, *I, "");
+        FunctionInnards << ", ";
+      printType(FunctionInnards, *I, "");
     }
     if (MTy->isVarArg()) {
       if (!MTy->getParamTypes().empty()) 
-    	FunctionInards << ", ...";
+    	FunctionInnards << ", ...";
     } else if (MTy->getParamTypes().empty()) {
-      FunctionInards << "void";
+      FunctionInnards << "void";
     }
-    FunctionInards << ")";
-    std::string tstr = FunctionInards.str();
+    FunctionInnards << ")";
+    std::string tstr = FunctionInnards.str();
     printType(Out, MTy->getReturnType(), tstr);
     return Out;
   }
@@ -650,9 +661,9 @@
     Out << "extern void * malloc(size_t);\n\n";
   }
 
-  // Output the global variable declerations
+  // Output the global variable declarations
   if (!M->gempty()) {
-    Out << "\n\n/* Global Variable Declerations */\n";
+    Out << "\n\n/* Global Variable Declarations */\n";
     for (Module::giterator I = M->gbegin(), E = M->gend(); I != E; ++I)
       if (!I->isExternal()) {
         Out << "extern ";
@@ -768,33 +779,36 @@
 
 
 void CWriter::printFunctionSignature(const Function *F, bool Prototype) {
-  // If the program provides it's own malloc prototype we don't need
+  // If the program provides its own malloc prototype we don't need
   // to include the general one.  
   if (getValueName(F) == "malloc")
     needsMalloc = false;
-  if (F->hasInternalLinkage()) Out << "static ";  
+
+  if (F->hasInternalLinkage()) Out << "static ";
+  if (F->hasLinkOnceLinkage()) Out << "inline ";
+  
   // Loop over the arguments, printing them...
   const FunctionType *FT = cast<FunctionType>(F->getFunctionType());
   
-  std::stringstream FunctionInards; 
+  std::stringstream FunctionInnards; 
     
   // Print out the name...
-  FunctionInards << getValueName(F) << "(";
+  FunctionInnards << getValueName(F) << "(";
     
   if (!F->isExternal()) {
     if (!F->aempty()) {
       std::string ArgName;
       if (F->abegin()->hasName() || !Prototype)
         ArgName = getValueName(F->abegin());
-      printType(FunctionInards, F->afront().getType(), ArgName);
+      printType(FunctionInnards, F->afront().getType(), ArgName);
       for (Function::const_aiterator I = ++F->abegin(), E = F->aend();
            I != E; ++I) {
-        FunctionInards << ", ";
+        FunctionInnards << ", ";
         if (I->hasName() || !Prototype)
           ArgName = getValueName(I);
         else 
           ArgName = "";
-        printType(FunctionInards, I->getType(), ArgName);
+        printType(FunctionInnards, I->getType(), ArgName);
       }
     }
   } else {
@@ -802,8 +816,8 @@
     for (FunctionType::ParamTypes::const_iterator I = 
 	   FT->getParamTypes().begin(),
 	   E = FT->getParamTypes().end(); I != E; ++I) {
-      if (I != FT->getParamTypes().begin()) FunctionInards << ", ";
-      printType(FunctionInards, *I);
+      if (I != FT->getParamTypes().begin()) FunctionInnards << ", ";
+      printType(FunctionInnards, *I);
     }
   }
 
@@ -811,12 +825,12 @@
   // unless there are no known types, in which case, we just emit ().
   //
   if (FT->isVarArg() && !FT->getParamTypes().empty()) {
-    if (FT->getParamTypes().size()) FunctionInards << ", ";
-    FunctionInards << "...";  // Output varargs portion of signature!
+    if (FT->getParamTypes().size()) FunctionInnards << ", ";
+    FunctionInnards << "...";  // Output varargs portion of signature!
   }
-  FunctionInards << ")";
+  FunctionInnards << ")";
   // Print out the return type and the entire signature for that matter
-  printType(Out, F->getReturnType(), FunctionInards.str());
+  printType(Out, F->getReturnType(), FunctionInnards.str());
   
 }
 




From lattner at cs.uiuc.edu  Wed Jun 25 09:32:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Jun 25 09:32:00 2003
Subject: [llvm-commits] CVS: llvm/www/docs/OpenProjects.html
Message-ID: <200306251431.JAA01106@tank.cs.uiuc.edu>


Changes in directory llvm/www/docs:

OpenProjects.html updated: 1.6 -> 1.7

---
Log message:

Remove completed tasks


---
Diffs of the changes:

Index: llvm/www/docs/OpenProjects.html
diff -u llvm/www/docs/OpenProjects.html:1.6 llvm/www/docs/OpenProjects.html:1.7
--- llvm/www/docs/OpenProjects.html:1.6	Tue May  6 15:47:11 2003
+++ llvm/www/docs/OpenProjects.html	Wed Jun 25 09:31:06 2003
@@ -130,7 +130,6 @@
 <li>Transform <tt>setjmp</tt> and <tt>longjmp</tt> calls to use the LLVM
     <tt>invoke</tt> mechanism.
 <li>Add support for a volatile attribute on loads and stores
-<li>Support for variable argument functions
 <li>Add a new conditional move instruction: <tt>X = select bool Cond, Y, Z</tt>
 <li>Add support for platform independent prefetch support.  The GCC <a
     href="http://gcc.gnu.org/projects/prefetch.html">prefetch project</a> page
@@ -192,7 +191,7 @@
 <li>Implement a flow-sensitive context-sensitive alias analysis algorithm<br>
   - Pick one of the somewhat efficient algorithms, but strive for maximum
     precision
-<li>Implement a flow-sensitive context-insensitive alias anlaysis algorithm<br>
+<li>Implement a flow-sensitive context-insensitive alias analysis algorithm<br>
   - Just an efficient local algorithm perhaps?
 
 <li>Implement an interface to update analyses in response to common code motion
@@ -200,15 +199,6 @@
 <li>Implement alias analysis based optimizations:
   <ul>
   <li>Dead store elimination
-  <li>Store+Reload or "store forwarding" elimination:<p>
-      Change:
-       <pre>
-        store int X, int* P
-        Y = load int* P</pre>
-      into:
-       <pre>
-        store int X, int *P
-        Y = X</pre>
   </ul>
 </ol>
 
@@ -290,6 +280,6 @@
 <address><a href="mailto:sabre at nondot.org">Chris Lattner</a></address>
 <!-- Created: Tue Aug  6 15:00:33 CDT 2002 -->
 <!-- hhmts start -->
-Last modified: Tue May  6 15:46:48 CDT 2003
+Last modified: Wed Jun 25 09:30:27 CDT 2003
 <!-- hhmts end -->
 </font></body></html>




From lattner at cs.uiuc.edu  Wed Jun 25 09:56:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Jun 25 09:56:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/Mem2Reg/2003-06-26-IterativePromote.ll
Message-ID: <200306251455.JAA07332@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/Mem2Reg:

2003-06-26-IterativePromote.ll added (r1.1)

---
Log message:

New testcase


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/Mem2Reg/2003-06-26-IterativePromote.ll
diff -c /dev/null llvm/test/Regression/Transforms/Mem2Reg/2003-06-26-IterativePromote.ll:1.1
*** /dev/null	Wed Jun 25 09:55:40 2003
--- llvm/test/Regression/Transforms/Mem2Reg/2003-06-26-IterativePromote.ll	Wed Jun 25 09:55:30 2003
***************
*** 0 ****
--- 1,19 ----
+ ; Promoting some values allows promotion of other values.
+ ; RUN: if as < %s | opt -mem2reg | dis | grep alloca
+ ; RUN: then exit 1
+ ; RUN: else exit 0
+ ; RUN: fi
+ 
+ int %test2() {
+         %result = alloca int             ; ty=int*
+         %a = alloca int          ; ty=int*
+         %p = alloca int*                 ; ty=int**
+         store int 0, int* %a
+         store int* %a, int** %p
+         %tmp.0 = load int** %p           ; ty=int*
+         %tmp.1 = load int* %tmp.0                ; ty=int
+         store int %tmp.1, int* %result
+         %tmp.2 = load int* %result               ; ty=int
+         ret int %tmp.2
+ }
+ 




From lattner at cs.uiuc.edu  Wed Jun 25 10:00:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Jun 25 10:00:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Scalar/Mem2Reg.cpp
Message-ID: <200306251459.JAA07487@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Scalar:

Mem2Reg.cpp updated: 1.3 -> 1.4

---
Log message:

Fix bug: Mem2Reg/2003-06-26-IterativePromote.ll


---
Diffs of the changes:

Index: llvm/lib/Transforms/Scalar/Mem2Reg.cpp
diff -u llvm/lib/Transforms/Scalar/Mem2Reg.cpp:1.3 llvm/lib/Transforms/Scalar/Mem2Reg.cpp:1.4
--- llvm/lib/Transforms/Scalar/Mem2Reg.cpp:1.3	Wed Apr 23 11:37:42 2003
+++ llvm/lib/Transforms/Scalar/Mem2Reg.cpp	Wed Jun 25 09:58:56 2003
@@ -40,19 +40,26 @@
 
   BasicBlock &BB = F.getEntryNode();  // Get the entry node for the function
 
-  // Find allocas that are safe to promote, by looking at all instructions in
-  // the entry node
-  for (BasicBlock::iterator I = BB.begin(), E = --BB.end(); I != E; ++I)
-    if (AllocaInst *AI = dyn_cast<AllocaInst>(I))       // Is it an alloca?
-      if (isAllocaPromotable(AI, TD))
-        Allocas.push_back(AI);
+  bool Changed  = false;
+  
+  while (1) {
+    Allocas.clear();
+
+    // Find allocas that are safe to promote, by looking at all instructions in
+    // the entry node
+    for (BasicBlock::iterator I = BB.begin(), E = --BB.end(); I != E; ++I)
+      if (AllocaInst *AI = dyn_cast<AllocaInst>(I))       // Is it an alloca?
+        if (isAllocaPromotable(AI, TD))
+          Allocas.push_back(AI);
+
+    if (Allocas.empty()) break;
 
-  if (!Allocas.empty()) {
     PromoteMemToReg(Allocas, getAnalysis<DominanceFrontier>(), TD);
     NumPromoted += Allocas.size();
-    return true;
+    Changed = true;
   }
-  return false;
+
+  return Changed;
 }
 
 // createPromoteMemoryToRegister - Provide an entry point to create this pass.




From jstanley at cs.uiuc.edu  Wed Jun 25 10:48:01 2003
From: jstanley at cs.uiuc.edu (Joel Stanley)
Date: Wed Jun 25 10:48:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Reoptimizer/BinInterface/sparcbin.cpp
Message-ID: <200306251547.KAA26697@trinity.cs.uiuc.edu>


Changes in directory llvm/lib/Reoptimizer/BinInterface:

sparcbin.cpp updated: 1.8 -> 1.9

---
Log message:

Removed warning.


---
Diffs of the changes:

Index: llvm/lib/Reoptimizer/BinInterface/sparcbin.cpp
diff -u llvm/lib/Reoptimizer/BinInterface/sparcbin.cpp:1.8 llvm/lib/Reoptimizer/BinInterface/sparcbin.cpp:1.9
--- llvm/lib/Reoptimizer/BinInterface/sparcbin.cpp:1.8	Tue Jun 24 14:47:23 2003
+++ llvm/lib/Reoptimizer/BinInterface/sparcbin.cpp	Wed Jun 25 10:46:52 2003
@@ -274,7 +274,7 @@
       if (instr->flags & IF_BRINTERNAL)
 	printf("{dest:s%02d} ", instr->ibranch.dest);
       else
-	printf("{dest:0x%lx} ", (intptr_t)instr->ebranch.dest);
+	printf("{dest:%p} ", instr->ebranch.dest);
       
       sparc_print_pseudo(instr->instr, 
 			 instr->alu.genrs1, 




From criswell at choi.cs.uiuc.edu  Wed Jun 25 11:16:01 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Wed Jun 25 11:16:01 2003
Subject: [llvm-commits] CVS: llvm/Makefile.SunOS
Message-ID: <200306251615.h5PGFTS14169@choi.cs.uiuc.edu>


Changes in directory llvm:

Makefile.SunOS (r1.9.2.1) removed

---
Log message:

Removed because autoconf takes care of platform specific options now.


---
Diffs of the changes:



From criswell at choi.cs.uiuc.edu  Wed Jun 25 11:16:06 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Wed Jun 25 11:16:06 2003
Subject: [llvm-commits] CVS: llvm/Makefile.Linux
Message-ID: <200306251615.h5PGFAm14159@choi.cs.uiuc.edu>


Changes in directory llvm:

Makefile.Linux (r1.10.2.1) removed

---
Log message:

Removed because autoconf takes care of this now.


---
Diffs of the changes:



From lattner at cs.uiuc.edu  Wed Jun 25 12:09:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Jun 25 12:09:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/LevelRaise/2003-06-25-ExprAnalysis.ll
Message-ID: <200306251708.MAA09570@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/LevelRaise:

2003-06-25-ExprAnalysis.ll added (r1.1)

---
Log message:

New testcase


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/LevelRaise/2003-06-25-ExprAnalysis.ll
diff -c /dev/null llvm/test/Regression/Transforms/LevelRaise/2003-06-25-ExprAnalysis.ll:1.1
*** /dev/null	Wed Jun 25 12:08:45 2003
--- llvm/test/Regression/Transforms/LevelRaise/2003-06-25-ExprAnalysis.ll	Wed Jun 25 12:08:34 2003
***************
*** 0 ****
--- 1,27 ----
+ ; The expr analysis routines were being too aggressive across cast instructions!
+ 
+ ; RUN: if as < %s | opt -raise | dis | grep 4294967295
+ ; RUN: then exit 1
+ ; RUN: else exit 0
+ ; RUN: fi
+ 
+ target endian = big
+ target pointersize = 64
+ 	%struct..istack_struct = type { %struct..istack_struct*, %struct..istk_entry*, uint }
+ 	%struct..istk_entry = type { double, int, int, double, double, sbyte* }
+ 
+ implementation   ; Functions:
+ bool %Intersection(%struct..istack_struct* %tmp.0, uint %tmp.12) {              ; No predecessors!
+         %tmp.8 = getelementptr %struct..istack_struct* %tmp.0, long 0, ubyte 1          ; <%struct..istk_entry**> [#uses=1]
+         %tmp.9 = load %struct..istk_entry** %tmp.8              ; <%struct..istk_entry*> [#uses=1]
+         %dec = sub uint %tmp.12, 1              ; <uint> [#uses=1]
+         %tmp.13 = cast uint %dec to ulong               ; <ulong> [#uses=1]
+         %tmp.14 = mul ulong %tmp.13, 40         ; <ulong> [#uses=1]
+         %tmp.16 = cast %struct..istk_entry* %tmp.9 to long              ; <long> [#uses=1]
+         %tmp.17 = cast ulong %tmp.14 to long            ; <long> [#uses=1]
+         %tmp.18 = add long %tmp.16, %tmp.17             ; <long> [#uses=1]
+         %tmp.19 = cast long %tmp.18 to %struct..istk_entry*             ; <%struct..istk_entry*> [#uses=1]
+         %tmp.21 = setne %struct..istk_entry* %tmp.19, null              ; <bool> [#uses=1]
+         ret bool %tmp.21
+ }
+ 




From lattner at cs.uiuc.edu  Wed Jun 25 12:10:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Jun 25 12:10:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Scalar/InstructionCombining.cpp
Message-ID: <200306251709.MAA09600@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Scalar:

InstructionCombining.cpp updated: 1.92 -> 1.93

---
Log message:

Instcombine:  X * -1 -> -X


---
Diffs of the changes:

Index: llvm/lib/Transforms/Scalar/InstructionCombining.cpp
diff -u llvm/lib/Transforms/Scalar/InstructionCombining.cpp:1.92 llvm/lib/Transforms/Scalar/InstructionCombining.cpp:1.93
--- llvm/lib/Transforms/Scalar/InstructionCombining.cpp:1.92	Mon Jun 23 16:59:52 2003
+++ llvm/lib/Transforms/Scalar/InstructionCombining.cpp	Wed Jun 25 12:09:20 2003
@@ -361,10 +361,11 @@
   if (Constant *Op1 = dyn_cast<Constant>(I.getOperand(1))) {
     if (ConstantInt *CI = dyn_cast<ConstantInt>(Op1)) {
       const Type *Ty = CI->getType();
-      uint64_t Val = Ty->isSigned() ?
-                          (uint64_t)cast<ConstantSInt>(CI)->getValue() : 
-                                    cast<ConstantUInt>(CI)->getValue();
+      int64_t Val = Ty->isSigned() ?        cast<ConstantSInt>(CI)->getValue() :
+                                   (int64_t)cast<ConstantUInt>(CI)->getValue();
       switch (Val) {
+      case -1:                               // X * -1 -> -X
+        return BinaryOperator::createNeg(Op0, I.getName());
       case 0:
         return ReplaceInstUsesWith(I, Op1);  // Eliminate 'mul double %X, 0'
       case 1:




From lattner at cs.uiuc.edu  Wed Jun 25 12:10:08 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Jun 25 12:10:08 2003
Subject: [llvm-commits] CVS: llvm/lib/Analysis/Expressions.cpp
Message-ID: <200306251709.MAA09589@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis:

Expressions.cpp updated: 1.31 -> 1.32

---
Log message:

Fix bug: LevelRaise/2003-06-25-ExprAnalysis.ll


---
Diffs of the changes:

Index: llvm/lib/Analysis/Expressions.cpp
diff -u llvm/lib/Analysis/Expressions.cpp:1.31 llvm/lib/Analysis/Expressions.cpp:1.32
--- llvm/lib/Analysis/Expressions.cpp:1.31	Tue May 20 13:45:29 2003
+++ llvm/lib/Analysis/Expressions.cpp	Wed Jun 25 12:09:03 2003
@@ -317,12 +317,10 @@
     if (isa<PointerType>(DestTy))
       DestTy = Type::ULongTy;  // Pointer types are represented as ulong
 
-    /*
     if (!Src.getExprType(0)->isLosslesslyConvertibleTo(DestTy)) {
       if (Src.ExprTy != ExprType::Constant)
         return I;  // Converting cast, and not a constant value...
     }
-    */
 
     const ConstantInt *Offset = Src.Offset;
     const ConstantInt *Scale  = Src.Scale;




From lattner at cs.uiuc.edu  Wed Jun 25 12:11:02 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Jun 25 12:11:02 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/InstCombine/mul.ll
Message-ID: <200306251710.MAA09618@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/InstCombine:

mul.ll updated: 1.5 -> 1.6

---
Log message:

Add test for instcombine -1*X -> -X


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/InstCombine/mul.ll
diff -u llvm/test/Regression/Transforms/InstCombine/mul.ll:1.5 llvm/test/Regression/Transforms/InstCombine/mul.ll:1.6
--- llvm/test/Regression/Transforms/InstCombine/mul.ll:1.5	Mon Mar 10 16:43:56 2003
+++ llvm/test/Regression/Transforms/InstCombine/mul.ll	Wed Jun 25 12:10:34 2003
@@ -33,8 +33,18 @@
 	ret int %B
 }
 
-int %test6(ubyte %A) {
+ubyte %test6(ubyte %A) {
 	%B = mul ubyte %A, 8
 	%C = mul ubyte %B, 13
 	ret ubyte %C
+}
+
+int %test7(int %i) {
+        %tmp = mul int %i, -1   ; %tmp = sub 0, %i
+        ret int %tmp
+}
+
+ulong %test8(ulong %i) {
+	%j = mul ulong %i, 18446744073709551615 ; tmp = sub 0, %i
+	ret ulong %j
 }




From lattner at cs.uiuc.edu  Wed Jun 25 12:30:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Jun 25 12:30:01 2003
Subject: [llvm-commits] CVS: llvm/test/Programs/TEST.nightly.Makefile
Message-ID: <200306251729.MAA09673@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Programs:

TEST.nightly.Makefile updated: 1.17 -> 1.18

---
Log message:

Hopefully this will finally get crafty working


---
Diffs of the changes:

Index: llvm/test/Programs/TEST.nightly.Makefile
diff -u llvm/test/Programs/TEST.nightly.Makefile:1.17 llvm/test/Programs/TEST.nightly.Makefile:1.18
--- llvm/test/Programs/TEST.nightly.Makefile:1.17	Tue Jun 24 12:31:24 2003
+++ llvm/test/Programs/TEST.nightly.Makefile	Wed Jun 25 12:29:27 2003
@@ -23,7 +23,7 @@
 Output/%.nightly.compile.report.txt: Output/%.llvm.bc $(LGCCAS)
 	@echo > $@
 	@echo '$(LGCCAS) Output/$*.linked.rll -o/dev/null $(TIMEOPT) >> $@ 2>&1'
-	@if ($(LGCCAS) Output/$*.linked.rll -o /dev/null $(TIMEOPT) >> $@ 2>&1)\
+	@-if ($(LGCCAS) Output/$*.linked.rll -o /dev/null $(TIMEOPT) >> $@ 2>&1)\
 	;then \
 	  echo "TEST-PASS: compile $(RELDIR)/$*" >> $@;\
 	  echo -n "TEST-RESULT-compile: " >> $@;\
@@ -52,7 +52,7 @@
 Output/%.nightly.llc.report.txt: Output/%.llvm.bc $(LLC)
 	@echo > $@
 	@echo 'time -p $(LLC) -f $(TIMEOPT) $< -o /dev/null) > $@ 2>&1'
-	@if (time -p $(LLC) -f $(TIMEOPT) $< -o /dev/null) >> $@ 2>&1; then \
+	@-if (time -p $(LLC) -f $(TIMEOPT) $< -o /dev/null) >> $@ 2>&1; then \
 	  echo "TEST-PASS: llc $(RELDIR)/$*" >> $@;\
 	  echo -n "TEST-RESULT-llc: " >> $@;\
 	  grep "Total Execution Time" $@.info >> $@;\
@@ -70,7 +70,7 @@
 Output/%.nightly.cbe.report.txt: Output/%.llvm.bc Output/%.exe-cbe $(LDIS)
 	@echo > $@
 	-head -n 100 Output/$*.exe-cbe >> $@
-	@if test -f Output/$*.exe-cbe; then \
+	@-if test -f Output/$*.exe-cbe; then \
 	  echo "TEST-PASS: cbe $(RELDIR)/$*" >> $@;\
           echo "TEST-RESULT-cbe: YES" >> $@;\
 	  echo -n "TEST-RESULT-cbe-time: " >> $@;\
@@ -85,7 +85,7 @@
 Output/%.nightly.jit.report.txt: Output/%.llvm.bc Output/%.exe-jit $(LLI)
 	@echo > $@
 	-head -n 100 Output/$*.exe-jit >> $@
-	@if test -f Output/$*.exe-jit; then \
+	@-if test -f Output/$*.exe-jit; then \
           echo "TEST-PASS: jit $(RELDIR)/$*" >> $@;\
 	  echo -n "TEST-RESULT-jit-time: " >> $@;\
 	  grep "^real" $(INFO_PREFIX)jit.time >> $@;\
@@ -113,6 +113,6 @@
 	@echo "---------------------------------------------------------------"
 	@echo ">>> ========= '$(RELDIR)/$*' Program"
 	@echo "---------------------------------------------------------------"
-	@cat $<
+	@-cat $<
 
 REPORT_DEPENDENCIES := $(LDIS) $(LLI) $(LLC)




From gaeke at cs.uiuc.edu  Wed Jun 25 13:02:02 2003
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Wed Jun 25 13:02:02 2003
Subject: [llvm-commits] CVS: llvm/lib/Target/X86/Printer.cpp
Message-ID: <200306251801.NAA01699@neo.cs.uiuc.edu>


Changes in directory llvm/lib/Target/X86:

Printer.cpp updated: 1.37 -> 1.38

---
Log message:

First draft of X86 LLC backend. This should be OK for small programs like
Shootout, but it has some issues with bigger programs. Work in progress.


---
Diffs of the changes:

Index: llvm/lib/Target/X86/Printer.cpp
diff -u llvm/lib/Target/X86/Printer.cpp:1.37 llvm/lib/Target/X86/Printer.cpp:1.38
--- llvm/lib/Target/X86/Printer.cpp:1.37	Thu Jun 19 14:58:32 2003
+++ llvm/lib/Target/X86/Printer.cpp	Wed Jun 25 13:01:07 2003
@@ -14,24 +14,37 @@
 #include "llvm/CodeGen/MachineConstantPool.h"
 #include "llvm/CodeGen/MachineInstr.h"
 #include "Support/Statistic.h"
+#include "Support/hash_map"
+#include "llvm/Type.h"
+#include "llvm/Constants.h"
+#include "llvm/Assembly/Writer.h"
+#include "llvm/DerivedTypes.h"
+#include "llvm/SlotCalculator.h"
+#include "Support/StringExtras.h"
+#include "llvm/Module.h"
 
 namespace {
   struct Printer : public MachineFunctionPass {
     std::ostream &O;
     unsigned ConstIdx;
     Printer(std::ostream &o) : O(o), ConstIdx(0) {}
+    const TargetData *TD;
 
     virtual const char *getPassName() const {
       return "X86 Assembly Printer";
     }
 
-    void printConstantPool(MachineConstantPool *MCP, const TargetData &TD);
-    bool runOnMachineFunction(MachineFunction &F);
-    
+    void printConstantPool(MachineConstantPool *MCP);
+    bool runOnMachineFunction(MachineFunction &F);    
+    std::string ConstantExprToString(const ConstantExpr* CE);
+    std::string valToExprString(const Value* V);
     bool doInitialization(Module &M);
     bool doFinalization(Module &M);
-
+    void PrintZeroBytesToPad(int numBytes);
+    void printConstantValueOnly(const Constant* CV, int numPadBytesAfter = 0);
+    void printSingleConstantValue(const Constant* CV);
   };
+  std::map<const Value *, unsigned> NumberForBB;
 }
 
 /// createX86CodePrinterPass - Print out the specified machine code function to
@@ -42,31 +55,317 @@
   return new Printer(O);
 }
 
+// valToExprString - Helper function for ConstantExprToString().
+// Appends result to argument string S.
+// 
+std::string Printer::valToExprString(const Value* V) {
+  std::string S;
+  bool failed = false;
+  if (const Constant* CV = dyn_cast<Constant>(V)) { // symbolic or known
+    if (const ConstantBool *CB = dyn_cast<ConstantBool>(CV))
+      S += std::string(CB == ConstantBool::True ? "1" : "0");
+    else if (const ConstantSInt *CI = dyn_cast<ConstantSInt>(CV))
+      S += itostr(CI->getValue());
+    else if (const ConstantUInt *CI = dyn_cast<ConstantUInt>(CV))
+      S += utostr(CI->getValue());
+    else if (const ConstantFP *CFP = dyn_cast<ConstantFP>(CV))
+      S += ftostr(CFP->getValue());
+    else if (isa<ConstantPointerNull>(CV))
+      S += "0";
+    else if (const ConstantPointerRef *CPR = dyn_cast<ConstantPointerRef>(CV))
+      S += valToExprString(CPR->getValue());
+    else if (const ConstantExpr *CE = dyn_cast<ConstantExpr>(CV))
+      S += ConstantExprToString(CE);
+    else
+      failed = true;
+  } else if (const GlobalValue* GV = dyn_cast<GlobalValue>(V)) {
+    // S += getID(GV);
+    assert (0 && "getID not implemented");
+  }
+  else
+    failed = true;
+
+  if (failed) {
+    assert(0 && "Cannot convert value to string");
+    S += "<illegal-value>";
+  }
+  return S;
+}
+
+// ConstantExprToString() - Convert a ConstantExpr to an asm expression
+// and return this as a string.
+std::string Printer::ConstantExprToString(const ConstantExpr* CE) {
+  std::string S;
+  switch(CE->getOpcode()) {
+  case Instruction::GetElementPtr:
+    { // generate a symbolic expression for the byte address
+      const Value* ptrVal = CE->getOperand(0);
+      std::vector<Value*> idxVec(CE->op_begin()+1, CE->op_end());
+      S += "(" + valToExprString(ptrVal) + ") + ("
+	+ utostr(TD->getIndexedOffset(ptrVal->getType(),idxVec)) + ")";
+      break;
+    }
+
+  case Instruction::Cast:
+    // Support only non-converting casts for now, i.e., a no-op.
+    // This assertion is not a complete check.
+    assert(TD->getTypeSize(CE->getType()) ==
+	   TD->getTypeSize(CE->getOperand(0)->getType()));
+    S += "(" + valToExprString(CE->getOperand(0)) + ")";
+    break;
+
+  case Instruction::Add:
+    S += "(" + valToExprString(CE->getOperand(0)) + ") + ("
+      + valToExprString(CE->getOperand(1)) + ")";
+    break;
+
+  default:
+    assert(0 && "Unsupported operator in ConstantExprToString()");
+    break;
+  }
+
+  return S;
+}
+
+// Print a single constant value.
+void
+Printer::printSingleConstantValue(const Constant* CV)
+{
+  assert(CV->getType() != Type::VoidTy &&
+         CV->getType() != Type::TypeTy &&
+         CV->getType() != Type::LabelTy &&
+         "Unexpected type for Constant");
+  
+  assert((!isa<ConstantArray>(CV) && ! isa<ConstantStruct>(CV))
+         && "Aggregate types should be handled outside this function");
+
+  const Type *type = CV->getType();
+  O << "\t";
+  switch(type->getPrimitiveID())
+    {
+    case Type::BoolTyID: case Type::UByteTyID: case Type::SByteTyID:
+      O << ".byte";
+      break;
+    case Type::UShortTyID: case Type::ShortTyID:
+      O << ".word";
+      break;
+    case Type::UIntTyID: case Type::IntTyID: case Type::PointerTyID:
+      O << ".long";
+      break;
+    case Type::ULongTyID: case Type::LongTyID:
+      O << ".quad";
+      break;
+    case Type::FloatTyID:
+      O << ".long";
+      break;
+    case Type::DoubleTyID:
+      O << ".quad";
+      break;
+    case Type::ArrayTyID:
+      if ((cast<ArrayType>(type)->getElementType() == Type::UByteTy) ||
+	  (cast<ArrayType>(type)->getElementType() == Type::SByteTy))
+	O << ".string";
+      else
+	assert (0 && "Can't handle printing this type of array");
+      break;
+    default:
+      assert (0 && "Can't handle printing this type of thing");
+      break;
+    }
+  O << "\t";
+  
+  if (type->isPrimitiveType())
+    {
+      if (type->isFloatingPoint()) {
+	// FP Constants are printed as integer constants to avoid losing
+	// precision...
+	double Val = cast<ConstantFP>(CV)->getValue();
+	if (type == Type::FloatTy) {
+	  float FVal = (float)Val;
+	  char *ProxyPtr = (char*)&FVal;        // Abide by C TBAA rules
+	  O << *(unsigned int*)ProxyPtr;            
+	} else if (type == Type::DoubleTy) {
+	  char *ProxyPtr = (char*)&Val;         // Abide by C TBAA rules
+	  O << *(uint64_t*)ProxyPtr;            
+	} else {
+	  assert(0 && "Unknown floating point type!");
+	}
+        
+	O << "\t# " << type->getDescription() << " value: " << Val << "\n";
+      } else {
+	WriteAsOperand(O, CV, false, false) << "\n";
+      }
+    }
+  else if (const ConstantPointerRef* CPR = dyn_cast<ConstantPointerRef>(CV))
+    {
+      // This is a constant address for a global variable or method.
+      // Use the name of the variable or method as the address value.
+      // O << getID(CPR->getValue()) << "\n";
+      assert (0 && "getID not implemented");
+
+    }
+  else if (isa<ConstantPointerNull>(CV))
+    {
+      // Null pointer value
+      O << "0\n";
+    }
+  else if (const ConstantExpr* CE = dyn_cast<ConstantExpr>(CV))
+    {
+      // Constant expression built from operators, constants, and
+      // symbolic addrs
+      O << ConstantExprToString(CE) << "\n";
+    }
+  else
+    {
+      assert(0 && "Unknown elementary type for constant");
+    }
+}
+
+// Can we treat the specified array as a string?  Only if it is an array of
+// ubytes or non-negative sbytes.
+//
+static bool isStringCompatible(const ConstantArray *CVA) {
+  const Type *ETy = cast<ArrayType>(CVA->getType())->getElementType();
+  if (ETy == Type::UByteTy) return true;
+  if (ETy != Type::SByteTy) return false;
+
+  for (unsigned i = 0; i < CVA->getNumOperands(); ++i)
+    if (cast<ConstantSInt>(CVA->getOperand(i))->getValue() < 0)
+      return false;
+
+  return true;
+}
+
+// toOctal - Convert the low order bits of X into an octal letter
+static inline char toOctal(int X) {
+  return (X&7)+'0';
+}
+
+// getAsCString - Return the specified array as a C compatible string, only if
+// the predicate isStringCompatible is true.
+//
+static std::string getAsCString(const ConstantArray *CVA) {
+  assert(isStringCompatible(CVA) && "Array is not string compatible!");
+
+  std::string Result;
+  const Type *ETy = cast<ArrayType>(CVA->getType())->getElementType();
+  Result = "\"";
+  for (unsigned i = 0; i < CVA->getNumOperands(); ++i) {
+    unsigned char C = (ETy == Type::SByteTy) ?
+      (unsigned char)cast<ConstantSInt>(CVA->getOperand(i))->getValue() :
+      (unsigned char)cast<ConstantUInt>(CVA->getOperand(i))->getValue();
+
+    if (C == '"') {
+      Result += "\\\"";
+    } else if (C == '\\') {
+      Result += "\\\\";
+    } else if (isprint(C)) {
+      Result += C;
+    } else {
+      switch(C) {
+      case '\a': Result += "\\a"; break;
+      case '\b': Result += "\\b"; break;
+      case '\f': Result += "\\f"; break;
+      case '\n': Result += "\\n"; break;
+      case '\r': Result += "\\r"; break;
+      case '\t': Result += "\\t"; break;
+      case '\v': Result += "\\v"; break;
+      default:
+        Result += '\\';
+        Result += toOctal(C >> 6);
+        Result += toOctal(C >> 3);
+        Result += toOctal(C >> 0);
+        break;
+      }
+    }
+  }
+  Result += "\"";
+  return Result;
+}
+
+// Print a constant value or values (it may be an aggregate).
+// Uses printSingleConstantValue() to print each individual value.
+void
+Printer::printConstantValueOnly(const Constant* CV,
+				int numPadBytesAfter /* = 0 */)
+{
+  const ConstantArray *CVA = dyn_cast<ConstantArray>(CV);
+
+  if (CVA && isStringCompatible(CVA))
+    { // print the string alone and return
+      O << "\t" << ".string" << "\t" << getAsCString(CVA) << "\n";
+    }
+  else if (CVA)
+    { // Not a string.  Print the values in successive locations
+      const std::vector<Use> &constValues = CVA->getValues();
+      for (unsigned i=0; i < constValues.size(); i++)
+        printConstantValueOnly(cast<Constant>(constValues[i].get()));
+    }
+  else if (const ConstantStruct *CVS = dyn_cast<ConstantStruct>(CV))
+    { // Print the fields in successive locations. Pad to align if needed!
+      const StructLayout *cvsLayout =
+        TD->getStructLayout(CVS->getType());
+      const std::vector<Use>& constValues = CVS->getValues();
+      unsigned sizeSoFar = 0;
+      for (unsigned i=0, N = constValues.size(); i < N; i++)
+        {
+          const Constant* field = cast<Constant>(constValues[i].get());
+
+          // Check if padding is needed and insert one or more 0s.
+          unsigned fieldSize = TD->getTypeSize(field->getType());
+          int padSize = ((i == N-1? cvsLayout->StructSize
+			  : cvsLayout->MemberOffsets[i+1])
+                         - cvsLayout->MemberOffsets[i]) - fieldSize;
+          sizeSoFar += (fieldSize + padSize);
+
+          // Now print the actual field value
+          printConstantValueOnly(field, padSize);
+        }
+      assert(sizeSoFar == cvsLayout->StructSize &&
+             "Layout of constant struct may be incorrect!");
+    }
+  else
+    printSingleConstantValue(CV);
+
+  if (numPadBytesAfter) {
+    unsigned numBytes = numPadBytesAfter;
+    for ( ; numBytes >= 8; numBytes -= 8)
+      printSingleConstantValue(Constant::getNullValue(Type::ULongTy));
+    if (numBytes >= 4)
+      {
+	printSingleConstantValue(Constant::getNullValue(Type::UIntTy));
+	numBytes -= 4;
+      }
+    while (numBytes--)
+      printSingleConstantValue(Constant::getNullValue(Type::UByteTy));
+  }
+}
 
 // printConstantPool - Print out any constants which have been spilled to
 // memory...
-void Printer::printConstantPool(MachineConstantPool *MCP, const TargetData &TD){
+void Printer::printConstantPool(MachineConstantPool *MCP){
   const std::vector<Constant*> &CP = MCP->getConstants();
   if (CP.empty()) return;
 
   for (unsigned i = 0, e = CP.size(); i != e; ++i) {
     O << "\t.section .rodata\n";
-    O << "\t.align " << (unsigned)TD.getTypeAlignment(CP[i]->getType()) << "\n";
+    O << "\t.align " << (unsigned)TD->getTypeAlignment(CP[i]->getType()) << "\n";
     O << ".CPI" << i+ConstIdx << ":\t\t\t\t\t#" << *CP[i] << "\n";
-    O << "\t*Constant output not implemented yet!*\n\n";
+    printConstantValueOnly (CP[i]);
   }
   ConstIdx += CP.size();  // Don't recycle constant pool index numbers
 }
 
-/// runOnFunction - This uses the X86InstructionInfo::print method
+/// runOnMachineFunction - This uses the X86InstructionInfo::print method
 /// to print assembly for each instruction.
 bool Printer::runOnMachineFunction(MachineFunction &MF) {
   static unsigned BBNumber = 0;
   const TargetMachine &TM = MF.getTarget();
   const TargetInstrInfo &TII = TM.getInstrInfo();
+  TD = &TM.getTargetData();
 
   // Print out constants referenced by the function
-  printConstantPool(MF.getConstantPool(), TM.getTargetData());
+  printConstantPool(MF.getConstantPool());
 
   // Print out labels for the function.
   O << "\t.text\n";
@@ -75,11 +374,18 @@
   O << "\t.type\t" << MF.getFunction()->getName() << ", @function\n";
   O << MF.getFunction()->getName() << ":\n";
 
+  NumberForBB.clear();
+  for (MachineFunction::const_iterator I = MF.begin(), E = MF.end();
+       I != E; ++I) {
+    NumberForBB[I->getBasicBlock()] = BBNumber++;
+  }
+
   // Print out code for the function.
   for (MachineFunction::const_iterator I = MF.begin(), E = MF.end();
        I != E; ++I) {
     // Print a label for the basic block.
-    O << ".BB" << BBNumber++ << ":\n";
+    O << ".BB" << NumberForBB[I->getBasicBlock()] << ":\t# "
+      << I->getBasicBlock()->getName() << "\n";
     for (MachineBasicBlock::const_iterator II = I->begin(), E = I->end();
 	 II != E; ++II) {
       // Print the assembly for the instruction.
@@ -94,20 +400,20 @@
 
 static bool isScale(const MachineOperand &MO) {
   return MO.isImmediate() &&
-           (MO.getImmedValue() == 1 || MO.getImmedValue() == 2 ||
-            MO.getImmedValue() == 4 || MO.getImmedValue() == 8);
+    (MO.getImmedValue() == 1 || MO.getImmedValue() == 2 ||
+     MO.getImmedValue() == 4 || MO.getImmedValue() == 8);
 }
 
 static bool isMem(const MachineInstr *MI, unsigned Op) {
   if (MI->getOperand(Op).isFrameIndex()) return true;
   if (MI->getOperand(Op).isConstantPoolIndex()) return true;
   return Op+4 <= MI->getNumOperands() &&
-         MI->getOperand(Op  ).isRegister() &&isScale(MI->getOperand(Op+1)) &&
-         MI->getOperand(Op+2).isRegister() &&MI->getOperand(Op+3).isImmediate();
+    MI->getOperand(Op  ).isRegister() &&isScale(MI->getOperand(Op+1)) &&
+    MI->getOperand(Op+2).isRegister() &&MI->getOperand(Op+3).isImmediate();
 }
 
 static void printOp(std::ostream &O, const MachineOperand &MO,
-                    const MRegisterInfo &RI) {
+                    const MRegisterInfo &RI, bool elideOffsetKeyword = false) {
   switch (MO.getType()) {
   case MachineOperand::MO_VirtualRegister:
     if (Value *V = MO.getVRegValueOrNull()) {
@@ -127,29 +433,30 @@
     O << (int)MO.getImmedValue();
     return;
   case MachineOperand::MO_PCRelativeDisp:
-    O << MO.getVRegValue()->getName();
+    O << ".BB" << NumberForBB[MO.getVRegValue()] << " # PC rel: "
+      << MO.getVRegValue()->getName();
     return;
   case MachineOperand::MO_GlobalAddress:
-    O << MO.getGlobal()->getName();
+    if (!elideOffsetKeyword) O << "OFFSET "; O << MO.getGlobal()->getName();
     return;
   case MachineOperand::MO_ExternalSymbol:
     O << MO.getSymbolName();
     return;
   default:
-    O << "<unknown op ty>"; return;    
+    O << "<unknown operand type>"; return;    
   }
 }
 
 static const std::string sizePtr(const TargetInstrDescriptor &Desc) {
   switch (Desc.TSFlags & X86II::ArgMask) {
-    default: assert(0 && "Unknown arg size!");
-    case X86II::Arg8:   return "BYTE PTR"; 
-    case X86II::Arg16:  return "WORD PTR"; 
-    case X86II::Arg32:  return "DWORD PTR"; 
-    case X86II::Arg64:  return "QWORD PTR"; 
-    case X86II::ArgF32:  return "DWORD PTR"; 
-    case X86II::ArgF64:  return "QWORD PTR"; 
-    case X86II::ArgF80:  return "XWORD PTR"; 
+  default: assert(0 && "Unknown arg size!");
+  case X86II::Arg8:   return "BYTE PTR"; 
+  case X86II::Arg16:  return "WORD PTR"; 
+  case X86II::Arg32:  return "DWORD PTR"; 
+  case X86II::Arg64:  return "QWORD PTR"; 
+  case X86II::ArgF32:  return "DWORD PTR"; 
+  case X86II::ArgF64:  return "QWORD PTR"; 
+  case X86II::ArgF80:  return "XWORD PTR"; 
   }
 }
 
@@ -264,7 +571,7 @@
     O << getName(MI->getOpcode()) << " ";
 
     if (MI->getNumOperands() == 1) {
-      printOp(O, MI->getOperand(0), RI);
+      printOp(O, MI->getOperand(0), RI, true); // Don't print "OFFSET"...
     }
     O << "\n";
     return;
@@ -417,7 +724,7 @@
             MI->getOperand(1).isRegister() || MI->getOperand(1).isImmediate())&&
            "Bad MRMSxR format!");
     assert((MI->getNumOperands() < 3 ||
-        (MI->getOperand(1).isRegister() && MI->getOperand(2).isImmediate())) &&
+	    (MI->getOperand(1).isRegister() && MI->getOperand(2).isImmediate())) &&
            "Bad MRMSxR format!");
 
     if (MI->getNumOperands() > 1 && MI->getOperand(1).isRegister() && 
@@ -476,7 +783,24 @@
 
 bool Printer::doFinalization(Module &M)
 {
-  // FIXME: We may have to print out constants here.
+  // Print out module-level global variables here.
+  for (Module::const_giterator I = M.gbegin(), E = M.gend(); I != E; ++I) {
+    if (I->hasInitializer()) {
+      Constant *C = I->getInitializer();
+      O << "\t.data\n";
+      O << "\t.globl " << I->getName() << "\n";
+      O << "\t.type " << I->getName() << ", at object\n";
+      O << "\t.size " << I->getName() << ","
+	<< (unsigned)TD->getTypeSize(I->getType()) << "\n";
+      O << "\t.align " << (unsigned)TD->getTypeAlignment(C->getType()) << "\n";
+      O << I->getName() << ":\t\t\t\t\t#" << *C << "\n";
+      printConstantValueOnly (C);
+    } else {
+      O << "\t.globl " << I->getName() << "\n";
+      O << "\t.comm " << I->getName() << ", "
+        << (unsigned)TD->getTypeSize(I->getType()) << ", "
+        << (unsigned)TD->getTypeAlignment(I->getType()) << "\n";
+    }
+  }
   return false; // success
 }
-




From criswell at choi.cs.uiuc.edu  Wed Jun 25 13:44:01 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Wed Jun 25 13:44:01 2003
Subject: [llvm-commits] CVS: llvm/Makefile.common
Message-ID: <200306251843.h5PIhho20924@choi.cs.uiuc.edu>


Changes in directory llvm:

Makefile.common updated: 1.88.2.3 -> 1.88.2.4

---
Log message:

Fixed the rules for linking optimized tools.
We were orignally linking against some profile libraries, but now we link
against the optimized libraries.


---
Diffs of the changes:

Index: llvm/Makefile.common
diff -u llvm/Makefile.common:1.88.2.3 llvm/Makefile.common:1.88.2.4
--- llvm/Makefile.common:1.88.2.3	Tue Jun 24 15:36:04 2003
+++ llvm/Makefile.common	Wed Jun 25 13:43:32 2003
@@ -560,7 +560,7 @@
 LLVM_LIBS_OPTIONS_P := $(patsubst %.o, $(LLVMLIBPROFILESOURCE)/%.o,$(LLVM_LIBS_OPTIONS))
 
 LIB_OPTS_G :=  $(LLVM_LIBS_OPTIONS_G) $(PROJ_LIBS_OPTIONS_G)
-LIB_OPTS_O :=  $(LLVM_LIBS_OPTIONS_O) $(PROJ_LIBS_OPTIONS_P)
+LIB_OPTS_O :=  $(LLVM_LIBS_OPTIONS_O) $(PROJ_LIBS_OPTIONS_O)
 LIB_OPTS_P :=  $(LLVM_LIBS_OPTIONS_P) $(PROJ_LIBS_OPTIONS_P)
 
 # USED_LIB_PATHS - Compute the path of the libraries used so that tools are




From criswell at choi.cs.uiuc.edu  Wed Jun 25 16:45:01 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Wed Jun 25 16:45:01 2003
Subject: [llvm-commits] CVS: llvm/www/docs/GettingStarted.html
Message-ID: <200306252144.h5PLiJT31190@choi.cs.uiuc.edu>


Changes in directory llvm/www/docs:

GettingStarted.html updated: 1.12 -> 1.12.2.1

---
Log message:

Added a requirements section so that people know what hardware and software
they will need.
Made more changes to reflect the new autoconf stuff.



---
Diffs of the changes:

Index: llvm/www/docs/GettingStarted.html
diff -u llvm/www/docs/GettingStarted.html:1.12 llvm/www/docs/GettingStarted.html:1.12.2.1
--- llvm/www/docs/GettingStarted.html:1.12	Sun Jun  8 10:33:25 2003
+++ llvm/www/docs/GettingStarted.html	Wed Jun 25 16:44:08 2003
@@ -19,6 +19,7 @@
       <li><a href="#overview">Overview</a>
       <li><a href="#starting">Getting started with LLVM</a>
         <ol>
+          <li><a href="#requirements">Requirements</a>
           <li><a href="#quickstart">Getting started quickly (a summary)</a>
           <li><a href="#checkout">Checkout LLVM from CVS</a>
           <li><a href="#terminology">Terminology and Notation</tt></a>
@@ -66,6 +67,57 @@
     </center>
     <!--=====================================================================-->
 
+    <!--=====================================================================-->
+    <h3><a name="requirements"><b>Requirements</b></a></h3>
+    <!--=====================================================================-->
+
+    LLVM is known to work on the following platforms:
+    <ul>
+        <li> Linux/x86
+        <li> Solaris/Sparc
+    </ul>
+
+    Compiling LLVM requires that you have several different software packages
+    installed:
+    <ul>
+        <li> GCC
+        <p>
+        The GNU Compiler Collection must be installed with C and C++ language
+        support.  GCC 3.x is supported, although some effort has been made to
+        support GCC 2.96.
+        </p>
+
+        <p>
+        Note that we currently do not support any other C++ compiler.
+        </p>
+
+        <li> GNU Make
+        <p>
+        The LLVM build system relies upon GNU make extensions.  Therefore, you
+        will need GNU Make (sometimes known as gmake) to build LLVM.
+        </p>
+
+        <li> Flex and Bison
+        <p>
+        The LLVM source code is built using flex and bison.  You will not be
+        able to configure and compile LLVM without them.
+        </p>
+    </ul>
+
+    <p>
+    There are some additional tools that you may want to have when working with
+    LLVM:
+    </p>
+
+    <ul>
+        <li> GNU Autoconf and GNU M4
+        <p>
+        If you want to make changes to the autoconf scripts which configure LLVM
+        for compilation, you will need GNU autoconf, and consequently, GNU M4.
+        LLVM was built with autoconf 2.53, so that release and any later
+        releases should work.
+        </p>
+    </ul>
 
     <!--=====================================================================-->
     <h3><a name="quickstart"><b>Getting Started Quickly (A Summary)</b></a></h3>
@@ -77,9 +129,17 @@
     <li><tt>cd <i>where-you-want-llvm-to-live</i></tt>
     <li><tt>cvs -d <i>CVSROOTDIR</i> checkout llvm</tt>
     <li><tt>cd llvm</tt>
-    <li>Edit <tt>Makefile.config</tt> to set local paths.  This includes
-        setting the install location of the C frontend, and the various paths
-        to the C and C++ compilers used to build LLVM itself.
+    <li>Run <tt>configure</tt> to configure the Makefiles and header files.
+    Useful options include:
+        <ul>
+            <li><tt>--with-objroot=<i>directory</i></tt>
+            <br>
+            Specifiy where object files should be placed during the build.
+
+            <li><tt>--with-llvmgccdir=<i>directory</i></tt>
+            <br>
+            Specifiy where the LLVM C frontend has been installed.
+        </ul>
     <li>Set your LLVM_LIB_SEARCH_PATH environment variable.
     <li><tt>gmake -k |& tee gnumake.out
 	    &nbsp;&nbsp;&nbsp;# this is csh or tcsh syntax</tt>
@@ -95,10 +155,10 @@
     <h3><a name="terminology">Terminology and Notation</a></h3>
     <!------------------------------------------------------------------------->
 
-    <p>Through this manual, the following names are used to denote paths
+    <p>Throughout this manual, the following names are used to denote paths
     specific to the local system and working environment.  <i>These are not
-    environment variables you need to set, but just strings used in the rest
-    of this document below.</i>.  In any of the examples below, simply replace
+    environment variables you need to set but just strings used in the rest
+    of this document below</i>.  In any of the examples below, simply replace
     each of these names with the appropriate pathname on your local system.
     All these paths are absolute:</p>
     <ul>
@@ -109,7 +169,7 @@
     <!------------------------------------------------------------------------->
 
     <p>Before checking out the source code, you will need to know the path to
-    CVS repository containing LLVM source code (we'll call this
+    the CVS repository containing the LLVM source code (we'll call this
     <i>CVSROOTDIR</i> below).  Ask the person responsible for your local LLVM
     installation to give you this path.
 
@@ -128,46 +188,71 @@
     <h3><a name="config">Local Configuration Options</a></h3>
     <!------------------------------------------------------------------------->
 
-    <p>The file <tt>llvm/Makefile.config</tt>
-    defines the following path variables,
-    which are specific to a particular installation of LLVM.
-    These should need to be modified only once after checking out a copy
-    of LLVM (if the default values do not already match your system):
+    <p>Once checked out from the CVS repository, options and pathnames specific
+    to an installation of LLVM can be set via the <tt>configure</tt> script.
+    This script sets variables in <tt>llvm/Makefile.config</tt> and
+    <tt>llvm/include/Config/config.h</tt>.
+
+    <p>
+    The following environment variables are used by <tt>configure</tt> to
+    configure Makefile.config:
+    </p>
 
     <ul>
-    <p><li><i>CXX</i> = Path to C++ compiler to use.
-    <p><li><i>LLVM_OBJ_DIR</i> = Path to the llvm directory where
-				 object files should be placed.
-				 (See the Section on <a href=#objfiles>
-				 The location for LLVM object files</a>
-				 for more information.)
-    <p><li><i>LLVMGCCDIR</i>   = Path to the location of the LLVM front-end
-				 binaries and associated libraries.
-    <p><li><i>PURIFY</i>       = Path to the purify program.
+    <p><li><i>CXX</i> = Pathname of the C++ compiler to use.
+    <p><li><i>CC</i> = Pathname of the C compiler to use.
     </ul>
 
-    In addition to settings in this file, you must set a
+    The following options can be used to set or enable LLVM specific options:
+
+    <ul>
+    <p><li><i>--with-objroot=LLVM_OBJ_ROOT</i> =
+        Path to the directory where
+		object files, libraries, and executables should be placed.
+		(See the Section on <a href=#objfiles>
+		The location for LLVM object files</a>
+		for more information.)
+    <p><li><i>--with-llvmgccdir=LLVMGCCDIR</i>   =
+        Path to the location of the LLVM front-end
+		binaries and associated libraries.
+    <p><li><i>--enable-optimized</i>       =
+        Enables optimized compilation (debugging symbols are removed and GCC
+        optimization flags are enabled).
+    <p><li><i>--enable-profiling</i>       =
+        Enables profiling compilation (compiler flags needed to add profiling
+        data are enabled).
+    <p><li><i>--enable-verbose</i>       =
+        Enables verbose messages during the compile.
+    <p><li><i>--enable-jit</i>       =
+        Compile the Just In Time (JIT) functionality.  This is not available
+        on all platforms.
+    </ul>
+
+    In addition to running <tt>configure</tt>, you must set the
     <tt>LLVM_LIB_SEARCH_PATH</tt> environment variable in your startup scripts.
     This environment variable is used to locate "system" libraries like
     "<tt>-lc</tt>" and "<tt>-lm</tt>" when linking.  This variable should be set
     to the absolute path for the bytecode-libs subdirectory of the C front-end
-    install.  For example,
+    install.  For example, <tt>LLVM_LIB_SEARCH_PATH</tt> is set to
     <tt>/home/vadve/lattner/local/x86/llvm-gcc/bytecode-libs</tt> for the X86
-    version of the C front-end, on our research machines.<p>
+    version of the C front-end on our research machines.<p>
 
     <!------------------------------------------------------------------------->
     <h3><a name="objfiles">The location for LLVM object files</a></h3>
     <!------------------------------------------------------------------------->
 
     <p>The LLVM make system sends most output files generated during the build
-    into the directory defined by the variable LLVM_OBJ_DIR in
+    into the directory defined by the variable OBJ_ROOT in
     <tt>llvm/Makefile.config</tt>.  This can be either just your normal LLVM
     source tree or some other directory writable by you.  You may wish to put
     object files on a different filesystem either to keep them from being backed
     up or to speed up local builds.
 
-    <p>If you do not wish to use a different location for object files (building
-    into the source tree directly), just set this variable to ".".<p>
+    <p>If you wish to place output files into a separate directory, use the
+    <tt>--with-objroot=<i>directory</i></tt> option of <tt>configure</tt> to
+    set the top level directory of where the object files will go.  Otherwise,
+    leave this option unspecified, and <tt>configure</tt> will place files
+    within the LLVM source tree.
 
     <!------------------------------------------------------------------------->
     <h3><a name="environment">Setting up your environment</a></h3>
@@ -185,13 +270,13 @@
        alias llvmgcc <i>LLVMGCCDIR</i><tt>/bin/llvm-gcc</tt>
 
        # Make the LLVM tools easy to use...
-       setenv PATH <i>LLVM_OBJ_DIR</i>/llvm/tools/Debug:${PATH}
+       setenv PATH <i>OBJ_ROOT</i>/llvm/tools/Debug:${PATH}
     </pre>
     The <tt>llvmgcc</tt> alias is useful because the C compiler is not
     included in the CVS tree you just checked out.
-    
+
     <p>The other <a href="#tools">LLVM tools</a> are part of the LLVM
-    source base, and built when compiling LLVM.  They will be built into the
+    source base and are built when compiling LLVM.  They will be built into the
     <tt><i>LLVM_OBJ_DIR</i>/tools/Debug</tt> directory.</p>
 
     <!------------------------------------------------------------------------->
@@ -199,13 +284,14 @@
     <!------------------------------------------------------------------------->
 
     <p>Every directory in the LLVM source tree includes a <tt>Makefile</tt> to
-    build it, and any subdirectories that it contains.  These makefiles require
+    build it and any subdirectories that it contains.  These makefiles require
     that you use <tt>gmake</tt>, instead of <tt>make</tt> to build them, but can
     otherwise be used freely.  To build the entire LLVM system, just enter the
     top level <tt>llvm</tt> directory and type <tt>gmake</tt>.  A few minutes
     later you will hopefully have a freshly compiled toolchain waiting for you
-    in <tt>llvm/tools/Debug</tt>.  If you want to look at the libraries that
-    were compiled, look in <tt>llvm/lib/Debug</tt>.</p>
+    in <tt><i>OBJ_ROOT</i>/llvm/tools/Debug</tt>.  If you want to look at the
+    libraries that
+    were compiled, look in <tt><i>OBJ_ROOT</i>/llvm/lib/Debug</tt>.</p>
 
     If you get an error talking about a <tt>/localhome</tt> directory, follow
     the instructions in the section about <a href="#environment">Setting Up Your
@@ -238,8 +324,10 @@
     directories</a></h3>
     <!------------------------------------------------------------------------->
 
-    If you are building with the "<tt>BUILD_ROOT=.</tt>" option enabled in the
-    <tt>Makefile.common</tt> file, most source directories will contain two
+    If you are building with the "<tt>OBJ_ROOT=.</tt>" option enabled in the
+    <tt>Makefile.config</tt> file (i.e. you did not specify
+    <tt>--with-objroot</tt> when you ran <tt>configure</tt>), most source
+    directories will contain two
     directories, <tt>Depend</tt> and <tt>Debug</tt>. The <tt>Depend</tt>
     directory contains automatically generated dependance files which are used
     during compilation to make sure that source files get rebuilt if a header
@@ -247,7 +335,7 @@
     files, library files and executables that are used for building a debug
     enabled build.  The <tt>Release</tt> directory is created to hold the same
     files when the <tt>ENABLE_OPTIMIZED=1</tt> flag is passed to <tt>gmake</tt>,
-    causing an optimized built to be performed.<p>
+    causing an optimized build to be performed.<p>
 
 
     <!------------------------------------------------------------------------->
@@ -255,7 +343,7 @@
     <!------------------------------------------------------------------------->
 
     This directory contains public header files exported from the LLVM
-    library. The two main subdirectories of this directory are:<p>
+    library. The three main subdirectories of this directory are:<p>
 
     <ol>
        <li><tt>llvm/include/llvm</tt> - This directory contains all of the LLVM
@@ -267,13 +355,20 @@
        support libraries that are independant of LLVM, but are used by LLVM.
        For example, some C++ STL utilities and a Command Line option processing
        library.
+
+       <li><tt>llvm/include/Config</tt> - This directory contains header files
+       configured by the <tt>configure</tt> script.  They wrap "standard" UNIX
+       and C header files.  Source code can include these header files which
+       automatically take care of the conditional #includes that the configure
+       script generates.
     </ol>
 
     <!------------------------------------------------------------------------->
     <h3><a name="lib"><tt>llvm/lib</tt></a></h3>
     <!------------------------------------------------------------------------->
 
-    This directory contains most source files of LLVM system. In LLVM almost all
+    This directory contains most of the source files of the LLVM system. In
+    LLVM almost all
     code exists in libraries, making it very easy to share code among the
     different <a href="#tools">tools</a>.<p>
 
@@ -362,7 +457,7 @@
         <dt><tt><b>gccas</b></tt><dd> This tool is invoked by the
         <tt>llvmgcc</tt> frontend as the "assembler" part of the compiler.  This
         tool actually assembles LLVM assembly to LLVM bytecode,
-	performs a variety of optimizations,
+	    performs a variety of optimizations,
         and outputs LLVM bytecode.  Thus when you invoke <tt>llvmgcc -c x.c -o
         x.o</tt>, you are causing <tt>gccas</tt> to be run, which writes the
         <tt>x.o</tt> file (which is an LLVM bytecode file that can be




From gaeke at cs.uiuc.edu  Wed Jun 25 17:01:03 2003
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Wed Jun 25 17:01:03 2003
Subject: [llvm-commits] CVS: llvm/lib/Target/X86/Printer.cpp
Message-ID: <200306252200.RAA04405@zion.cs.uiuc.edu>


Changes in directory llvm/lib/Target/X86:

Printer.cpp updated: 1.38 -> 1.39

---
Log message:

Integrate the C writer's name mangler.


---
Diffs of the changes:

Index: llvm/lib/Target/X86/Printer.cpp
diff -u llvm/lib/Target/X86/Printer.cpp:1.38 llvm/lib/Target/X86/Printer.cpp:1.39
--- llvm/lib/Target/X86/Printer.cpp:1.38	Wed Jun 25 13:01:07 2003
+++ llvm/lib/Target/X86/Printer.cpp	Wed Jun 25 17:00:39 2003
@@ -24,7 +24,11 @@
 #include "llvm/Module.h"
 
 namespace {
+
+
+  std::set<const Value*> MangledGlobals;
   struct Printer : public MachineFunctionPass {
+
     std::ostream &O;
     unsigned ConstIdx;
     Printer(std::ostream &o) : O(o), ConstIdx(0) {}
@@ -55,6 +59,49 @@
   return new Printer(O);
 }
 
+// We dont want identifier names with ., space, -  in them. 
+// So we replace them with _
+static std::string makeNameProper(std::string x) {
+  std::string tmp;
+  for (std::string::iterator sI = x.begin(), sEnd = x.end(); sI != sEnd; sI++)
+    switch (*sI) {
+    case '.': tmp += "d_"; break;
+    case ' ': tmp += "s_"; break;
+    case '-': tmp += "D_"; break;
+    default:  tmp += *sI;
+    }
+
+  return tmp;
+}
+
+std::string getValueName(const Value *V) {
+  if (V->hasName()) { // Print out the label if it exists...
+    
+    // Name mangling occurs as follows:
+    // - If V is not a global, mangling always occurs.
+    // - Otherwise, mangling occurs when any of the following are true:
+    //   1) V has internal linkage
+    //   2) V's name would collide if it is not mangled.
+    //
+    
+    if(const GlobalValue* gv = dyn_cast<GlobalValue>(V)) {
+      if(!gv->hasInternalLinkage() && !MangledGlobals.count(gv)) {
+        // No internal linkage, name will not collide -> no mangling.
+        return makeNameProper(gv->getName());
+      }
+    }
+    
+    // Non-global, or global with internal linkage / colliding name -> mangle.
+    return "l" + utostr(V->getType()->getUniqueID()) + "_" +
+      makeNameProper(V->getName());      
+  }
+
+  static int Count = 0;
+  Count++;
+  return "ltmp_" + itostr(Count) + "_" + utostr(V->getType()->getUniqueID());
+}
+
+
 // valToExprString - Helper function for ConstantExprToString().
 // Appends result to argument string S.
 // 
@@ -79,8 +126,7 @@
     else
       failed = true;
   } else if (const GlobalValue* GV = dyn_cast<GlobalValue>(V)) {
-    // S += getID(GV);
-    assert (0 && "getID not implemented");
+    S += getValueName(GV);
   }
   else
     failed = true;
@@ -200,9 +246,7 @@
     {
       // This is a constant address for a global variable or method.
       // Use the name of the variable or method as the address value.
-      // O << getID(CPR->getValue()) << "\n";
-      assert (0 && "getID not implemented");
-
+      O << getValueName(CPR->getValue()) << "\n";
     }
   else if (isa<ConstantPointerNull>(CV))
     {
@@ -370,9 +414,9 @@
   // Print out labels for the function.
   O << "\t.text\n";
   O << "\t.align 16\n";
-  O << "\t.globl\t" << MF.getFunction()->getName() << "\n";
-  O << "\t.type\t" << MF.getFunction()->getName() << ", @function\n";
-  O << MF.getFunction()->getName() << ":\n";
+  O << "\t.globl\t" << getValueName(MF.getFunction()) << "\n";
+  O << "\t.type\t" << getValueName(MF.getFunction()) << ", @function\n";
+  O << getValueName(MF.getFunction()) << ":\n";
 
   NumberForBB.clear();
   for (MachineFunction::const_iterator I = MF.begin(), E = MF.end();
@@ -437,7 +481,7 @@
       << MO.getVRegValue()->getName();
     return;
   case MachineOperand::MO_GlobalAddress:
-    if (!elideOffsetKeyword) O << "OFFSET "; O << MO.getGlobal()->getName();
+    if (!elideOffsetKeyword) O << "OFFSET "; O << getValueName(MO.getGlobal());
     return;
   case MachineOperand::MO_ExternalSymbol:
     O << MO.getSymbolName();
@@ -778,6 +822,27 @@
   // Tell gas we are outputting Intel syntax (not AT&T syntax) assembly,
   // with no % decorations on register names.
   O << "\t.intel_syntax noprefix\n";
+
+  // Ripped from CWriter:
+  // Calculate which global values have names that will collide when we throw
+  // away type information.
+  {  // Scope to delete the FoundNames set when we are done with it...
+    std::set<std::string> FoundNames;
+    for (Module::iterator I = M.begin(), E = M.end(); I != E; ++I)
+      if (I->hasName())                      // If the global has a name...
+        if (FoundNames.count(I->getName()))  // And the name is already used
+          MangledGlobals.insert(I);          // Mangle the name
+        else
+          FoundNames.insert(I->getName());   // Otherwise, keep track of name
+
+    for (Module::giterator I = M.gbegin(), E = M.gend(); I != E; ++I)
+      if (I->hasName())                      // If the global has a name...
+        if (FoundNames.count(I->getName()))  // And the name is already used
+          MangledGlobals.insert(I);          // Mangle the name
+        else
+          FoundNames.insert(I->getName());   // Otherwise, keep track of name
+  }
+
   return false; // success
 }
 
@@ -785,22 +850,24 @@
 {
   // Print out module-level global variables here.
   for (Module::const_giterator I = M.gbegin(), E = M.gend(); I != E; ++I) {
+    std::string name(getValueName(I));
     if (I->hasInitializer()) {
       Constant *C = I->getInitializer();
       O << "\t.data\n";
-      O << "\t.globl " << I->getName() << "\n";
-      O << "\t.type " << I->getName() << ", at object\n";
-      O << "\t.size " << I->getName() << ","
+      O << "\t.globl " << name << "\n";
+      O << "\t.type " << name << ", at object\n";
+      O << "\t.size " << name << ","
 	<< (unsigned)TD->getTypeSize(I->getType()) << "\n";
       O << "\t.align " << (unsigned)TD->getTypeAlignment(C->getType()) << "\n";
-      O << I->getName() << ":\t\t\t\t\t#" << *C << "\n";
+      O << name << ":\t\t\t\t\t#" << *C << "\n";
       printConstantValueOnly (C);
     } else {
-      O << "\t.globl " << I->getName() << "\n";
-      O << "\t.comm " << I->getName() << ", "
+      O << "\t.globl " << name << "\n";
+      O << "\t.comm " << name << ", "
         << (unsigned)TD->getTypeSize(I->getType()) << ", "
         << (unsigned)TD->getTypeAlignment(I->getType()) << "\n";
     }
   }
+  MangledGlobals.clear();
   return false; // success
 }




From lattner at cs.uiuc.edu  Wed Jun 25 23:21:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Jun 25 23:21:00 2003
Subject: [llvm-commits] CVS: llvm/test/Libraries/crtend/crtend.c
Message-ID: <200306260420.XAA13390@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Libraries/crtend:

crtend.c updated: 1.1 -> 1.2

---
Log message:

Don't register the destructors if the list is empty


---
Diffs of the changes:

Index: llvm/test/Libraries/crtend/crtend.c
diff -u llvm/test/Libraries/crtend/crtend.c:1.1 llvm/test/Libraries/crtend/crtend.c:1.2
--- llvm/test/Libraries/crtend/crtend.c:1.1	Wed May 14 08:09:41 2003
+++ llvm/test/Libraries/crtend/crtend.c	Wed Jun 25 23:20:38 2003
@@ -34,8 +34,12 @@
   /* Loop over all of the constructor records, calling each function pointer. */
   TorRec *R = __llvm_getGlobalCtors();
 
-  if (atexit(run_destructors))
-    abort();  /* Should be able to install ONE atexit handler! */
+  /* Only register the global dtor handler if there is at least one global
+   * dtor!
+   */
+  if (__llvm_getGlobalDtors()[0].FP)
+    if (atexit(run_destructors))
+      abort();  /* Should be able to install ONE atexit handler! */
 
   /* FIXME: This should sort the list by priority! */
   for (; R->FP; ++R)




From lattner at cs.uiuc.edu  Wed Jun 25 23:33:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Jun 25 23:33:00 2003
Subject: [llvm-commits] CVS: llvm/tools/gccld/gccld.cpp
Message-ID: <200306260432.XAA13730@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/gccld:

gccld.cpp updated: 1.35 -> 1.36

---
Log message:

Move the instcombine pass before globaldce, so that if globals are made dead by instcombine, that they can be deleted.


---
Diffs of the changes:

Index: llvm/tools/gccld/gccld.cpp
diff -u llvm/tools/gccld/gccld.cpp:1.35 llvm/tools/gccld/gccld.cpp:1.36
--- llvm/tools/gccld/gccld.cpp:1.35	Thu Jun 19 12:03:51 2003
+++ llvm/tools/gccld/gccld.cpp	Wed Jun 25 23:32:31 2003
@@ -400,14 +400,14 @@
   //
   Passes.add(createDeadArgEliminationPass());
 
-  // Now that we have optimized the program, discard unreachable functions...
-  //
-  Passes.add(createGlobalDCEPass());
-
   // The FuncResolve pass may leave cruft around if functions were prototyped
   // differently than they were defined.  Remove this cruft.
   //
   Passes.add(createInstructionCombiningPass());
+
+  // Now that we have optimized the program, discard unreachable functions...
+  //
+  Passes.add(createGlobalDCEPass());
 
   // Add the pass that writes bytecode to the output file...
   std::string RealBytecodeOutput = OutputFilename;




From lattner at cs.uiuc.edu  Thu Jun 26 00:07:02 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Jun 26 00:07:02 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Scalar/InstructionCombining.cpp
Message-ID: <200306260506.AAA14775@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Scalar:

InstructionCombining.cpp updated: 1.93 -> 1.94

---
Log message:

Add support for elimination of load instruction from global constants


---
Diffs of the changes:

Index: llvm/lib/Transforms/Scalar/InstructionCombining.cpp
diff -u llvm/lib/Transforms/Scalar/InstructionCombining.cpp:1.93 llvm/lib/Transforms/Scalar/InstructionCombining.cpp:1.94
--- llvm/lib/Transforms/Scalar/InstructionCombining.cpp:1.93	Wed Jun 25 12:09:20 2003
+++ llvm/lib/Transforms/Scalar/InstructionCombining.cpp	Thu Jun 26 00:06:25 2003
@@ -22,6 +22,7 @@
 #include "llvm/Constants.h"
 #include "llvm/ConstantHandling.h"
 #include "llvm/DerivedTypes.h"
+#include "llvm/GlobalVariable.h"
 #include "llvm/Support/InstIterator.h"
 #include "llvm/Support/InstVisitor.h"
 #include "llvm/Support/CallSite.h"
@@ -79,6 +80,7 @@
     Instruction *visitPHINode(PHINode &PN);
     Instruction *visitGetElementPtrInst(GetElementPtrInst &GEP);
     Instruction *visitAllocationInst(AllocationInst &AI);
+    Instruction *visitLoadInst(LoadInst &LI);
     Instruction *visitBranchInst(BranchInst &BI);
 
     // visitInstruction - Specify what to return for unhandled instructions...
@@ -1260,6 +1262,52 @@
     }
   return 0;
 }
+
+/// GetGEPGlobalInitializer - Given a constant, and a getelementptr
+/// constantexpr, return the constant value being addressed by the constant
+/// expression, or null if something is funny.
+///
+static Constant *GetGEPGlobalInitializer(Constant *C, ConstantExpr *CE) {
+  if (CE->getOperand(1) != Constant::getNullValue(Type::LongTy))
+    return 0;  // Do not allow stepping over the value!
+
+  // Loop over all of the operands, tracking down which value we are
+  // addressing...
+  for (unsigned i = 2, e = CE->getNumOperands(); i != e; ++i)
+    if (ConstantUInt *CU = dyn_cast<ConstantUInt>(CE->getOperand(i))) {
+      ConstantStruct *CS = cast<ConstantStruct>(C);
+      if (CU->getValue() >= CS->getValues().size()) return 0;
+      C = cast<Constant>(CS->getValues()[CU->getValue()]);
+    } else if (ConstantSInt *CS = dyn_cast<ConstantSInt>(CE->getOperand(i))) {
+      ConstantArray *CA = cast<ConstantArray>(C);
+      if ((uint64_t)CS->getValue() >= CA->getValues().size()) return 0;
+      C = cast<Constant>(CA->getValues()[CS->getValue()]);
+    } else 
+      return 0;
+  return C;
+}
+
+Instruction *InstCombiner::visitLoadInst(LoadInst &LI) {
+  Value *Op = LI.getOperand(0);
+  if (ConstantPointerRef *CPR = dyn_cast<ConstantPointerRef>(Op))
+    Op = CPR->getValue();
+
+  // Instcombine load (constant global) into the value loaded...
+  if (GlobalVariable *GV = dyn_cast<GlobalVariable>(Op))
+    if (GV->isConstant())
+      return ReplaceInstUsesWith(LI, GV->getInitializer());
+
+  // Instcombine load (constantexpr_GEP global, 0, ...) into the value loaded...
+  if (ConstantExpr *CE = dyn_cast<ConstantExpr>(Op))
+    if (CE->getOpcode() == Instruction::GetElementPtr)
+      if (ConstantPointerRef *G=dyn_cast<ConstantPointerRef>(CE->getOperand(0)))
+        if (GlobalVariable *GV = dyn_cast<GlobalVariable>(G->getValue()))
+          if (GV->isConstant())
+            if (Constant *V = GetGEPGlobalInitializer(GV->getInitializer(), CE))
+              return ReplaceInstUsesWith(LI, V);
+  return 0;
+}
+
 
 Instruction *InstCombiner::visitBranchInst(BranchInst &BI) {
   // Change br (not X), label True, label False to: br X, label False, True




From lattner at cs.uiuc.edu  Thu Jun 26 00:07:10 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Jun 26 00:07:10 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/InstCombine/load.ll
Message-ID: <200306260506.AAA14731@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/InstCombine:

load.ll added (r1.1)

---
Log message:

add tests for elimination of loads



---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/InstCombine/load.ll
diff -c /dev/null llvm/test/Regression/Transforms/InstCombine/load.ll:1.1
*** /dev/null	Thu Jun 26 00:06:01 2003
--- llvm/test/Regression/Transforms/InstCombine/load.ll	Thu Jun 26 00:05:51 2003
***************
*** 0 ****
--- 1,29 ----
+ ; This test makes sure that these instructions are properly eliminated.
+ ;
+ 
+ ; RUN: if as < %s | opt -instcombine | dis | grep load
+ ; RUN: then exit 1
+ ; RUN: else exit 0
+ ; RUN: fi
+ 
+ %X = constant int 42
+ %Y = constant [2 x { int, float }] [ { int, float } { int 12, float 1.0 }, 
+                                      { int, float } { int 37, float 1.2312 } ]
+ int %test1() {
+ 	%B = load int* %X
+ 	ret int %B
+ }
+ 
+ float %test2() {
+ 	%A = getelementptr [2 x { int, float}]* %Y, long 0, long 1, ubyte 1
+ 	%B = load float* %A
+ 	ret float %B
+ }
+ 
+ int %test3() {
+ 	%A = getelementptr [2 x { int, float}]* %Y, long 0, long 0, ubyte 0
+ 	%B = load int* %A
+ 	ret int %B
+ }
+ 
+ 




From lattner at cs.uiuc.edu  Thu Jun 26 00:23:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Jun 26 00:23:01 2003
Subject: [llvm-commits] CVS: llvm/lib/VMCore/ConstantHandling.cpp
Message-ID: <200306260522.AAA15542@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/VMCore:

ConstantHandling.cpp updated: 1.31 -> 1.32

---
Log message:

Implement more aggressive folding of constant GEP instructions


---
Diffs of the changes:

Index: llvm/lib/VMCore/ConstantHandling.cpp
diff -u llvm/lib/VMCore/ConstantHandling.cpp:1.31 llvm/lib/VMCore/ConstantHandling.cpp:1.32
--- llvm/lib/VMCore/ConstantHandling.cpp:1.31	Tue May 27 14:16:07 2003
+++ llvm/lib/VMCore/ConstantHandling.cpp	Thu Jun 26 00:22:45 2003
@@ -141,12 +141,30 @@
 
   // FIXME: Implement folding of GEP constant exprs the same as instcombine does
 
-  // Implement folding of:
-  //    int* getelementptr ([2 x int]* cast ([3 x int]* %X to [2 x int]*),
-  //                        long 0, long 0)
-  // To: int* getelementptr ([3 x int]* %X, long 0, long 0)
-  //
-  if (const ConstantExpr *CE = dyn_cast<ConstantExpr>(C))
+  if (const ConstantExpr *CE = dyn_cast<ConstantExpr>(C)) {
+    // Implement folding of:
+    //    void ()** getelementptr (%struct..TorRec* getelementptr
+    //              ([N x %struct..TorRec]* %llvm.global_dtors, long 0, long 0),
+    //                 long 0, ubyte 1)
+    // Into:
+    //    %struct..TorRec* getelementptr ([N x %struct..TorRec]*
+    //                      %llvm.global_dtors, long 0, long 0, ubyte 1)
+    //
+    if (CE->getOpcode() == Instruction::GetElementPtr)
+      if (IdxList[0] == Constant::getNullValue(Type::LongTy)) {
+        std::vector<Constant*> NewIndices;
+        NewIndices.reserve(IdxList.size() + CE->getNumOperands());
+        for (unsigned i = 1, e = CE->getNumOperands(); i != e; ++i)
+          NewIndices.push_back(cast<Constant>(CE->getOperand(i)));
+        NewIndices.insert(NewIndices.end(), IdxList.begin()+1, IdxList.end());
+        return ConstantExpr::getGetElementPtr(CE->getOperand(0), NewIndices);
+      }
+
+    // Implement folding of:
+    //    int* getelementptr ([2 x int]* cast ([3 x int]* %X to [2 x int]*),
+    //                        long 0, long 0)
+    // To: int* getelementptr ([3 x int]* %X, long 0, long 0)
+    //
     if (CE->getOpcode() == Instruction::Cast && IdxList.size() > 1 &&
         IdxList[0]->isNullValue())
       if (const PointerType *SPT = 
@@ -157,6 +175,7 @@
             if (CAT->getElementType() == SAT->getElementType())
               return ConstantExpr::getGetElementPtr(
                       (Constant*)CE->getOperand(0), IdxList);
+  }
   return 0;
 }
 




From lattner at cs.uiuc.edu  Thu Jun 26 00:29:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Jun 26 00:29:01 2003
Subject: [llvm-commits] CVS: llvm/test/Programs/Makefile.programs
Message-ID: <200306260528.AAA15987@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Programs:

Makefile.programs updated: 1.60 -> 1.61

---
Log message:

Add dependency on GCCLD


---
Diffs of the changes:

Index: llvm/test/Programs/Makefile.programs
diff -u llvm/test/Programs/Makefile.programs:1.60 llvm/test/Programs/Makefile.programs:1.61
--- llvm/test/Programs/Makefile.programs:1.60	Sun Jun 22 15:21:29 2003
+++ llvm/test/Programs/Makefile.programs	Thu Jun 26 00:28:23 2003
@@ -176,10 +176,10 @@
 # Link the program to the libraries it uses, then perform postlink
 # optimization...
 $(PROGRAMS_TO_TEST:%=Output/%.llvm.bc): \
-Output/%.llvm.bc: Output/%.linked.bc
+Output/%.llvm.bc: Output/%.linked.bc $(LGCCLDPROG)
 	$(LGCCLD) $(STATS) $< -lgcc -lc $(LIBS) crtend.o -o Output/$*.llvm
 $(PROGRAMS_TO_TEST:%=Output/%.llvm): \
-Output/%.llvm: Output/%.linked.bc
+Output/%.llvm: Output/%.linked.bc $(LGCCLDPROG)
 	$(LGCCLD) $(STATS) $< -lgcc -lc $(LIBS) crtend.o -o Output/$*.llvm
 
 # Rule to get the tracing version of the llvm.bc file for tracing:




From lattner at cs.uiuc.edu  Thu Jun 26 00:29:08 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Jun 26 00:29:08 2003
Subject: [llvm-commits] CVS: llvm/test/Makefile.tests
Message-ID: <200306260528.AAA15879@apoc.cs.uiuc.edu>


Changes in directory llvm/test:

Makefile.tests updated: 1.65 -> 1.66

---
Log message:

New variable which is just the GCCLD tool


---
Diffs of the changes:

Index: llvm/test/Makefile.tests
diff -u llvm/test/Makefile.tests:1.65 llvm/test/Makefile.tests:1.66
--- llvm/test/Makefile.tests:1.65	Wed Jun 18 14:39:06 2003
+++ llvm/test/Makefile.tests	Thu Jun 26 00:28:00 2003
@@ -43,6 +43,7 @@
 LAS      = $(TOOLS)/as
 LGCCAS   = $(TOOLS)/gccas
 LGCCLD   = $(TOOLS)/gccld -L$(LLVMGCCDIR)/lib/gcc-lib/$(LLVMGCCARCH) -L$(LLVMGCCDIR)/lib
+LGCCLDPROG = $(TOOLS)/gccld
 LDIS     = $(TOOLS)/dis 
 LOPT     = $(TOOLS)/opt
 LLINK    = $(TOOLS)/link




From lattner at cs.uiuc.edu  Thu Jun 26 00:31:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Jun 26 00:31:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/IPO/Internalize.cpp
Message-ID: <200306260530.AAA16245@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/IPO:

Internalize.cpp updated: 1.16 -> 1.17

---
Log message:

When internalizing global ctor/dtor list, also mark it constant.  This is gross, but 
until DSA is working all of the time and is totally reliable, we do this.


---
Diffs of the changes:

Index: llvm/lib/Transforms/IPO/Internalize.cpp
diff -u llvm/lib/Transforms/IPO/Internalize.cpp:1.16 llvm/lib/Transforms/IPO/Internalize.cpp:1.17
--- llvm/lib/Transforms/IPO/Internalize.cpp:1.16	Thu May 22 15:27:13 2003
+++ llvm/lib/Transforms/IPO/Internalize.cpp	Thu Jun 26 00:30:40 2003
@@ -87,6 +87,14 @@
       for (Module::giterator I = M.gbegin(), E = M.gend(); I != E; ++I)
         if (!I->isExternal() && !I->hasInternalLinkage() &&
             !ExternalNames.count(I->getName())) {
+          // Special case handling of the global ctor and dtor list.  When we
+          // internalize it, we mark it constant, which allows elimination of
+          // the list if it's empty.
+          //
+          if (I->hasAppendingLinkage() && (I->getName() == "llvm.global_ctors"||
+                                           I->getName() == "llvm.global_dtors"))
+            I->setConstant(true);
+
           I->setLinkage(GlobalValue::InternalLinkage);
           Changed = true;
           ++NumGlobals;




From lattner at cs.uiuc.edu  Thu Jun 26 00:31:06 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Jun 26 00:31:06 2003
Subject: [llvm-commits] CVS: llvm/tools/gccld/gccld.cpp
Message-ID: <200306260530.AAA16234@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/gccld:

gccld.cpp updated: 1.36 -> 1.37

---
Log message:

Run the simplify CFG pass after instcombine which has the effect of deleting 
ALL of the global ctor/dtor stuff if it is not used!


---
Diffs of the changes:

Index: llvm/tools/gccld/gccld.cpp
diff -u llvm/tools/gccld/gccld.cpp:1.36 llvm/tools/gccld/gccld.cpp:1.37
--- llvm/tools/gccld/gccld.cpp:1.36	Wed Jun 25 23:32:31 2003
+++ llvm/tools/gccld/gccld.cpp	Thu Jun 26 00:29:50 2003
@@ -405,6 +405,10 @@
   //
   Passes.add(createInstructionCombiningPass());
 
+  // Delete basic blocks, which optimization passes may have killed...
+  //
+  Passes.add(createCFGSimplificationPass());
+
   // Now that we have optimized the program, discard unreachable functions...
   //
   Passes.add(createGlobalDCEPass());




From lattner at cs.uiuc.edu  Thu Jun 26 00:42:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Jun 26 00:42:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/IPO/GlobalDCE.cpp
Message-ID: <200306260541.AAA17393@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/IPO:

GlobalDCE.cpp updated: 1.23 -> 1.24

---
Log message:

Add support to globaldce for deleting dead function prototypes


---
Diffs of the changes:

Index: llvm/lib/Transforms/IPO/GlobalDCE.cpp
diff -u llvm/lib/Transforms/IPO/GlobalDCE.cpp:1.23 llvm/lib/Transforms/IPO/GlobalDCE.cpp:1.24
--- llvm/lib/Transforms/IPO/GlobalDCE.cpp:1.23	Sun Oct 13 12:12:47 2002
+++ llvm/lib/Transforms/IPO/GlobalDCE.cpp	Thu Jun 26 00:41:18 2003
@@ -46,12 +46,18 @@
     // Nothing to do if no unreachable functions have been found...
     if (FunctionsToDelete.empty()) return false;
     
-    // Unreachables functions have been found and should have no references to
+    // Unreachable functions have been found and should have no references to
     // them, delete them now.
     //
     for (std::vector<CallGraphNode*>::iterator I = FunctionsToDelete.begin(),
            E = FunctionsToDelete.end(); I != E; ++I)
       delete CallGraph.removeFunctionFromModule(*I);
+
+    // Walk the function list, removing prototypes for functions which are not
+    // used.
+    for (Module::iterator I = M.begin(), E = M.end(); I != E; ++I)
+      if (I->use_size() == 0 && I->isExternal())
+        delete CallGraph.removeFunctionFromModule(I);
     
     return true;
   }




From lattner at cs.uiuc.edu  Thu Jun 26 09:57:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Jun 26 09:57:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/C++Frontend/ConstructorDestructorAttributes.c
Message-ID: <200306261456.JAA21028@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/C++Frontend:

ConstructorDestructorAttributes.c added (r1.1)

---
Log message:

Add a testcase to ensure that ctor/dtor attributes work in C


---
Diffs of the changes:

Index: llvm/test/Regression/C++Frontend/ConstructorDestructorAttributes.c
diff -c /dev/null llvm/test/Regression/C++Frontend/ConstructorDestructorAttributes.c:1.1
*** /dev/null	Thu Jun 26 09:56:47 2003
--- llvm/test/Regression/C++Frontend/ConstructorDestructorAttributes.c	Thu Jun 26 09:56:36 2003
***************
*** 0 ****
--- 1,14 ----
+ #include <stdio.h>
+ 
+ void ctor() __attribute__((constructor));
+ 
+ void ctor() {
+    printf("Create!\n");
+ }
+ void dtor() __attribute__((destructor));
+ 
+ void dtor() {
+    printf("Create!\n");
+ }
+ 
+ int main() { return 0; }




From criswell at choi.cs.uiuc.edu  Thu Jun 26 10:52:01 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 10:52:01 2003
Subject: [llvm-commits] CVS: llvm/samples/README sample.tar
Message-ID: <200306261551.h5QFpcf31919@choi.cs.uiuc.edu>


Changes in directory llvm/samples:

README added (r1.1.2.1)
sample.tar added (r1.1.2.1)

---
Log message:

Add a sample project tree as a tar archive.


---
Diffs of the changes:







From criswell at choi.cs.uiuc.edu  Thu Jun 26 10:52:06 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 10:52:06 2003
Subject: [llvm-commits] CVS: llvm/samples/
Message-ID: <200306261551.h5QFpB231900@choi.cs.uiuc.edu>


Changes in directory llvm/samples:


---
Log message:

Directory /home/vadve/vadve/Research/DynOpt/CVSRepository/llvm/samples added to the repository
--> Using per-directory sticky tag `autoconf'


---
Diffs of the changes:



From criswell at choi.cs.uiuc.edu  Thu Jun 26 10:55:01 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 10:55:01 2003
Subject: [llvm-commits] CVS: llvm/Makefile.common
Message-ID: <200306261554.h5QFs6Z31939@choi.cs.uiuc.edu>


Changes in directory llvm:

Makefile.common updated: 1.88.2.4 -> 1.88.2.5

---
Log message:

Fixed the rules for building dynamic objects.  This involved making each shell
command its own line (since the $VERB was causing /bin/sh grief).
Also modified the makefile so that the target all builds a dynamically linked
library if SHARED_LIBRARY is defined.  This fixes a problems where recursive
builds didn't build the library.
Fixed the distclean target so that it cleans LLVM proper instead of your
projects directory.



---
Diffs of the changes:

Index: llvm/Makefile.common
diff -u llvm/Makefile.common:1.88.2.4 llvm/Makefile.common:1.88.2.5
--- llvm/Makefile.common:1.88.2.4	Wed Jun 25 13:43:32 2003
+++ llvm/Makefile.common	Thu Jun 26 10:53:56 2003
@@ -130,7 +130,7 @@
 
 ifdef SHARED_LIBRARY
 # if SHARED_LIBRARY is specified, the default is to build the dynamic lib
-dynamic ::
+all:: dynamic
 endif
 
 # Default Rule:  Make sure it's also a :: rule
@@ -467,35 +467,35 @@
 # Rules for building dynamically linked libraries.
 #
 $(LIBNAME_O): $(LObjectsO) $(LibSubDirs) $(DESTLIBRELEASE)/.dir
-	@echo ======= Linking $(LIBRARYNAME) release library =======
-	$(VERB) $(Link) -o $*.la $(LObjectsO) $(LibSubDirs) $(LibLinkOpts); \
+	@echo ======= Linking $(LIBRARYNAME) dynamic release library =======
+	$(VERB) $(Link) -o $*.la $(LObjectsO) $(LibSubDirs) $(LibLinkOpts);
 	$(VERB) $(LIBTOOL) --mode=install $(INSTALL) $*.la $(DESTLIBCURRENT);
 
 $(LIBNAME_P): $(LObjectsP) $(LibSubDirs) $(DESTLIBPROFILE)/.dir
-	@echo ======= Linking $(LIBRARYNAME) profile library =======
-	$(VERB) $(Link) -o $*.la $(LObjectsP) $(LibSubDirs) $(LibLinkOpts); \
+	@echo ======= Linking $(LIBRARYNAME) dynamic profile library =======
+	$(VERB) $(Link) -o $*.la $(LObjectsP) $(LibSubDirs) $(LibLinkOpts);
 	$(VERB) $(LIBTOOL) --mode=install $(INSTALL) $*.la $(DESTLIBCURRENT);
 
 $(LIBNAME_G): $(LObjectsG) $(LibSubDirs) $(DESTLIBDEBUG)/.dir
-	@echo ======= Linking $(LIBRARYNAME) debug library =======
-	$(VERB) $(Link) -o $*.la $(LObjectsG) $(LibSubDirs) $(LibLinkOpts); \
+	@echo ======= Linking $(LIBRARYNAME) dynamic debug library =======
+	$(VERB) $(Link) -o $*.la $(LObjectsG) $(LibSubDirs) $(LibLinkOpts);
 	$(VERB) $(LIBTOOL) --mode=install $(INSTALL) $*.la $(DESTLIBCURRENT);
 
 #
 # Rules for building static archive libraries.
 #
 $(LIBNAME_AO): $(LObjectsO) $(LibSubDirs) $(DESTLIBRELEASE)/.dir
-	@echo ======= Linking $(LIBRARYNAME) release library =======
+	@echo ======= Linking $(LIBRARYNAME) archive release library =======
 	@$(RM) -f $@
 	$(VERB) $(Link) -03 -o $@ $(LObjectsO) $(LibSubDirs) -static
 
 $(LIBNAME_AP): $(LObjectsP) $(LibSubDirs) $(DESTLIBPROFILE)/.dir
-	@echo ======= Linking $(LIBRARYNAME) profile library =======
+	@echo ======= Linking $(LIBRARYNAME) archive profile library =======
 	@$(RM) -f $@
 	$(VERB) $(Link) -03 $(PROFILE) -o $@ $(LObjectsP) $(LibSubDirs) -static
 
 $(LIBNAME_AG): $(ObjectsG) $(LibSubDirs) $(DESTLIBDEBUG)/.dir
-	@echo ======= Linking $(LIBRARYNAME) debug library =======
+	@echo ======= Linking $(LIBRARYNAME) archive debug library =======
 	@$(RM) -f $@
 	$(VERB) $(Link) -g $(STRIP) -o $@ $(LObjectsG) $(LibSubDirs) -static
 
@@ -712,10 +712,10 @@
 	$(VERB) $(RM) -f $(LEX_OUTPUT) $(YACC_OUTPUT)
 
 distclean:: clean
-	$(VERB) $(RM) -rf $(LEVEL)/Makefile.config \
-	                  $(LEVEL)/include/Config/config.h \
-	                  $(LEVEL)/autom4te.cache \
-	                  $(LEVEL)/config.log
+	$(VERB) (cd $(LLVM_SRC_ROOT); $(RM) -rf $(LEVEL)/Makefile.config \
+	            $(LEVEL)/include/Config/config.h \
+	            $(LEVEL)/autom4te.cache \
+	            $(LEVEL)/config.log)
 
 ###########################################################################
 # C/C++ Dependencies




From criswell at choi.cs.uiuc.edu  Thu Jun 26 10:56:02 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 10:56:02 2003
Subject: [llvm-commits] CVS: llvm/www/docs/GettingStarted.html
Message-ID: <200306261555.h5QFt1a31966@choi.cs.uiuc.edu>


Changes in directory llvm/www/docs:

GettingStarted.html updated: 1.12.2.1 -> 1.12.2.2

---
Log message:

Fixed some small typos.
Added a section on disk usage requirements.


---
Diffs of the changes:

Index: llvm/www/docs/GettingStarted.html
diff -u llvm/www/docs/GettingStarted.html:1.12.2.1 llvm/www/docs/GettingStarted.html:1.12.2.2
--- llvm/www/docs/GettingStarted.html:1.12.2.1	Wed Jun 25 16:44:08 2003
+++ llvm/www/docs/GettingStarted.html	Thu Jun 26 10:54:51 2003
@@ -20,6 +20,10 @@
       <li><a href="#starting">Getting started with LLVM</a>
         <ol>
           <li><a href="#requirements">Requirements</a>
+          <ol>
+              <li><a href="#hardware">Hardware</a>
+              <li><a href="#software">Software</a>
+          </ol>
           <li><a href="#quickstart">Getting started quickly (a summary)</a>
           <li><a href="#checkout">Checkout LLVM from CVS</a>
           <li><a href="#terminology">Terminology and Notation</tt></a>
@@ -71,14 +75,39 @@
     <h3><a name="requirements"><b>Requirements</b></a></h3>
     <!--=====================================================================-->
 
+    <!--=====================================================================-->
+    <h4><a name="hardware"><b>Hardware</b></a></h4>
+    <!--=====================================================================-->
     LLVM is known to work on the following platforms:
     <ul>
-        <li> Linux/x86
-        <li> Solaris/Sparc
+        <li> Linux on x86
+        <ul>
+            <li> Approximately 700 MB of Free Disk Space
+            <ul>
+                <li>Source code: 30 MB
+                <li>Object code: 670 MB
+            </ul>
+        </ul>
+        <li> Solaris on Sparc
+        <ul>
+            <li> Approximately 1.3 GB of Free Disk Space
+            <ul>
+                <li>Source code: 30 MB
+                <li>Object code: 1000 MB
+            </ul>
+        </ul>
     </ul>
 
+    LLVM may compile on other platforms.  While the LLVM utilities should work,
+    they will only generate Sparc or x86 machine code.
+
+    <!--=====================================================================-->
+    <h4><a name="software"><b>Software</b></a></h4>
+    <!--=====================================================================-->
+    <p>
     Compiling LLVM requires that you have several different software packages
     installed:
+
     <ul>
         <li> GCC
         <p>
@@ -93,7 +122,7 @@
 
         <li> GNU Make
         <p>
-        The LLVM build system relies upon GNU make extensions.  Therefore, you
+        The LLVM build system relies upon GNU Make extensions.  Therefore, you
         will need GNU Make (sometimes known as gmake) to build LLVM.
         </p>
 
@@ -115,7 +144,7 @@
         If you want to make changes to the autoconf scripts which configure LLVM
         for compilation, you will need GNU autoconf, and consequently, GNU M4.
         LLVM was built with autoconf 2.53, so that release and any later
-        releases should work.
+        release should work.
         </p>
     </ul>
 




From brukman at cs.uiuc.edu  Thu Jun 26 11:17:00 2003
From: brukman at cs.uiuc.edu (Misha Brukman)
Date: Thu Jun 26 11:17:00 2003
Subject: [llvm-commits] CVS: llvm/www/docs/GettingStarted.html
Message-ID: <200306261616.LAA04412@zion.cs.uiuc.edu>


Changes in directory llvm/www/docs:

GettingStarted.html updated: 1.12.2.2 -> 1.12.2.3

---
Log message:

1000 MB + 30 MB = 1.03 GB, not 1.3 


---
Diffs of the changes:

Index: llvm/www/docs/GettingStarted.html
diff -u llvm/www/docs/GettingStarted.html:1.12.2.2 llvm/www/docs/GettingStarted.html:1.12.2.3
--- llvm/www/docs/GettingStarted.html:1.12.2.2	Thu Jun 26 10:54:51 2003
+++ llvm/www/docs/GettingStarted.html	Thu Jun 26 11:15:54 2003
@@ -90,7 +90,7 @@
         </ul>
         <li> Solaris on Sparc
         <ul>
-            <li> Approximately 1.3 GB of Free Disk Space
+            <li> Approximately 1.03 GB of Free Disk Space
             <ul>
                 <li>Source code: 30 MB
                 <li>Object code: 1000 MB




From jstanley at cs.uiuc.edu  Thu Jun 26 11:47:01 2003
From: jstanley at cs.uiuc.edu (Joel Stanley)
Date: Thu Jun 26 11:47:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Reoptimizer/Inst/lib/Phase1/Phase1.cpp
Message-ID: <200306261646.LAA00466@trinity.cs.uiuc.edu>


Changes in directory llvm/lib/Reoptimizer/Inst/lib/Phase1:

Phase1.cpp updated: 1.27 -> 1.28

---
Log message:



---
Diffs of the changes:

Index: llvm/lib/Reoptimizer/Inst/lib/Phase1/Phase1.cpp
diff -u llvm/lib/Reoptimizer/Inst/lib/Phase1/Phase1.cpp:1.27 llvm/lib/Reoptimizer/Inst/lib/Phase1/Phase1.cpp:1.28
--- llvm/lib/Reoptimizer/Inst/lib/Phase1/Phase1.cpp:1.27	Tue Jun 24 10:39:53 2003
+++ llvm/lib/Reoptimizer/Inst/lib/Phase1/Phase1.cpp	Thu Jun 26 11:45:59 2003
@@ -194,8 +194,20 @@
     // be the pair of interval-declaration sigfuns.  TODO: Clean up this approach, and do
     // the same for point-site declarations.
 
-    Function* intervalStartDeclFunc = m_module->getNamedFunction("pp_declareIntervalStart");
-    Function* intervalEndDeclFunc = m_module->getNamedFunction("pp_declareIntervalEnd");
+    params.push_back(PointerType::get(Type::DoubleTy));
+    params.push_back(Type::UIntTy);
+    //Function* intervalStartDeclFunc = m_module->getNamedFunction("pp_declareIntervalStart");
+
+    Function* intervalStartDeclFunc =
+        m_module->getOrInsertFunction("pp_declareIntervalStart",
+                                      FunctionType::get(Type::VoidTy, params, false));
+        
+    params.clear();
+    params.push_back(Type::UIntTy);
+    params.push_back(PointerType::get(Type::DoubleTy));
+    Function* intervalEndDeclFunc =
+        m_module->getOrInsertFunction("pp_declareIntervalEnd",
+                                      FunctionType::get(Type::VoidTy, params, false));
 
     rpairs.push_back(std::make_pair(intervalStartDeclFunc, intervalEndDeclFunc));
 
@@ -272,6 +284,7 @@
     Function* startFunc = rpair.first;
     Function* endFunc = rpair.second;
 
+    assert(startFunc && endFunc && "Invalid start/end Function*'s");
     assert(startFunc->use_size() == endFunc->use_size() &&
            "Expect region functions to be used in pairs");
 




From gaeke at cs.uiuc.edu  Thu Jun 26 13:03:01 2003
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Thu Jun 26 13:03:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Target/X86/Printer.cpp
Message-ID: <200306261802.NAA13206@neo.cs.uiuc.edu>


Changes in directory llvm/lib/Target/X86:

Printer.cpp updated: 1.39 -> 1.40

---
Log message:

Number constants from constant pool as CPIf_i where f is the function index
and i is the constant pool index.


---
Diffs of the changes:

Index: llvm/lib/Target/X86/Printer.cpp
diff -u llvm/lib/Target/X86/Printer.cpp:1.39 llvm/lib/Target/X86/Printer.cpp:1.40
--- llvm/lib/Target/X86/Printer.cpp:1.39	Wed Jun 25 17:00:39 2003
+++ llvm/lib/Target/X86/Printer.cpp	Thu Jun 26 13:02:30 2003
@@ -24,16 +24,12 @@
 #include "llvm/Module.h"
 
 namespace {
-
-
+  unsigned fnIndex;
   std::set<const Value*> MangledGlobals;
   struct Printer : public MachineFunctionPass {
-
     std::ostream &O;
-    unsigned ConstIdx;
-    Printer(std::ostream &o) : O(o), ConstIdx(0) {}
+    Printer(std::ostream &o) : O(o) {}
     const TargetData *TD;
-
     virtual const char *getPassName() const {
       return "X86 Assembly Printer";
     }
@@ -393,11 +389,11 @@
 
   for (unsigned i = 0, e = CP.size(); i != e; ++i) {
     O << "\t.section .rodata\n";
-    O << "\t.align " << (unsigned)TD->getTypeAlignment(CP[i]->getType()) << "\n";
-    O << ".CPI" << i+ConstIdx << ":\t\t\t\t\t#" << *CP[i] << "\n";
+    O << "\t.align " << (unsigned)TD->getTypeAlignment(CP[i]->getType())
+      << "\n";
+    O << ".CPI" << fnIndex << "_" << i << ":\t\t\t\t\t#" << *CP[i] << "\n";
     printConstantValueOnly (CP[i]);
   }
-  ConstIdx += CP.size();  // Don't recycle constant pool index numbers
 }
 
 /// runOnMachineFunction - This uses the X86InstructionInfo::print method
@@ -438,6 +434,7 @@
     }
   }
 
+  fnIndex++;
   // We didn't modify anything.
   return false;
 }
@@ -515,7 +512,8 @@
     O << "]";
     return;
   } else if (MI->getOperand(Op).isConstantPoolIndex()) {
-    O << "[.CPI" << MI->getOperand(Op).getConstantPoolIndex();
+    O << "[.CPI" << fnIndex << "_"
+      << MI->getOperand(Op).getConstantPoolIndex();
     if (MI->getOperand(Op+3).getImmedValue())
       O << " + " << MI->getOperand(Op+3).getImmedValue();
     O << "]";
@@ -822,6 +820,9 @@
   // Tell gas we are outputting Intel syntax (not AT&T syntax) assembly,
   // with no % decorations on register names.
   O << "\t.intel_syntax noprefix\n";
+
+  // Start function index at 0
+  fnIndex = 0;
 
   // Ripped from CWriter:
   // Calculate which global values have names that will collide when we throw




From lattner at cs.uiuc.edu  Thu Jun 26 15:12:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Jun 26 15:12:01 2003
Subject: [llvm-commits] CVS: llvm/utils/NightlyTest.pl
Message-ID: <200306262011.PAA32729@zion.cs.uiuc.edu>


Changes in directory llvm/utils:

NightlyTest.pl updated: 1.13 -> 1.14

---
Log message:

Modify script to support checkins on branches


---
Diffs of the changes:

Index: llvm/utils/NightlyTest.pl
diff -u llvm/utils/NightlyTest.pl:1.13 llvm/utils/NightlyTest.pl:1.14
--- llvm/utils/NightlyTest.pl:1.13	Wed Jun 18 13:26:22 2003
+++ llvm/utils/NightlyTest.pl	Thu Jun 26 15:11:07 2003
@@ -229,7 +229,7 @@
 # Loop over every record from the CVS history, filling in the hashes.
 foreach $File (@CVSHistory) {
   my ($Type, $Date, $UID, $Rev, $Filename);
-  if ($File =~ /([AMRUGC])\s($DateRE)\s([^\s]+)\s+([0-9.]+)\s+([^\s]+)\s+([^\s]+)/) {
+  if ($File =~ /([AMRUGC])\s($DateRE)\s([^\s].......) (.[0-9.]*)\s+([^\s]+)\s+([^\s]+)/) {
     ($Type, $Date, $UID, $Rev, $Filename) = ($1, $2, $3, $4, "$6/$5");
   } elsif ($File =~ /([W])\s($DateRE)\s([^\s]+) +([^\s]+)\s+([^\s]+)/) {
     ($Type, $Date, $UID, $Rev, $Filename) = ($1, $2, $3, $4, "$6/$5");




From lattner at cs.uiuc.edu  Thu Jun 26 15:27:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Jun 26 15:27:01 2003
Subject: [llvm-commits] CVS: llvm/test/Programs/TEST.nightly.Makefile
Message-ID: <200306262026.PAA23993@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Programs:

TEST.nightly.Makefile updated: 1.18 -> 1.19

---
Log message:

Count size of llvm.bc file, not gccas output


---
Diffs of the changes:

Index: llvm/test/Programs/TEST.nightly.Makefile
diff -u llvm/test/Programs/TEST.nightly.Makefile:1.18 llvm/test/Programs/TEST.nightly.Makefile:1.19
--- llvm/test/Programs/TEST.nightly.Makefile:1.18	Wed Jun 25 12:29:27 2003
+++ llvm/test/Programs/TEST.nightly.Makefile	Thu Jun 26 15:26:14 2003
@@ -30,10 +30,11 @@
 	  grep "Total Execution Time" $@.info >> $@;\
 	  echo >> $@;\
 	  echo -n "TEST-RESULT-compile: " >> $@;\
-	  grep "Number of bytecode bytes written" $@.info >> $@;\
+	  wc -c $< >> $@;\
 	  echo >> $@;\
+	  $(LANALYZE) -stats -instcount $< -info-output-file=Output/$*.ic;\
 	  echo -n "TEST-RESULT-compile: " >> $@;\
-	  grep 'Number of instructions$$' $@.info >> $@;\
+	  grep 'Number of instructions (of all types)$$' Output/$*.ic >> $@;\
 	  echo >> $@;\
 	else \
 	  echo "TEST-FAIL: compile $(RELDIR)/$*" >> $@;\




From lattner at cs.uiuc.edu  Thu Jun 26 15:27:06 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Jun 26 15:27:06 2003
Subject: [llvm-commits] CVS: llvm/test/Programs/TEST.nightly.report
Message-ID: <200306262026.PAA23988@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Programs:

TEST.nightly.report updated: 1.7 -> 1.8

---
Log message:

Count size of llvm.bc file, not gccas output


---
Diffs of the changes:

Index: llvm/test/Programs/TEST.nightly.report
diff -u llvm/test/Programs/TEST.nightly.report:1.7 llvm/test/Programs/TEST.nightly.report:1.8
--- llvm/test/Programs/TEST.nightly.report:1.7	Sat Jun 21 15:09:13 2003
+++ llvm/test/Programs/TEST.nightly.report	Thu Jun 26 15:26:25 2003
@@ -18,7 +18,7 @@
  [],
 # Times
  ["GCCAS"    , "TEST-RESULT-compile: $WallTimeRE"],
- ["Bytecode" , 'TEST-RESULT-compile: *([0-9]+).*Number of bytecode bytes'],
+ ["Bytecode" , 'TEST-RESULT-compile: *([0-9]+)'],
  ["Instrs"   , 'TEST-RESULT-compile: *([0-9]+).*Number of inst'],
  ["LLC"      , "TEST-RESULT-llc: $WallTimeRE"],
  [],




From gaeke at cs.uiuc.edu  Thu Jun 26 15:38:01 2003
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Thu Jun 26 15:38:01 2003
Subject: [llvm-commits] CVS: llvm/www/docs/HistoricalNotes/2003-06-25-Reoptimizer1.txt 2003-06-26-Reoptimizer2.txt
Message-ID: <200306262037.PAA16214@neo.cs.uiuc.edu>


Changes in directory llvm/www/docs/HistoricalNotes:

2003-06-25-Reoptimizer1.txt added (r1.1)
2003-06-26-Reoptimizer2.txt added (r1.1)

---
Log message:

Here are the notes from our Reoptimizer meetings.


---
Diffs of the changes:

Index: llvm/www/docs/HistoricalNotes/2003-06-25-Reoptimizer1.txt
diff -c /dev/null llvm/www/docs/HistoricalNotes/2003-06-25-Reoptimizer1.txt:1.1
*** /dev/null	Thu Jun 26 15:37:53 2003
--- llvm/www/docs/HistoricalNotes/2003-06-25-Reoptimizer1.txt	Thu Jun 26 15:37:42 2003
***************
*** 0 ****
--- 1,137 ----
+ Wed Jun 25 15:13:51 CDT 2003
+ 
+ First-level instrumentation
+ ---------------------------
+ 
+ We use opt to do Bytecode-to-bytecode instrumentation. Look at
+ back-edges and insert llvm_first_trigger() function call which takes
+ no arguments and no return value. This instrumentation is designed to
+ be easy to remove, for instance by writing a NOP over the function
+ call instruction.
+ 
+ Keep count of every call to llvm_first_trigger(), and maintain
+ counters in a map indexed by return address. If the trigger count
+ exceeds a threshold, we identify a hot loop and perform second-level
+ instrumentation on the hot loop region (the instructions between the
+ target of the back-edge and the branch that causes the back-edge).  We
+ do not move code across basic-block boundaries.
+ 
+ 
+ Second-level instrumentation
+ ---------------------------
+ 
+ We remove the first-level instrumentation by overwriting the CALL to
+ llvm_first_trigger() with a NOP.
+ 
+ The reoptimizer maintains a map between machine-code basic blocks and
+ LLVM BasicBlock*s.  We only keep track of paths that start at the
+ first machine-code basic block of the hot loop region.
+ 
+ How do we keep track of which edges to instrument, and which edges are
+ exits from the hot region? 3 step process.
+ 
+ 1) Do a DFS from the first machine-code basic block of the hot loop
+ region and mark reachable edges.
+ 
+ 2) Do a DFS from the last machine-code basic block of the hot loop
+ region IGNORING back edges, and mark the edges which are reachable in
+ 1) and also in 2) (i.e., must be reachable from both the start BB and
+ the end BB of the hot region).
+ 
+ 3) Mark BBs which end in edges that exit the hot region; we need to
+ instrument these differently.
+ 
+ Assume that there is 1 free register. On SPARC we use %g1, which LLC
+ has agreed not to use.  Shift a 1 into it at the beginning. At every
+ edge which corresponds to a conditional branch, we shift 0 for not
+ taken and 1 for taken into a register. This uniquely numbers the paths
+ through the hot region. Silently fail if we need more than 64 bits.
+ 
+ At the end BB we call countPath and increment the counter based on %g1
+ and the return address of the countPath call.  We keep track of the
+ number of iterations and the number of paths.  We only run this
+ version 30 or 40 times.
+ 
+ Find the BBs that total 90% or more of execution, and aggregate them
+ together to form our trace. But we do not allow more than 5 paths; if
+ we have more than 5 we take the ones that are executed the most.  We
+ verify our assumption that we picked a hot back-edge in first-level
+ instrumentation, by making sure that the number of times we took an
+ exit edge from the hot trace is less than 10% of the number of
+ iterations.
+ 
+ LLC has been taught to recognize llvm_first_trigger() calls and NOT
+ generate saves and restores of caller-saved registers around these
+ calls.
+ 
+ 
+ Phase behavior
+ --------------
+ 
+ We turn off llvm_first_trigger() calls with NOPs, but this would hide
+ phase behavior from us (when some funcs/traces stop being hot and
+ others become hot.)
+ 
+ We have a SIGALRM timer that counts time for us. Every time we get a
+ SIGALRM we look at our priority queue of locations where we have
+ removed llvm_first_trigger() calls. Each location is inserted along
+ with a time when we will next turn instrumentation back on for that
+ call site. If the time has arrived for a particular call site, we pop
+ that off the prio. queue and turn instrumentation back on for that
+ call site.
+ 
+ 
+ Generating traces
+ -----------------
+ 
+ When we finally generate an optimized trace we first copy the code
+ into the trace cache. This leaves us with 3 copies of the code: the
+ original code, the instrumented code, and the optimized trace. The
+ optimized trace does not have instrumentation. The original code and
+ the instrumented code are modified to have a branch to the trace
+ cache, where the optimized traces are kept.
+ 
+ We copy the code from the original to the instrumentation version
+ by tracing the LLVM-to-Machine code basic block map and then copying
+ each machine code basic block we think is in the hot region into the
+ trace cache. Then we instrument that code. The process is similar for
+ generating the final optimized trace; we copy the same basic blocks
+ because we might need to put in fixup code for exit BBs.
+ 
+ LLVM basic blocks are not typically used in the Reoptimizer except
+ for the mapping information.
+ 
+ We are restricted to using single instructions to branch between the
+ original code, trace, and instrumented code. So we have to keep the
+ code copies in memory near the original code (they can't be far enough
+ away that a single pc-relative branch would not work.) Malloc() or
+ data region space is too far away. this impacts the design of the 
+ trace cache.
+ 
+ We use a dummy function that is full of a bunch of for loops which we
+ overwrite with trace-cache code. The trace manager keeps track of
+ whether or not we have enough space in the trace cache, etc.
+ 
+ The trace insertion routine takes an original start address, a vector
+ of machine instructions representing the trace, index of branches and
+ their corresponding absolute targets, and index of calls and their
+ corresponding absolute targets.
+ 
+ The trace insertion routine is responsible for inserting branches from
+ the beginning of the original code to the beginning of the optimized
+ trace. This is because at some point the trace cache may run out of
+ space and it may have to evict a trace, at which point the branch to
+ the trace would also have to be removed. It uses a round-robin
+ replacement policy; we have found that this is almost as good as LRU
+ and better than random (especially because of problems fitting the new
+ trace in.)
+ 
+ We cannot deal with discontiguous trace cache areas.  The trace cache
+ is supposed to be cache-line-aligned, but it is not page-aligned.
+ 
+ We generate instrumentation traces and optimized traces into separate
+ trace caches. We keep the instrumented code around because you don't
+ want to delete a trace when you still might have to return to it
+ (i.e., return from a llvm_first_trigger() or countPath() call.)
+ 
+ 


Index: llvm/www/docs/HistoricalNotes/2003-06-26-Reoptimizer2.txt
diff -c /dev/null llvm/www/docs/HistoricalNotes/2003-06-26-Reoptimizer2.txt:1.1
*** /dev/null	Thu Jun 26 15:37:53 2003
--- llvm/www/docs/HistoricalNotes/2003-06-26-Reoptimizer2.txt	Thu Jun 26 15:37:42 2003
***************
*** 0 ****
--- 1,110 ----
+ Thu Jun 26 14:43:04 CDT 2003
+ 
+ Information about BinInterface
+ ------------------------------
+ 
+ Take in a set of instructions with some particular register
+ allocation. It allows you to add, modify, or delete some instructions,
+ in SSA form (kind of like LLVM's MachineInstrs.) Then re-allocate
+ registers. It assumes that the transformations you are doing are safe.
+ It does not update the mapping information or the LLVM representation
+ for the modified trace (so it would not, for instance, support
+ multiple optimization passes; passes have to be aware of and update
+ manually the mapping information.)
+ 
+ The way you use it is you take the original code and provide it to
+ BinInterface; then you do optimizations to it, then you put it in the
+ trace cache.
+ 
+ The BinInterface tries to find live-outs for traces so that it can do
+ register allocation on just the trace, and stitch the trace back into
+ the original code. It has to preserve the live-ins and live-outs when
+ it does its register allocation.  (On exits from the trace we have
+ epilogues that copy live-outs back into the right registers, but
+ live-ins have to be in the right registers.)
+ 
+ 
+ Limitations of BinInterface
+ ---------------------------
+ 
+ It does copy insertions for PHIs, which it infers from the machine
+ code. The mapping info inserted by LLC is not sufficient to determine
+ the PHIs.
+ 
+ It does not handle integer or floating-point condition codes and it
+ does not handle floating-point register allocation.
+ 
+ It is not aggressively able to use lots of registers.
+ 
+ There is a problem with alloca: we cannot find our spill space for
+ spilling registers, normally allocated on the stack, if the trace
+ follows an alloca(). What might be an acceptable solution would be to
+ disable trace generation on functions that have variable-sized
+ alloca()s. Variable-sized allocas in the trace would also probably
+ screw things up.
+ 
+ Because of the FP and alloca limitations, the BinInterface is
+ completely disabled right now.
+ 
+ 
+ Demo
+ ----
+ 
+ This is a demo of the Ball & Larus version that does NOT use 2-level
+ profiling.
+ 
+ 1. Compile program with llvm-gcc.
+ 2. Run opt -lowerswitch -paths -emitfuncs on the bytecode.
+    -lowerswitch change switch statements to branches
+    -paths       Ball & Larus path-profiling algorithm
+    -emitfuncs   emit the table of functions
+ 3. Run llc to generate SPARC assembly code for the result of step 2.
+ 4. Use g++ to link the (instrumented) assembly code.
+ 
+ We use a script to do all this:
+ ------------------------------------------------------------------------------
+ #!/bin/sh
+ llvm-gcc $1.c -o $1
+ opt -lowerswitch -paths -emitfuncs $1.bc > $1.run.bc
+ llc -f $1.run.bc 
+ LIBS=$HOME/llvm_sparc/lib/Debug
+ GXX=/usr/dcs/software/evaluation/bin/g++
+ $GXX -g -L $LIBS $1.run.s -o $1.run.llc \
+ $LIBS/tracecache.o \
+ $LIBS/mapinfo.o \
+ $LIBS/trigger.o \
+ $LIBS/profpaths.o \
+ $LIBS/bininterface.o \
+ $LIBS/support.o \
+ $LIBS/vmcore.o \
+ $LIBS/transformutils.o \
+ $LIBS/bcreader.o \
+ -lscalaropts -lscalaropts -lanalysis \
+ -lmalloc -lcpc -lm -ldl
+ ------------------------------------------------------------------------------
+ 
+ 5. Run the resulting binary.  You will see output from BinInterface
+ (described below) intermixed with the output from the program.
+ 
+ 
+ Output from BinInterface
+ ------------------------
+ 
+ BinInterface's debugging code prints out the following stuff in order:
+ 
+ 1. Initial code provided to BinInterface with original register
+ allocation.
+ 
+ 2. Section 0 is the trace prolog, consisting mainly of live-ins and
+ register saves which will be restored in epilogs.
+ 
+ 3. Section 1 is the trace itself, in SSA form used by BinInterface,
+ along with the PHIs that are inserted.
+ PHIs are followed by the copies that implement them.
+ Each branch (i.e., out of the trace) is annotated with the
+ section number that represents the epilog it branches to.
+ 
+ 4. All the other sections starting with Section 2 are trace epilogs.
+ Every branch from the trace has to go to some epilog.
+ 
+ 5. After the last section is the register allocation output.




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:36:02 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:36:02 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Reoptimizer/Mapping/FInfo.h
Message-ID: <200306262135.h5QLZ1R32250@choi.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Reoptimizer/Mapping:

FInfo.h (r1.1) removed

---
Log message:

Merged with mainline on Thursday, June 26, 2003.
Kept includes of "Config/assert.h" to avoid implicit dependencies on header
files.


---
Diffs of the changes:



From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:38:01 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:38:01 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/CodeGen/SparcRegInfo.h InstrForest.h InstrSelection.h LiveRange.h
Message-ID: <200306262137.h5QLbPC32642@choi.cs.uiuc.edu>


Changes in directory llvm/include/llvm/CodeGen:

SparcRegInfo.h added (r1.5.2.1)
InstrForest.h updated: 1.26.2.1 -> 1.26.2.2
InstrSelection.h updated: 1.23.2.1 -> 1.23.2.2
LiveRange.h updated: 1.19.2.1 -> 1.19.2.2

---
Log message:

Merged with mainline on Thursday, June 26, 2003.
Kept includes of "Config/assert.h" to avoid implicit dependencies on header
files.


---
Diffs of the changes:

Index: llvm/include/llvm/CodeGen/SparcRegInfo.h
diff -c /dev/null llvm/include/llvm/CodeGen/SparcRegInfo.h:1.1
*** /dev/null	Thu Jun 26 16:37:24 2003
--- llvm/include/llvm/CodeGen/SparcRegInfo.h	Fri Aug 31 15:30:42 2001
***************
*** 0 ****
--- 1,169 ----
+ /* Title:   SparcRegClassInfo.h
+    Author:  Ruchira Sasanka
+    Date:    Aug 20, 01
+    Purpose: Contains the description of integer register class of Sparc
+ */
+ 
+ 
+ #ifndef SPARC_INT_REG_CLASS_H
+ #define SPARC_INT_REG_CLASS_H
+ 
+ 
+ #include "llvm/CodeGen/TargetMachine.h"
+ #include <string>
+ 
+ 
+ //-----------------------------------------------------------------------------
+ // Integer Register Class
+ //-----------------------------------------------------------------------------
+ 
+ 
+ // Int register names in same order as enum in class SparcIntRegOrder
+ 
+ static string const IntRegNames[] = 
+   {       "g1", "g2", "g3", "g4", "g5", "g6", "g7",
+     "o0", "o1", "o2", "o3", "o4", "o5",       "o7",
+     "l0", "l1", "l2", "l3", "l4", "l5", "l6", "l7",
+     "i0", "i1", "i2", "i3", "i4", "i5", 
+     "g0", "i6", "i7",  "o6" }; 
+ 
+ 
+ 
+ class SparcIntRegOrder{ 
+ 
+  public:
+ 
+   enum RegsInPrefOrder   // colors possible for a LR (in preferred order)
+    { 
+      // --- following colors are volatile across function calls
+      // %g0 can't be used for coloring - always 0
+                      
+      g1, g2, g3, g4, g5, g6, g7,  //%g1-%g7  
+      o0, o1, o2, o3, o4, o5, o7,  // %o0-%o5, 
+ 
+      // %o6 is sp, 
+      // all %0's can get modified by a call
+ 
+      // --- following colors are NON-volatile across function calls
+       
+      l0, l1, l2, l3, l4, l5, l6, l7,    //  %l0-%l7
+      i0, i1, i2, i3, i4, i5,            // %i0-%i5: i's need not be preserved 
+       
+      // %i6 is the fp - so not allocated
+      // %i7 is the ret address - can be used if saved
+ 
+      // max # of colors reg coloring  can allocate (NumOfAvailRegs)
+ 
+      // --- following colors are not available for allocation within this phase
+      // --- but can appear for pre-colored ranges 
+ 
+      g0, i6, i7,  o6,   
+ 
+      NumOfAllRegs   // place holder to count all possilbe colors
+ 
+    };
+ 
+   // max # of colors reg coloring  can allocate
+   static unsigned int const NumOfAvailRegs = g0;
+ 
+   static unsigned int const StartOfNonVolatileRegs = l0;
+   static unsigned int const StartOfAllRegs = g1;
+ 
+ 
+   static const string  getRegName(const unsigned reg) {
+     assert( reg < NumOfAllRegs );
+     return IntRegNames[reg];
+   }
+ 
+ };
+ 
+ 
+ 
+ class SparcIntRegClass : public MachineRegClassInfo
+ {
+  public:
+ 
+   SparcIntRegClass(unsigned ID) 
+     : MachineRegClassInfo(0, 
+ 			  SparcIntRegOrder::NumOfAvailRegs,
+ 			  SparcIntRegOrder::NumOfAllRegs)
+     {  }
+ 
+   void colorIGNode(IGNode * Node, bool IsColorUsedArr[] ) const;
+ 
+ };
+ 
+ //-----------------------------------------------------------------------------
+ // Float Register Class
+ //-----------------------------------------------------------------------------
+ 
+ static string const FloatRegNames[] = 
+   {    
+     "f0",  "f1",  "f2",  "f3",  "f4",  "f5",  "f6",  "f7",  "f8",  "f9", 
+     "f10", "f11", "f12", "f13", "f14", "f15", "f16", "f17", "f18", "f19",
+     "f20", "f21", "f22", "f23", "f24", "f25", "f26", "f27", "f28", "f29",
+     "f30", "f31", "f32", "f33", "f34", "f35", "f36", "f37", "f38", "f39",
+     "f40", "f41", "f42", "f43", "f44", "f45", "f46", "f47", "f48", "f49",
+     "f50", "f51", "f52", "f53", "f54", "f55", "f56", "f57", "f58", "f59",
+     "f60", "f61", "f62", "f63"
+   };
+ 
+ 
+ class SparcFloatRegOrder{ 
+ 
+  public:
+ 
+   enum RegsInPrefOrder {
+ 
+     f0, f1, f2, f3, f4, f5, f6, f7, f8, f9, 
+     f10, f11, f12, f13, f14, f15, f16, f17, f18, f19,
+     f20, f21, f22, f23, f24, f25, f26, f27, f28, f29,
+     f30, f31, f32, f33, f34, f35, f36, f37, f38, f39,
+     f40, f41, f42, f43, f44, f45, f46, f47, f48, f49,
+     f50, f51, f52, f53, f54, f55, f56, f57, f58, f59,
+     f60, f61, f62, f63
+ 
+   };
+ 
+   // there are 64 regs alltogether but only 32 regs can be allocated at
+   // a time.
+ 
+   static unsigned int const NumOfAvailRegs = 32;
+   static unsigned int const NumOfAllRegs = 64;
+ 
+   static unsigned int const StartOfNonVolatileRegs = f6;
+   static unsigned int const StartOfAllRegs = f0;
+ 
+ 
+   static const string  getRegName(const unsigned reg) {
+     assert( reg < NumOfAllRegs );
+     return FloatRegNames[reg];
+   }
+ 
+ 
+ 
+ };
+ 
+ 
+ class SparcFloatRegClass : public MachineRegClassInfo
+ {
+  private:
+ 
+   int findFloatColor(const IGNode *const Node, unsigned Start,
+ 		     unsigned End, bool IsColorUsedArr[] ) const;
+ 
+  public:
+ 
+   SparcFloatRegClass(unsigned ID) 
+     : MachineRegClassInfo(1, 
+ 			  SparcFloatRegOrder::NumOfAvailRegs,
+ 			  SparcFloatRegOrder::NumOfAllRegs)
+     {  }
+ 
+   void colorIGNode(IGNode * Node, bool IsColorUsedArr[] ) const;
+ 
+ };
+ 
+ 
+ 
+ #endif


Index: llvm/include/llvm/CodeGen/InstrForest.h
diff -u llvm/include/llvm/CodeGen/InstrForest.h:1.26.2.1 llvm/include/llvm/CodeGen/InstrForest.h:1.26.2.2
--- llvm/include/llvm/CodeGen/InstrForest.h:1.26.2.1	Mon Jun 23 13:51:50 2003
+++ llvm/include/llvm/CodeGen/InstrForest.h	Thu Jun 26 16:34:44 2003
@@ -24,7 +24,6 @@
 #include "Config/assert.h"
 
 class Constant;
-class BasicBlock;
 class Function;
 class InstrTreeNode;
 class InstrForest;


Index: llvm/include/llvm/CodeGen/InstrSelection.h
diff -u llvm/include/llvm/CodeGen/InstrSelection.h:1.23.2.1 llvm/include/llvm/CodeGen/InstrSelection.h:1.23.2.2
--- llvm/include/llvm/CodeGen/InstrSelection.h:1.23.2.1	Mon Jun 23 13:51:50 2003
+++ llvm/include/llvm/CodeGen/InstrSelection.h	Thu Jun 26 16:34:44 2003
@@ -9,6 +9,7 @@
 
 #include "llvm/Instruction.h"
 #include "Config/assert.h"
+
 class Function;
 class InstrForest;
 class MachineInstr;


Index: llvm/include/llvm/CodeGen/LiveRange.h
diff -u llvm/include/llvm/CodeGen/LiveRange.h:1.19.2.1 llvm/include/llvm/CodeGen/LiveRange.h:1.19.2.2
--- llvm/include/llvm/CodeGen/LiveRange.h:1.19.2.1	Mon Jun 23 13:51:50 2003
+++ llvm/include/llvm/CodeGen/LiveRange.h	Thu Jun 26 16:34:44 2003
@@ -1,4 +1,4 @@
-//===-- LiveRange.h - Store info about a live range --------------*- C++ -*--=//
+//===-- LiveRange.h - Store info about a live range -------------*- C++ -*-===//
 //
 // Implements a live range using a ValueSet. A LiveRange is a simple set
 // of Values. 
@@ -18,7 +18,6 @@
 
 class RegClass;
 class IGNode;
-class Type;
 
 class LiveRange : public ValueSet {
   RegClass *MyRegClass;       // register classs (e.g., int, FP) for this LR




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:38:07 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:38:07 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/Interpreter/ExternalFunctions.cpp
Message-ID: <200306262137.h5QLbPl32648@choi.cs.uiuc.edu>


Changes in directory llvm/tools/lli/Interpreter:

ExternalFunctions.cpp updated: 1.52.2.1 -> 1.52.2.2

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/tools/lli/Interpreter/ExternalFunctions.cpp
diff -u llvm/tools/lli/Interpreter/ExternalFunctions.cpp:1.52.2.1 llvm/tools/lli/Interpreter/ExternalFunctions.cpp:1.52.2.2
--- llvm/tools/lli/Interpreter/ExternalFunctions.cpp:1.52.2.1	Mon Jun 23 14:04:52 2003
+++ llvm/tools/lli/Interpreter/ExternalFunctions.cpp	Thu Jun 26 16:36:15 2003
@@ -19,7 +19,7 @@
 #include <map>
 #include "Config/dlfcn.h"
 #include "Config/link.h"
-#include <math.h>
+#include <cmath>
 #include "Config/stdio.h"
 using std::vector;
 
@@ -217,14 +217,6 @@
   return GV;
 }
 
-// int isnan(double value);
-GenericValue lle_X_isnan(FunctionType *F, const vector<GenericValue> &Args) {
-  assert(Args.size() == 1);
-  GenericValue GV;
-  GV.IntVal = isnan(Args[0].DoubleVal);
-  return GV;
-}
-
 // double floor(double)
 GenericValue lle_X_floor(FunctionType *M, const vector<GenericValue> &Args) {
   assert(Args.size() == 1);
@@ -744,7 +736,6 @@
   FuncNames["lle_X_pow"]          = lle_X_pow;
   FuncNames["lle_X_exp"]          = lle_X_exp;
   FuncNames["lle_X_log"]          = lle_X_log;
-  FuncNames["lle_X_isnan"]        = lle_X_isnan;
   FuncNames["lle_X_floor"]        = lle_X_floor;
   FuncNames["lle_X_srand"]        = lle_X_srand;
   FuncNames["lle_X_drand48"]      = lle_X_drand48;




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:38:13 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:38:13 2003
Subject: [llvm-commits] CVS: llvm/tools/link/link.cpp
Message-ID: <200306262137.h5QLboL00625@choi.cs.uiuc.edu>


Changes in directory llvm/tools/link:

link.cpp updated: 1.24 -> 1.24.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/tools/link/link.cpp
diff -u llvm/tools/link/link.cpp:1.24 llvm/tools/link/link.cpp:1.24.2.1
--- llvm/tools/link/link.cpp:1.24	Thu May 22 15:13:15 2003
+++ llvm/tools/link/link.cpp	Thu Jun 26 16:36:10 2003
@@ -4,9 +4,6 @@
 // This utility may be invoked in the following manner:
 //  link a.bc b.bc c.bc -o x.bc
 //
-// Alternatively, this can be used as an 'ar' tool as well.  If invoked as
-// either 'ar' or 'llvm-ar', it accepts a 'rc' parameter as well.
-//
 //===----------------------------------------------------------------------===//
 
 #include "llvm/Transforms/Utils/Linker.h"
@@ -89,13 +86,6 @@
 
   unsigned BaseArg = 0;
   std::string ErrorMessage;
-
-  // TODO: TEST argv[0] for llvm-ar forms... for now, this is a huge hack.
-  if (InputFilenames.size() >= 3 && InputFilenames[0] == "rc" &&
-      OutputFilename == "-") {
-    BaseArg = 2;
-    OutputFilename = InputFilenames[1];
-  }
 
   std::auto_ptr<Module> Composite(LoadFile(InputFilenames[BaseArg]));
   if (Composite.get() == 0) return 1;




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:38:18 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:38:18 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/CBackend/2003-06-23-PromotedExprs.llx Makefile
Message-ID: <200306262137.h5QLbPr32654@choi.cs.uiuc.edu>


Changes in directory llvm/test/Regression/CBackend:

2003-06-23-PromotedExprs.llx added (r1.1.2.1)
Makefile updated: 1.7 -> 1.7.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/test/Regression/CBackend/2003-06-23-PromotedExprs.llx
diff -c /dev/null llvm/test/Regression/CBackend/2003-06-23-PromotedExprs.llx:1.1
*** /dev/null	Thu Jun 26 16:37:25 2003
--- llvm/test/Regression/CBackend/2003-06-23-PromotedExprs.llx	Mon Jun 23 14:59:17 2003
***************
*** 0 ****
--- 1,16 ----
+ 
+ ; RUN: as < %s | dis -c > Output/%s.cbe.c
+ ; RUN: gcc -B/usr/bin/ Output/%s.cbe.c -o Output/%s.cbe
+ ; RUN: Output/%s.cbe
+ 
+ bool %doTest(ubyte %x) {
+ 	%dec.0 = add ubyte %x, 255
+     %tmp.1001 = cast ubyte %dec.0 to bool
+     ret bool %tmp.1001
+ }
+ 
+ int %main () {
+     %result = call bool %doTest(ubyte 1)
+     %p = cast bool %result to int
+     ret int %p
+ }


Index: llvm/test/Regression/CBackend/Makefile
diff -u llvm/test/Regression/CBackend/Makefile:1.7 llvm/test/Regression/CBackend/Makefile:1.7.2.1
--- llvm/test/Regression/CBackend/Makefile:1.7	Mon May 12 10:33:52 2003
+++ llvm/test/Regression/CBackend/Makefile	Thu Jun 26 16:35:45 2003
@@ -10,9 +10,9 @@
 .PRECIOUS: Output/%.c
 
 TESTS := $(wildcard *.ll)
+FTESTS := $(wildcard *.llx)   # Freeform tests
 
-all:: $(addprefix Output/, $(TESTS:%.ll=%.to))
-
+all:: $(addprefix Output/, $(TESTS:%.ll=%.to)) $(addprefix Output/, $(FTESTS:%.llx=%.llx.out))
 
 Output/%.to: Output/%.c
 	$(CC) -c -Werror $< -o $@ || \
@@ -21,3 +21,7 @@
 Output/%.c: %.ll Output/.dir $(LAS) $(LDIS)
 	$(LAS) < $< | $(LDIS) -c > $@ || \
 		(rm -f $@; $(FAILURE) $@ )
+
+Output/%.llx.out: %.llx Output/.dir $(LAS) $(LDIS)
+	-$(TESTRUNR) $<
+




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:38:23 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:38:23 2003
Subject: [llvm-commits] CVS: llvm/lib/CWriter/Writer.cpp
Message-ID: <200306262137.h5QLbPA32662@choi.cs.uiuc.edu>


Changes in directory llvm/lib/CWriter:

Writer.cpp updated: 1.92 -> 1.92.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/lib/CWriter/Writer.cpp
diff -u llvm/lib/CWriter/Writer.cpp:1.92 llvm/lib/CWriter/Writer.cpp:1.92.2.1
--- llvm/lib/CWriter/Writer.cpp:1.92	Fri Jun  6 02:10:24 2003
+++ llvm/lib/CWriter/Writer.cpp	Thu Jun 26 16:35:15 2003
@@ -97,6 +97,20 @@
       return I.getParent() == cast<Instruction>(I.use_back())->getParent();
     }
 
+    // isDirectAlloca - Define fixed sized allocas in the entry block as direct
+    // variables which are accessed with the & operator.  This causes GCC to
+    // generate significantly better code than to emit alloca calls directly.
+    //
+    static const AllocaInst *isDirectAlloca(const Value *V) {
+      const AllocaInst *AI = dyn_cast<AllocaInst>(V);
+      if (!AI) return false;
+      if (AI->isArrayAllocation())
+        return 0;   // FIXME: we can also inline fixed size array allocas!
+      if (AI->getParent() != &AI->getParent()->getParent()->getEntryNode())
+        return 0;
+      return AI;
+    }
+
     // Instruction visitation functions
     friend class InstVisitor<CWriter>;
 
@@ -150,14 +164,25 @@
 }
 
 std::string CWriter::getValueName(const Value *V) {
-  if (V->hasName()) {              // Print out the label if it exists...
-    if (isa<GlobalValue>(V) &&     // Do not mangle globals...
-        (cast<GlobalValue>(V)->hasExternalLinkage() &&// Unless it's internal or
-         !MangledGlobals.count(V))) // Unless the name would collide if we don't
-      return makeNameProper(V->getName());
-
+  if (V->hasName()) { // Print out the label if it exists...
+    
+    // Name mangling occurs as follows:
+    // - If V is not a global, mangling always occurs.
+    // - Otherwise, mangling occurs when any of the following are true:
+    //   1) V has internal linkage
+    //   2) V's name would collide if it is not mangled.
+    //
+    
+    if(const GlobalValue* gv = dyn_cast<GlobalValue>(V)) {
+      if(!gv->hasInternalLinkage() && !MangledGlobals.count(gv)) {
+        // No internal linkage, name will not collide -> no mangling.
+        return makeNameProper(gv->getName());
+      }
+    }
+    
+    // Non-global, or global with internal linkage / colliding name -> mangle.
     return "l" + utostr(V->getType()->getUniqueID()) + "_" +
-           makeNameProper(V->getName());      
+      makeNameProper(V->getName());      
   }
 
   int Slot = Table->getValSlot(V);
@@ -206,23 +231,23 @@
   switch (Ty->getPrimitiveID()) {
   case Type::FunctionTyID: {
     const FunctionType *MTy = cast<FunctionType>(Ty);
-    std::stringstream FunctionInards; 
-    FunctionInards << " (" << NameSoFar << ") (";
+    std::stringstream FunctionInnards; 
+    FunctionInnards << " (" << NameSoFar << ") (";
     for (FunctionType::ParamTypes::const_iterator
            I = MTy->getParamTypes().begin(),
            E = MTy->getParamTypes().end(); I != E; ++I) {
       if (I != MTy->getParamTypes().begin())
-        FunctionInards << ", ";
-      printType(FunctionInards, *I, "");
+        FunctionInnards << ", ";
+      printType(FunctionInnards, *I, "");
     }
     if (MTy->isVarArg()) {
       if (!MTy->getParamTypes().empty()) 
-    	FunctionInards << ", ...";
+    	FunctionInnards << ", ...";
     } else if (MTy->getParamTypes().empty()) {
-      FunctionInards << "void";
+      FunctionInnards << "void";
     }
-    FunctionInards << ")";
-    std::string tstr = FunctionInards.str();
+    FunctionInnards << ")";
+    std::string tstr = FunctionInnards.str();
     printType(Out, MTy->getReturnType(), tstr);
     return Out;
   }
@@ -285,24 +310,35 @@
   bool isString = (ETy == Type::SByteTy || ETy == Type::UByteTy);
 
   // Make sure the last character is a null char, as automatically added by C
-  if (CPA->getNumOperands() == 0 ||
-      !cast<Constant>(*(CPA->op_end()-1))->isNullValue())
+  if (isString && (CPA->getNumOperands() == 0 ||
+                   !cast<Constant>(*(CPA->op_end()-1))->isNullValue()))
     isString = false;
   
   if (isString) {
     Out << "\"";
+    // Keep track of whether the last number was a hexadecimal escape
+    bool LastWasHex = false;
+
     // Do not include the last character, which we know is null
     for (unsigned i = 0, e = CPA->getNumOperands()-1; i != e; ++i) {
       unsigned char C = (ETy == Type::SByteTy) ?
         (unsigned char)cast<ConstantSInt>(CPA->getOperand(i))->getValue() :
         (unsigned char)cast<ConstantUInt>(CPA->getOperand(i))->getValue();
       
-      if (isprint(C)) {
+      // Print it out literally if it is a printable character.  The only thing
+      // to be careful about is when the last letter output was a hex escape
+      // code, in which case we have to be careful not to print out hex digits
+      // explicitly (the C compiler thinks it is a continuation of the previous
+      // character, sheesh...)
+      //
+      if (isprint(C) && (!LastWasHex || !isxdigit(C))) {
+        LastWasHex = false;
         if (C == '"' || C == '\\')
           Out << "\\" << C;
         else
           Out << C;
       } else {
+        LastWasHex = false;
         switch (C) {
         case '\n': Out << "\\n"; break;
         case '\t': Out << "\\t"; break;
@@ -313,8 +349,9 @@
         case '\'': Out << "\\\'"; break;           
         default:
           Out << "\\x";
-          Out << ( C/16  < 10) ? ( C/16 +'0') : ( C/16 -10+'A');
-          Out << ((C&15) < 10) ? ((C&15)+'0') : ((C&15)-10+'A');
+          Out << (char)(( C/16  < 10) ? ( C/16 +'0') : ( C/16 -10+'A'));
+          Out << (char)(((C&15) < 10) ? ((C&15)+'0') : ((C&15)-10+'A'));
+          LastWasHex = true;
           break;
         }
       }
@@ -334,6 +371,27 @@
   }
 }
 
+/// FPCSafeToPrint - Returns true if we may assume that CFP may be
+/// written out textually as a double (rather than as a reference to a
+/// stack-allocated variable). We decide this by converting CFP to a
+/// string and back into a double, and then checking whether the
+/// conversion results in a bit-equal double to the original value of
+/// CFP. This depends on us and the target C compiler agreeing on the
+/// conversion process (which is pretty likely since we only deal in
+/// IEEE FP.) This is adapted from similar code in
+/// lib/VMCore/AsmWriter.cpp:WriteConstantInt().
+static bool FPCSafeToPrint (const ConstantFP *CFP) {
+  std::string StrVal = ftostr(CFP->getValue());
+  // Check to make sure that the stringized number is not some string like
+  // "Inf" or NaN, that atof will accept, but the lexer will not.  Check that
+  // the string matches the "[-+]?[0-9]" regex.
+  if ((StrVal[0] >= '0' && StrVal[0] <= '9') ||
+      ((StrVal[0] == '-' || StrVal[0] == '+') &&
+       (StrVal[1] >= '0' && StrVal[1] <= '9')))
+    // Reparse stringized version!
+    return (atof(StrVal.c_str()) == CFP->getValue());
+  return false;
+}
 
 // printConstant - The LLVM Constant to C Constant converter.
 void CWriter::printConstant(Constant *CPV) {
@@ -409,7 +467,11 @@
       Out << "(*(" << (FPC->getType() == Type::FloatTy ? "float" : "double")
           << "*)&FloatConstant" << I->second << ")";
     } else {
-      Out << FPC->getValue();
+      if (FPCSafeToPrint (FPC)) {
+	Out << ftostr (FPC->getValue ());
+      } else {
+	Out << FPC->getValue(); // Who knows? Give it our best shot...
+      }
     }
     break;
   }
@@ -451,7 +513,7 @@
 
 void CWriter::writeOperandInternal(Value *Operand) {
   if (Instruction *I = dyn_cast<Instruction>(Operand))
-    if (isInlinableInst(*I)) {
+    if (isInlinableInst(*I) && !isDirectAlloca(I)) {
       // Should we inline this instruction to build a tree?
       Out << "(";
       visit(*I);
@@ -471,12 +533,12 @@
 }
 
 void CWriter::writeOperand(Value *Operand) {
-  if (isa<GlobalVariable>(Operand))
+  if (isa<GlobalVariable>(Operand) || isDirectAlloca(Operand))
     Out << "(&";  // Global variables are references as their addresses by llvm
 
   writeOperandInternal(Operand);
 
-  if (isa<GlobalVariable>(Operand))
+  if (isa<GlobalVariable>(Operand) || isDirectAlloca(Operand))
     Out << ")";
 }
 
@@ -516,7 +578,9 @@
       << "extern void *__builtin_alloca(unsigned long);\n"
       << "#define alloca(x) __builtin_alloca(x)\n"
       << "#else\n"
+      << "#ifndef __FreeBSD__\n"
       << "#include <alloca.h>\n"
+      << "#endif\n"
       << "#endif\n\n";
 }
 
@@ -597,9 +661,9 @@
     Out << "extern void * malloc(size_t);\n\n";
   }
 
-  // Output the global variable declerations
+  // Output the global variable declarations
   if (!M->gempty()) {
-    Out << "\n\n/* Global Variable Declerations */\n";
+    Out << "\n\n/* Global Variable Declarations */\n";
     for (Module::giterator I = M->gbegin(), E = M->gend(); I != E; ++I)
       if (!I->isExternal()) {
         Out << "extern ";
@@ -715,33 +779,36 @@
 
 
 void CWriter::printFunctionSignature(const Function *F, bool Prototype) {
-  // If the program provides it's own malloc prototype we don't need
+  // If the program provides its own malloc prototype we don't need
   // to include the general one.  
   if (getValueName(F) == "malloc")
     needsMalloc = false;
-  if (F->hasInternalLinkage()) Out << "static ";  
+
+  if (F->hasInternalLinkage()) Out << "static ";
+  if (F->hasLinkOnceLinkage()) Out << "inline ";
+  
   // Loop over the arguments, printing them...
   const FunctionType *FT = cast<FunctionType>(F->getFunctionType());
   
-  std::stringstream FunctionInards; 
+  std::stringstream FunctionInnards; 
     
   // Print out the name...
-  FunctionInards << getValueName(F) << "(";
+  FunctionInnards << getValueName(F) << "(";
     
   if (!F->isExternal()) {
     if (!F->aempty()) {
       std::string ArgName;
       if (F->abegin()->hasName() || !Prototype)
         ArgName = getValueName(F->abegin());
-      printType(FunctionInards, F->afront().getType(), ArgName);
+      printType(FunctionInnards, F->afront().getType(), ArgName);
       for (Function::const_aiterator I = ++F->abegin(), E = F->aend();
            I != E; ++I) {
-        FunctionInards << ", ";
+        FunctionInnards << ", ";
         if (I->hasName() || !Prototype)
           ArgName = getValueName(I);
         else 
           ArgName = "";
-        printType(FunctionInards, I->getType(), ArgName);
+        printType(FunctionInnards, I->getType(), ArgName);
       }
     }
   } else {
@@ -749,8 +816,8 @@
     for (FunctionType::ParamTypes::const_iterator I = 
 	   FT->getParamTypes().begin(),
 	   E = FT->getParamTypes().end(); I != E; ++I) {
-      if (I != FT->getParamTypes().begin()) FunctionInards << ", ";
-      printType(FunctionInards, *I);
+      if (I != FT->getParamTypes().begin()) FunctionInnards << ", ";
+      printType(FunctionInnards, *I);
     }
   }
 
@@ -758,16 +825,15 @@
   // unless there are no known types, in which case, we just emit ().
   //
   if (FT->isVarArg() && !FT->getParamTypes().empty()) {
-    if (FT->getParamTypes().size()) FunctionInards << ", ";
-    FunctionInards << "...";  // Output varargs portion of signature!
+    if (FT->getParamTypes().size()) FunctionInnards << ", ";
+    FunctionInnards << "...";  // Output varargs portion of signature!
   }
-  FunctionInards << ")";
+  FunctionInnards << ")";
   // Print out the return type and the entire signature for that matter
-  printType(Out, F->getReturnType(), FunctionInards.str());
+  printType(Out, F->getReturnType(), FunctionInnards.str());
   
 }
 
-
 void CWriter::printFunction(Function *F) {
   if (F->isExternal()) return;
 
@@ -778,11 +844,15 @@
 
   // print local variable information for the function
   for (inst_iterator I = inst_begin(F), E = inst_end(F); I != E; ++I)
-    if ((*I)->getType() != Type::VoidTy && !isInlinableInst(**I)) {
+    if (const AllocaInst *AI = isDirectAlloca(*I)) {
+      Out << "  ";
+      printType(Out, AI->getAllocatedType(), getValueName(AI));
+      Out << ";    /* Address exposed local */\n";
+    } else if ((*I)->getType() != Type::VoidTy && !isInlinableInst(**I)) {
       Out << "  ";
       printType(Out, (*I)->getType(), getValueName(*I));
       Out << ";\n";
-
+      
       if (isa<PHINode>(*I)) {  // Print out PHI node temporaries as well...
         Out << "  ";
         printType(Out, (*I)->getType(), getValueName(*I)+"__PHI_TEMPORARY");
@@ -794,12 +864,14 @@
 
   // Scan the function for floating point constants.  If any FP constant is used
   // in the function, we want to redirect it here so that we do not depend on
-  // the precision of the printed form.
+  // the precision of the printed form, unless the printed form preserves
+  // precision.
   //
   unsigned FPCounter = 0;
   for (constant_iterator I = constant_begin(F), E = constant_end(F); I != E;++I)
     if (const ConstantFP *FPC = dyn_cast<ConstantFP>(*I))
-      if (FPConstantMap.find(FPC) == FPConstantMap.end()) {
+      if ((!FPCSafeToPrint(FPC)) // Do not put in FPConstantMap if safe.
+	  && (FPConstantMap.find(FPC) == FPConstantMap.end())) {
         double Val = FPC->getValue();
         
         FPConstantMap[FPC] = FPCounter;  // Number the FP constants
@@ -842,7 +914,7 @@
 
     // Output all of the instructions in the basic block...
     for (BasicBlock::iterator II = BB->begin(), E = --BB->end(); II != E; ++II){
-      if (!isInlinableInst(*II)) {
+      if (!isInlinableInst(*II) && !isDirectAlloca(II)) {
         if (II->getType() != Type::VoidTy)
           outputLValue(II);
         else
@@ -971,6 +1043,17 @@
 void CWriter::visitBinaryOperator(Instruction &I) {
   // binary instructions, shift instructions, setCond instructions.
   assert(!isa<PointerType>(I.getType()));
+
+  // We must cast the results of binary operations which might be promoted.
+  bool needsCast = false;
+  if ((I.getType() == Type::UByteTy) || (I.getType() == Type::SByteTy)
+      || (I.getType() == Type::UShortTy) || (I.getType() == Type::ShortTy)
+      || (I.getType() == Type::FloatTy)) {
+    needsCast = true;
+    Out << "((";
+    printType(Out, I.getType(), "", false, false);
+    Out << ")(";
+  }
       
   writeOperand(I.getOperand(0));
 
@@ -995,6 +1078,10 @@
   }
 
   writeOperand(I.getOperand(1));
+
+  if (needsCast) {
+    Out << "))";
+  }
 }
 
 void CWriter::visitCastInst(CastInst &I) {
@@ -1118,6 +1205,8 @@
   } else if (ConstantPointerRef *CPR = dyn_cast<ConstantPointerRef>(Ptr)) {
     HasImplicitAddress = true;
     Ptr = CPR->getValue();         // Get to the global...
+  } else if (isDirectAlloca(Ptr)) {
+    HasImplicitAddress = true;
   }
 
   if (I == E) {




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:38:29 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:38:29 2003
Subject: [llvm-commits] CVS: llvm/test/Programs/External/SPEC/Sandbox.sh
Message-ID: <200306262137.h5QLbmr00619@choi.cs.uiuc.edu>


Changes in directory llvm/test/Programs/External/SPEC:

Sandbox.sh updated: 1.7 -> 1.7.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/test/Programs/External/SPEC/Sandbox.sh
diff -u llvm/test/Programs/External/SPEC/Sandbox.sh:1.7 llvm/test/Programs/External/SPEC/Sandbox.sh:1.7.2.1
--- llvm/test/Programs/External/SPEC/Sandbox.sh:1.7	Fri Jun 20 15:42:46 2003
+++ llvm/test/Programs/External/SPEC/Sandbox.sh	Thu Jun 26 16:35:38 2003
@@ -22,7 +22,7 @@
 cp -r $INPUT_DIR/* .
 
 # If there is an "all" input directory, take it as well...
-cp -r $INPUT_DIR/../../all/input/* . 2>&1 > /dev/null
+cp -r $INPUT_DIR/../../all/input/* . > /dev/null 2>&1 
 
 # Run the program now...
 echo Running: $*




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:38:34 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:38:34 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/JIT/Emitter.cpp JIT.cpp
Message-ID: <200306262137.h5QLbQQ32672@choi.cs.uiuc.edu>


Changes in directory llvm/tools/lli/JIT:

Emitter.cpp updated: 1.14.2.1 -> 1.14.2.2
JIT.cpp updated: 1.7 -> 1.7.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/tools/lli/JIT/Emitter.cpp
diff -u llvm/tools/lli/JIT/Emitter.cpp:1.14.2.1 llvm/tools/lli/JIT/Emitter.cpp:1.14.2.2
--- llvm/tools/lli/JIT/Emitter.cpp:1.14.2.1	Mon Jun 23 14:04:54 2003
+++ llvm/tools/lli/JIT/Emitter.cpp	Thu Jun 26 16:36:16 2003
@@ -74,6 +74,9 @@
   static const long pageSize = sysconf(_SC_PAGESIZE);
 
 #if defined(i386) || defined(__i386__) || defined(__x86__)
+#if defined(MAP_ANON) && !defined(MAP_ANONYMOUS)
+# define MAP_ANONYMOUS MAP_ANON
+#endif
   pa = mmap(0, pageSize*NumPages, PROT_READ|PROT_WRITE|PROT_EXEC,
             MAP_PRIVATE|MAP_ANONYMOUS, 0, 0);  /* fd = 0  */
 #elif defined(sparc) || defined(__sparc__) || defined(__sparcv9)


Index: llvm/tools/lli/JIT/JIT.cpp
diff -u llvm/tools/lli/JIT/JIT.cpp:1.7 llvm/tools/lli/JIT/JIT.cpp:1.7.2.1
--- llvm/tools/lli/JIT/JIT.cpp:1.7	Fri Jun  6 01:59:55 2003
+++ llvm/tools/lli/JIT/JIT.cpp	Thu Jun 26 16:36:16 2003
@@ -15,19 +15,23 @@
 #include "llvm/PassManager.h"
 
 namespace {
-  cl::opt<std::string>
-  Arch("march", cl::desc("Architecture: `x86' or `sparc'"), cl::Prefix,
-       cl::value_desc("machine architecture"));
-  
-  static std::string DefaultArch = 
+  enum ArchName { nojit, x86, Sparc };
+
+  cl::opt<ArchName>
+  Arch("march", cl::desc("Architecture to JIT to:"), cl::Prefix,
+       cl::values(clEnumVal(x86, "  IA-32 (pentium and above)"),
+#if defined(sparc) || defined(__sparc__) || defined(__sparcv9)
+                  clEnumValN(Sparc, "sparc", "  Sparc-V9"),
+#endif
+                  0),
 #if defined(i386) || defined(__i386__) || defined(__x86__)
-  "x86";
+  cl::init(x86)
 #elif defined(sparc) || defined(__sparc__) || defined(__sparcv9)
-  "sparc";
+  cl::init(Sparc)
 #else
-  "";
+  cl::init(nojit)
 #endif
-
+       );
 }
 
 /// createJIT - Create an return a new JIT compiler if there is one available
@@ -36,28 +40,31 @@
 ExecutionEngine *ExecutionEngine::createJIT(Module *M, unsigned Config) {
   
   TargetMachine* (*TargetMachineAllocator)(unsigned) = 0;
-  if (Arch == "")
-    Arch = DefaultArch;
 
   // Allow a command-line switch to override what *should* be the default target
   // machine for this platform. This allows for debugging a Sparc JIT on X86 --
   // our X86 machines are much faster at recompiling LLVM and linking lli.
-  if (Arch == "x86") {
+  switch (Arch) {
+  case x86:
     TargetMachineAllocator = allocateX86TargetMachine;
-  } else if (Arch == "sparc") {
+    break;
+#if defined(sparc) || defined(__sparc__) || defined(__sparcv9)
+  case Sparc:
     TargetMachineAllocator = allocateSparcTargetMachine;
-  }
-
-  if (TargetMachineAllocator) {
-    // Allocate a target...
-    TargetMachine *Target = (*TargetMachineAllocator)(Config);
-    assert(Target && "Could not allocate target machine!");
-
-    // Create the virtual machine object...
-    return new VM(M, Target);
-  } else {
+    break;
+#endif
+  default:
+    assert(0 && "-march flag not supported on this host!");
+  case nojit:
     return 0;
   }
+
+  // Allocate a target...
+  TargetMachine *Target = (*TargetMachineAllocator)(Config);
+  assert(Target && "Could not allocate target machine!");
+  
+  // Create the virtual machine object...
+  return new VM(M, Target);
 }
 
 VM::VM(Module *M, TargetMachine *tm) : ExecutionEngine(M), TM(*tm) {
@@ -68,6 +75,7 @@
 
   setupPassManager();
 
+#if defined(sparc) || defined(__sparc__) || defined(__sparcv9)
   // THIS GOES BEYOND UGLY HACKS
   if (TM.getName() == "UltraSparc-Native") {
     extern Pass *createPreSelectionPass(TargetMachine &TM);
@@ -77,6 +85,7 @@
     PM.add(createPreSelectionPass(TM));
     PM.run(*M);
   }
+#endif
 
   emitGlobals();
 }




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:38:40 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:38:40 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/CFrontend/2003-06-22-UnionCrash.c 2003-06-23-GCC-fold-infinite-recursion.c
Message-ID: <200306262137.h5QLbQC32676@choi.cs.uiuc.edu>


Changes in directory llvm/test/Regression/CFrontend:

2003-06-22-UnionCrash.c added (r1.1.2.1)
2003-06-23-GCC-fold-infinite-recursion.c added (r1.1.2.1)

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/test/Regression/CFrontend/2003-06-22-UnionCrash.c
diff -c /dev/null llvm/test/Regression/CFrontend/2003-06-22-UnionCrash.c:1.1
*** /dev/null	Thu Jun 26 16:37:26 2003
--- llvm/test/Regression/CFrontend/2003-06-22-UnionCrash.c	Sun Jun 22 16:40:38 2003
***************
*** 0 ****
--- 1,11 ----
+ struct Blend_Map_Entry {
+   union {
+    float Colour[5];
+    double Point_Slope[2];
+   } Vals;
+ };
+ 
+ void test(struct Blend_Map_Entry* Foo)
+ {
+ }
+ 


Index: llvm/test/Regression/CFrontend/2003-06-23-GCC-fold-infinite-recursion.c
diff -c /dev/null llvm/test/Regression/CFrontend/2003-06-23-GCC-fold-infinite-recursion.c:1.1
*** /dev/null	Thu Jun 26 16:37:26 2003
--- llvm/test/Regression/CFrontend/2003-06-23-GCC-fold-infinite-recursion.c	Mon Jun 23 13:06:21 2003
***************
*** 0 ****
--- 1,4 ----
+ double Test(double A, double B, double C, double D) {
+   return -(A-B) - (C-D);
+ }
+ 




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:38:45 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:38:45 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/AbstractTypeUser.h Constants.h DerivedTypes.h GlobalVariable.h Instruction.h SymbolTable.h User.h Value.h iOther.h iPHINode.h iTerminators.h
Message-ID: <200306262137.h5QLbmg00613@choi.cs.uiuc.edu>


Changes in directory llvm/include/llvm:

AbstractTypeUser.h updated: 1.6.2.1 -> 1.6.2.2
Constants.h updated: 1.26.2.1 -> 1.26.2.2
DerivedTypes.h updated: 1.29.2.1 -> 1.29.2.2
GlobalVariable.h updated: 1.20.2.1 -> 1.20.2.2
Instruction.h updated: 1.40.2.1 -> 1.40.2.2
SymbolTable.h updated: 1.21.2.1 -> 1.21.2.2
User.h updated: 1.18.2.1 -> 1.18.2.2
Value.h updated: 1.40 -> 1.40.2.1
iOther.h updated: 1.35.2.1 -> 1.35.2.2
iPHINode.h updated: 1.9.2.1 -> 1.9.2.2
iTerminators.h updated: 1.27.2.1 -> 1.27.2.2

---
Log message:

Merged with mainline on Thursday, June 26, 2003.
Kept includes of "Config/assert.h" to avoid implicit dependencies on header
files.


---
Diffs of the changes:

Index: llvm/include/llvm/AbstractTypeUser.h
diff -u llvm/include/llvm/AbstractTypeUser.h:1.6.2.1 llvm/include/llvm/AbstractTypeUser.h:1.6.2.2
--- llvm/include/llvm/AbstractTypeUser.h:1.6.2.1	Mon Jun 23 13:51:40 2003
+++ llvm/include/llvm/AbstractTypeUser.h	Thu Jun 26 16:34:37 2003
@@ -58,9 +58,8 @@
 // example.  This class is a simple class used to keep the use list of abstract
 // types up-to-date.
 //
-template <class TypeSubClass>
 class PATypeHandle {
-  const TypeSubClass *Ty;
+  const Type *Ty;
   AbstractTypeUser * const User;
 
   // These functions are defined at the bottom of Type.h.  See the comment there
@@ -69,7 +68,7 @@
   inline void removeUser();
 public:
   // ctor - Add use to type if abstract.  Note that Ty must not be null
-  inline PATypeHandle(const TypeSubClass *ty, AbstractTypeUser *user) 
+  inline PATypeHandle(const Type *ty, AbstractTypeUser *user) 
     : Ty(ty), User(user) {
     addUser();
   }
@@ -83,11 +82,11 @@
   inline ~PATypeHandle() { removeUser(); }
 
   // Automatic casting operator so that the handle may be used naturally
-  inline operator const TypeSubClass *() const { return Ty; }
-  inline const TypeSubClass *get() const { return Ty; }
+  inline operator const Type *() const { return Ty; }
+  inline const Type *get() const { return Ty; }
 
   // operator= - Allow assignment to handle
-  inline const TypeSubClass *operator=(const TypeSubClass *ty) {
+  inline const Type *operator=(const Type *ty) {
     if (Ty != ty) {   // Ensure we don't accidentally drop last ref to Ty
       removeUser();
       Ty = ty;
@@ -97,16 +96,16 @@
   }
 
   // operator= - Allow assignment to handle
-  inline const TypeSubClass *operator=(const PATypeHandle &T) {
+  inline const Type *operator=(const PATypeHandle &T) {
     return operator=(T.Ty);
   }
 
-  inline bool operator==(const TypeSubClass *ty) {
+  inline bool operator==(const Type *ty) {
     return Ty == ty;
   }
 
   // operator-> - Allow user to dereference handle naturally...
-  inline const TypeSubClass *operator->() const { return Ty; }
+  inline const Type *operator->() const { return Ty; }
 
   // removeUserFromConcrete - This function should be called when the User is
   // notified that our type is refined... and the type is being refined to
@@ -122,10 +121,10 @@
 // as both a handle (as above) and an AbstractTypeUser.  It uses the callback to
 // keep its pointer member updated to the current version of the type.
 //
-struct PATypeHolder : public AbstractTypeUser, public PATypeHandle<Type> {
-  inline PATypeHolder(const Type *ty) : PATypeHandle<Type>(ty, this) {}
+struct PATypeHolder : public AbstractTypeUser, public PATypeHandle {
+  inline PATypeHolder(const Type *ty) : PATypeHandle(ty, this) {}
   inline PATypeHolder(const PATypeHolder &T)
-    : AbstractTypeUser(T), PATypeHandle<Type>(T, this) {}
+    : AbstractTypeUser(T), PATypeHandle(T, this) {}
 
   // refineAbstractType - All we do is update our PATypeHandle member to point
   // to the new type.
@@ -138,20 +137,20 @@
     removeUserFromConcrete();
 
     if ((const Type*)OldTy != NewTy)
-      PATypeHandle<Type>::operator=(NewTy);
+      PATypeHandle::operator=(NewTy);
   }
 
   // operator= - Allow assignment to handle
   inline const Type *operator=(const Type *ty) {
-    return PATypeHandle<Type>::operator=(ty);
+    return PATypeHandle::operator=(ty);
   }
 
   // operator= - Allow assignment to handle
-  inline const Type *operator=(const PATypeHandle<Type> &T) {
-    return PATypeHandle<Type>::operator=(T);
+  inline const Type *operator=(const PATypeHandle &T) {
+    return PATypeHandle::operator=(T);
   }
   inline const Type *operator=(const PATypeHolder &H) {
-    return PATypeHandle<Type>::operator=(H);
+    return PATypeHandle::operator=(H);
   }
 
   void dump() const;


Index: llvm/include/llvm/Constants.h
diff -u llvm/include/llvm/Constants.h:1.26.2.1 llvm/include/llvm/Constants.h:1.26.2.2
--- llvm/include/llvm/Constants.h:1.26.2.1	Mon Jun 23 13:51:40 2003
+++ llvm/include/llvm/Constants.h	Thu Jun 26 16:34:37 2003
@@ -9,7 +9,6 @@
 #define LLVM_CONSTANTS_H
 
 #include "Config/assert.h"
-
 #include "llvm/Constant.h"
 #include "Support/DataTypes.h"
 


Index: llvm/include/llvm/DerivedTypes.h
diff -u llvm/include/llvm/DerivedTypes.h:1.29.2.1 llvm/include/llvm/DerivedTypes.h:1.29.2.2
--- llvm/include/llvm/DerivedTypes.h:1.29.2.1	Mon Jun 23 13:51:40 2003
+++ llvm/include/llvm/DerivedTypes.h	Thu Jun 26 16:34:37 2003
@@ -12,7 +12,6 @@
 #define LLVM_DERIVED_TYPES_H
 
 #include "Config/assert.h"
-
 #include "llvm/Type.h"
 
 class DerivedType : public Type {
@@ -86,9 +85,9 @@
 
 class FunctionType : public DerivedType {
 public:
-  typedef std::vector<PATypeHandle<Type> > ParamTypes;
+  typedef std::vector<PATypeHandle> ParamTypes;
 private:
-  PATypeHandle<Type> ResultType;
+  PATypeHandle ResultType;
   ParamTypes ParamTys;
   bool isVarArgs;
 
@@ -181,7 +180,7 @@
 
 class StructType : public CompositeType {
 public:
-  typedef std::vector<PATypeHandle<Type> > ElementTypes;
+  typedef std::vector<PATypeHandle> ElementTypes;
 
 private:
   ElementTypes ETypes;                              // Element types of struct
@@ -245,10 +244,10 @@
   SequentialType(const SequentialType &);                  // Do not implement!
   const SequentialType &operator=(const SequentialType &); // Do not implement!
 protected:
-  PATypeHandle<Type> ElementType;
+  PATypeHandle ElementType;
 
   SequentialType(PrimitiveID TID, const Type *ElType)
-    : CompositeType(TID), ElementType(PATypeHandle<Type>(ElType, this)) {
+    : CompositeType(TID), ElementType(PATypeHandle(ElType, this)) {
   }
 public:
 
@@ -390,18 +389,17 @@
 // contains an AbstractTypeUser instance, so there is no good way to factor out
 // the code.  Hence this bit of uglyness.
 //
-template <class TypeSubClass> void PATypeHandle<TypeSubClass>::addUser() {
+inline void PATypeHandle::addUser() {
   assert(Ty && "Type Handle has a null type!");
   if (Ty->isAbstract())
     cast<DerivedType>(Ty)->addAbstractTypeUser(User);
 }
-template <class TypeSubClass> void PATypeHandle<TypeSubClass>::removeUser() {
+inline void PATypeHandle::removeUser() {
   if (Ty->isAbstract())
     cast<DerivedType>(Ty)->removeAbstractTypeUser(User);
 }
 
-template <class TypeSubClass>
-void PATypeHandle<TypeSubClass>::removeUserFromConcrete() {
+inline void PATypeHandle::removeUserFromConcrete() {
   if (!Ty->isAbstract())
     cast<DerivedType>(Ty)->removeAbstractTypeUser(User);
 }


Index: llvm/include/llvm/GlobalVariable.h
diff -u llvm/include/llvm/GlobalVariable.h:1.20.2.1 llvm/include/llvm/GlobalVariable.h:1.20.2.2
--- llvm/include/llvm/GlobalVariable.h:1.20.2.1	Mon Jun 23 13:51:40 2003
+++ llvm/include/llvm/GlobalVariable.h	Thu Jun 26 16:34:37 2003
@@ -14,8 +14,8 @@
 #define LLVM_GLOBAL_VARIABLE_H
 
 #include "Config/assert.h"
-
 #include "llvm/GlobalValue.h"
+
 class Module;
 class Constant;
 class PointerType;


Index: llvm/include/llvm/Instruction.h
diff -u llvm/include/llvm/Instruction.h:1.40.2.1 llvm/include/llvm/Instruction.h:1.40.2.2
--- llvm/include/llvm/Instruction.h:1.40.2.1	Mon Jun 23 13:51:40 2003
+++ llvm/include/llvm/Instruction.h	Thu Jun 26 16:34:37 2003
@@ -9,8 +9,8 @@
 #define LLVM_INSTRUCTION_H
 
 #include "Config/assert.h"
-
 #include "llvm/User.h"
+
 template<typename SC> struct ilist_traits;
 template<typename ValueSubClass, typename ItemParentClass, typename SymTabClass,
          typename SubClass> class SymbolTableListTraits;


Index: llvm/include/llvm/SymbolTable.h
diff -u llvm/include/llvm/SymbolTable.h:1.21.2.1 llvm/include/llvm/SymbolTable.h:1.21.2.2
--- llvm/include/llvm/SymbolTable.h:1.21.2.1	Mon Jun 23 13:51:40 2003
+++ llvm/include/llvm/SymbolTable.h	Thu Jun 26 16:34:37 2003
@@ -17,7 +17,6 @@
 #define LLVM_SYMBOL_TABLE_H
 
 #include "Config/assert.h"
-
 #include "llvm/Value.h"
 #include <map>
 


Index: llvm/include/llvm/User.h
diff -u llvm/include/llvm/User.h:1.18.2.1 llvm/include/llvm/User.h:1.18.2.2
--- llvm/include/llvm/User.h:1.18.2.1	Mon Jun 23 13:51:40 2003
+++ llvm/include/llvm/User.h	Thu Jun 26 16:34:37 2003
@@ -13,7 +13,6 @@
 #define LLVM_USER_H
 
 #include "Config/assert.h"
-
 #include "llvm/Value.h"
 
 class User : public Value {
@@ -48,6 +47,11 @@
   inline const_op_iterator op_begin() const { return Operands.begin(); }
   inline op_iterator       op_end()         { return Operands.end(); }
   inline const_op_iterator op_end()   const { return Operands.end(); }
+
+  /// op_erase - This method is used to remove one of the arguments from the
+  /// operands list.  Only use this if you know what you are doing.
+  ///
+  op_iterator op_erase(op_iterator I) { return Operands.erase(I); }
 
   // dropAllReferences() - This function is in charge of "letting go" of all
   // objects that this User refers to.  This allows one to


Index: llvm/include/llvm/Value.h
diff -u llvm/include/llvm/Value.h:1.40 llvm/include/llvm/Value.h:1.40.2.1
--- llvm/include/llvm/Value.h:1.40	Tue Jan 14 15:29:58 2003
+++ llvm/include/llvm/Value.h	Thu Jun 26 16:34:37 2003
@@ -50,7 +50,7 @@
 private:
   std::vector<User *> Uses;
   std::string Name;
-  PATypeHandle<Type> Ty;
+  PATypeHandle Ty;
   ValueTy VTy;
 
   void operator=(const Value &);     // Do not implement


Index: llvm/include/llvm/iOther.h
diff -u llvm/include/llvm/iOther.h:1.35.2.1 llvm/include/llvm/iOther.h:1.35.2.2
--- llvm/include/llvm/iOther.h:1.35.2.1	Mon Jun 23 13:51:40 2003
+++ llvm/include/llvm/iOther.h	Thu Jun 26 16:34:37 2003
@@ -9,7 +9,6 @@
 #define LLVM_IOTHER_H
 
 #include "Config/assert.h"
-
 #include "llvm/InstrTypes.h"
 
 //===----------------------------------------------------------------------===//


Index: llvm/include/llvm/iPHINode.h
diff -u llvm/include/llvm/iPHINode.h:1.9.2.1 llvm/include/llvm/iPHINode.h:1.9.2.2
--- llvm/include/llvm/iPHINode.h:1.9.2.1	Mon Jun 23 13:51:40 2003
+++ llvm/include/llvm/iPHINode.h	Thu Jun 26 16:34:37 2003
@@ -8,7 +8,6 @@
 #define LLVM_IPHINODE_H
 
 #include "Config/assert.h"
-
 #include "llvm/Instruction.h"
 class BasicBlock;
 


Index: llvm/include/llvm/iTerminators.h
diff -u llvm/include/llvm/iTerminators.h:1.27.2.1 llvm/include/llvm/iTerminators.h:1.27.2.2
--- llvm/include/llvm/iTerminators.h:1.27.2.1	Mon Jun 23 13:51:40 2003
+++ llvm/include/llvm/iTerminators.h	Thu Jun 26 16:34:37 2003
@@ -10,12 +10,11 @@
 #define LLVM_ITERMINATORS_H
 
 #include "Config/assert.h"
-
 #include "llvm/InstrTypes.h"
 
 //===---------------------------------------------------------------------------
-// ReturnInst - Return a value (possibly void), from a method.  Execution does
-//              not continue in this method any longer.
+// ReturnInst - Return a value (possibly void), from a function.  Execution does
+//              not continue in this function any longer.
 //
 class ReturnInst : public TerminatorInst {
   ReturnInst(const ReturnInst &RI) : TerminatorInst(Instruction::Ret) {
@@ -194,7 +193,7 @@
 class InvokeInst : public TerminatorInst {
   InvokeInst(const InvokeInst &BI);
 public:
-  InvokeInst(Value *Meth, BasicBlock *IfNormal, BasicBlock *IfException,
+  InvokeInst(Value *Fn, BasicBlock *IfNormal, BasicBlock *IfException,
 	     const std::vector<Value*> &Params, const std::string &Name = "",
              Instruction *InsertBefore = 0);
 
@@ -212,7 +211,7 @@
     return dyn_cast<Function>(Operands[0].get());
   }
 
-  // getCalledValue - Get a pointer to a method that is invoked by this inst.
+  // getCalledValue - Get a pointer to a function that is invoked by this inst.
   inline const Value *getCalledValue() const { return Operands[0]; }
   inline       Value *getCalledValue()       { return Operands[0]; }
 




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:38:51 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:38:51 2003
Subject: [llvm-commits] CVS: llvm/lib/CodeGen/InstrSelection/InstrForest.cpp
Message-ID: <200306262137.h5QLbQ632684@choi.cs.uiuc.edu>


Changes in directory llvm/lib/CodeGen/InstrSelection:

InstrForest.cpp updated: 1.40.2.1 -> 1.40.2.2

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/lib/CodeGen/InstrSelection/InstrForest.cpp
diff -u llvm/lib/CodeGen/InstrSelection/InstrForest.cpp:1.40.2.1 llvm/lib/CodeGen/InstrSelection/InstrForest.cpp:1.40.2.2
--- llvm/lib/CodeGen/InstrSelection/InstrForest.cpp:1.40.2.1	Mon Jun 23 14:07:05 2003
+++ llvm/lib/CodeGen/InstrSelection/InstrForest.cpp	Thu Jun 26 16:35:16 2003
@@ -22,8 +22,6 @@
 #include "llvm/CodeGen/MachineInstr.h"
 #include "Support/STLExtras.h"
 #include "Config/alloca.h"
-using std::cerr;
-using std::vector;
 
 //------------------------------------------------------------------------ 
 // class InstrTreeNode
@@ -117,9 +115,9 @@
 InstructionNode::dumpNode(int indent) const
 {
   for (int i=0; i < indent; i++)
-    cerr << "    ";
-  cerr << getInstruction()->getOpcodeName()
-       << " [label " << getOpLabel() << "]" << "\n";
+    std::cerr << "    ";
+  std::cerr << getInstruction()->getOpcodeName()
+            << " [label " << getOpLabel() << "]" << "\n";
 }
 
 
@@ -127,9 +125,9 @@
 VRegListNode::dumpNode(int indent) const
 {
   for (int i=0; i < indent; i++)
-    cerr << "    ";
+    std::cerr << "    ";
   
-  cerr << "List" << "\n";
+  std::cerr << "List" << "\n";
 }
 
 
@@ -137,29 +135,29 @@
 VRegNode::dumpNode(int indent) const
 {
   for (int i=0; i < indent; i++)
-    cerr << "    ";
+    std::cerr << "    ";
   
-  cerr << "VReg " << getValue() << "\t(type "
-       << (int) getValue()->getValueType() << ")" << "\n";
+  std::cerr << "VReg " << getValue() << "\t(type "
+            << (int) getValue()->getValueType() << ")" << "\n";
 }
 
 void
 ConstantNode::dumpNode(int indent) const
 {
   for (int i=0; i < indent; i++)
-    cerr << "    ";
+    std::cerr << "    ";
   
-  cerr << "Constant " << getValue() << "\t(type "
-       << (int) getValue()->getValueType() << ")" << "\n";
+  std::cerr << "Constant " << getValue() << "\t(type "
+            << (int) getValue()->getValueType() << ")" << "\n";
 }
 
 void
 LabelNode::dumpNode(int indent) const
 {
   for (int i=0; i < indent; i++)
-    cerr << "    ";
+    std::cerr << "    ";
   
-  cerr << "Label " << getValue() << "\n";
+  std::cerr << "Label " << getValue() << "\n";
 }
 
 //------------------------------------------------------------------------
@@ -258,8 +256,7 @@
   // if a fixed array is too small.
   // 
   int numChildren = 0;
-  InstrTreeNode **childArray =
-    (InstrTreeNode **)alloca(instr->getNumOperands()*sizeof(InstrTreeNode *));
+  InstrTreeNode** childArray = new InstrTreeNode*[instr->getNumOperands()];
   
   //
   // Walk the operands of the instruction
@@ -366,6 +363,7 @@
       assert(n == 1);
       setRightChild(parent, childArray[numChildren - 1]);
     }
-  
+
+  delete [] childArray;
   return treeNode;
 }




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:38:57 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:38:57 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Reoptimizer/inst/Makefile Test1.c
Message-ID: <200306262137.h5QLbRm32698@choi.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Reoptimizer/inst:

Makefile updated: 1.5 -> 1.5.4.1
Test1.c updated: 1.4 -> 1.4.4.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/test/Regression/Reoptimizer/inst/Makefile
diff -u llvm/test/Regression/Reoptimizer/inst/Makefile:1.5 llvm/test/Regression/Reoptimizer/inst/Makefile:1.5.4.1
--- llvm/test/Regression/Reoptimizer/inst/Makefile:1.5	Thu May 22 22:25:43 2003
+++ llvm/test/Regression/Reoptimizer/inst/Makefile	Thu Jun 26 16:35:47 2003
@@ -1,4 +1,5 @@
-LEVEL = /localhome/$(USER)/Research/llvm
+LEVEL = ../../../../
+#LEVEL = /localhome/$(USER)/Research/llvm
 
 CC = /usr/dcs/software/evaluation/bin/gcc
 CXX = /usr/dcs/software/evaluation/bin/g++
@@ -10,8 +11,8 @@
 	   $(LEVEL)/lib/Debug/libtracecache.a   \
 	   $(LEVEL)/lib/Debug/libbininterface.a
 
-
-LIBINSTRBC = /home/vadve/lattner/cvs/gcc_install_sparc/llvm/lib/libinstr.bc
+CRTEND = $(LLVM_LIB_SEARCH_PATH)/crtend.o
+LIBINSTRBC = $(LEVEL)/test/Libraries/Output/libinstr.bc
 
 Test1: Test1.o $(PERFOBJS)
         # NB: The only reason we don't include the PAPI library in the list of exclusions
@@ -19,8 +20,8 @@
         #     tables, and thus will never try to instrument the PAPI functions, which are
         #     dynamically linked.
 
-	mkexcl $(PERFOBJS) libinstr.o > mkexcl.cpp
-	$(CXX) -o Test1 $^ mkexcl.cpp -L$(PAPIDIR) -lpapi -lelf
+	mkexcl $(PERFOBJS) libinstr.o crtend.o > mkexcl.cpp
+	$(CXX) -o Test1 $^ mkexcl.cpp crtend.o -L$(PAPIDIR) -lpapi -lelf
 
 Test1.o: Test1.opt.c
 	$(CC) -c Test1.opt.c -o Test1.o
@@ -28,12 +29,16 @@
 Test1.opt.c: Test1.opt.bc
 	dis -c < Test1.opt.bc > Test1.opt.c
 
-Test1.opt.bc: Test1.bc libinstr.o $(LEVEL)/lib/Debug/libperf.so
+Test1.opt.bc: Test1.bc libinstr.o crtend.o $(LEVEL)/lib/Debug/libperf.so
 	opt -load $(LEVEL)/lib/Debug/libperf.so -pp1 -emitfuncs -deadtypeelim < Test1.bc > Test1.opt.o
 	gccld -disable-internalize $(LIBINSTRBC) Test1.opt.o -o Test1.opt
 
 Test1.bc: Test1.c
 	$(LLVMGCC) -I$(LOCAL_CVSROOT)/llvm/lib/Reoptimizer/Inst/rtl -c Test1.c -o Test1.bc
+
+crtend.o: $(CRTEND)
+	dis -c < $^ > $(@:.o=.c)
+	$(CC) -c $(@:.o=.c) 2>/dev/null
 
 libinstr.o: $(LIBINSTRBC)
 	dis -c < $^ > $(@:.o=.c)


Index: llvm/test/Regression/Reoptimizer/inst/Test1.c
diff -u llvm/test/Regression/Reoptimizer/inst/Test1.c:1.4 llvm/test/Regression/Reoptimizer/inst/Test1.c:1.4.4.1
--- llvm/test/Regression/Reoptimizer/inst/Test1.c:1.4	Thu May 22 22:25:43 2003
+++ llvm/test/Regression/Reoptimizer/inst/Test1.c	Thu Jun 26 16:35:47 2003
@@ -3,41 +3,30 @@
 // The following is actually the semantic result of processing a metric
 // declaration (need more sigfun support)
 
-double elapsedTime = 0;
-double elapsedTime2 = 0;
+double elapsedTime = 3.14;
 int l1cachemiss = 0;
 
 #include <stdio.h>
 
-void phase2();
-
 int fib(int n);
 void fibs();
 
-void someOtherFunction()
-{
-    int x;
-    pp_L1_cache_miss_start(&x);
-    printf("this statement resides within a region!\n");
-    pp_L1_cache_miss_end(&l1cachemiss, &x);
-    printf("someOtherFunction complete: retval is %d\n", l1cachemiss);
-}
-
 int main(int argc, char** argv) 
 {
     phase2();
 
-    printf("Just about to call fibs()...\n");
+    registerIntervalInst(42, pp_elapsed_time_start,
+                         pp_elapsed_time_end, sizeof(double),
+                         &elapsedTime);
+
+    printf("Just about to call fibs() for the first time...\n");
     fibs();
 
-    printf("Just about to call someOtherFunction()...\n");
-    someOtherFunction();
+    printf("Just about to call fibs() for the second time...\n");
+    fibs();
 
     printf("Leaving main...\n");
 
-    pp_regionPair(pp_elapsed_time_start, pp_elapsed_time_end);
-    pp_regionPair(pp_L1_cache_miss_start, pp_L1_cache_miss_end);
-
     return 0;
 }
 
@@ -49,39 +38,12 @@
         return fib(n-2) + fib(n-1);
 }
 
-void functionToBeInlined()
-{
-    double y;
-    pp_elapsed_time_start(&y);
-    printf("This function should be inlined...");
-    pp_elapsed_time_end(&elapsedTime2, &y);
-}
-
 void fibs() 
 {
     int i;
+    double tmp;
 
-    printf("Inside fibs now, calling functionToBeInlined...\n");
-    
-    functionToBeInlined();
-
-    ////////////////
-    // The following is the "scope" that the sampling applies to, represented by
-    // sigfuns. We need better sigfun support.
-    //
-    // That is, the below should look like:
-    //
-    // pp_region_start(elapsedTime);
-    // <for loop>
-    // pp_region_end(elaspedTime);
-    //
-    // or something similar. We should not be putting the actual invoked functions below,
-    // although the post-phase1 transformed code should be "similar" (calls removed and
-    // volatile loads put in their place, of course!)
-
-    long x;
-    double y;
-    pp_elapsed_time_start(&y);
+    pp_declareIntervalStart(&tmp, 42);
 
     for(i = 10; i > 0; i--) {
         printf("fib(%d) = %d\n", i, fib(i));
@@ -90,9 +52,14 @@
     printf("after for loop...\n");
     fflush(stdout);
 
-    pp_elapsed_time_end(&elapsedTime, &y);
+    registerIntervalInst(42, pp_L1_cache_miss_start,
+                         pp_L1_cache_miss_end, sizeof(int),
+                         &l1cachemiss);
+
+    pp_declareIntervalEnd(42, &tmp);
 
-    printf("ElapsedTime (metric variable) has address %lx, value %f\n", &elapsedTime, elapsedTime);
+    printf("value of variable elapsedTime: %f\n", elapsedTime);
+    printf("value of variable l1cachemiss: %d\n", l1cachemiss);
     
     //
     ////////////////




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:39:03 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:39:03 2003
Subject: [llvm-commits] CVS: llvm/test/Programs/Makefile.programs TEST.nightly.Makefile
Message-ID: <200306262137.h5QLblf00586@choi.cs.uiuc.edu>


Changes in directory llvm/test/Programs:

Makefile.programs updated: 1.60 -> 1.60.2.1
TEST.nightly.Makefile updated: 1.15 -> 1.15.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/test/Programs/Makefile.programs
diff -u llvm/test/Programs/Makefile.programs:1.60 llvm/test/Programs/Makefile.programs:1.60.2.1
--- llvm/test/Programs/Makefile.programs:1.60	Sun Jun 22 15:21:29 2003
+++ llvm/test/Programs/Makefile.programs	Thu Jun 26 16:35:37 2003
@@ -176,10 +176,10 @@
 # Link the program to the libraries it uses, then perform postlink
 # optimization...
 $(PROGRAMS_TO_TEST:%=Output/%.llvm.bc): \
-Output/%.llvm.bc: Output/%.linked.bc
+Output/%.llvm.bc: Output/%.linked.bc $(LGCCLDPROG)
 	$(LGCCLD) $(STATS) $< -lgcc -lc $(LIBS) crtend.o -o Output/$*.llvm
 $(PROGRAMS_TO_TEST:%=Output/%.llvm): \
-Output/%.llvm: Output/%.linked.bc
+Output/%.llvm: Output/%.linked.bc $(LGCCLDPROG)
 	$(LGCCLD) $(STATS) $< -lgcc -lc $(LIBS) crtend.o -o Output/$*.llvm
 
 # Rule to get the tracing version of the llvm.bc file for tracing:


Index: llvm/test/Programs/TEST.nightly.Makefile
diff -u llvm/test/Programs/TEST.nightly.Makefile:1.15 llvm/test/Programs/TEST.nightly.Makefile:1.15.2.1
--- llvm/test/Programs/TEST.nightly.Makefile:1.15	Sat Jun 21 21:38:29 2003
+++ llvm/test/Programs/TEST.nightly.Makefile	Thu Jun 26 16:35:37 2003
@@ -21,8 +21,9 @@
 # Compilation tests
 $(PROGRAMS_TO_TEST:%=Output/%.nightly.compile.report.txt): \
 Output/%.nightly.compile.report.txt: Output/%.llvm.bc $(LGCCAS)
-	@echo '$(LGCCAS) Output/$*.linked.rll -o /dev/null $(TIMEOPT) > $@ 2>&1'
-	@if ($(LGCCAS) Output/$*.linked.rll -o /dev/null $(TIMEOPT) > $@ 2>&1)\
+	@echo > $@
+	@echo '$(LGCCAS) Output/$*.linked.rll -o/dev/null $(TIMEOPT) >> $@ 2>&1'
+	@-if ($(LGCCAS) Output/$*.linked.rll -o /dev/null $(TIMEOPT) >> $@ 2>&1)\
 	;then \
 	  echo "TEST-PASS: compile $(RELDIR)/$*" >> $@;\
 	  echo -n "TEST-RESULT-compile: " >> $@;\
@@ -42,14 +43,16 @@
 # NAT tests
 $(PROGRAMS_TO_TEST:%=Output/%.nightly.nat.report.txt): \
 Output/%.nightly.nat.report.txt: Output/%.out-nat
+	@echo > $@
 	echo -n "TEST-RESULT-nat-time: " >> $@
 	-grep "^real" Output/$*.out-nat.time >> $@
 
 # LLC tests
 $(PROGRAMS_TO_TEST:%=Output/%.nightly.llc.report.txt): \
 Output/%.nightly.llc.report.txt: Output/%.llvm.bc $(LLC)
+	@echo > $@
 	@echo 'time -p $(LLC) -f $(TIMEOPT) $< -o /dev/null) > $@ 2>&1'
-	@if (time -p $(LLC) -f $(TIMEOPT) $< -o /dev/null) > $@ 2>&1; then \
+	@-if (time -p $(LLC) -f $(TIMEOPT) $< -o /dev/null) >> $@ 2>&1; then \
 	  echo "TEST-PASS: llc $(RELDIR)/$*" >> $@;\
 	  echo -n "TEST-RESULT-llc: " >> $@;\
 	  grep "Total Execution Time" $@.info >> $@;\
@@ -65,8 +68,9 @@
 # CBE tests
 $(PROGRAMS_TO_TEST:%=Output/%.nightly.cbe.report.txt): \
 Output/%.nightly.cbe.report.txt: Output/%.llvm.bc Output/%.exe-cbe $(LDIS)
-	-head -n 100 Output/$*.exe-cbe > $@
-	@if test -f Output/$*.exe-cbe; then \
+	@echo > $@
+	-head -n 100 Output/$*.exe-cbe >> $@
+	@-if test -f Output/$*.exe-cbe; then \
 	  echo "TEST-PASS: cbe $(RELDIR)/$*" >> $@;\
           echo "TEST-RESULT-cbe: YES" >> $@;\
 	  echo -n "TEST-RESULT-cbe-time: " >> $@;\
@@ -79,8 +83,9 @@
 # JIT tests
 $(PROGRAMS_TO_TEST:%=Output/%.nightly.jit.report.txt): \
 Output/%.nightly.jit.report.txt: Output/%.llvm.bc Output/%.exe-jit $(LLI)
-	-head -n 100 Output/$*.exe-jit > $@
-	@if test -f Output/$*.exe-jit; then \
+	@echo > $@
+	-head -n 100 Output/$*.exe-jit >> $@
+	@-if test -f Output/$*.exe-jit; then \
           echo "TEST-PASS: jit $(RELDIR)/$*" >> $@;\
 	  echo -n "TEST-RESULT-jit-time: " >> $@;\
 	  grep "^real" $(INFO_PREFIX)jit.time >> $@;\
@@ -99,7 +104,7 @@
 # Overall tests: just run subordinate tests
 $(PROGRAMS_TO_TEST:%=Output/%.$(TEST).report.txt): \
 Output/%.$(TEST).report.txt: $(addprefix Output/%.nightly., $(REPORTS_SUFFIX))
-	cat $(addprefix Output/$*.nightly., $(REPORTS_SUFFIX)) > $@
+	-cat $(addprefix Output/$*.nightly., $(REPORTS_SUFFIX)) > $@
 
 
 
@@ -108,6 +113,6 @@
 	@echo "---------------------------------------------------------------"
 	@echo ">>> ========= '$(RELDIR)/$*' Program"
 	@echo "---------------------------------------------------------------"
-	@cat $<
+	@-cat $<
 
 REPORT_DEPENDENCIES := $(LDIS) $(LLI) $(LLC)




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:39:08 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:39:08 2003
Subject: [llvm-commits] CVS: llvm/tools/opt/Makefile
Message-ID: <200306262137.h5QLbRt32690@choi.cs.uiuc.edu>


Changes in directory llvm/tools/opt:

Makefile updated: 1.39.2.1 -> 1.39.2.2

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/tools/opt/Makefile
diff -u llvm/tools/opt/Makefile:1.39.2.1 llvm/tools/opt/Makefile:1.39.2.2
--- llvm/tools/opt/Makefile:1.39.2.1	Mon Jun 23 14:04:55 2003
+++ llvm/tools/opt/Makefile	Thu Jun 26 16:36:17 2003
@@ -6,4 +6,6 @@
            ipo ipa.a datastructure transforms target.a analysis \
            transformutils vmcore support
 
+TOOLLINKOPTS = $(PLATFORMLIBDL)
+
 include $(LEVEL)/Makefile.common




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:39:14 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:39:14 2003
Subject: [llvm-commits] CVS: llvm/Makefile.common
Message-ID: <200306262137.h5QLbRo32704@choi.cs.uiuc.edu>


Changes in directory llvm:

Makefile.common updated: 1.88.2.5 -> 1.88.2.6

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/Makefile.common
diff -u llvm/Makefile.common:1.88.2.5 llvm/Makefile.common:1.88.2.6
--- llvm/Makefile.common:1.88.2.5	Thu Jun 26 10:53:56 2003
+++ llvm/Makefile.common	Thu Jun 26 16:33:47 2003
@@ -81,18 +81,24 @@
 # Set the source root directory.
 #
 ifndef BUILD_SRC_ROOT
-BUILD_SRC_ROOT = $(BUILD_SRC_DIR)/$(LEVEL)
+BUILD_SRC_ROOT = $(shell cd $(BUILD_SRC_DIR)/$(LEVEL); pwd)
 endif
 
 #
+# Determine the path of the source tree relative from $HOME (the mythical
+# home directory).
+#
+HOME_OBJ_ROOT := $(OBJ_ROOT)/$(patsubst $(HOME)%,%,$(BUILD_SRC_ROOT))
+
+#
 # Set the object build directory.  Its location depends upon the source path
 # and where object files should go.
 #
 ifndef BUILD_OBJ_DIR
 ifeq ($(OBJ_ROOT),.)
-BUILD_OBJ_DIR = $(shell pwd)
+BUILD_OBJ_DIR = $(BUILD_SRC_DIR)
 else
-BUILD_OBJ_DIR := $(OBJ_ROOT)$(patsubst $(HOME)%,%,$(shell cd $(BUILD_SRC_DIR); pwd))
+BUILD_OBJ_DIR := $(HOME_OBJ_ROOT)$(patsubst $(BUILD_SRC_ROOT)%,%,$(BUILD_SRC_DIR))
 endif
 endif
 
@@ -101,9 +107,9 @@
 #
 ifndef BUILD_OBJ_ROOT
 ifeq ($(OBJ_ROOT),.)
-BUILD_OBJ_ROOT = $(shell cd $(LEVEL); pwd)
+BUILD_OBJ_ROOT = $(BUILD_SRC_ROOT)
 else
-BUILD_OBJ_ROOT := $(OBJ_ROOT)$(patsubst $(HOME)%,%,$(shell cd $(LEVEL); pwd))
+BUILD_OBJ_ROOT := $(HOME_OBJ_ROOT)
 endif
 endif
 
@@ -141,6 +147,16 @@
 
 # Default rule for test.  It ensures everything has a test rule
 test::
+
+# Print out the directories used for building
+prdirs::
+	echo "Home Offset      : " $(HOME_OBJ_ROOT);
+	echo "Build Source Root: " $(BUILD_SRC_ROOT);
+	echo "Build Source Dir : " $(BUILD_SRC_DIR);
+	echo "Build Object Root: " $(BUILD_OBJ_ROOT);
+	echo "Build Object Dir : " $(BUILD_OBJ_DIR);
+	echo "LLVM  Source Root: " $(LLVM_SRC_ROOT);
+	echo "LLVM  Object Root: " $(LLVM_OBJ_ROOT);
 
 ###########################################################################
 # Miscellaneous paths and commands:




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:39:20 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:39:20 2003
Subject: [llvm-commits] CVS: llvm/test/Libraries/libinstr/tracelib.c
Message-ID: <200306262137.h5QLbkU00576@choi.cs.uiuc.edu>


Changes in directory llvm/test/Libraries/libinstr:

tracelib.c updated: 1.3 -> 1.3.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/test/Libraries/libinstr/tracelib.c
diff -u llvm/test/Libraries/libinstr/tracelib.c:1.3 llvm/test/Libraries/libinstr/tracelib.c:1.3.2.1
--- llvm/test/Libraries/libinstr/tracelib.c:1.3	Tue May 27 16:43:14 2003
+++ llvm/test/Libraries/libinstr/tracelib.c	Thu Jun 26 16:35:36 2003
@@ -10,7 +10,9 @@
 #include <stdlib.h>
 #include <stdio.h>
 #include <string.h>
+#ifndef sun
 #include <stdint.h>
+#endif
 
 /*===---------------------------------------------------------------------=====
  * HASH FUNCTIONS




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:39:26 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:39:26 2003
Subject: [llvm-commits] CVS: llvm/lib/Reoptimizer/BinInterface/sparcbin.cpp
Message-ID: <200306262137.h5QLbSG32711@choi.cs.uiuc.edu>


Changes in directory llvm/lib/Reoptimizer/BinInterface:

sparcbin.cpp updated: 1.7 -> 1.7.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/lib/Reoptimizer/BinInterface/sparcbin.cpp
diff -u llvm/lib/Reoptimizer/BinInterface/sparcbin.cpp:1.7 llvm/lib/Reoptimizer/BinInterface/sparcbin.cpp:1.7.2.1
--- llvm/lib/Reoptimizer/BinInterface/sparcbin.cpp:1.7	Wed Jun  4 04:40:54 2003
+++ llvm/lib/Reoptimizer/BinInterface/sparcbin.cpp	Thu Jun 26 16:35:18 2003
@@ -274,7 +274,7 @@
       if (instr->flags & IF_BRINTERNAL)
 	printf("{dest:s%02d} ", instr->ibranch.dest);
       else
-	printf("{dest:0x%x} ", (intptr_t)instr->ebranch.dest);
+	printf("{dest:%p} ", instr->ebranch.dest);
       
       sparc_print_pseudo(instr->instr, 
 			 instr->alu.genrs1, 




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:39:32 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:39:32 2003
Subject: [llvm-commits] CVS: llvm/utils/NightlyTest.pl NightlyTestTemplate.html
Message-ID: <200306262137.h5QLbSj32719@choi.cs.uiuc.edu>


Changes in directory llvm/utils:

NightlyTest.pl updated: 1.12 -> 1.12.2.1
NightlyTestTemplate.html updated: 1.7 -> 1.7.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/utils/NightlyTest.pl
diff -u llvm/utils/NightlyTest.pl:1.12 llvm/utils/NightlyTest.pl:1.12.2.1
--- llvm/utils/NightlyTest.pl:1.12	Sun May 11 10:23:10 2003
+++ llvm/utils/NightlyTest.pl	Thu Jun 26 16:36:18 2003
@@ -155,9 +155,9 @@
 if (!$NOCHECKOUT) {
   # Change the Makefile.config to build into the local directory...
   rename "Makefile.config", "Makefile.config.orig";
-  system "sed '/^LLVM_OBJ_DIR/d' < Makefile.config.orig > Makefile.config";
+  system "sed '/^OBJ_ROOT/d' < Makefile.config.orig > Makefile.config";
   system "echo >> Makefile.config";
-  system "echo 'LLVM_OBJ_DIR := .' >> Makefile.config";
+  system "echo 'OBJ_ROOT := .' >> Makefile.config";
 
   # Change the Makefile.config to not strip executables...
   system "echo 'KEEP_SYMBOLS := 1' >> Makefile.config";


Index: llvm/utils/NightlyTestTemplate.html
diff -u llvm/utils/NightlyTestTemplate.html:1.7 llvm/utils/NightlyTestTemplate.html:1.7.2.1
--- llvm/utils/NightlyTestTemplate.html:1.7	Sat May 10 15:54:55 2003
+++ llvm/utils/NightlyTestTemplate.html	Thu Jun 26 16:36:18 2003
@@ -126,15 +126,11 @@
 <li><a name="Bytecode">Bytecode - The size of the bytecode for the program
 <li><a name="Instrs">Instrs - The number of LLVM instructions in the
     compiled bytecode
-<li><a name="LLC">LLC - The time taken compile with the Sparc Backend.  Note
-    that this justs tests to see if the backend runs successfully, not if the
-    generated code works.
+<li><a name="LLC">LLC - The time taken compile with LLC (the static X86 backend)
+<li><a name="NAT">NAT - The time taken to execute the program when compiled
+    with GCC -O3.
 <li><a name="CBE">CBE - The time taken to execute the program after compilation
     through the C Backend.  If the diff fails, this a *.
-<li><a name="LLI">LLI - How long does the program take to execute in
-    LLI's interpreter (note that this is capped to ~300s).
-<li><a name="DynInstrs">DynInstrs - How many dynamic instructions are executed
-    by the interpreter, if it finishes.
 <li><a name="JIT">JIT - How long does the program take to execute in the
     Just-In-Time compiler.
 <li><a name="MachCode">MachCode - The number of bytes of machine code




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:39:38 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:39:38 2003
Subject: [llvm-commits] CVS: llvm/lib/Analysis/AliasAnalysisCounter.cpp BasicAliasAnalysis.cpp Expressions.cpp LoadValueNumbering.cpp
Message-ID: <200306262137.h5QLbkc00570@choi.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis:

AliasAnalysisCounter.cpp updated: 1.3 -> 1.3.2.1
BasicAliasAnalysis.cpp updated: 1.10 -> 1.10.2.1
Expressions.cpp updated: 1.31 -> 1.31.2.1
LoadValueNumbering.cpp updated: 1.2 -> 1.2.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/lib/Analysis/AliasAnalysisCounter.cpp
diff -u llvm/lib/Analysis/AliasAnalysisCounter.cpp:1.3 llvm/lib/Analysis/AliasAnalysisCounter.cpp:1.3.2.1
--- llvm/lib/Analysis/AliasAnalysisCounter.cpp:1.3	Wed Feb 26 13:27:05 2003
+++ llvm/lib/Analysis/AliasAnalysisCounter.cpp	Thu Jun 26 16:35:05 2003
@@ -32,23 +32,27 @@
           << "\n===== Alias Analysis Counter Report =====\n"
           << "  Analysis counted: " << Name << "\n"
           << "  " << AASum << " Total Alias Queries Performed\n";
-        printLine("no alias",     No, AASum);
-        printLine("may alias",   May, AASum);
-        printLine("must alias", Must, AASum);
-        std::cerr
-          << "  Alias Analysis Counter Summary: " << No*100/AASum << "%/"
-          << May*100/AASum << "%/" << Must*100/AASum<<"%\n\n";
+        if (AASum) {
+          printLine("no alias",     No, AASum);
+          printLine("may alias",   May, AASum);
+          printLine("must alias", Must, AASum);
+          std::cerr
+            << "  Alias Analysis Counter Summary: " << No*100/AASum << "%/"
+            << May*100/AASum << "%/" << Must*100/AASum<<"%\n\n";
+        }
 
         std::cerr
           << "  " << MRSum    << " Total Mod/Ref Queries Performed\n";
-        printLine("no mod/ref",    NoMR, MRSum);
-        printLine("ref",        JustRef, MRSum);
-        printLine("mod",        JustMod, MRSum);
-        printLine("mod/ref",         MR, MRSum);
-        std::cerr
-          << "  Mod/Ref Analysis Counter Summary: " << NoMR*100/MRSum<< "%/"
-          << JustRef*100/MRSum << "%/" << JustMod*100/MRSum << "%/" 
-          << MR*100/MRSum <<"%\n\n";
+        if (MRSum) {
+          printLine("no mod/ref",    NoMR, MRSum);
+          printLine("ref",        JustRef, MRSum);
+          printLine("mod",        JustMod, MRSum);
+          printLine("mod/ref",         MR, MRSum);
+          std::cerr
+            << "  Mod/Ref Analysis Counter Summary: " << NoMR*100/MRSum<< "%/"
+            << JustRef*100/MRSum << "%/" << JustMod*100/MRSum << "%/" 
+            << MR*100/MRSum <<"%\n\n";
+        }
       }
     }
 


Index: llvm/lib/Analysis/BasicAliasAnalysis.cpp
diff -u llvm/lib/Analysis/BasicAliasAnalysis.cpp:1.10 llvm/lib/Analysis/BasicAliasAnalysis.cpp:1.10.2.1
--- llvm/lib/Analysis/BasicAliasAnalysis.cpp:1.10	Mon Jun  2 00:42:39 2003
+++ llvm/lib/Analysis/BasicAliasAnalysis.cpp	Thu Jun 26 16:35:05 2003
@@ -69,6 +69,12 @@
   if (const Instruction *I = dyn_cast<Instruction>(V)) {
     if (isa<CastInst>(I) || isa<GetElementPtrInst>(I))
       return getUnderlyingObject(I->getOperand(0));
+  } else if (const ConstantExpr *CE = dyn_cast<ConstantExpr>(V)) {
+    if (CE->getOpcode() == Instruction::Cast ||
+        CE->getOpcode() == Instruction::GetElementPtr)
+      return getUnderlyingObject(CE->getOperand(0));
+  } else if (const ConstantPointerRef *CPR = dyn_cast<ConstantPointerRef>(V)) {
+    return CPR->getValue();
   }
   return 0;
 }


Index: llvm/lib/Analysis/Expressions.cpp
diff -u llvm/lib/Analysis/Expressions.cpp:1.31 llvm/lib/Analysis/Expressions.cpp:1.31.2.1
--- llvm/lib/Analysis/Expressions.cpp:1.31	Tue May 20 13:45:29 2003
+++ llvm/lib/Analysis/Expressions.cpp	Thu Jun 26 16:35:05 2003
@@ -317,12 +317,10 @@
     if (isa<PointerType>(DestTy))
       DestTy = Type::ULongTy;  // Pointer types are represented as ulong
 
-    /*
     if (!Src.getExprType(0)->isLosslesslyConvertibleTo(DestTy)) {
       if (Src.ExprTy != ExprType::Constant)
         return I;  // Converting cast, and not a constant value...
     }
-    */
 
     const ConstantInt *Offset = Src.Offset;
     const ConstantInt *Scale  = Src.Scale;


Index: llvm/lib/Analysis/LoadValueNumbering.cpp
diff -u llvm/lib/Analysis/LoadValueNumbering.cpp:1.2 llvm/lib/Analysis/LoadValueNumbering.cpp:1.2.2.1
--- llvm/lib/Analysis/LoadValueNumbering.cpp:1.2	Wed Feb 26 13:27:35 2003
+++ llvm/lib/Analysis/LoadValueNumbering.cpp	Thu Jun 26 16:35:06 2003
@@ -24,7 +24,7 @@
 #include <set>
 
 namespace {
-  // FIXME: This should not be a functionpass.
+  // FIXME: This should not be a FunctionPass.
   struct LoadVN : public FunctionPass, public ValueNumbering {
     
     /// Pass Implementation stuff.  This doesn't do any analysis.
@@ -51,6 +51,8 @@
     ///
     bool haveEqualValueNumber(LoadInst *LI, LoadInst *LI2, AliasAnalysis &AA,
                               DominatorSet &DomSetInfo) const;
+    bool haveEqualValueNumber(LoadInst *LI, StoreInst *SI, AliasAnalysis &AA,
+                              DominatorSet &DomSetInfo) const;
   };
 
   // Register this pass...
@@ -83,13 +85,13 @@
                                  std::vector<Value*> &RetVals) const {
 
   if (LoadInst *LI = dyn_cast<LoadInst>(V)) {
-    // If we have a load instruction, find all of the load instructions that use
-    // the same source operand.  We implement this recursively, because there
-    // could be a load of a load of a load that are all identical.  We are
-    // guaranteed that this cannot be an infinite recursion because load
-    // instructions would have to pass through a PHI node in order for there to
-    // be a cycle.  The PHI node would be handled by the else case here,
-    // breaking the infinite recursion.
+    // If we have a load instruction, find all of the load and store
+    // instructions that use the same source operand.  We implement this
+    // recursively, because there could be a load of a load of a load that are
+    // all identical.  We are guaranteed that this cannot be an infinite
+    // recursion because load instructions would have to pass through a PHI node
+    // in order for there to be a cycle.  The PHI node would be handled by the
+    // else case here, breaking the infinite recursion.
     //
     std::vector<Value*> PointerSources;
     getEqualNumberNodes(LI->getOperand(0), PointerSources);
@@ -98,30 +100,40 @@
     Function *F = LI->getParent()->getParent();
 
     // Now that we know the set of equivalent source pointers for the load
-    // instruction, look to see if there are any load candiates that are
-    // identical.
+    // instruction, look to see if there are any load or store candiates that
+    // are identical.
     //
     std::vector<LoadInst*> CandidateLoads;
+    std::vector<StoreInst*> CandidateStores;
+
     while (!PointerSources.empty()) {
       Value *Source = PointerSources.back();
       PointerSources.pop_back();                // Get a source pointer...
 
       for (Value::use_iterator UI = Source->use_begin(), UE = Source->use_end();
            UI != UE; ++UI)
-        if (LoadInst *Cand = dyn_cast<LoadInst>(*UI))  // Is a load of source?
+        if (LoadInst *Cand = dyn_cast<LoadInst>(*UI)) {// Is a load of source?
           if (Cand->getParent()->getParent() == F &&   // In the same function?
               Cand != LI)                              // Not LI itself?
             CandidateLoads.push_back(Cand);     // Got one...
+        } else if (StoreInst *Cand = dyn_cast<StoreInst>(*UI)) {
+          if (Cand->getParent()->getParent() == F &&
+              Cand->getOperand(1) == Source)  // It's a store THROUGH the ptr...
+            CandidateStores.push_back(Cand);
+        }
     }
 
     // Remove duplicates from the CandidateLoads list because alias analysis
     // processing may be somewhat expensive and we don't want to do more work
     // than neccesary.
     //
+    unsigned OldSize = CandidateLoads.size();
     std::sort(CandidateLoads.begin(), CandidateLoads.end());
     CandidateLoads.erase(std::unique(CandidateLoads.begin(),
                                      CandidateLoads.end()),
                          CandidateLoads.end());
+    // FIXME: REMOVE THIS SORTING AND UNIQUING IF IT CAN'T HAPPEN
+    assert(CandidateLoads.size() == OldSize && "Shrunk the candloads list?");
 
     // Get Alias Analysis...
     AliasAnalysis &AA = getAnalysis<AliasAnalysis>();
@@ -133,9 +145,11 @@
     for (unsigned i = 0, e = CandidateLoads.size(); i != e; ++i)
       if (haveEqualValueNumber(LI, CandidateLoads[i], AA, DomSetInfo))
         RetVals.push_back(CandidateLoads[i]);
-    
+    for (unsigned i = 0, e = CandidateStores.size(); i != e; ++i)
+      if (haveEqualValueNumber(LI, CandidateStores[i], AA, DomSetInfo))
+        RetVals.push_back(CandidateStores[i]->getOperand(0));
+      
   } else {
-    // Make sure passmanager doesn't try to fulfill our request with ourself!
     assert(&getAnalysis<ValueNumbering>() != (ValueNumbering*)this &&
            "getAnalysis() returned this!");
 
@@ -215,7 +229,7 @@
     return true;
   } else {
     // Make sure that there are no store instructions between L1 and the end of
-    // it's basic block...
+    // its basic block...
     //
     if (AA.canInstructionRangeModify(*L1, *BB1->getTerminator(), LoadAddress,
                                      LoadSize))
@@ -225,6 +239,74 @@
     // and the second load instruction...
     //
     if (AA.canInstructionRangeModify(BB2->front(), *L2, LoadAddress, LoadSize))
+      return false;   // Cannot eliminate load
+
+    // Do a depth first traversal of the inverse CFG starting at L2's block,
+    // looking for L1's block.  The inverse CFG is made up of the predecessor
+    // nodes of a block... so all of the edges in the graph are "backward".
+    //
+    std::set<BasicBlock*> VisitedSet;
+    for (pred_iterator PI = pred_begin(BB2), PE = pred_end(BB2); PI != PE; ++PI)
+      if (CheckForInvalidatingInst(*PI, BB1, LoadAddress, LoadSize, AA,
+                                   VisitedSet))
+        return false;
+
+    // If we passed all of these checks then we are sure that the two loads
+    // produce the same value.
+    return true;
+  }
+}
+
+
+/// haveEqualValueNumber - Given a load instruction and a store instruction,
+/// determine if the stored value reaches the loaded value unambiguously on
+/// every execution of the program.  This uses the AliasAnalysis implementation
+/// to invalidate the stored value when stores or function calls occur that
+/// could modify the value produced by the load.
+///
+bool LoadVN::haveEqualValueNumber(LoadInst *Load, StoreInst *Store,
+                                  AliasAnalysis &AA,
+                                  DominatorSet &DomSetInfo) const {
+  // If the store does not dominate the load, we cannot do anything...
+  if (!DomSetInfo.dominates(Store, Load)) 
+    return false;
+
+  BasicBlock *BB1 = Store->getParent(), *BB2 = Load->getParent();
+  Value *LoadAddress = Load->getOperand(0);
+
+  assert(LoadAddress->getType() == Store->getOperand(1)->getType() &&
+         "How could the same source pointer return different types?");
+
+  // Find out how many bytes of memory are loaded by the load instruction...
+  unsigned LoadSize = getAnalysis<TargetData>().getTypeSize(Load->getType());
+
+  // Compute a basic block iterator pointing to the instruction after the store.
+  BasicBlock::iterator StoreIt = Store; ++StoreIt;
+
+  // Check to see if the intervening instructions between the two store and load
+  // include a store or call...
+  //
+  if (BB1 == BB2) {  // In same basic block?
+    // In this degenerate case, no checking of global basic blocks has to occur
+    // just check the instructions BETWEEN Store & Load...
+    //
+    if (AA.canInstructionRangeModify(*StoreIt, *Load, LoadAddress, LoadSize))
+      return false;   // Cannot eliminate load
+
+    // No instructions invalidate the stored value, they produce the same value!
+    return true;
+  } else {
+    // Make sure that there are no store instructions between the Store and the
+    // end of its basic block...
+    //
+    if (AA.canInstructionRangeModify(*StoreIt, *BB1->getTerminator(),
+                                     LoadAddress, LoadSize))
+      return false;   // Cannot eliminate load
+
+    // Make sure that there are no store instructions between the start of BB2
+    // and the second load instruction...
+    //
+    if (AA.canInstructionRangeModify(BB2->front(), *Load, LoadAddress,LoadSize))
       return false;   // Cannot eliminate load
 
     // Do a depth first traversal of the inverse CFG starting at L2's block,




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:39:44 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:39:44 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/Makefile
Message-ID: <200306262137.h5QLbTb32725@choi.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms:

Makefile updated: 1.23 -> 1.23.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/Makefile
diff -u llvm/test/Regression/Transforms/Makefile:1.23 llvm/test/Regression/Transforms/Makefile:1.23.2.1
--- llvm/test/Regression/Transforms/Makefile:1.23	Tue May 27 11:10:03 2003
+++ llvm/test/Regression/Transforms/Makefile	Thu Jun 26 16:35:48 2003
@@ -23,6 +23,7 @@
 	Reassociate \
 	SCCP \
 	ScalarRepl \
-	SimplifyCFG
+	SimplifyCFG \
+	TailDup
 
 include $(LEVEL)/Makefile.common




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:39:50 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:39:50 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Reoptimizer/BinInterface/bitmath.h sparc9.h
Message-ID: <200306262137.h5QLbT132733@choi.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Reoptimizer/BinInterface:

bitmath.h updated: 1.9 -> 1.9.2.1
sparc9.h updated: 1.14 -> 1.14.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.
Kept includes of "Config/assert.h" to avoid implicit dependencies on header
files.


---
Diffs of the changes:

Index: llvm/include/llvm/Reoptimizer/BinInterface/bitmath.h
diff -u llvm/include/llvm/Reoptimizer/BinInterface/bitmath.h:1.9 llvm/include/llvm/Reoptimizer/BinInterface/bitmath.h:1.9.2.1
--- llvm/include/llvm/Reoptimizer/BinInterface/bitmath.h:1.9	Sat May 31 17:16:41 2003
+++ llvm/include/llvm/Reoptimizer/BinInterface/bitmath.h	Thu Jun 26 16:34:49 2003
@@ -16,6 +16,8 @@
 #ifndef __BITMATH_H__
 #define __BITMATH_H__
 
+#include "Support/DataTypes.h"
+
 //*********************************
 //	Misc functions
 //*********************************
@@ -29,9 +31,9 @@
 //*********************************
 #define FLD_UPPER(FLD_DEF) (1 ? FLD_DEF)
 #define FLD_LOWER(FLD_DEF) (0 ? FLD_DEF)
-#define MASKBELOW(V)       ((1 << V) - 1)
+#define MASKBELOW(V)       ((((uint64_t) 1) << V) - 1) // 64-bit '1' constant here to avoid overflow warnings
 #define MASKEQBELOW(V)	   ((1 << V) | MASKBELOW(V)) //masks off everything ABOVE
-#define RD_FLD(x, FLD)     ((x & MASKEQBELOW(FLD_UPPER(FLD)))  >> FLD_LOWER(FLD))
+#define RD_FLD(x, FLD)     (unsigned)(((x & MASKEQBELOW(FLD_UPPER(FLD)))  >> FLD_LOWER(FLD)))
 #define MK_FLD(FLD, val)   ((val << FLD_LOWER(FLD)) & MASKEQBELOW(FLD_UPPER(FLD)))
 #define MASK_FLD(FLD)      (MASKEQBELOW(FLD_UPPER(FLD)) & ~MASKBELOW(FLD_LOWER(FLD)))
 #define RM_FLD(FLD, val)   (val & ~(MASK_FLD(FLD)))
@@ -129,7 +131,9 @@
    return ((0xFFFFAA50 >> w) & 3)+n;   // this is a LUT
 }
 
-
+static unsigned LOW10(unsigned value)    { return value & 0x000003ff;       }
+static unsigned HIGH22(unsigned value)   { return value >> 10;              }
+static unsigned HIGHWORD(uint64_t value) { return (unsigned) (value >> 32); }
+static unsigned LOWWORD(uint64_t value)  { return (unsigned) value;         }
 
 #endif
-


Index: llvm/include/llvm/Reoptimizer/BinInterface/sparc9.h
diff -u llvm/include/llvm/Reoptimizer/BinInterface/sparc9.h:1.14 llvm/include/llvm/Reoptimizer/BinInterface/sparc9.h:1.14.2.1
--- llvm/include/llvm/Reoptimizer/BinInterface/sparc9.h:1.14	Sat May 31 17:16:52 2003
+++ llvm/include/llvm/Reoptimizer/BinInterface/sparc9.h	Thu Jun 26 16:34:49 2003
@@ -430,12 +430,111 @@
    MK_FLD(INSTR_I, 1)             |            \
    MK_FLD(INSTR_OP3, OP3_ADD))
 
-#define MK_INSTR_BRANCH_RS1(cond)               \
+#define MK_INSTR_BRANCH_RS1(cond)              \
    (MK_FLD(INSTR_OP, OP_BRANCH)   |            \
     MK_FLD(INSTR_BPR, BPR)        |            \
     MK_FLD(INSTR_BPR_COND, cond))
 
 #define MK_INSTR_BRANCH_NO_RS1(cond)           \
    (MK_FLD(INSTR_OP, OP_BRANCH)   |            \
-    MK_FLD(INSTR_INT_BR, Bicc)        |            \
+    MK_FLD(INSTR_INT_BR, Bicc)    |            \
     MK_FLD(INSTR_INTBR_COND, cond))
+
+// Add a register to a register
+
+#define MK_ADD_R_R(dreg, sreg1, sreg2)         \
+   (MK_FLD(INSTR_OP, OP_2)        |            \
+    MK_FLD(INSTR_RD, dreg)        |            \
+    MK_FLD(INSTR_OP3, OP3_ADD)    |            \
+    MK_FLD(INSTR_RS1, sreg1)      |            \
+    MK_FLD(INSTR_I, 0)            |            \
+    MK_FLD(INSTR_RS2, sreg2))
+
+// Construct immediate-valued logical operation
+
+#define MK_LOGIC_IMM(op3, dreg, sreg, imm)  \
+ (MK_FLD(INSTR_OP,     OP_2) |              \
+  MK_FLD(INSTR_RD,     dreg) |              \
+  MK_FLD(INSTR_RS1,    sreg) |              \
+  MK_FLD(INSTR_SIMM13, imm)  |              \
+  MK_FLD(INSTR_I,      1)    |              \
+  MK_FLD(INSTR_OP3,    op3))
+
+// Construct reg-to-reg logical operation
+
+#define MK_LOGIC(op3, dreg, sreg1, sreg2) \
+ (MK_FLD(INSTR_OP,  OP_2)  |              \
+  MK_FLD(INSTR_RD,  dreg)  |              \
+  MK_FLD(INSTR_RS1, sreg1) |              \
+  MK_FLD(INSTR_RS2, sreg2) |              \
+  MK_FLD(INSTR_I,   0)     |              \
+  MK_FLD(INSTR_OP3, op3))
+
+// Construct sethi instruction
+
+#define MK_SETHI(dreg, imm)                \
+  (MK_FLD(INSTR_OP, OP_BRANCH)  |          \
+   MK_FLD(INSTR_RD, dreg)       |          \
+   MK_FLD(INSTR_OP2, OP2_SETHI) |          \
+   MK_FLD(INSTR_IMM22, imm))
+
+// Construct S{LL,RL,RA}X shift instruction. 
+
+#define MK_SHIFTX(op3, dreg, sreg, shcnt)   \
+  (MK_FLD(INSTR_OP, OP_2)  |                \
+   MK_FLD(INSTR_RD, dreg)  |                \
+   MK_FLD(INSTR_OP3, op3)  |                \
+   MK_FLD(INSTR_RS1, sreg) |                \
+   MK_FLD(INSTR_I, 1)      |                \
+   MK_FLD(INSTR_X, 1)      |                \
+   MK_FLD(INSTR_SHCNT64, shcnt))
+
+#define MK_STORE_IMM(srcreg, basereg, offset)  \
+  (MK_FLD(INSTR_OP, OP_3)         |            \
+   MK_FLD(INSTR_RD, srcreg)       |            \
+   MK_FLD(INSTR_OP3, OP3_STX)     |            \
+   MK_FLD(INSTR_RS1, basereg)     |            \
+   MK_FLD(INSTR_I, 1)             |            \
+   MK_FLD(INSTR_SIMM13, offset))
+
+#define MK_LOAD_IMM(destreg, basereg, offset)  \
+  (MK_FLD(INSTR_OP, OP_3)         |            \
+   MK_FLD(INSTR_RD, destreg)      |            \
+   MK_FLD(INSTR_OP3, OP3_LDX)     |            \
+   MK_FLD(INSTR_RS1, basereg)     |            \
+   MK_FLD(INSTR_I, 1)             |            \
+   MK_FLD(INSTR_SIMM13, offset))
+
+// Construct save instruction
+
+#define MK_SAVE_IMM(dreg, sreg, imm)        \
+  (MK_FLD(INSTR_OP, OP_2)         |         \
+   MK_FLD(INSTR_RD, dreg)         |         \
+   MK_FLD(INSTR_OP3, OP3_SAVE)    |         \
+   MK_FLD(INSTR_RS1, sreg)        |         \
+   MK_FLD(INSTR_I, 1)             |         \
+   MK_FLD(INSTR_SIMM13, imm))
+
+#define MK_SAVE_REG(dreg, sreg1, sreg2)     \
+  (MK_FLD(INSTR_OP, OP_2)         |         \
+   MK_FLD(INSTR_RD, dreg)         |         \
+   MK_FLD(INSTR_OP3, OP3_SAVE)    |         \
+   MK_FLD(INSTR_RS1, sreg1)       |         \
+   MK_FLD(INSTR_I, 0)             |         \
+   MK_FLD(INSTR_RS2, sreg2))
+
+#define MK_RESTORE_IMM(dreg, sreg, imm)     \
+  (MK_FLD(INSTR_OP, OP_2)         |         \
+   MK_FLD(INSTR_RD, dreg)         |         \
+   MK_FLD(INSTR_OP3, OP3_RESTORE) |         \
+   MK_FLD(INSTR_RS1, sreg)        |         \
+   MK_FLD(INSTR_I, 1)             |         \
+   MK_FLD(INSTR_SIMM13, imm))
+
+#define MK_JMPL_INDIRECT(dreg, sreg, imm)   \
+  (MK_FLD(INSTR_OP, OP_2)         |         \
+   MK_FLD(INSTR_RD, dreg)         |         \
+   MK_FLD(INSTR_OP3, OP3_JMPL)    |         \
+   MK_FLD(INSTR_RS1, sreg)        |         \
+   MK_FLD(INSTR_I, 1)             |         \
+   MK_FLD(INSTR_SIMM13, imm))




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:39:56 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:39:56 2003
Subject: [llvm-commits] CVS: llvm/test/Libraries/crtend/crtend.c
Message-ID: <200306262137.h5QLbjl00558@choi.cs.uiuc.edu>


Changes in directory llvm/test/Libraries/crtend:

crtend.c updated: 1.1 -> 1.1.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/test/Libraries/crtend/crtend.c
diff -u llvm/test/Libraries/crtend/crtend.c:1.1 llvm/test/Libraries/crtend/crtend.c:1.1.2.1
--- llvm/test/Libraries/crtend/crtend.c:1.1	Wed May 14 08:09:41 2003
+++ llvm/test/Libraries/crtend/crtend.c	Thu Jun 26 16:35:35 2003
@@ -34,8 +34,12 @@
   /* Loop over all of the constructor records, calling each function pointer. */
   TorRec *R = __llvm_getGlobalCtors();
 
-  if (atexit(run_destructors))
-    abort();  /* Should be able to install ONE atexit handler! */
+  /* Only register the global dtor handler if there is at least one global
+   * dtor!
+   */
+  if (__llvm_getGlobalDtors()[0].FP)
+    if (atexit(run_destructors))
+      abort();  /* Should be able to install ONE atexit handler! */
 
   /* FIXME: This should sort the list by priority! */
   for (; R->FP; ++R)




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:40:02 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:40:02 2003
Subject: [llvm-commits] CVS: llvm/lib/Reoptimizer/Inst/lib/PhaseInfo.h Phases.cpp SparcInstManip.cpp design.txt
Message-ID: <200306262137.h5QLbTU32745@choi.cs.uiuc.edu>


Changes in directory llvm/lib/Reoptimizer/Inst/lib:

PhaseInfo.h updated: 1.9 -> 1.9.4.1
Phases.cpp updated: 1.35 -> 1.35.4.1
SparcInstManip.cpp updated: 1.16 -> 1.16.4.1
design.txt updated: 1.15 -> 1.15.4.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/lib/Reoptimizer/Inst/lib/PhaseInfo.h
diff -u llvm/lib/Reoptimizer/Inst/lib/PhaseInfo.h:1.9 llvm/lib/Reoptimizer/Inst/lib/PhaseInfo.h:1.9.4.1
--- llvm/lib/Reoptimizer/Inst/lib/PhaseInfo.h:1.9	Sun May 18 15:33:46 2003
+++ llvm/lib/Reoptimizer/Inst/lib/PhaseInfo.h	Thu Jun 26 16:35:19 2003
@@ -27,12 +27,9 @@
 typedef std::pair<uint64_t, uint64_t> AddressRange;
 
 typedef struct GBTElem {
-    unsigned gbtType;
-    unsigned short* loadVar;
-    unsigned gbtStartIdx;
-    unsigned paramSize;
-    void*    retVal;
-    void*    instFunc;
+    unsigned      siteID;
+    unsigned      gbtType;
+    unsigned      short* loadVar;
 };
 
 class Phase3Info 


Index: llvm/lib/Reoptimizer/Inst/lib/Phases.cpp
diff -u llvm/lib/Reoptimizer/Inst/lib/Phases.cpp:1.35 llvm/lib/Reoptimizer/Inst/lib/Phases.cpp:1.35.4.1
--- llvm/lib/Reoptimizer/Inst/lib/Phases.cpp:1.35	Thu May 22 22:26:12 2003
+++ llvm/lib/Reoptimizer/Inst/lib/Phases.cpp	Thu Jun 26 16:35:19 2003
@@ -45,10 +45,13 @@
 //
 //      2. Change the branch to the phase 4 slot to branch to a (new) phase 5 slot. See
 //      appropriate InstManip instance for detailed information about phase 5 slot
-//      contents.
+//      contents.  If there is no registered instrumentation for the phase 5 slot, leave a
+//      nop in place of the branch to reduce runtime overhead.
 //
 //      3. Deallocate the slot that originated this invocation of phase4().
 //
+// PHASE 5: Phase 5 isn't like the other phases; rather, it simply invokes all
+//          registered instrumentation functions for a particular site.
 
 #include <algorithm>
 #include <iomanip>
@@ -56,6 +59,7 @@
 #include <set>
 #include <stdlib.h>
 #include <vector>
+#include <list>
 
 #include "llvm/Reoptimizer/Inst/ElfReader.h"
 #include "llvm/Reoptimizer/MemoryManager.h"
@@ -146,11 +150,173 @@
     uint64_t    m_tag;         // Entry to look for in the GBT
 };
 
+// InstFunctionInfo is the class used to represent information (e.g., address of return
+// value, pointer-to-instrumentation function) about particular registered instrumentation
+// functions.  In the case of end-interval functions, the link 'm_pStart' is filled in to
+// refer to the InstFunctionInfo instance that contains information about the
+// corresponding start-interval function.
+
+class InstFunctionInfo 
+{
+  public:
+    InstFunctionInfo(void* pRetVal,
+                     void* pInstFunc,
+                     InstFunctionInfo* pStart = 0):
+        m_pRetVal(pRetVal),
+        m_pInstFunc(pInstFunc),
+        m_pStart(pStart),
+        m_invoked(false)
+    {
+    }
+
+    InstFunctionInfo():
+        m_pRetVal(0),
+        m_pInstFunc(0),
+        m_pStart(0),
+        m_invoked(false)
+    {
+    }
+
+    void invoke();
+
+  protected:
+    void*             m_pRetVal;
+    void*             m_pInstFunc;
+    InstFunctionInfo* m_pStart;   // Info about start-interval function
+    bool              m_invoked;  // Has this function been invoked yet?
+};
+
+// InstSiteInfo instances contain information about the state of particular
+// instrumentation sites.  More specifically, it holds the instrumentation
+// status (e.g. whether or not the site has been handled by phase 4 yet) of the
+// sites as well as the list of (pointers to) the InstFunctionInfo instances
+// registered with a particular site.
+
+class InstSiteInfo 
+{
+  public:
+    InstSiteInfo():
+        m_branchInstalled(false),
+        m_instrumented(false)
+    {
+    }
+
+    // For start-interval sites only -- allocates memory for the return value of
+    // the instrumentation function (size of allocated memory is retValBytes).
+    // Returns a pointer to the InstFunctionInfo* that corresponds to the
+    // instrumentation function, or null if the function has already been
+    // registered.
+
+    InstFunctionInfo* push_back(unsigned retValBytes,
+                                void* func);
+
+    // For end-interval sites (w/ optional link to corresponding start); other
+    // site types should use this routine as well.  The provided retVal ptr is
+    // used as the return-value parameter of the instrumentation function.
+
+    void push_back(void* retVal,
+                   void* func,
+                   InstFunctionInfo* startInfo = 0);
+
+    void invokeFunctions();
+
+  protected:
+    bool                      m_branchInstalled;     // Installed branch to slot yet?
+    bool                      m_instrumented;        // Has phase 4 instrumented site?
+    uint64_t                  m_brInstallAddr;       // Address to install branch inst
+    unsigned                  m_branchInst;          // The branch inst to install
+    std::set<void*>           m_registeredFuncs;     // Set of func-ptrs registered here
+    vector<InstFunctionInfo*> m_instFuncInfos;       // Info for all registered funcs
+
+    friend void Phase4::transform();
+
+    void installBranch();
+
+    bool isRegistered(void* func) 
+    {
+        return m_registeredFuncs.find(func) != m_registeredFuncs.end();
+    }
+
+    void setInstrumented() { m_instrumented = true; }
+
+    void setBranchInst(uint64_t addr, unsigned branchInst)
+    {
+        m_brInstallAddr = addr;
+        m_branchInst = branchInst;
+    }
+};
+
+// InstInfo is the class that holds data about the instrumentation that gets
+// bound to instrumentation sites and intervals at runtime.  There should only
+// be on instance of this class (i.e., it is a singleton class).  The
+// implementation hides an STL map that maps the unique identifier associated
+// with an instrumentation interval/site to a pair of InstSiteInfo instances,
+// which contains the information about the instrumentations registered for the
+// given interval or point site.  In the case of intervals, the first element of
+// the pair is the InstSiteInfo instance that contains data about the start
+// site, whereas the second element of the pair contains data about the end
+// site.  For point sites, only the first element of the pair contains valid
+// data.
+
+class InstInfo 
+{
+  public:
+    typedef std::pair<InstSiteInfo, InstSiteInfo> SiteInfoPair;
+
+    static InstInfo* instance() 
+    {
+        if(!m_pInstance)
+            m_pInstance = new InstInfo;
+        return m_pInstance;
+    }
+
+    static InstSiteInfo* findSiteInfo(unsigned siteID, unsigned gbtType) 
+    {
+        SiteInfoPair* sip = instance()->findSiteInfo(siteID);
+        InstSiteInfo* siteInfo;
+
+        switch(gbtType) {
+            case pp::GBT_INTERVAL_START: siteInfo = &sip->first; break;
+            case pp::GBT_INTERVAL_END:   siteInfo = &sip->second; break;
+            default: assert(0 && "Unhandled gbtType encountered"); break;
+        }
+
+        return siteInfo;
+    }
+
+    SiteInfoPair* findSiteInfo(unsigned siteID)
+    {
+        SiteInfoMap::iterator i = m_siteInfoMap.find(siteID);
+        if(i == m_siteInfoMap.end())
+            return &m_siteInfoMap[siteID];
+        return &i->second;
+    }
+
+    void setVM(VirtualMem* vm) { m_pVM = vm;   }
+    VirtualMem* getVM() const  { return m_pVM; }
+
+  protected:
+    typedef std::map<unsigned, SiteInfoPair> SiteInfoMap;
+
+    InstInfo(): m_pVM(0)
+    {
+    }
+    
+    SiteInfoMap m_siteInfoMap;
+    VirtualMem* m_pVM;
+
+  private:
+    static InstInfo* m_pInstance;
+};
+
+InstInfo* InstInfo::m_pInstance = 0;
+
 //////////////// Phase 2 implementation ////////////////
 
 extern "C" void phase2() 
 {
     TraceCache* pTC = new TraceCache();
+    InstInfo::instance()->setVM(pTC->getVM());
     Phase2 ph(pTC, new SparcInstManip(pTC));
     ph.transform();
 }
@@ -192,7 +358,7 @@
         if(m_excludeSet.find(i->first) == m_excludeSet.end()) {
             // Function is not in exclude set, so go ahead and transform it
 
-            DEBUG_MSG(1, "Transforming function " << i->first
+            DEBUG_MSG(4, "Transforming function " << i->first
                       << "[" << HEX(i->second.first)
                       << ", " << HEX(i->second.second) << "]...\n");
             
@@ -217,17 +383,30 @@
     ::doFlush(slotBase, slotBase + im->getInstWidth() * snippet.size());
 }
 
+static uint64_t makeNewSlot(uint64_t srcAddr,
+                            unsigned slotSize,
+                            unsigned& branchInst,
+                            TraceCache* tc,
+                            InstManip* im)
+{
+    // Return a branch instruction to the new slot via branchInst.
+    uint64_t slotBase = tc->getMemMgr()->getMemory(slotSize);
+    assert(slotBase && "Unable to obtain memory from MemoryManager instance");
+
+    branchInst = im->getBranchAlways(slotBase, srcAddr);
+    return slotBase;
+}
+
 static uint64_t replaceInstWithBrToSlot(uint64_t srcAddr,
                                         unsigned slotSize,
                                         TraceCache* tc,
                                         InstManip* im) 
 {
-    // Obtain a new slot of the given size
-    uint64_t slotBase = tc->getMemMgr()->getMemory(slotSize);
-    assert(slotBase && "Unable to obtain memory from MemoryManager instance");
+    unsigned branchInst;
+    uint64_t slotBase = makeNewSlot(srcAddr, slotSize, branchInst, tc, im);
 
     // Replace instruction at srcAddr with branch to start of new slot
-    tc->getVM()->writeInstToVM(srcAddr, im->getBranchAlways(slotBase, srcAddr));
+    tc->getVM()->writeInstToVM(srcAddr, branchInst);
     ::doFlush(srcAddr, srcAddr + im->getInstWidth());
 
     return slotBase;
@@ -363,8 +542,8 @@
         vector<unsigned> snippet;
         m_pIM->buildSlot(p4info, snippet);
 
-        DEBUG_MSG(3, "phase4 slot instructions:\n");
-#if VERBOSE > 2
+        DEBUG_MSG(4, "phase4 slot instructions:\n");
+#if VERBOSE > 3
         dumpSnippet(snippet, m_pIM);
 #endif
 
@@ -430,7 +609,7 @@
 
     for(unsigned i = 0; i < ppGBTSize; ++i) {
         ostr << "[pp] ppGBT[" << i << "]: " << ppGBT[i].gbtType << ", "
-             << ppGBT[i].loadVar << ", " << ppGBT[i].gbtStartIdx << endl;
+             << ppGBT[i].loadVar << endl;
     }
 }
 
@@ -460,29 +639,44 @@
                && "Unexpected number of instructions in candidate");
 
         // Write NOPs over the original instructions that were associated with the elected
-        // candidate.  No need to no-op over the candidate load instruction itself since
-        // we're about to write over it with a branch to the phase 5 slot.
+        // candidate.
 
         VirtualMem* vm = m_pTC->getVM();
-        for(vector<std::pair<uint64_t, unsigned> >::const_iterator i = cand.getInsts().begin() + 1,
+        for(vector<std::pair<uint64_t, unsigned> >::const_iterator i = cand.getInsts().begin(),
                 e = cand.getInsts().end(); i != e; ++i)
             vm->writeInstToVM(i->first, m_pIM->getNOP());
 
-        // Obtain memory (& rewrite branch) to the phase 5 jump slot.
+        // Obtain new slot, the phase 5 jump slot.
         
         unsigned slotSize = m_pIM->getSlotSize(this);
         uint64_t repAddr = cand.front().first;
-        uint64_t slotBase = replaceInstWithBrToSlot(repAddr, slotSize, m_pTC, m_pIM);
+        unsigned branchInst;
+        uint64_t slotBase = makeNewSlot(repAddr, slotSize, branchInst, m_pTC, m_pIM);
 
         vector<unsigned> snippet;
         m_pIM->buildSlot(gbte, slotBase, repAddr, m_pPhase4Info->getRange(), snippet);
 
-        DEBUG_MSG(3, "phase 5 slot contents:\n");
-#if VERBOSE > 2
+        DEBUG_MSG(4, "phase 5 slot contents:\n");
+#if VERBOSE > 3
         dumpSnippet(snippet, m_pIM);
 #endif
 
         copySnippetToSlot(snippet, slotBase, m_pTC->getVM(), m_pIM);
+
+        // Grab the information about this particular site.
+
+        assert(gbte->gbtType == pp::GBT_INTERVAL_START ||
+               gbte->gbtType == pp::GBT_INTERVAL_END &&
+               "Unhandled gbtType encountered (must implement)");
+        
+        InstSiteInfo* siteInfo = InstInfo::findSiteInfo(gbte->siteID, gbte->gbtType);
+
+        // Take steps to install the branch; note that the InstSiteInfo instance
+        // knows whether or not to actually write the branch instruction, etc.
+
+        siteInfo->setInstrumented();
+        siteInfo->setBranchInst(repAddr, branchInst);
+        siteInfo->installBranch();
     }
     else {
         DEBUG_MSG(1, "does not match\n");
@@ -495,14 +689,6 @@
                           m_pPhase4Info->getCandidate().front().second);
     }
 
-#if 0
-    // (TEMP) For now, restore the candidate load to its original position for debugging
-    // purposes.
-
-    m_pPhase4Info->getTraceCache()->getVM()->writeInstToVM(m_pPhase4Info->getCandidate().front().first,
-                                                           m_pPhase4Info->getCandidate().front().second);
-#endif
-
     DEBUG_MSG(1, "================ End Phase 4 ================\n");
 }
 
@@ -510,27 +696,131 @@
 
 void phase5(GBTElem* gbte)
 {
-    switch(gbte->gbtType){
-        case pp::GBT_INTERVAL_START: {
-            DEBUG_MSG(1, "--- phase 5 start site invocation ---\n");
-            DEBUG_MSG(2, "retVal address is " << HEX(gbte->retVal) << endl);
-
-            void (*instFunc)(void*) = (void (*)(void*)) gbte->instFunc;
-            instFunc(gbte->retVal);
-            break;
-        }
-        case pp::GBT_INTERVAL_END: {
-            DEBUG_MSG(1, "--- phase 5 end site invocation ---\n");
-            DEBUG_MSG(2, "start parameter is at gbt index " << gbte->gbtStartIdx << endl);
-            DEBUG_MSG(2, "start parameter addr is "
-                      << HEX(ppGBT[gbte->gbtStartIdx].retVal) << endl);
+    DEBUG_MSG(1, "================ Begin Phase 5 ================\n");    
+    InstSiteInfo* siteInfo = InstInfo::findSiteInfo(gbte->siteID, gbte->gbtType);
+    siteInfo->invokeFunctions();
+    DEBUG_MSG(1, "================ End Phase 5 ================\n");
+}
+
+//////////////// InstSiteInfo implementation ////////////////
+
+InstFunctionInfo* InstSiteInfo::push_back(unsigned retValBytes,
+                                          void* func)
+{
+    DEBUG_MSG(3, "Inside InstSiteInfo::push_back, registering func w/ address "
+              << HEX(func) << ", new retVal of size " << retValBytes << endl);
+
+    InstFunctionInfo* fi = 0;
+
+    if(!isRegistered(func)) {
+        DEBUG_MSG(3, "not yet registered, registering...\n");
+
+        void* retVal = static_cast<void*>(new char[retValBytes]);
+        m_registeredFuncs.insert(func);
+        m_instFuncInfos.push_back(fi = new InstFunctionInfo(retVal, func));
+        installBranch();
+    }
+    else
+        DEBUG_MSG(3, "WARNING: Attempt to register instrumentation at site that was already instrumented\n");
+
+    return fi;
+}
 
-            void (*instFunc)(void*, void*) = (void (*)(void*, void*)) gbte->instFunc;
-            instFunc(gbte->retVal, ppGBT[gbte->gbtStartIdx].retVal);
+void InstSiteInfo::push_back(void* retVal,
+                             void* func,
+                             InstFunctionInfo* startInfo)
+{
+    DEBUG_MSG(3, "Inside InstSiteInfo::push_back, registering func w/ address "
+              << HEX(func) << ", retVal addr " << HEX(retVal) << endl);
+
+    if(!isRegistered(func)) {
+        DEBUG_MSG(3, "not yet registered, registering...\n");
+
+        m_registeredFuncs.insert(func);
+        m_instFuncInfos.push_back(new InstFunctionInfo(retVal, func, startInfo));
+        installBranch();
+    }
+    else
+        DEBUG_MSG(3, "WARNING: Attempt to register instrumentation at site that was already instrumented\n");
+}
+
+void InstSiteInfo::invokeFunctions()
+{
+    assert(m_branchInstalled && m_instrumented && "Invoking functions not permitted");
+
+    for(int i = 0, e = m_instFuncInfos.size(); i < e; ++i) {
+        DEBUG_MSG(3, "Invoking function " << i << endl);
+        m_instFuncInfos[i]->invoke();
+    }
+}
+
+void InstSiteInfo::installBranch()
+{
+    if(!m_branchInstalled && m_instrumented && m_instFuncInfos.size() > 0) {
+        DEBUG_MSG(3, "(InstSiteInfo::installBranch) Installing branch...\n");
+        VirtualMem* vm = InstInfo::instance()->getVM();
+        vm->writeInstToVM(m_brInstallAddr, m_branchInst);
+        m_branchInstalled = true;
+    }
+    else
+        DEBUG_MSG(3, "(InstSiteInfo::installBranch) Not installing branch, it's not needed\n");
+}
+
+//////////////// InstFunctionInfo implementation ////////////////
 
-            break;
+void InstFunctionInfo::invoke()
+{
+    // In the case of start-interval functions, the boolean m_invoked is set to true upon
+    // completion of the invocation.  This same boolean (i.e, the one associated with the
+    // start-interval InstFunctionInfo instance) is cleared by the corresponding
+    // end-interval function invocation.  This is to ensure that no end-interval
+    // instrumentation function is ever invoked unless its corresponding start-interval
+    // instrumentation function has been invoked (which implies that the start-interval
+    // site has been handled properly and all of the registration mechanisms).
+
+    DEBUG_MSG(3, "(InstFunctionInfo::invoke) retVal address is: " << HEX(m_pRetVal) << endl);
+    if(m_pStart) {
+        DEBUG_MSG(3, "End-interval inst func detected\n");
+        void (*instFunc)(void*, void*) = (void (*)(void*, void*)) m_pInstFunc;
+
+        if(m_pStart->m_invoked) {
+            DEBUG_MSG(4, "Corresponding start function has been invoked, so invoking this inst func\n");
+            instFunc(m_pRetVal, m_pStart->m_pRetVal);
+            m_pStart->m_invoked = false;
         }
+        else
+            DEBUG_MSG(4, "Corresponding start func has not been invoked, so not invoking\n");
+    }
+    else {
+        DEBUG_MSG(3, "Start-interval or non-end inst func detected\n");
+        void (*instFunc)(void*) = (void (*)(void*)) m_pInstFunc;
+        instFunc(m_pRetVal);
+        m_invoked = true;
     }
+        
+    DEBUG_MSG(3, "(InstFunctionInfo::invoke) instrumentation function returned\n");
+}
+
+//////////////// register{Interval,Point}Inst implementation ////////////////
 
-    DEBUG_MSG(1, "--- phase 5 invocation completed ---\n" << std::flush);
+extern "C" void registerIntervalInst(unsigned siteID,
+                                     void* startFunc,
+                                     void* endFunc,
+                                     unsigned paramSize,
+                                     void* retVal)
+{
+    InstInfo::SiteInfoPair* sip = InstInfo::instance()->findSiteInfo(siteID);
+    assert(sip && "Unable to obtain SiteInfoPair");
+    
+    // Handle start function
+    DEBUG_MSG(3, "registerIntervalInst: Registering start function...\n");
+    InstFunctionInfo* fi = sip->first.push_back(paramSize, startFunc);
+
+    if(fi) {
+        // Handle end function
+        DEBUG_MSG(3, "registerIntervalInst: Registering end function...\n");
+        sip->second.push_back(retVal, endFunc, fi);
+    }
+    else
+        DEBUG_MSG(3, "WARNING: Register-start-function returned 0, which implies redundant registration");
 }


Index: llvm/lib/Reoptimizer/Inst/lib/SparcInstManip.cpp
diff -u llvm/lib/Reoptimizer/Inst/lib/SparcInstManip.cpp:1.16 llvm/lib/Reoptimizer/Inst/lib/SparcInstManip.cpp:1.16.4.1
--- llvm/lib/Reoptimizer/Inst/lib/SparcInstManip.cpp:1.16	Wed May 28 08:52:41 2003
+++ llvm/lib/Reoptimizer/Inst/lib/SparcInstManip.cpp	Thu Jun 26 16:35:19 2003
@@ -242,14 +242,6 @@
 
     DEBUG_MSG(2, "buildPhase5HeapSlot completed\n");
 
-    // If we're dealing with a start-interval instrumentation function, heap-allocate
-    // its parameter memory
-
-    if(gbte->gbtType == pp::GBT_INTERVAL_START) {
-        assert(!gbte->retVal && "Expected null retVal value");
-        gbte->retVal = static_cast<void*>(new char[gbte->paramSize]);
-    }
-
     ////////////////
     // Construct the phase 5 jump slot
 
@@ -363,7 +355,7 @@
                         unsigned sizeBytes,
                         int protBits)
 {
-    DEBUG_MSG(3, "Setting access bits on heap slot page(s)" << endl);
+    DEBUG_MSG(4, "Setting access bits on heap slot page(s)" << endl);
 
     int rc = mprotect(pageBase, sizeBytes, protBits);
     if(rc < 0) {
@@ -416,10 +408,10 @@
 
     setPageBits(heapSlot, numHeapBytes, PROT_READ | PROT_WRITE | PROT_EXEC);
 
-#if VERBOSE > 2
-    DEBUG_MSG(3, "Dumping contents of heap-slot memory...\n");
+#if VERBOSE > 3
+    DEBUG_MSG(4, "Dumping contents of heap-slot memory...\n");
     dumpHeapSlot(heapSlot, getPhase5HeapSize(), this);
-    DEBUG_MSG(3, "Done with heap region construction, moving on to jump slot\n");
+    DEBUG_MSG(4, "Done with heap region construction, moving on to jump slot\n");
 #endif
 
     return heapSlot;
@@ -462,7 +454,7 @@
         return addr + getInstWidth();
     }
     else {
-        DEBUG_MSG(2, "WARNING: Non-save instruction at function entry\n");
+        DEBUG_MSG(4, "WARNING: Non-save instruction at function entry\n");
         return addr;
     }
     
@@ -569,7 +561,7 @@
                                                m_logicalToActualReg[useForIndirect],
                                                offset));
 
-    DEBUG_MSG(3, "JMPL instruction word is " << HEX(m_pCurrSnippet->back()) << endl);
+    DEBUG_MSG(4, "JMPL instruction word is " << HEX(m_pCurrSnippet->back()) << endl);
     m_pCurrSnippet->push_back(getNOP());
 
     assert(m_pCurrSnippet->size() - initSize == GEN_JMPL_SIZE &&


Index: llvm/lib/Reoptimizer/Inst/lib/design.txt
diff -u llvm/lib/Reoptimizer/Inst/lib/design.txt:1.15 llvm/lib/Reoptimizer/Inst/lib/design.txt:1.15.4.1
--- llvm/lib/Reoptimizer/Inst/lib/design.txt:1.15	Sun May 18 12:45:26 2003
+++ llvm/lib/Reoptimizer/Inst/lib/design.txt	Thu Jun 26 16:35:19 2003
@@ -880,91 +880,70 @@
 
 {{{ MILESTONES
 
+  - Experiments
+  - The Paper
+  - The Thesis
 
 }}}
 
 {{{ TODO
 
-    - Move statically-sized spill regions so that they are internal to SparcInstManip.
-      (do not need variable-sized spill region except for phase5 invocations)
+    In priority order:
 
-    - Start table-of-stacks implementation for phase4 authorship of phase 5 slots.
-      - Placed on hold temporary because of "alloca-finding" approach. However, see the 
-        following e-mail for the current state of things:
-
-        {{{ E-mail regarding alloca-finding and table-of-stacks approach
-Okay, this is starting to seem intractable. I have another problem that
-I don't think can be resolved without resorting to a custom-stack
-mechanism that will incur prohibitive overhead.
-
-Everything is working for start-region instrumentation sites.  For
-end-region instrumentation sites, however, there's a problem. In order
-to write the slot for end sites, I have to know (or know how to compute)
-the address of the return value of the corresponding start site. I had
-originally thought that I would just store this in the GBT, or
-"something", but I clearly didn't think through the problem well enough.
-
-There are only two ways I can think of that this can occur:
-
-(a) Write the effective address of the return value of the start inst
-func, so that it gets passed to the end inst func.
-
-or
-
-(b) Somehow encode the stack offset to the return value from the start
-inst, where the offset is from the %sp *at the end-region site*
-
-Both of these have problems.
-
-First, I don't think (b) can work at all, given that there may be
-alloca's present in the original application that would change the %sp,
-and thus the offset value that we'd need, and we can't determine the
-exact allocas that are executed statically.
-
-For (a), the effective address isn't known until runtime. We can store
-this address in some global table where the phase 4 invocation for the
-end site can find it, but it is not sufficient to have a single scalar
-address here -- we must have a stack, due to potential recursive
-invocations. I think that this is clear, please let me know if I'm not
-making sense. :) 
-
-Hence, we'd need to maintain a stack of effective addresses, which was
-pushed during the execution of phase 5 for the start site, and then read
-and popped during the execution of phase 5 for the end site.  We're
-already really bloated with how many instructions we've got going on for
-all of the spills, etc, and I'm concerned about the effect that this
-stack manipulation will have on our overhead, as we talked about before.
-
-The way I see it, we only have two options if we're to make forward
-progress and not obliterate our chances of having lower overhead
-numbers. Hopefully we have some better choices. In the interests of
-short-term forward progress, I'm going to go with #1 for now.
-
-#1 - Make another common-case assumption that there will be no allocas
-between start and end sites, on *any* control path. If this is the case,
-then we know that the stack pointer will not have been manipulated (I
-think) between the start and end sites, and so the %sp offsets to the
-requisite data will be unchanged since when the phase 5 step occurred
-for the start site.  
-
-#2 - Just implement our fall-back solution that everything seems to be
-pointing to. I'm not sure exactly what other logistic nightmares might
-be entailed in this, though, because I've only a sketch of the idea.
-
-I wanted to point out, also, that the so-called "fall back" approach we
-discussed previous also involves manipulation of a stack at runtime
-(push/pop actions still have to occur at runtime), so perhaps the stack
-of effective addresses is less prohibitive than I thought, if only in
-the sense that we cannot avoid it. :(
-        }}}
-
-    - Write phase 5 stuff for end-region sites -- will assume that not allocas lie between
-    the start and end sites, which is not particularly a fair assumption.
-
-    - Optimizations:
-        - No need to save registers (other than those clobbered) in phase 3 slot, since phase 3
-          is invoked at the start of the function. Must still spill/restore shared, though.
-        - No need to save registers (other than those clobbered) in general.
+    1) New implementation for instrumenting at function-level granularity
+    2) Apache through LLVM, experiments
+    3) Writing, writing, writing: ICS version 2 paper, do
+       a) outline
+       b) intro
+       c) language section
+       d) compiler section
+
+    The three top-level items above are more-or-less interchangable. However, the
+    experiments will not be able to be completed unless the "instrumentation @
+    function-level granularity" implementation is done, and there should be an emphasis on
+    getting Apache through LLVM in the short-term because Chris is leaving for Norway in
+    early July.
+
+    However, *all* of the writing (including experimentation sections) for ICS v2 must be
+    done by the end of June, so that Adve can approve it, make the desired corrections,
+    etc., so that I can get started on thesis authorship and submission.  Realistically,
+    the timeframe should look something like this:
+
+    Week of 6/10 (5 days): Implementation, Apache w/ bug reports
+    Week of 6/16 (6 days): Implementation, Small example, continue Apache & do tests
+
+    -- At this point, all experiments should be more-or-less completed --
+
+    Week of 6/23 (6 days): Write, write, write. 3 days nonstop for content, 3
+                           days of Adve making corrections, etc.
+
+    Monday, 6/30 is D-Day...
+
+    Schedule revision 13 Jun 2003: Apache is in stasis waiting to hear back from
+    Chris. Some effort might be expended to see if any more work can be done on compiling
+    Apache even though the build process currently fails altogether.  Implementation for
+    instrumentation at function-level granularity is in stasis until I hear back from
+    Vikram.  This leaves three immediate options open until either of the two are
+    resolved, in which case forward progress on either the implementation or Apache should
+    be made.
+
+       Option 1) POV-Ray through LLVM.
+       Option 2) Writing
+       Option 3) "Small example"
+
+    Option 3 really shouldn't be undertaken until we know if obtaining things like I/O
+    elapsed time (via function-level) is possible, or until I can talk with Vikram in our
+    next meeting about what the heck this nebulous example should look like...this leaves
+    only options 1 and 2 above as viable.
+
+    POV-Ray should be an easy thing to start, and would be good both as a fall-back if
+    Apache isn't possible as well as a useful additional example if the latter is
+    possible.  This would also give V & C some time to respond to the pending queries.
+
+    - Optimizations: - No need to save registers (other than those clobbered) in
+    phase 3 slot, since phase 3 is invoked at the start of the function. Must
+    still spill/restore shared, though.  - No need to save registers (other than
+    those clobbered) in general.
 
 }}}
 
@@ -1460,5 +1439,226 @@
 Also, Chris remarked that any novel page-management mechanisms, etc., (for the
 code that is jumped to in newly-allocated pages) that I devise should perhaps be
 integrated into the LLVM JIT if they are suitable.
+
+}}}
+
+{{{ Experiments
+
+We must devise experiments that demonstrate the novel aspects of the work.  We
+are currently planning on using Apache and/or POV-Ray, and demonstrating how a
+"deep" performance analysis can be encoded using performance primitives. A
+"deep" performance analysis is one which essentially (using Hollingsworth's
+terminology from his PhD thesis) gets at the "what, where, and when" aspects of
+performance bottlenecks.  However, instead of doing this at the "arbitrary
+program during execution" level, as the W^3 search model does, we will encode
+these performance aspects at the application level itself.
+
+Here's what Vikram suggested for a good start:
+
+  I just thought that the examples of performance issues he explores in his
+  automatic search would give you (a) some insights into what performance issues
+  make sense to consider, and (b) some ideas about how to do a systematic
+  diagnosis, albeit at the application level.
+  
+  But isn't detecting the "what, why, and when" of performance bottlenecks
+  pretty closely related to the goal of performance diagnosis?  We'd be looking
+  for bottlenecks too, except that we can use application domain information, we
+  can look for bottlenecks at the algorithmic level instead of the general
+  system level, and we can record it permanently in the program as a first class
+  feature of the program.
+  
+  Anyway, about your second question: here's a way to do what I suggesting:
+  
+  -- think about 2-3 key performance issues with Apache (or POV-Ray) that you'd
+  want to diagnose e.g., cache misses, TLB misses, thread overhead (estimating
+  that could be interesting), I/O delays
+  
+  -- if those issues make sense with a small sort example, try to diagnose those
+  issues in the small example first.  e.g., I think cache misses, TLB misses,
+  and I/O delays would all be issues if you were sorting a huge file of some
+  kind.
+  
+  This is purely to give you a small, well-understood code to try out before
+  going to the big ones where it may be difficult to know, when one diagnosis
+  attempt fails, whether it failed because you misunderstood the performance
+  issues or because the guess was wrong or both.
+
+TLB misses aren't an option, because we cannot get at them with PAPI. Cache
+misses are available, so that'd certainly be a good place to start.  As for I/O
+delays, I have no idea how we'd measure this, either.  The following is the
+comprehensive list of the low-level metrics that are exposed to us via PAPI.  On
+our own, we can support simple things like elapsed time, load average, etc.  I'm
+not altogether clear how we'd determine how much time (for a particular region)
+was spent doing I/O-bound activities...
+
+Number of hardware counters: 2
+Name: PAPI_L1_ICM 	Description: Level 1 instruction cache misses
+Name: PAPI_L2_TCM 	Description: Level 2 cache misses
+Name: PAPI_CA_SNP 	Description: Requests for a snoop
+Name: PAPI_CA_INV 	Description: Requests for cache line invalidation
+Name: PAPI_L1_LDM 	Description: Level 1 load misses
+Name: PAPI_L1_STM 	Description: Level 1 store misses
+Name: PAPI_BR_MSP 	Description: Conditional branch instructions mispredicted
+Name: PAPI_TOT_IIS 	Description: Instructions issued
+Name: PAPI_TOT_INS 	Description: Instructions completed
+Name: PAPI_LD_INS 	Description: Load instructions
+Name: PAPI_SR_INS 	Description: Store instructions
+Name: PAPI_TOT_CYC 	Description: Total cycles
+Name: PAPI_IPS    	Description: Instructions per second
+Name: PAPI_L1_DCR 	Description: Level 1 data cache reads
+Name: PAPI_L1_DCW 	Description: Level 1 data cache writes
+Name: PAPI_L1_ICH 	Description: Level 1 instruction cache hits
+Name: PAPI_L2_ICH 	Description: Level 2 instruction cache hits
+Name: PAPI_L1_ICA 	Description: Level 1 instruction cache accesses
+Name: PAPI_L2_TCH 	Description: Level 2 total cache hits
+Name: PAPI_L2_TCA 	Description: Level 2 total cache accesses
+
+So, in the short-term, we have two outstanding problems. First, what kinds of
+metric would we want to apply to Apache/POV-Ray/Simple example? Second, of those
+metrics, which can we actually realize with the current system? Third, we should
+create the simple program (such as sorting large amounts of data from a file,
+etc.) so that it can use these metrics in such a way that "models" or
+"anticipates" the way they will be used in the bigger applications.
+
+This is an outstanding issue, and I don't really know where to go with it yet.
+
+More notes about this as of 4 Jun 2003:
+
+One way to obtain metric values that are based on the elapsed times of
+particular functions is to somehow register instrumentation for those particular
+functions, and for a particular region -- Vikram argues that we have the ability
+to do this dynamically and we don't need any markers or phase-1 actions because
+we're operating at function-level granularity. 
+
+Here is a sample scenario: We have defined an interval I over some scoped region
+of code.  During phase 1 and phase 2, no instrumentation is registered for this
+interval.  Later on, we construct a metric that is qualified by a list of
+functions that (for example) are to have their runtimes measured and added to
+some running total.  Let's call this the "measure_functions_start" and
+"measure_functions_end" metric, and have it yield a value of type double which
+is the aggregate runtime of the list of functions when they get executed within
+interval I.  The metric registration function will have to have some way
+(varargs?) of denoting what the functions are: perhaps it can simply pass in an
+array of function names together with a size.
+
+Example: 
+
+pp_registerIntervalInst(intervalID, measure_functions_start,
+                        measure_functions_end, &retVal, sizeof(double), 
+                        func1, func2, func3, ...);
+
+However, what do "measure_functions_start" have to do with anything? More than
+likely, what we need to do is specify a particular metric to apply to a
+paritcular function, such that the value will be sampled each time that function
+gets executed.  Then, since there can be multiple invocations (and hence,
+multiple samples) for this selected function within I, we will have to have some
+default or user-specified way of aggregating the data :(.  This is gross.  In
+other words, we should probably simplify the above to something like:
+
+pp_registerIntervalInst(intervalID, some_metric_start, some_metric_end, 
+                        &retVal, sizeof(double), HOW_TO_AGGREGATE, func1);
+
+Where func1 is the function to be instrumented and HOW_TO_AGGREGATE is some
+value that specifies one from a couple of ways of combining the data.  For now,
+HOW_TO_AGGREGATE will not exist and will implicitly sum all return
+values...hence, if the some_metric_{start,end} function ptrs above were to
+elapsed time, at the end of the interval I with interval id intervalID, retVal
+would contain the combined elapsed time of all time spent in function func1.
+
+Enabling this measurement at the start of the interval, disabling at the end of
+the interval; clearly, the start-instrumentation site will need to transform
+func1 (to compute the metric value) when crossed, and the end-insturmentation
+site must remove such instrumentation when crossed. This doesn't follow the
+normal model of what occurs at the instrumentation sites, in terms of just
+ripping down a list of functions...or does it? Perhaps one of the functions in
+the list is just this function that does the transformation on the target
+function...in this case, the process would look something like:
+
+1. Register the transformation function (for the start point) -- call this
+xform_start -- as a regular instrumentation function.  The runtime call to do
+the registration will build the appropriate data structures which will encode
+what metric to associated with the target function, aggregation method, return
+value, etc.
+
+2. Register the transformation function (for the end point) -- call this
+xform_end -- as a regular instrumentation function.
+
+3. When xform_start is invoked as a regular instrumentation function, it will
+instrument the target function with the selected instrumentation.
+
+This is the hardest step to conceptualize and realize.  The problem is that,
+without any placeholders from phase 1, it's not clear that we can instrument the
+target function easily.  Clearly, our instrumentation points are at the start
+and end points of the function (entry and exit).  But this is not really true.
+Our instrumentation points are really at the entry and *all* function exits.
+
+The important question is, what if we have all of the exit points at our
+disposal? Would that change anything?
+
+It would. The entry point together with all exit points would form a set of
+instrumentation points. For each of these instrumentation points, we could
+over-write a branch to a new slot that would call the desired instrumentation
+function, restore the replaced instruction, and return to the instrumentation
+point to continue execution.  This would potentially work. One major problem
+that comes to mind is that for system calls (such as read()), the body of the
+function is highly likely to be out of short-jmp range to the
+tracecache-allocated slot.  The only way around this would be to create a heap
+region and copy the target function into it, etc.  We don't have any code to do
+this yet, so again there is no code reuse or (much) leveraging of existing
+functionality.  Additionally, work will have to be done to make the data
+structure that maps the address of a function to its extents (which come from
+the ELf information).
+
+--
+
+The other alternative is to make sure a "wrapper" function (i.e., for the
+function read()):
+
+int read_wrapper(args for read) {
+  start_inst();
+  int rc = read(args for read);
+  end_inst();
+  return rc;
+}
+
+But this isn't an option because we cannot locate the calls to read() to replace
+them with the wrapper.  We could do the following:
+
+3a. Copy the entire target function to a heap region, and instrument it to our
+heart's content.  However, finding exit points may not be easy without a CFG,
+etc.
+
+3b. Replace the body of the real target function with a call to the modified
+duplicate of the target function, returning whatever the modified duplicate
+returns.
+
+This works, I think, but is incredibly cumbersome and, contrary to what was
+previously discussed, we do *not* possess all of the mechanisms.
+
+--
+
+Notes on POV-Ray experiment 23 Jun 2003
+
+We propose using a user-defined metric, flops/display pixel to compute the
+computational cost per pixel.
+
+1. Compute flops/dpixel and report using pp mechanisms.
+  1a. flops can be obtained via PAPI, IIRC.
+
+2. Create a moving average of flops/dpixel.
+
+3. When the flops/dpixel exceeds mavg flops/dpixel by some multiplicative
+threshold (e.g.), a performance assertion is violated.  When the PA is violated,
+we report other metrics (perhaps ranked in some manner) that had been being
+recorded but not reported.  Initial suggestions for the other metrics to measure
+_over the same region_ (say, the main routine for a single ray-trace) are:
+
+    1) MAverage L1 icache misses vs. this-ray L1 icache misses
+    2) MAverage L2 cache misses vs. this-ray L2 cache misses
+    3) MAverage load count vs. this-ray load count
+    4) MAverage store count vs. this-ray store count
+
+Immediate problem: I don't think we can monitor more than 2 of these values
+concurrently due to hardware limitations on the SPARC...
 
 }}}




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:40:09 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:40:09 2003
Subject: [llvm-commits] CVS: llvm/utils/Burg/Makefile
Message-ID: <200306262137.h5QLbUL32751@choi.cs.uiuc.edu>


Changes in directory llvm/utils/Burg:

Makefile updated: 1.13.2.1 -> 1.13.2.2

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/utils/Burg/Makefile
diff -u llvm/utils/Burg/Makefile:1.13.2.1 llvm/utils/Burg/Makefile:1.13.2.2
--- llvm/utils/Burg/Makefile:1.13.2.1	Mon Jun 23 14:02:36 2003
+++ llvm/utils/Burg/Makefile	Thu Jun 26 16:36:19 2003
@@ -1,7 +1,6 @@
-# $Id: Makefile,v 1.13.2.1 2003/06/23 19:02:36 criswell Exp $
 LEVEL = ../..
 TOOLNAME = burg
-EXTRASOURCES = gram.tab.c
+ExtraSource = gram.tab.c
 
 include $(LEVEL)/Makefile.common
 




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:40:16 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:40:16 2003
Subject: [llvm-commits] CVS: llvm/test/Feature/Makefile TestAsmDisasm.sh TestOptimizer.sh
Message-ID: <200306262137.h5QLbhF00552@choi.cs.uiuc.edu>


Changes in directory llvm/test/Feature:

Makefile updated: 1.7 -> 1.7.2.1
TestAsmDisasm.sh updated: 1.10 -> 1.10.2.1
TestOptimizer.sh updated: 1.17 -> 1.17.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/test/Feature/Makefile
diff -u llvm/test/Feature/Makefile:1.7 llvm/test/Feature/Makefile:1.7.2.1
--- llvm/test/Feature/Makefile:1.7	Tue Jun 17 15:55:04 2003
+++ llvm/test/Feature/Makefile	Thu Jun 26 16:35:33 2003
@@ -24,10 +24,10 @@
 testcbe     : $(OTESTS:%.ll=%.tc)
 
 Output/%.asmdis: % $(LAS) $(LDIS) Output/.dir
-	@./TestAsmDisasm.sh $< $(LLVMTOOLCURRENT) $(LIBDEBUG)
+	@./TestAsmDisasm.sh $< $(LLVMTOOLCURRENT)
 
 Output/%.opt: % $(LAS) $(LDIS) $(LOPT) Output/.dir
-	@./TestOptimizer.sh $< $(LLVMTOOLCURRENT) $(LIBDEBUG)
+	@./TestOptimizer.sh $< $(LLVMTOOLCURRENT)
 
 Output/%.tc: Output/%.bc $(LDIS)
 	@echo "======== Generating C code for $<"


Index: llvm/test/Feature/TestAsmDisasm.sh
diff -u llvm/test/Feature/TestAsmDisasm.sh:1.10 llvm/test/Feature/TestAsmDisasm.sh:1.10.2.1
--- llvm/test/Feature/TestAsmDisasm.sh:1.10	Fri Jun 20 11:14:27 2003
+++ llvm/test/Feature/TestAsmDisasm.sh	Thu Jun 26 16:35:33 2003
@@ -4,19 +4,6 @@
 
 rm -f test.bc.temp[12]
 
-#
-# Split the current LD_LIBRARY_PATH into two separate components.
-#
-FirstLDP=`echo $LD_LIBRARY_PATH | cut -d\; -f1`
-SecondLDP=`echo $LD_LIBRARY_PATH | cut -d\; -f2`
-
-#
-# Now create a new LD_LIBRARY_PATH with our command line options as
-# the first section.
-#
-LD_LIBRARY_PATH="$3:${FirstLDP}\;${SecondLDP}"
-export LD_LIBRARY_PATH
-
 AS=$2/as
 DIS=$2/dis
 


Index: llvm/test/Feature/TestOptimizer.sh
diff -u llvm/test/Feature/TestOptimizer.sh:1.17 llvm/test/Feature/TestOptimizer.sh:1.17.2.1
--- llvm/test/Feature/TestOptimizer.sh:1.17	Fri Jun 20 11:14:49 2003
+++ llvm/test/Feature/TestOptimizer.sh	Thu Jun 26 16:35:33 2003
@@ -1,18 +1,5 @@
 #!/bin/sh
 
-#
-# Split the current LD_LIBRARY_PATH into two separate components.
-#
-FirstLDP=`echo $LD_LIBRARY_PATH | cut -d\; -f1`
-SecondLDP=`echo $LD_LIBRARY_PATH | cut -d\; -f2`
-
-#
-# Now create a new LD_LIBRARY_PATH with our command line options as
-# the first section.
-#
-LD_LIBRARY_PATH="$3:${FirstLDP}\;${SecondLDP}"
-export LD_LIBRARY_PATH
-
 AS=$2/as
 DIS=$2/dis
 OPT=$2/opt




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:40:23 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:40:23 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/ADCE/2003-06-24-BadSuccessor.ll 2003-06-24-BasicFunctionality.ll 2002-05-23-ZeroArgPHITest.ll 2002-05-28-Crash.ll basictest1.ll basictest2.ll
Message-ID: <200306262137.h5QLbU200300@choi.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/ADCE:

2003-06-24-BadSuccessor.ll added (r1.1.2.1)
2003-06-24-BasicFunctionality.ll added (r1.1.2.1)
2002-05-23-ZeroArgPHITest.ll updated: 1.2 -> 1.2.2.1
2002-05-28-Crash.ll updated: 1.2 -> 1.2.2.1
basictest1.ll updated: 1.2 -> 1.2.2.1
basictest2.ll updated: 1.2 -> 1.2.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/ADCE/2003-06-24-BadSuccessor.ll
diff -c /dev/null llvm/test/Regression/Transforms/ADCE/2003-06-24-BadSuccessor.ll:1.1
*** /dev/null	Thu Jun 26 16:37:30 2003
--- llvm/test/Regression/Transforms/ADCE/2003-06-24-BadSuccessor.ll	Tue Jun 24 17:58:05 2003
***************
*** 0 ****
--- 1,94 ----
+ ; RUN: as < %s | opt -adce -disable-output
+ 
+ target endian = little
+ target pointersize = 32
+ 	%struct..CppObjTypeDesc = type { uint, ushort, ushort }
+ 	%struct..TypeToken = type { uint, ushort, ushort }
+ 
+ implementation   ; Functions:
+ 
+ uint %C_ReFaxToDb() {
+ entry:		; No predecessors!
+ 	br bool false, label %endif.0, label %then.0
+ 
+ then.0:		; preds = %entry
+ 	ret uint 0
+ 
+ endif.0:		; preds = %entry
+ 	br bool false, label %then.11, label %then.4
+ 
+ then.4:		; preds = %endif.0
+ 	ret uint 0
+ 
+ then.11:		; preds = %endif.0
+ 	br bool false, label %loopentry.0, label %else.2
+ 
+ loopentry.0:		; preds = %then.11, %endif.14, %loopentry.1
+ 	br bool false, label %endif.14, label %loopexit.0
+ 
+ endif.14:		; preds = %loopentry.0
+ 	br bool false, label %loopentry.1, label %loopentry.0
+ 
+ loopentry.1:		; preds = %endif.14, %then.53, %then.53, %then.53, %then.53, %then.53
+ 	%SubArrays.10 = phi uint* [ %SubArrays.8, %then.53 ] , [ null, %endif.14 ]		; <uint*> [#uses=3]
+ 	br bool false, label %no_exit.1, label %loopentry.0
+ 
+ no_exit.1:		; preds = %loopentry.1
+ 	switch uint 0, label %label.17 [
+ 		 uint 2, label %label.11
+ 		 uint 19, label %label.10
+ 	]
+ 
+ label.10:		; preds = %no_exit.1
+ 	br bool false, label %then.43, label %endif.43
+ 
+ then.43:		; preds = %label.10
+ 	br bool false, label %then.44, label %endif.44
+ 
+ then.44:		; preds = %then.43
+ 	br bool false, label %shortcirc_next.4, label %endif.45
+ 
+ shortcirc_next.4:		; preds = %then.44
+ 	br bool false, label %no_exit.2, label %loopexit.2
+ 
+ no_exit.2:		; preds = %shortcirc_next.4
+ 	%tmp.897 = getelementptr uint* %SubArrays.10, long 0		; <uint*> [#uses=1]
+ 	%tmp.899 = load uint* %tmp.897		; <uint> [#uses=1]
+ 	store uint %tmp.899, uint* null
+ 	ret uint 0
+ 
+ loopexit.2:		; preds = %shortcirc_next.4
+ 	ret uint 0
+ 
+ endif.45:		; preds = %then.44
+ 	ret uint 0
+ 
+ endif.44:		; preds = %then.43
+ 	ret uint 0
+ 
+ endif.43:		; preds = %label.10
+ 	ret uint 0
+ 
+ label.11:		; preds = %no_exit.1
+ 	ret uint 0
+ 
+ label.17:		; preds = %no_exit.1, %no_exit.1, %no_exit.1, %no_exit.1, %no_exit.1, %no_exit.1
+ 	br bool false, label %then.53, label %shortcirc_next.7
+ 
+ shortcirc_next.7:		; preds = %label.17
+ 	br bool false, label %then.53, label %shortcirc_next.8
+ 
+ shortcirc_next.8:		; preds = %shortcirc_next.7
+ 	ret uint 0
+ 
+ then.53:		; preds = %shortcirc_next.7, %label.17
+ 	%SubArrays.8 = phi uint* [ %SubArrays.10, %shortcirc_next.7 ], [ %SubArrays.10, %label.17 ]		; <uint*> [#uses=5]
+ 	%tmp.1023 = load uint* null		; <uint> [#uses=1]
+ 	switch uint %tmp.1023, label %loopentry.1 []
+ 
+ loopexit.0:		; preds = %loopentry.0
+ 	ret uint 0
+ 
+ else.2:		; preds = %then.11
+ 	ret uint 0
+ }


Index: llvm/test/Regression/Transforms/ADCE/2003-06-24-BasicFunctionality.ll
diff -c /dev/null llvm/test/Regression/Transforms/ADCE/2003-06-24-BasicFunctionality.ll:1.1
*** /dev/null	Thu Jun 26 16:37:30 2003
--- llvm/test/Regression/Transforms/ADCE/2003-06-24-BasicFunctionality.ll	Tue Jun 24 16:22:09 2003
***************
*** 0 ****
--- 1,43 ----
+ ; RUN: if as < %s | opt -adce -simplifycfg | dis | grep then:
+ ; RUN: then exit 1
+ ; RUN: else exit 0
+ ; RUN: fi
+ 
+ void %dead_test8(int* %data.1, int %idx.1) {
+ entry:		; No predecessors!
+ 	%tmp.1 = load int* %data.1		; <int> [#uses=2]
+ 	%tmp.41 = setgt int %tmp.1, 0		; <bool> [#uses=1]
+ 	br bool %tmp.41, label %no_exit.preheader, label %return
+ 
+ no_exit.preheader:		; preds = %entry
+ 	%tmp.11 = getelementptr int* %data.1, long 1		; <int*> [#uses=1]
+ 	%tmp.22-idxcast = cast int %idx.1 to long		; <long> [#uses=1]
+ 	%tmp.28 = getelementptr int* %data.1, long %tmp.22-idxcast		; <int*> [#uses=1]
+ 	br label %no_exit
+ 
+ no_exit:		; preds = %no_exit.preheader, %endif
+ 	%k.1 = phi int [ %k.0, %endif ], [ 0, %no_exit.preheader ]		; <int> [#uses=3]
+ 	%i.0 = phi int [ %inc.1, %endif ], [ 0, %no_exit.preheader ]		; <int> [#uses=1]
+ 	%tmp.12 = load int* %tmp.11		; <int> [#uses=1]
+ 	%tmp.14 = sub int 0, %tmp.12		; <int> [#uses=1]
+ 	%tmp.161 = setne int %k.1, %tmp.14		; <bool> [#uses=1]
+ 	br bool %tmp.161, label %then, label %else
+ 
+ then:		; preds = %no_exit
+ 	%inc.0 = add int %k.1, 1		; <int> [#uses=1]
+ 	br label %endif
+ 
+ else:		; preds = %no_exit
+ 	%dec = add int %k.1, -1		; <int> [#uses=1]
+ 	br label %endif
+ 
+ endif:		; preds = %else, %then
+ 	%k.0 = phi int [ %dec, %else ], [ %inc.0, %then ]		; <int> [#uses=1]
+ 	store int 2, int* %tmp.28
+ 	%inc.1 = add int %i.0, 1		; <int> [#uses=2]
+ 	%tmp.4 = setlt int %inc.1, %tmp.1		; <bool> [#uses=1]
+ 	br bool %tmp.4, label %no_exit, label %return
+ 
+ return:		; preds = %entry, %endif
+ 	ret void
+ }


Index: llvm/test/Regression/Transforms/ADCE/2002-05-23-ZeroArgPHITest.ll
diff -u llvm/test/Regression/Transforms/ADCE/2002-05-23-ZeroArgPHITest.ll:1.2 llvm/test/Regression/Transforms/ADCE/2002-05-23-ZeroArgPHITest.ll:1.2.2.1
--- llvm/test/Regression/Transforms/ADCE/2002-05-23-ZeroArgPHITest.ll:1.2	Thu Aug 22 11:39:04 2002
+++ llvm/test/Regression/Transforms/ADCE/2002-05-23-ZeroArgPHITest.ll	Thu Jun 26 16:35:49 2003
@@ -23,7 +23,7 @@
 
 bb2:					;[#uses=3]
 	%reg109 = phi %node_t* [ %reg110, %bb2 ], [ %reg107, %bb1 ]		; <%node_t*> [#uses=1]
-	%reg212 = getelementptr %node_t* %reg109, uint 0, ubyte 1		; <%node_t**> [#uses=1]
+	%reg212 = getelementptr %node_t* %reg109, long 0, ubyte 1		; <%node_t**> [#uses=1]
 	%reg110 = load %node_t** %reg212		; <%node_t*> [#uses=2]
 	%cond213 = setne %node_t* %reg110, null		; <bool> [#uses=1]
 	br bool %cond213, label %bb2, label %bb3


Index: llvm/test/Regression/Transforms/ADCE/2002-05-28-Crash.ll
diff -u llvm/test/Regression/Transforms/ADCE/2002-05-28-Crash.ll:1.2 llvm/test/Regression/Transforms/ADCE/2002-05-28-Crash.ll:1.2.2.1
--- llvm/test/Regression/Transforms/ADCE/2002-05-28-Crash.ll:1.2	Tue May 28 16:29:28 2002
+++ llvm/test/Regression/Transforms/ADCE/2002-05-28-Crash.ll	Thu Jun 26 16:35:50 2003
@@ -26,7 +26,8 @@
 	%reg129 = add int %cast109, -1		; <int> [#uses=1]
 	%reg114-idxcast = cast ulong %reg114 to uint		; <uint> [#uses=1]
 	%reg114-idxcast-offset = add uint %reg114-idxcast, 1073741823		; <uint> [#uses=1]
-	%reg124 = getelementptr uint* %set, uint %reg114-idxcast-offset		; <uint*> [#uses=1]
+	%reg114-idxcast-offset = cast uint %reg114-idxcast-offset to long
+	%reg124 = getelementptr uint* %set, long %reg114-idxcast-offset		; <uint*> [#uses=1]
 	%reg125 = load uint* %reg124		; <uint> [#uses=1]
 	%cond232 = setne uint %reg125, 0		; <bool> [#uses=1]
 	br bool %cond232, label %bb3, label %bb2
@@ -38,7 +39,8 @@
 	%add1-indvar = add int %cann-indvar, 1		; <int> [#uses=1]
 	%reg130-idxcast = cast int %reg130 to uint		; <uint> [#uses=1]
 	%reg130-idxcast-offset = add uint %reg130-idxcast, 1073741823		; <uint> [#uses=1]
-	%reg118 = getelementptr uint* %set, uint %reg130-idxcast-offset		; <uint*> [#uses=1]
+	%reg130-idxcast-offset = cast uint %reg130-idxcast-offset to long
+	%reg118 = getelementptr uint* %set, long %reg130-idxcast-offset		; <uint*> [#uses=1]
 	%reg119 = load uint* %reg118		; <uint> [#uses=1]
 	%cond233 = seteq uint %reg119, 0		; <bool> [#uses=1]
 	br bool %cond233, label %bb2, label %bb3


Index: llvm/test/Regression/Transforms/ADCE/basictest1.ll
diff -u llvm/test/Regression/Transforms/ADCE/basictest1.ll:1.2 llvm/test/Regression/Transforms/ADCE/basictest1.ll:1.2.2.1
--- llvm/test/Regression/Transforms/ADCE/basictest1.ll:1.2	Thu Aug 22 11:39:04 2002
+++ llvm/test/Regression/Transforms/ADCE/basictest1.ll	Thu Jun 26 16:35:50 2003
@@ -31,7 +31,7 @@
 	br bool %cond266, label %bb3, label %bb2
 
 bb2:					;[#uses=1]
-	%cast273 = getelementptr [17 x sbyte]* %.LC12, uint 0, uint 0		; <sbyte*> [#uses=0]
+	%cast273 = getelementptr [17 x sbyte]* %.LC12, long 0, long 0		; <sbyte*> [#uses=0]
 	br label %bb3
 
 bb3:					;[#uses=2]
@@ -39,19 +39,19 @@
 	br bool %cond267, label %bb5, label %bb4
 
 bb4:					;[#uses=2]
-	%reg111 = getelementptr [20 x %FILE]* %__iob, uint 0, uint 1, ubyte 3		; <ubyte*> [#uses=1]
-	%cast274 = getelementptr [34 x sbyte]* %.LC9, uint 0, uint 0		; <sbyte*> [#uses=0]
+	%reg111 = getelementptr [20 x %FILE]* %__iob, long 0, long 1, ubyte 3		; <ubyte*> [#uses=1]
+	%cast274 = getelementptr [34 x sbyte]* %.LC9, long 0, long 0		; <sbyte*> [#uses=0]
 	%cast282 = cast ubyte* %reg111 to %FILE*		; <%FILE*> [#uses=0]
 	call void %exit( int 1 )
 	br label %UnifiedExitNode
 
 bb5:					;[#uses=1]
-	%reg107-idxcast1 = cast int %fd to uint		; <uint> [#uses=2]
-	%reg107-idxcast2 = cast int %fd to uint		; <uint> [#uses=1]
-	%reg1311 = getelementptr [3 x %spec_fd_t]* %spec_fd, uint 0, uint %reg107-idxcast2		; <%spec_fd_t*> [#uses=1]
-	%idx1 = getelementptr [3 x %spec_fd_t]* %spec_fd, uint 0, uint %reg107-idxcast1, ubyte 2		; <int> [#uses=3]
+	%reg107-idxcast1 = cast int %fd to long		; <long> [#uses=2]
+	%reg107-idxcast2 = cast int %fd to long		; <long> [#uses=1]
+	%reg1311 = getelementptr [3 x %spec_fd_t]* %spec_fd, long 0, long %reg107-idxcast2		; <%spec_fd_t*> [#uses=1]
+	%idx1 = getelementptr [3 x %spec_fd_t]* %spec_fd, long 0, long %reg107-idxcast1, ubyte 2		; <int> [#uses=3]
 	%reg1321 = load int* %idx1
-	%idx2 = getelementptr %spec_fd_t* %reg1311, uint 0, ubyte 1		; <int> [#uses=1]
+	%idx2 = getelementptr %spec_fd_t* %reg1311, long 0, ubyte 1		; <int> [#uses=1]
 	%reg1331 = load int* %idx2
 	%cond270 = setlt int %reg1321, %reg1331		; <bool> [#uses=1]
 	br bool %cond270, label %bb9, label %bb6
@@ -62,29 +62,29 @@
 	br bool %cond271, label %bb8, label %bb7
 
 bb7:					;[#uses=1]
-	%cast277 = getelementptr [4 x sbyte]* %.LC10, uint 0, uint 0		; <sbyte*> [#uses=0]
+	%cast277 = getelementptr [4 x sbyte]* %.LC10, long 0, long 0		; <sbyte*> [#uses=0]
 	br label %bb8
 
 bb8:					;[#uses=3]
 	br label %UnifiedExitNode
 
 bb9:					;[#uses=1]
-	%reg107-idxcast3 = cast int %fd to uint		; <uint> [#uses=1]
-	%idx3 = getelementptr [3 x %spec_fd_t]* %spec_fd, uint 0, uint %reg107-idxcast3, ubyte 3		; <ubyte*> [#uses=1]
+	%reg107-idxcast3 = cast int %fd to long		; <long> [#uses=1]
+	%idx3 = getelementptr [3 x %spec_fd_t]* %spec_fd, long 0, long %reg107-idxcast3, ubyte 3		; <ubyte*> [#uses=1]
 	%reg1601 = load ubyte** %idx3
-	%reg132-idxcast1 = cast int %reg1321 to uint		; <uint> [#uses=1]
-	%idx4 = getelementptr ubyte* %reg1601, uint %reg132-idxcast1		; <ubyte> [#uses=2]
+	%reg132-idxcast1 = cast int %reg1321 to long		; <long> [#uses=1]
+	%idx4 = getelementptr ubyte* %reg1601, long %reg132-idxcast1		; <ubyte> [#uses=2]
 	%reg1621 = load ubyte* %idx4
-	%cast108 = cast ubyte %reg1621 to uint		; <uint> [#uses=0]
+	%cast108 = cast ubyte %reg1621 to long		; <long> [#uses=0]
 	%reg157 = add int %reg1321, 1		; <int> [#uses=1]
-	%idx5 = getelementptr [3 x %spec_fd_t]* %spec_fd, uint 0, uint %reg107-idxcast1, ubyte 2
+	%idx5 = getelementptr [3 x %spec_fd_t]* %spec_fd, long 0, long %reg107-idxcast1, ubyte 2
 	store int %reg157, int* %idx5
 	%reg163 = load int* %dbglvl		; <int> [#uses=1]
 	%cond272 = setle int %reg163, 4		; <bool> [#uses=1]
 	br bool %cond272, label %bb11, label %bb10
 
 bb10:					;[#uses=1]
-	%cast279 = getelementptr [4 x sbyte]* %.LC11, uint 0, uint 0		; <sbyte*> [#uses=0]
+	%cast279 = getelementptr [4 x sbyte]* %.LC11, long 0, long 0		; <sbyte*> [#uses=0]
 	br label %bb11
 
 bb11:					;[#uses=3]


Index: llvm/test/Regression/Transforms/ADCE/basictest2.ll
diff -u llvm/test/Regression/Transforms/ADCE/basictest2.ll:1.2 llvm/test/Regression/Transforms/ADCE/basictest2.ll:1.2.2.1
--- llvm/test/Regression/Transforms/ADCE/basictest2.ll:1.2	Thu Aug 22 11:39:04 2002
+++ llvm/test/Regression/Transforms/ADCE/basictest2.ll	Thu Jun 26 16:35:50 2003
@@ -31,7 +31,7 @@
 	br bool %cond266, label %bb3, label %bb2
 
 bb2:					;[#uses=1]
-	%cast273 = getelementptr [17 x sbyte]* %.LC12, uint 0, uint 0		; <sbyte*> [#uses=0]
+	%cast273 = getelementptr [17 x sbyte]* %.LC12, long 0, long 0		; <sbyte*> [#uses=0]
 	br label %bb3
 
 bb3:					;[#uses=2]
@@ -40,19 +40,19 @@
 	br label %bb5
 
 bb4:					;[#uses=2]
-	%reg111 = getelementptr [20 x %FILE]* %__iob, uint 0, uint 1, ubyte 3		; <ubyte*> [#uses=1]
-	%cast274 = getelementptr [34 x sbyte]* %.LC9, uint 0, uint 0		; <sbyte*> [#uses=0]
+	%reg111 = getelementptr [20 x %FILE]* %__iob, long 0, long 1, ubyte 3		; <ubyte*> [#uses=1]
+	%cast274 = getelementptr [34 x sbyte]* %.LC9, long 0, long 0		; <sbyte*> [#uses=0]
 	%cast282 = cast ubyte* %reg111 to %FILE*		; <%FILE*> [#uses=0]
 	call void %exit( int 1 )
 	br label %UnifiedExitNode
 
 bb5:					;[#uses=1]
-	%reg107-idxcast1 = cast int %fd to uint		; <uint> [#uses=2]
-	%reg107-idxcast2 = cast int %fd to uint		; <uint> [#uses=1]
-	%reg1311 = getelementptr [3 x %spec_fd_t]* %spec_fd, uint 0, uint %reg107-idxcast2		; <%spec_fd_t*> [#uses=1]
-	%idx1 = getelementptr [3 x %spec_fd_t]* %spec_fd, uint 0, uint %reg107-idxcast1, ubyte 2		; <int> [#uses=3]
+	%reg107-idxcast1 = cast int %fd to long		; <long> [#uses=2]
+	%reg107-idxcast2 = cast int %fd to long		; <long> [#uses=1]
+	%reg1311 = getelementptr [3 x %spec_fd_t]* %spec_fd, long 0, long %reg107-idxcast2		; <%spec_fd_t*> [#uses=1]
+	%idx1 = getelementptr [3 x %spec_fd_t]* %spec_fd, long 0, long %reg107-idxcast1, ubyte 2		; <int> [#uses=3]
 	%reg1321 = load int* %idx1
-	%idx2 = getelementptr %spec_fd_t* %reg1311, uint 0, ubyte 1		; <int> [#uses=1]
+	%idx2 = getelementptr %spec_fd_t* %reg1311, long 0, ubyte 1		; <int> [#uses=1]
 	%reg1331 = load int* %idx2
 	%cond270 = setlt int %reg1321, %reg1331		; <bool> [#uses=1]
 	br bool %cond270, label %bb9, label %bb6
@@ -63,29 +63,29 @@
 	br bool %cond271, label %bb8, label %bb7
 
 bb7:					;[#uses=1]
-	%cast277 = getelementptr [4 x sbyte]* %.LC10, uint 0, uint 0		; <sbyte*> [#uses=0]
+	%cast277 = getelementptr [4 x sbyte]* %.LC10, long 0, long 0		; <sbyte*> [#uses=0]
 	br label %bb8
 
 bb8:					;[#uses=3]
 	br label %UnifiedExitNode
 
 bb9:					;[#uses=1]
-	%reg107-idxcast3 = cast int %fd to uint		; <uint> [#uses=1]
-	%idx3 = getelementptr [3 x %spec_fd_t]* %spec_fd, uint 0, uint %reg107-idxcast3, ubyte 3		; <ubyte*> [#uses=1]
+	%reg107-idxcast3 = cast int %fd to long		; <uint> [#uses=1]
+	%idx3 = getelementptr [3 x %spec_fd_t]* %spec_fd, long 0, long %reg107-idxcast3, ubyte 3		; <ubyte*> [#uses=1]
 	%reg1601 = load ubyte** %idx3
-	%reg132-idxcast1 = cast int %reg1321 to uint		; <uint> [#uses=1]
-	%idx4 = getelementptr ubyte* %reg1601, uint %reg132-idxcast1		; <ubyte> [#uses=2]
+	%reg132-idxcast1 = cast int %reg1321 to long		; <long> [#uses=1]
+	%idx4 = getelementptr ubyte* %reg1601, long %reg132-idxcast1		; <ubyte> [#uses=2]
 	%reg1621 = load ubyte* %idx4
-	%cast108 = cast ubyte %reg1621 to uint		; <uint> [#uses=0]
+	%cast108 = cast ubyte %reg1621 to long		; <long> [#uses=0]
 	%reg157 = add int %reg1321, 1		; <int> [#uses=1]
-	%idx5 = getelementptr [3 x %spec_fd_t]* %spec_fd, uint 0, uint %reg107-idxcast1, ubyte 2
+	%idx5 = getelementptr [3 x %spec_fd_t]* %spec_fd, long 0, long %reg107-idxcast1, ubyte 2
 	store int %reg157, int* %idx5
 	%reg163 = load int* %dbglvl		; <int> [#uses=1]
 	%cond272 = setle int %reg163, 4		; <bool> [#uses=1]
 	br bool %cond272, label %bb11, label %bb10
 
 bb10:					;[#uses=1]
-	%cast279 = getelementptr [4 x sbyte]* %.LC11, uint 0, uint 0		; <sbyte*> [#uses=0]
+	%cast279 = getelementptr [4 x sbyte]* %.LC11, long 0, long 0		; <sbyte*> [#uses=0]
 	br label %bb11
 
 bb11:					;[#uses=3]




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:40:29 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:40:29 2003
Subject: [llvm-commits] CVS: llvm/www/pubs/2002-08-09-LLVMCompilationStrategy.html 2003-05-05-LCTES03-CodeSafety.html
Message-ID: <200306262137.h5QLbUL00313@choi.cs.uiuc.edu>


Changes in directory llvm/www/pubs:

2002-08-09-LLVMCompilationStrategy.html updated: 1.1 -> 1.1.2.1
2003-05-05-LCTES03-CodeSafety.html updated: 1.2 -> 1.2.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/www/pubs/2002-08-09-LLVMCompilationStrategy.html
diff -u llvm/www/pubs/2002-08-09-LLVMCompilationStrategy.html:1.1 llvm/www/pubs/2002-08-09-LLVMCompilationStrategy.html:1.1.2.1
--- llvm/www/pubs/2002-08-09-LLVMCompilationStrategy.html:1.1	Mon May 19 10:32:47 2003
+++ llvm/www/pubs/2002-08-09-LLVMCompilationStrategy.html	Thu Jun 26 16:37:20 2003
@@ -35,6 +35,20 @@
   Technical Report #UIUCDCS-R-2002-2292, Computer Science Dept., Univ. of Illinois, Aug. 2002.
 </ul>
 
+<h2>Update:</h2>
+<ul>
+  Since this document was published, one <b>significant</b> change has been made
+  to LLVM: the GCC C front-end described in the document has been completely
+  rewritten from scratch.  The new C front-end is based on the mainline GCC CVS
+  tree (what will become GCC 3.4), and expands type-safe LLVM code from the GCC
+  AST representation, instead of from the untyped GCC RTL representation.<p>
+
+  This change dramatically improved the quality of code generated and the
+  stability of the system as a whole.<p>
+</ul>
+
+
+
 <h2>Download:</h2>
 <ul>
 <li><a href="2002-08-09-LLVMCompilationStrategy.ps">The LLVM Instruction Set and Compilation Strategy</a> (PS)


Index: llvm/www/pubs/2003-05-05-LCTES03-CodeSafety.html
diff -u llvm/www/pubs/2003-05-05-LCTES03-CodeSafety.html:1.2 llvm/www/pubs/2003-05-05-LCTES03-CodeSafety.html:1.2.2.1
--- llvm/www/pubs/2003-05-05-LCTES03-CodeSafety.html:1.2	Wed May 21 12:15:41 2003
+++ llvm/www/pubs/2003-05-05-LCTES03-CodeSafety.html	Thu Jun 26 16:37:20 2003
@@ -58,6 +58,7 @@
 <ul>
 <li><a href="2003-05-05-LCTES03-CodeSafety.ps">Memory Safety Without Runtime Checks or Garbage Collection</a> (PS)
 <li><a href="2003-05-05-LCTES03-CodeSafety.pdf">Memory Safety Without Runtime Checks or Garbage Collection</a> (PDF)
+<li><a href="2003-06-12-LCTES03-CodeSafety.ppt">Memory Safety Without Runtime Checks or Garbage Collection - Presentation</a> (PPT)
 </ul>
 
 <h2>Bibtex Entry:</h2>
@@ -72,5 +73,9 @@
     URL = {http://llvm.cs.uiuc.edu/pubs/2003-05-05-LCTES03-CodeSafety.html}
   }
 </pre>
+
+<h2> Links: </h2>
+
+<a href="http://safecode.cs.uiuc.edu/">SAFECode project</a> 
 
 </body></html>




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:40:36 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:40:36 2003
Subject: [llvm-commits] CVS: llvm/include/Support/Alloca.h BitSetVector.h CommandLine.h DataTypes.h DepthFirstIterator.h EquivalenceClasses.h GraphTraits.h HashExtras.h LeakDetector.h MathExtras.h NonCopyable.h PostOrderIterator.h STLExtras.h SetOperations.h StringExtras.h TarjanSCCIterator.h Timer.h Tree.h TypeInfo.h hash_map hash_set ilist iterator slist
Message-ID: <200306262137.h5QLbhL00542@choi.cs.uiuc.edu>


Changes in directory llvm/include/Support:

Alloca.h added (r1.2.2.1)
BitSetVector.h updated: 1.5.2.1 -> 1.5.2.2
CommandLine.h updated: 1.12.2.1 -> 1.12.2.2
DataTypes.h updated: 1.8.2.1 -> 1.8.2.2
DepthFirstIterator.h updated: 1.4 -> 1.4.2.1
EquivalenceClasses.h updated: 1.2 -> 1.2.2.1
GraphTraits.h updated: 1.2 -> 1.2.2.1
HashExtras.h updated: 1.5 -> 1.5.2.1
LeakDetector.h updated: 1.1 -> 1.1.2.1
MathExtras.h updated: 1.5 -> 1.5.2.1
NonCopyable.h updated: 1.2 -> 1.2.2.1
PostOrderIterator.h updated: 1.8 -> 1.8.2.1
STLExtras.h updated: 1.5 -> 1.5.2.1
SetOperations.h updated: 1.1 -> 1.1.2.1
StringExtras.h updated: 1.3 -> 1.3.2.1
TarjanSCCIterator.h updated: 1.5.2.1 -> 1.5.2.2
Timer.h updated: 1.6.2.1 -> 1.6.2.2
Tree.h updated: 1.3.2.1 -> 1.3.2.2
TypeInfo.h updated: 1.1 -> 1.1.2.1
hash_map updated: 1.6.2.1 -> 1.6.2.2
hash_set updated: 1.6.2.1 -> 1.6.2.2
ilist updated: 1.6.2.1 -> 1.6.2.2
iterator updated: 1.2.2.1 -> 1.2.2.2
slist updated: 1.2.2.1 -> 1.2.2.2

---
Log message:

Merged with mainline on Thursday, June 26, 2003.
Kept includes of "Config/assert.h" to avoid implicit dependencies on header
files.


---
Diffs of the changes:

Index: llvm/include/Support/Alloca.h
diff -c /dev/null llvm/include/Support/Alloca.h:1.1
*** /dev/null	Thu Jun 26 16:37:43 2003
--- llvm/include/Support/Alloca.h	Mon Jun 16 16:53:55 2003
***************
*** 0 ****
--- 1,30 ----
+ //===-- include/Support/Alloca.h - Support for alloca header -----*- C++ -*--=//
+ //
+ // Some platforms do not have alloca.h; others do. You can include this
+ // file instead of <alloca.h> and it will include <alloca.h> on the platforms
+ // that require you to do so to use alloca().
+ //
+ //===----------------------------------------------------------------------===//
+ 
+ #ifndef LLVM_SUPPORT_ALLOCA_H
+ #define LLVM_SUPPORT_ALLOCA_H
+ 
+ // TODO: Determine HAVE_ALLOCA_H based on autoconf results.
+ // The following method is too brittle.
+ #if defined(HAVE_ALLOCA_H)
+ #undef HAVE_ALLOCA_H
+ #endif
+ 
+ #if defined(__linux__)
+ #define HAVE_ALLOCA_H 1
+ #elif defined(__sparc__)
+ #define HAVE_ALLOCA_H 1
+ #elif defined(__FreeBSD__)
+ // not defined here
+ #endif
+ 
+ #if HAVE_ALLOCA_H
+ #include <alloca.h>
+ #endif
+ 
+ #endif  /* LLVM_SUPPORT_ALLOCA_H */


Index: llvm/include/Support/BitSetVector.h
diff -u llvm/include/Support/BitSetVector.h:1.5.2.1 llvm/include/Support/BitSetVector.h:1.5.2.2
--- llvm/include/Support/BitSetVector.h:1.5.2.1	Mon Jun 23 13:50:38 2003
+++ llvm/include/Support/BitSetVector.h	Thu Jun 26 16:34:32 2003
@@ -9,12 +9,6 @@
 // We therefore use a vector of bitsets.  The maxmimum size of our sets
 // (i.e., the size of the universal set) can be chosen at creation time.
 //
-// The size of each Bitset is defined by the macro WORDSIZE.
-// 
-// NOTE: The WORDSIZE macro should be made machine-dependent, in order to use
-// 64-bit words or whatever gives most efficient Bitsets on each platform.
-// 
-// 
 // External functions:
 // 
 // bool Disjoint(const BitSetVector& set1, const BitSetVector& set2):
@@ -23,22 +17,20 @@
 // 
 //===----------------------------------------------------------------------===//
 
-#ifndef LLVM_SUPPORT_BITVECTORSET_H
-#define LLVM_SUPPORT_BITVECTORSET_H
+#ifndef SUPPORT_BITSETVECTOR_H
+#define SUPPORT_BITSETVECTOR_H
 
 #include <bitset>
 #include <vector>
 #include <functional>
 #include <iostream>
-
 #include "Config/assert.h"
 
-#define WORDSIZE (32U)
-
-
 class BitSetVector {
+  enum { BITSET_WORDSIZE = sizeof(long)*8 };
+
   // Types used internal to the representation
-  typedef std::bitset<WORDSIZE> bitword;
+  typedef std::bitset<BITSET_WORDSIZE> bitword;
   typedef bitword::reference reference;
   class iterator;
 
@@ -48,11 +40,13 @@
 
 private:
   // Utility functions for the representation
-  static unsigned NumWords(unsigned Size) { return (Size+WORDSIZE-1)/WORDSIZE;} 
-  static unsigned LastWordSize(unsigned Size) { return Size % WORDSIZE; }
+  static unsigned NumWords(unsigned Size) {
+    return (Size+BITSET_WORDSIZE-1)/BITSET_WORDSIZE;
+  } 
+  static unsigned LastWordSize(unsigned Size) { return Size % BITSET_WORDSIZE; }
 
   // Clear the unused bits in the last word.
-  // The unused bits are the high (WORDSIZE - LastWordSize()) bits
+  // The unused bits are the high (BITSET_WORDSIZE - LastWordSize()) bits
   void ClearUnusedBits() {
     unsigned long usedBits = (1U << LastWordSize(size())) - 1;
     bitsetVec.back() &= bitword(usedBits);
@@ -91,7 +85,7 @@
   }
   reference operator[](unsigned n) {
     assert(n  < size() && "BitSetVector: Bit number out of range");
-    unsigned ndiv = n / WORDSIZE, nmod = n % WORDSIZE;
+    unsigned ndiv = n / BITSET_WORDSIZE, nmod = n % BITSET_WORDSIZE;
     return bitsetVec[ndiv][nmod];
   }
   iterator begin() { return iterator::begin(*this); }
@@ -116,7 +110,7 @@
   ///  
   bool test(unsigned n) const {
     assert(n  < size() && "BitSetVector: Bit number out of range");
-    unsigned ndiv = n / WORDSIZE, nmod = n % WORDSIZE;
+    unsigned ndiv = n / BITSET_WORDSIZE, nmod = n % BITSET_WORDSIZE;
     return bitsetVec[ndiv].test(nmod);
   }
   bool any() const {
@@ -203,13 +197,13 @@
 
     // Increment and decrement operators (pre and post)
     iterator& operator++() {
-      if (++currentBit == WORDSIZE)
+      if (++currentBit == BITSET_WORDSIZE)
         { currentBit = 0; if (currentWord < bitvec->size()) ++currentWord; }
       return *this;
     }
     iterator& operator--() {
       if (currentBit == 0) {
-        currentBit = WORDSIZE-1;
+        currentBit = BITSET_WORDSIZE-1;
         currentWord = (currentWord == 0)? bitvec->size() : --currentWord;
       }
       else


Index: llvm/include/Support/CommandLine.h
diff -u llvm/include/Support/CommandLine.h:1.12.2.1 llvm/include/Support/CommandLine.h:1.12.2.2
--- llvm/include/Support/CommandLine.h:1.12.2.1	Mon Jun 23 13:50:38 2003
+++ llvm/include/Support/CommandLine.h	Thu Jun 26 16:34:32 2003
@@ -10,8 +10,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef LLVM_SUPPORT_COMMANDLINE_H
-#define LLVM_SUPPORT_COMMANDLINE_H
+#ifndef SUPPORT_COMMANDLINE_H
+#define SUPPORT_COMMANDLINE_H
 
 #include <string>
 #include <vector>


Index: llvm/include/Support/DataTypes.h
diff -u llvm/include/Support/DataTypes.h:1.8.2.1 llvm/include/Support/DataTypes.h:1.8.2.2
--- llvm/include/Support/DataTypes.h:1.8.2.1	Mon Jun 23 13:50:38 2003
+++ llvm/include/Support/DataTypes.h	Thu Jun 26 16:34:33 2003
@@ -16,8 +16,8 @@
 // TODO: This file sucks.  Not only does it not work, but this stuff should be
 // autoconfiscated anyways. Major FIXME
 
-#ifndef LLVM_SUPPORT_DATATYPES_H
-#define LLVM_SUPPORT_DATATYPES_H
+#ifndef SUPPORT_DATATYPES_H
+#define SUPPORT_DATATYPES_H
 
 //
 // Un-define these constants so that we can re-define only the one that we want
@@ -44,4 +44,4 @@
 #error "include/Support/DataTypes.h could not determine endianness!"
 #endif
 
-#endif  /* LLVM_SUPPORT_DATATYPES_H */
+#endif  /* SUPPORT_DATATYPES_H */


Index: llvm/include/Support/DepthFirstIterator.h
diff -u llvm/include/Support/DepthFirstIterator.h:1.4 llvm/include/Support/DepthFirstIterator.h:1.4.2.1
--- llvm/include/Support/DepthFirstIterator.h:1.4	Sun Oct 27 20:11:16 2002
+++ llvm/include/Support/DepthFirstIterator.h	Thu Jun 26 16:34:33 2003
@@ -5,8 +5,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef LLVM_SUPPORT_DEPTH_FIRST_ITERATOR_H
-#define LLVM_SUPPORT_DEPTH_FIRST_ITERATOR_H
+#ifndef SUPPORT_DEPTHFIRSTITERATOR_H
+#define SUPPORT_DEPTHFIRSTITERATOR_H
 
 #include "Support/GraphTraits.h"
 #include "Support/iterator"


Index: llvm/include/Support/EquivalenceClasses.h
diff -u llvm/include/Support/EquivalenceClasses.h:1.2 llvm/include/Support/EquivalenceClasses.h:1.2.2.1
--- llvm/include/Support/EquivalenceClasses.h:1.2	Wed Jun  4 03:00:05 2003
+++ llvm/include/Support/EquivalenceClasses.h	Thu Jun 26 16:34:33 2003
@@ -8,8 +8,8 @@
 // 
 //===------------------------------------------------------------------===//
 
-#ifndef LLVM_SUPPORT_EQUIVALENCE_CLASSES_H
-#define LLVM_SUPPORT_EQUIVALENCE_CLASSES_H
+#ifndef SUPPORT_EQUIVALENCECLASSES_H
+#define SUPPORT_EQUIVALENCECLASSES_H
 
 #include <map>
 #include <vector>


Index: llvm/include/Support/GraphTraits.h
diff -u llvm/include/Support/GraphTraits.h:1.2 llvm/include/Support/GraphTraits.h:1.2.2.1
--- llvm/include/Support/GraphTraits.h:1.2	Sun Oct 13 12:12:05 2002
+++ llvm/include/Support/GraphTraits.h	Thu Jun 26 16:34:33 2003
@@ -8,8 +8,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef LLVM_SUPPORT_GRAPH_TRAITS_H
-#define LLVM_SUPPORT_GRAPH_TRAITS_H
+#ifndef SUPPORT_GRAPHTRAITS_H
+#define SUPPORT_GRAPHTRAITS_H
 
 // GraphTraits - This class should be specialized by different graph types...
 // which is why the default version is empty.


Index: llvm/include/Support/HashExtras.h
diff -u llvm/include/Support/HashExtras.h:1.5 llvm/include/Support/HashExtras.h:1.5.2.1
--- llvm/include/Support/HashExtras.h:1.5	Sun Oct 27 20:11:16 2002
+++ llvm/include/Support/HashExtras.h	Thu Jun 26 16:34:33 2003
@@ -7,8 +7,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef LLVM_SUPPORT_HASHEXTRAS_H
-#define LLVM_SUPPORT_HASHEXTRAS_H
+#ifndef SUPPORT_HASHEXTRAS_H
+#define SUPPORT_HASHEXTRAS_H
 
 #include <string>
 #include "Support/hash_map"


Index: llvm/include/Support/LeakDetector.h
diff -u llvm/include/Support/LeakDetector.h:1.1 llvm/include/Support/LeakDetector.h:1.1.2.1
--- llvm/include/Support/LeakDetector.h:1.1	Sun Sep  8 13:51:12 2002
+++ llvm/include/Support/LeakDetector.h	Thu Jun 26 16:34:33 2003
@@ -12,8 +12,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef SUPPORT_LEAK_DETECTOR_H
-#define SUPPORT_LEAK_DETECTOR_H
+#ifndef SUPPORT_LEAKDETECTOR_H
+#define SUPPORT_LEAKDETECTOR_H
 
 #include <string>
 class Value;


Index: llvm/include/Support/MathExtras.h
diff -u llvm/include/Support/MathExtras.h:1.5 llvm/include/Support/MathExtras.h:1.5.2.1
--- llvm/include/Support/MathExtras.h:1.5	Sun Oct 27 20:11:16 2002
+++ llvm/include/Support/MathExtras.h	Thu Jun 26 16:34:33 2003
@@ -4,8 +4,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef SUPPORT_MATH_EXTRAS_H
-#define SUPPORT_MATH_EXTRAS_H
+#ifndef SUPPORT_MATHEXTRAS_H
+#define SUPPORT_MATHEXTRAS_H
 
 #include "Support/DataTypes.h"
 


Index: llvm/include/Support/NonCopyable.h
diff -u llvm/include/Support/NonCopyable.h:1.2 llvm/include/Support/NonCopyable.h:1.2.2.1
--- llvm/include/Support/NonCopyable.h:1.2	Tue Jun  3 10:30:48 2003
+++ llvm/include/Support/NonCopyable.h	Thu Jun 26 16:34:33 2003
@@ -9,8 +9,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef LLVM_SUPPORT_NONCOPYABLE_H
-#define LLVM_SUPPORT_NONCOPYABLE_H
+#ifndef SUPPORT_NONCOPYABLE_H
+#define SUPPORT_NONCOPYABLE_H
 
 class NonCopyable {
   // Disable the copy constructor and the assignment operator


Index: llvm/include/Support/PostOrderIterator.h
diff -u llvm/include/Support/PostOrderIterator.h:1.8 llvm/include/Support/PostOrderIterator.h:1.8.2.1
--- llvm/include/Support/PostOrderIterator.h:1.8	Fri Mar 21 15:40:39 2003
+++ llvm/include/Support/PostOrderIterator.h	Thu Jun 26 16:34:33 2003
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef LLVM_SUPPORT_POSTORDER_ITERATOR_H
-#define LLVM_SUPPORT_POSTORDER_ITERATOR_H
+#ifndef SUPPORT_POSTORDERITERATOR_H
+#define SUPPORT_POSTORDERITERATOR_H
 
 #include "Support/GraphTraits.h"
 #include "Support/iterator"


Index: llvm/include/Support/STLExtras.h
diff -u llvm/include/Support/STLExtras.h:1.5 llvm/include/Support/STLExtras.h:1.5.2.1
--- llvm/include/Support/STLExtras.h:1.5	Sun Oct 27 13:16:27 2002
+++ llvm/include/Support/STLExtras.h	Thu Jun 26 16:34:33 2003
@@ -7,8 +7,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef LLVM_SUPPORT_STL_EXTRAS_H
-#define LLVM_SUPPORT_STL_EXTRAS_H
+#ifndef SUPPORT_STLEXTRAS_H
+#define SUPPORT_STLEXTRAS_H
 
 #include <functional>
 #include "Support/iterator"


Index: llvm/include/Support/SetOperations.h
diff -u llvm/include/Support/SetOperations.h:1.1 llvm/include/Support/SetOperations.h:1.1.2.1
--- llvm/include/Support/SetOperations.h:1.1	Mon Feb  4 21:35:10 2002
+++ llvm/include/Support/SetOperations.h	Thu Jun 26 16:34:33 2003
@@ -5,8 +5,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef LLVM_SUPPORT_SET_OPERATIONS_H
-#define LLVM_SUPPORT_SET_OPERATIONS_H
+#ifndef SUPPORT_SETOPERATIONS_H
+#define SUPPORT_SETOPERATIONS_H
 
 // set_union(A, B) - Compute A := A u B, return whether A changed.
 //


Index: llvm/include/Support/StringExtras.h
diff -u llvm/include/Support/StringExtras.h:1.3 llvm/include/Support/StringExtras.h:1.3.2.1
--- llvm/include/Support/StringExtras.h:1.3	Sun Apr  7 03:36:19 2002
+++ llvm/include/Support/StringExtras.h	Thu Jun 26 16:34:33 2003
@@ -4,8 +4,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef SUPPORT_STRING_EXTRAS_H
-#define SUPPORT_STRING_EXTRAS_H
+#ifndef SUPPORT_STRINGEXTRAS_H
+#define SUPPORT_STRINGEXTRAS_H
 
 #include "Support/DataTypes.h"
 #include <string>


Index: llvm/include/Support/TarjanSCCIterator.h
diff -u llvm/include/Support/TarjanSCCIterator.h:1.5.2.1 llvm/include/Support/TarjanSCCIterator.h:1.5.2.2
--- llvm/include/Support/TarjanSCCIterator.h:1.5.2.1	Mon Jun 23 13:50:38 2003
+++ llvm/include/Support/TarjanSCCIterator.h	Thu Jun 26 16:34:33 2003
@@ -1,4 +1,4 @@
-//===-- Support/TarjanSCCIterator.h -Generic Tarjan SCC iterator -*- C++ -*--=//
+//===-- Support/TarjanSCCIterator.h - Tarjan SCC iterator -------*- C++ -*-===//
 //
 // This builds on the Support/GraphTraits.h file to find the strongly 
 // connected components (SCCs) of a graph in O(N+E) time using
@@ -9,18 +9,18 @@
 // 
 // To visit S1 *before* S2, use the TarjanSCCIterator on the Inverse graph.
 // (NOTE: This requires some simple wrappers and is not supported yet.)
+//
 //===----------------------------------------------------------------------===//
 
-#ifndef LLVM_SUPPORT_TARJANSCC_ITERATOR_H
-#define LLVM_SUPPORT_TARJANSCC_ITERATOR_H
+#ifndef SUPPORT_TARJANSCCITERATOR_H
+#define SUPPORT_TARJANSCCITERATOR_H
 
 #include "Support/GraphTraits.h"
-#include <Support/Statistic.h>
-#include <Support/iterator>
+#include "Support/Statistic.h"
+#include "Support/iterator"
 #include <vector>
 #include <stack>
 #include <map>
-
 #include "Config/assert.h"
 
 //--------------------------------------------------------------------------


Index: llvm/include/Support/Timer.h
diff -u llvm/include/Support/Timer.h:1.6.2.1 llvm/include/Support/Timer.h:1.6.2.2
--- llvm/include/Support/Timer.h:1.6.2.1	Mon Jun 23 13:50:38 2003
+++ llvm/include/Support/Timer.h	Thu Jun 26 16:34:33 2003
@@ -64,7 +64,7 @@
     PeakMemBase = T.PeakMemBase;
     Name = T.Name;
     Started = T.Started;
-    assert (TG == T.TG && "Can only assign timers in the same TimerGroup!");
+    assert(TG == T.TG && "Can only assign timers in the same TimerGroup!");
     return *this;
   }
 


Index: llvm/include/Support/Tree.h
diff -u llvm/include/Support/Tree.h:1.3.2.1 llvm/include/Support/Tree.h:1.3.2.2
--- llvm/include/Support/Tree.h:1.3.2.1	Mon Jun 23 13:50:38 2003
+++ llvm/include/Support/Tree.h	Thu Jun 26 16:34:33 2003
@@ -5,11 +5,10 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef LLVM_SUPPORT_TREE_H
-#define LLVM_SUPPORT_TREE_H
+#ifndef SUPPORT_TREE_H
+#define SUPPORT_TREE_H
 
 #include <vector>
-
 #include "Config/assert.h"
 
 template<class ConcreteTreeNode, class Payload>


Index: llvm/include/Support/TypeInfo.h
diff -u llvm/include/Support/TypeInfo.h:1.1 llvm/include/Support/TypeInfo.h:1.1.2.1
--- llvm/include/Support/TypeInfo.h:1.1	Tue Jul 23 12:56:53 2002
+++ llvm/include/Support/TypeInfo.h	Thu Jun 26 16:34:33 2003
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef LLVM_SUPPORT_TYPEINFO_H
-#define LLVM_SUPPORT_TYPEINFO_H
+#ifndef SUPPORT_TYPEINFO_H
+#define SUPPORT_TYPEINFO_H
 
 #include <typeinfo>
 


Index: llvm/include/Support/hash_map
diff -u llvm/include/Support/hash_map:1.6.2.1 llvm/include/Support/hash_map:1.6.2.2
--- llvm/include/Support/hash_map:1.6.2.1	Mon Jun 23 13:50:38 2003
+++ llvm/include/Support/hash_map	Thu Jun 26 16:34:33 2003
@@ -7,8 +7,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef SUPPORT_HASHMAP_H
-#define SUPPORT_HASHMAP_H
+#ifndef SUPPORT_HASH_MAP
+#define SUPPORT_HASH_MAP
 
 // Compiler Support Matrix
 //


Index: llvm/include/Support/hash_set
diff -u llvm/include/Support/hash_set:1.6.2.1 llvm/include/Support/hash_set:1.6.2.2
--- llvm/include/Support/hash_set:1.6.2.1	Mon Jun 23 13:50:38 2003
+++ llvm/include/Support/hash_set	Thu Jun 26 16:34:33 2003
@@ -7,8 +7,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef SUPPORT_HASHSET_H
-#define SUPPORT_HASHSET_H
+#ifndef SUPPORT_HASH_SET
+#define SUPPORT_HASH_SET
 
 // Compiler Support Matrix
 //


Index: llvm/include/Support/ilist
diff -u llvm/include/Support/ilist:1.6.2.1 llvm/include/Support/ilist:1.6.2.2
--- llvm/include/Support/ilist:1.6.2.1	Mon Jun 23 13:50:38 2003
+++ llvm/include/Support/ilist	Thu Jun 26 16:34:33 2003
@@ -28,8 +28,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef INCLUDED_SUPPORT_ILIST
-#define INCLUDED_SUPPORT_ILIST
+#ifndef SUPPORT_ILIST
+#define SUPPORT_ILIST
 
 #include "Config/assert.h"
 #include <algorithm>


Index: llvm/include/Support/iterator
diff -u llvm/include/Support/iterator:1.2.2.1 llvm/include/Support/iterator:1.2.2.2
--- llvm/include/Support/iterator:1.2.2.1	Mon Jun 23 13:50:38 2003
+++ llvm/include/Support/iterator	Thu Jun 26 16:34:33 2003
@@ -16,8 +16,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef SUPPORT_ITERATOR_H
-#define SUPPORT_ITERATOR_H
+#ifndef SUPPORT_ITERATOR
+#define SUPPORT_ITERATOR
 
 #include "Config/config.h"
 


Index: llvm/include/Support/slist
diff -u llvm/include/Support/slist:1.2.2.1 llvm/include/Support/slist:1.2.2.2
--- llvm/include/Support/slist:1.2.2.1	Mon Jun 23 13:50:38 2003
+++ llvm/include/Support/slist	Thu Jun 26 16:34:33 2003
@@ -7,8 +7,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef SUPPORT_SLIST_H
-#define SUPPORT_SLIST_H
+#ifndef SUPPORT_SLIST
+#define SUPPORT_SLIST
 
 #include "Config/config.h"
 




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:40:43 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:40:43 2003
Subject: [llvm-commits] CVS: llvm/lib/Reoptimizer/Inst/lib/Phase1/Phase1.cpp PrimInfo.cpp PrimInfo.h
Message-ID: <200306262137.h5QLbU700318@choi.cs.uiuc.edu>


Changes in directory llvm/lib/Reoptimizer/Inst/lib/Phase1:

Phase1.cpp updated: 1.26 -> 1.26.4.1
PrimInfo.cpp updated: 1.15 -> 1.15.4.1
PrimInfo.h updated: 1.10 -> 1.10.4.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/lib/Reoptimizer/Inst/lib/Phase1/Phase1.cpp
diff -u llvm/lib/Reoptimizer/Inst/lib/Phase1/Phase1.cpp:1.26 llvm/lib/Reoptimizer/Inst/lib/Phase1/Phase1.cpp:1.26.4.1
--- llvm/lib/Reoptimizer/Inst/lib/Phase1/Phase1.cpp:1.26	Thu May 22 22:26:13 2003
+++ llvm/lib/Reoptimizer/Inst/lib/Phase1/Phase1.cpp	Thu Jun 26 16:35:20 2003
@@ -12,7 +12,7 @@
 #include <set>
 #include <algorithm>
 #include <map>
-#include <typeinfo>
+#include <string>
 
 #include "llvm/Pass.h"
 #include "llvm/Module.h"
@@ -78,6 +78,12 @@
     Module*          m_module;             // Current module
     vector<PrimInfo> m_primInfos;          // Deferred documentation structures
     TargetData       m_targetData;         // For obtaining target-specific info
+
+    static const unsigned NUM_START_OPDS = 3;
+    static const unsigned NUM_END_OPDS = 3;
+    static const unsigned NUM_STOPD_USES = 2;
+    static const unsigned ENDCALL_ID_OPIDX = 1;
+    static const unsigned POINTCALL_ID_OPIDX = 1;
 };
 
 //////////////// Phase1 implementation ////////////////
@@ -112,7 +118,7 @@
                                      m_module);
 
     // Invoke PrimInfo's static initializer
-    PrimInfo::buildStructType(&m);
+    PrimInfo::buildStructTypes(&m);
 
     //
     ////////////////
@@ -181,12 +187,17 @@
 
     // Populate with builtins
 
-    // TODO: Replace the hard-coded mechanism for the point metrics with a
-    // sigfun registration mechanism like that which exists for region metrics
-    // (pp_regionPair()).
+    // The new approach: pp_declareInterval{Start,End} are the only sigfuns used to
+    // declare an interval, instead of the interval being implicit by the given locations
+    // of the calls to (registered) region sigfuns.  For now, we will still use the old
+    // code for processing the contents of 'rpairs', but the only element of rpairs will
+    // be the pair of interval-declaration sigfuns.  TODO: Clean up this approach, and do
+    // the same for point-site declarations.
+
+    Function* intervalStartDeclFunc = m_module->getNamedFunction("pp_declareIntervalStart");
+    Function* intervalEndDeclFunc = m_module->getNamedFunction("pp_declareIntervalEnd");
 
-    // Find user-declared region-pairs and add them to the list
-    findRegionPairs(rpairs);
+    rpairs.push_back(std::make_pair(intervalStartDeclFunc, intervalEndDeclFunc));
 
     ////////////////
     // Build list of point metric spec sigfuns, finding ppMetricSpec will yield
@@ -218,26 +229,10 @@
     for(unsigned i = 0, e = rpairs.size(); i < e; ++i)
         transformSites(rpairs[i]);
 
-    // Build the interval PrimInfo instance -> GBT index map. This maps the address of a
-    // particular interval PrimInfo instance to the index in the GBT that corresponds to
-    // the record that represents the start-interval global-volatile and related
-    // information.
-
-    std::map<PrimInfo*, unsigned> gbtIdxMap;
-    int idx = 0;
-    for(vector<PrimInfo>::iterator i = m_primInfos.begin(), e = m_primInfos.end(); i != e; ++i){
-        if(i->getType() == PrimInfo::REGION) {
-            gbtIdxMap[&*i] = idx;
-            idx += 2;
-        }
-        else
-            idx++;
-    }
-
     // Build the global bookkeeping table (GBT) contents
     vector<Constant*> gbtElems;
     for(vector<PrimInfo>::iterator i = m_primInfos.begin(), e = m_primInfos.end(); i != e; ++i)
-        i->buildStructInstances(gbtElems, gbtIdxMap, &m_targetData);
+        i->buildStructInstances(gbtElems);
 
     // Make the GBT itself, and the corresponding global variable.
     ArrayType* gbtType = ArrayType::get(PrimInfo::getStructType(), gbtElems.size());
@@ -295,19 +290,19 @@
 
         CallInst* startCall = dyn_cast<CallInst>(*u);
         assert(startCall && "Use of a registered pp sigfun not in a call");
-        
+
         // Find call to endpoint: The first operand of the start function is used in only
         // two places: the call to the start function, and the call to the end
         // function. Find the call to the end function.
 
-        assert(startCall->getNumOperands() == 2 &&
-               "Start-region call insts should only have 2 operands");
+        assert(startCall->getNumOperands() == NUM_START_OPDS &&
+               "Start-region call inst has unexpected number of operands");
 
         Value* stOpd = startCall->getOperand(1);
 
         assert(isa<AllocaInst>(stOpd) && "Unexpected first operand of start call");
-        assert(stOpd->use_size() == 2 &&
-               "Expect only two uses of start-region operand");
+        assert(stOpd->use_size() == NUM_STOPD_USES &&
+               "Unexpected number of uses of of start-region (\"connector\") operand");
         
         CallInst* endCall;
         if((endCall = dyn_cast<CallInst>(stOpd->use_back()))) {
@@ -338,11 +333,21 @@
 
     return cast<GlobalVariable>(ci->getOperand(1));
 }
-    
+
+static unsigned extractUnsignedValue(CallInst* ci, unsigned nth) 
+{
+    // extract the literal value of the nth actual of ci, which must be of type
+    // unsigned.
+
+    if(ConstantUInt* ui = dyn_cast<ConstantUInt>(ci->getOperand(nth)))
+        return ui->getValue();
+
+    assert(0 && "Failed to find unsigned value as call actual");
+    return 0; // squash warning
+}
+
 void Phase1::transformSite(CallInst* startCall, CallInst* endCall) 
 {
-    GlobalVariable* metricVar = GetLHSGlobal(endCall);
-    
     // Insert placeholder code
     PlaceholderPair startPair, endPair;
     generatePlaceholder(startCall, startPair);
@@ -350,8 +355,9 @@
 
     // Create info about this site for later documentation
 
-    m_primInfos.push_back(PrimInfo(PrimInfo::REGION, startPair.first, endPair.first,
-                                   metricVar, m_module, 
+    unsigned siteID = extractUnsignedValue(endCall, ENDCALL_ID_OPIDX);
+    m_primInfos.push_back(PrimInfo(PrimInfo::REGION, siteID,
+                                   startPair.first, endPair.first, m_module,
                                    startCall->getCalledFunction(),
                                    endCall->getCalledFunction()));
 
@@ -379,8 +385,6 @@
 
 void Phase1::transformSite(CallInst* pointCall)
 {
-    GlobalVariable* metricVar = GetLHSGlobal(pointCall);
-
     // Insert placeholder code
     PlaceholderPair php;
     generatePlaceholder(pointCall, php);
@@ -391,7 +395,8 @@
     PrimInfo::PrimType type = pointCall->getCalledFunction() == m_counterPrimFunc ?
         PrimInfo::COUNTER : PrimInfo::POINT;
 
-    m_primInfos.push_back(PrimInfo(type, php.first, metricVar, m_module,
+    unsigned siteID = extractUnsignedValue(pointCall, POINTCALL_ID_OPIDX);
+    m_primInfos.push_back(PrimInfo(type, siteID, php.first, m_module,
                                    pointCall->getCalledFunction()));
 
     // Remove the call


Index: llvm/lib/Reoptimizer/Inst/lib/Phase1/PrimInfo.cpp
diff -u llvm/lib/Reoptimizer/Inst/lib/Phase1/PrimInfo.cpp:1.15 llvm/lib/Reoptimizer/Inst/lib/Phase1/PrimInfo.cpp:1.15.4.1
--- llvm/lib/Reoptimizer/Inst/lib/Phase1/PrimInfo.cpp:1.15	Tue May 13 14:43:03 2003
+++ llvm/lib/Reoptimizer/Inst/lib/Phase1/PrimInfo.cpp	Thu Jun 26 16:35:20 2003
@@ -2,10 +2,8 @@
 // programmer: Joel Stanley
 //       date: Tue Jan 28 14:07:05 CST 2003
 //     fileid: PrimInfo.cpp
-//    purpose: Captures information about an instance of a performance primitive.  The
-//    information is intended to be carried between phases by reconstructing PrimInfo
-//    contents that have been placed in global static data structures by Phase1's
-//    invocation of PrimInfo::Document.
+//    purpose: Captures information about an instance of a performance primitive "marker"
+//             in the code.
 
 #include "llvm/Module.h"
 #include "llvm/Type.h"
@@ -15,7 +13,6 @@
 #include "llvm/GlobalVariable.h"
 #include "llvm/Constant.h"
 #include "Support/VectorExtras.h"
-#include "llvm/Target/TargetData.h"
 
 #include "PrimInfo.h"
 #include "Intraphase.h"
@@ -31,16 +28,16 @@
 StructType* PrimInfo::sm_structType = 0;
 
 PrimInfo::PrimInfo(PrimType type,
+                   unsigned siteID,
                    GlobalVariable* startGlob,
                    GlobalVariable* endGlob,
-                   GlobalVariable* metricVar,
                    Module* module,
                    Function* startFunc,
                    Function* endFunc):
     m_type(type),
+    m_siteID(siteID),
     m_globVol(startGlob),
     m_globVolEnd(endGlob),
-    m_metricVar(metricVar),
     m_module(module),
     m_startFunc(startFunc),
     m_endFunc(endFunc)
@@ -48,126 +45,75 @@
 }
 
 PrimInfo::PrimInfo(PrimType type,
+                   unsigned siteID,
                    GlobalVariable* globVol,
-                   GlobalVariable* metricVar,
                    Module* module,
                    Function* calledFunc):
     m_type(type),
+    m_siteID(siteID),
     m_globVol(globVol),
-    m_metricVar(metricVar),
     m_module(module),
     m_startFunc(calledFunc)
 {
     assert(type != POINT || calledFunc && "point type --> non-null called function");
 }
 
-void PrimInfo::buildStructType(Module* module)
+void PrimInfo::buildStructTypes(Module* module)
 {
-    // The structure deposited into the LLVM bytecode essentially looks like:
-    // struct PrimInfo
+    static bool init = false;
+
+    assert(!init && "Expect only one invocation of this function");
+    init = true;
+
+    // {{{ struct PrimInfo
     // {
+    //   // Site ID for this site (one value identifies one entire interval as well)
+    //   unsigned siteID;
+    //   
     //   // Type of GBT entry type (i.e., member of enum GBTEntryType)
     //   unsigned gbtType;
     //
     //   // Address of global variable corresponding to this primitive.
     //   unsigned short* loadVar;
     //
-    //   // Index in GBT of struct of corresponding GBT_INTERVAL_START entry
-    //   // (NB: Only valid if gbtType == GBT_INTERVAL_END)
-    //   unsigned gbtStartIdx;
-    //
-    //   // Size, in bytes, of the parameter to the instrumentation function that must be
-    //   // invoked at the instrumentation that corresponds to this structure instance.
-    //   unsigned paramSize;
-    //
-    //   // Pointer to memory for instFunc's return value
-    //   // This is initalized to null for start-interval sites, and filled in at runtime
-    //   // by a pointer to heap-allocated memory of the appropriate size. For all other
-    //   // site types, this is initialized to the metric variable address.
-    //   void* retVal;
-    //   
-    //   // Pointer to instrumentation function
-    //   void* instFunc;
-    //
-    // Other stuff will need to go here...(TODO)
     // };
-
-    static bool init = false;
-
-    assert(!init && "Expect only one invocation of this function");
-    init = true;
+    // }}}
 
     PointerType* uspt = PointerType::get(Type::UShortTy);
-    PointerType* vpt = PointerType::get(Type::VoidTy);
+
     sm_structType =
         StructType::get(make_vector<const Type*>(
-          Type::UIntTy, uspt, Type::UIntTy, Type::UIntTy, vpt, vpt, 0));
-    module->addTypeName("PrimInfo", sm_structType);
-}
-
+          Type::UIntTy, Type::UIntTy, uspt, 0));
 
-// For start sites
-static ConstantStruct* makeConstStruct(StructType* st,
-                                       unsigned gbtType,
-                                       GlobalVariable* loadVar,
-                                       unsigned paramSize,
-                                       Function* instFunc)
-{
-    std::vector<Constant*> init;
-    init.push_back(ConstantUInt::get(Type::UIntTy, gbtType));
-    init.push_back(ConstantPointerRef::get(loadVar));
-    init.push_back(ConstantUInt::get(Type::UIntTy, 0));
-    init.push_back(ConstantUInt::get(Type::UIntTy, paramSize));
-    init.push_back(Constant::getNullValue(PointerType::get(Type::VoidTy)));
-    init.push_back(ConstantExpr::getCast(ConstantPointerRef::get(instFunc),
-                                         PointerType::get(Type::VoidTy)));
-    return ConstantStruct::get(st, init);
+    module->addTypeName("PrimInfo", sm_structType);
 }
 
-// For non-start sites
 static ConstantStruct* makeConstStruct(StructType* st,
+                                       unsigned siteID,
                                        unsigned gbtType,
-                                       GlobalVariable* loadVar,
-                                       unsigned startLinkIdx,
-                                       unsigned paramSize,
-                                       GlobalVariable* metricVar,
-                                       Function* instFunc)
+                                       GlobalVariable* loadVar)
 {
     std::vector<Constant*> init;
-    init.push_back(ConstantUInt::get(Type::UIntTy, gbtType));
-    init.push_back(ConstantPointerRef::get(loadVar));
-    init.push_back(ConstantUInt::get(Type::UIntTy, startLinkIdx));
-    init.push_back(ConstantUInt::get(Type::UIntTy, paramSize));
-    init.push_back(ConstantExpr::getCast(ConstantPointerRef::get(metricVar),
-                                         PointerType::get(Type::VoidTy)));
-    init.push_back(ConstantExpr::getCast(ConstantPointerRef::get(instFunc),
-                                         PointerType::get(Type::VoidTy)));
+    init.push_back(ConstantUInt::get(Type::UIntTy, siteID));  // siteID
+    init.push_back(ConstantUInt::get(Type::UIntTy, gbtType)); // gbtType
+    init.push_back(ConstantPointerRef::get(loadVar));         // loadVar
     return ConstantStruct::get(st, init);
 }
 
-void PrimInfo::buildStructInstances(std::vector<Constant*>& gbtElems,
-                                    std::map<PrimInfo*, unsigned>& gbtIdxMap,
-                                    const TargetData* targetData)
+void PrimInfo::buildStructInstances(std::vector<Constant*>& gbtElems)
 {
     StructType* st = getStructType();
 
     switch(m_type) {
         case REGION: {
-            
-            std::map<PrimInfo*, unsigned>::iterator i = gbtIdxMap.find(this);
-            assert(i != gbtIdxMap.end() && "No entry in map for this PrimInfo instance");
-            
+
             // Create the struct for the start of the region
-            assert(m_startFunc->asize() == 1 && "Unexpected # args for start func");
-            unsigned psize = targetData->getTypeSize(m_startFunc->afront().getType());
-            gbtElems.push_back(makeConstStruct(st, GBT_INTERVAL_START, m_globVol,
-                                               psize, m_startFunc));
+            assert(m_startFunc->asize() == 2 && "Unexpected # args for start func");
+            gbtElems.push_back(makeConstStruct(st, m_siteID, GBT_INTERVAL_START, m_globVol));
 
             // Create the struct for the end of the region
             assert(m_endFunc->asize() == 2 && "Unexpected # args for end func");
-            psize = targetData->getTypeSize(m_endFunc->afront().getType());
-            gbtElems.push_back(makeConstStruct(st, GBT_INTERVAL_END, m_globVolEnd,
-                                               i->second, psize, m_metricVar, m_endFunc));
+            gbtElems.push_back(makeConstStruct(st, m_siteID, GBT_INTERVAL_END, m_globVolEnd));
 
             break;
         }


Index: llvm/lib/Reoptimizer/Inst/lib/Phase1/PrimInfo.h
diff -u llvm/lib/Reoptimizer/Inst/lib/Phase1/PrimInfo.h:1.10 llvm/lib/Reoptimizer/Inst/lib/Phase1/PrimInfo.h:1.10.4.1
--- llvm/lib/Reoptimizer/Inst/lib/Phase1/PrimInfo.h:1.10	Mon May 12 21:00:24 2003
+++ llvm/lib/Reoptimizer/Inst/lib/Phase1/PrimInfo.h	Thu Jun 26 16:35:20 2003
@@ -13,8 +13,6 @@
 
 #include <map>
 
-class TargetData;
-
 namespace pp
 {
 
@@ -26,7 +24,6 @@
     PrimInfo(Module* m):
         m_globVol(0),
         m_globVolEnd(0),
-        m_metricVar(0),
         m_module(m),
         m_startFunc(0),
         m_endFunc(0)
@@ -35,16 +32,16 @@
 
     // Phase 1 construction -- point/counter computation site
     PrimInfo(PrimType type,
+             unsigned siteID,
              GlobalVariable* globVol,
-             GlobalVariable* metricVar,
              Module* m,
              Function* calledFunc = 0);    
 
     // Phase 1 construction -- region computation site
     PrimInfo(PrimType type,
+             unsigned siteID,
              GlobalVariable* startGlob,
              GlobalVariable* endGlob,
-             GlobalVariable* metricVar,
              Module* m,
              Function* startFunc,
              Function* endFunc);
@@ -57,15 +54,11 @@
     const Function*       getStartFunc() const  { return m_startFunc;  }
     Function*             getEndFunc()          { return m_endFunc;    }
     const Function*       getEndFunc() const    { return m_endFunc;    }
-    GlobalVariable*       getMetricVar()        { return m_metricVar;  }
-    const GlobalVariable* getMetricVar() const  { return m_metricVar;  }
     PrimType              getType() const       { return m_type;       }
 
-    void buildStructInstances(std::vector<Constant*>& gbtElems,
-                              std::map<PrimInfo*, unsigned>& gbtIdxMap,
-                              const TargetData* targetData);
+    void buildStructInstances(std::vector<Constant*>& gbtElems);
 
-    static void           buildStructType(Module* module);
+    static void           buildStructTypes(Module* module);
     static StructType*    getStructType() { return sm_structType; }
     
   private:
@@ -74,9 +67,9 @@
     // supplied configuration parameters.
 
     PrimType           m_type;
+    unsigned           m_siteID;
     GlobalVariable*    m_globVol;
     GlobalVariable*    m_globVolEnd;
-    GlobalVariable*    m_metricVar;
     Module*            m_module;
     Function*          m_startFunc;
     Function*          m_endFunc;




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:40:50 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:40:50 2003
Subject: [llvm-commits] CVS: llvm/www/www-index.html
Message-ID: <200306262137.h5QLbVs00326@choi.cs.uiuc.edu>


Changes in directory llvm/www:

www-index.html updated: 1.39 -> 1.39.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/www/www-index.html
diff -u llvm/www/www-index.html:1.39 llvm/www/www-index.html:1.39.2.1
--- llvm/www/www-index.html:1.39	Fri Jun  6 13:29:01 2003
+++ llvm/www/www-index.html	Thu Jun 26 16:36:20 2003
@@ -43,42 +43,38 @@
                 <td>
                 Low Level Virtual Machine (LLVM) is:<ol>
                     <li><a href="pubs/2002-08-09-LLVMCompilationStrategy.html">
-		        A compilation strategy</a> - Fundamentally,
-                        LLVM is a compilation strategy designed
-                        to enable effective program optimization
-                        across the entire lifetime of a program.
-                        LLVM supports effective optimization at
-                        compile time, link-time (particularly
-                        interprocedural), run-time and offline
-                        (i.e., after software is installed),
-                        while maintaining compatibility with
-                        existing build scripts.<p></li>
+		        A compilation strategy</a> - Fundamentally, LLVM is a
+		        compilation strategy designed to enable effective
+		        program optimization across the entire lifetime of a
+		        program.  LLVM supports effective optimization at
+		        compile time, link-time (particularly interprocedural),
+		        run-time and offline (i.e., after software is
+		        installed), while maintaining compatibility with
+		        existing build scripts.<p></li>
+
                     <li><a href="docs/LangRef.html"><u>A virtual
-                        instruction set</u></a> - LLVM is a
-                        low-level object code representation that
-                        uses simple RISC-like instructions, but
-                        provides high-level type and dataflow
-                        (SSA) information about operands. This
-                        combination enables sophisticated
-                        transformations on object code, while
-                        remaining light-weight enough to be
-                        attached to the executable. This
-                        combination is key to allowing link-time,
-                        run-time, and offline transformations.<p></li>
+                        instruction set</u></a> - LLVM is a low-level object
+                        code representation that uses simple RISC-like
+                        instructions, but provides high-level type and dataflow
+                        (SSA) information about operands. This combination
+                        enables sophisticated transformations on object code,
+                        while remaining light-weight enough to be attached to
+                        the executable. This combination is key to allowing
+                        link-time, run-time, and offline
+                        transformations.<p></li>
+
                     <li><a href="pubs/2002-12-LattnerMSThesis.html">A compiler 
-                        infrastructure</a> - LLVM
-                        is also a collection of source code that
-                        implements the language and compilation
-                        strategy. The primary components of the
-                        LLVM infrastructure are a GCC-based C
-                        front-end, a link-time optimization
-                        framework with a growing set of global
-                        and interprocedural analyses and
-                        transformations, and back ends for the
-                        SPARC v9 architecture and for C. See
-                        &quot;<a href="#subprojects">Current
-                        Projects</a>&quot; for information about
-                        other components under development.<p></li>
+                        infrastructure</a> - LLVM is also a collection of source
+                        code that implements the language and compilation
+                        strategy. The primary components of the LLVM
+                        infrastructure are a GCC-based C front-end, a link-time
+                        optimization framework with a growing set of global and
+                        interprocedural analyses and transformations, static
+                        back-ends for the SPARC v9 architecture and for C, and a
+                        Just-In-Time compiler for X86 processors. See &quot;<a
+                        href="#subprojects">Current Projects</a>&quot; for
+                        information about other components under
+                        development.<p></li>
                 </ol>
                 <p>LLVM is a product of the <a
                 href="http://www-faculty.cs.uiuc.edu/~vadve/lcoproject.html">Lifelong




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:40:57 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:40:57 2003
Subject: [llvm-commits] CVS: llvm/test/Makefile.tests
Message-ID: <200306262137.h5QLbfR00490@choi.cs.uiuc.edu>


Changes in directory llvm/test:

Makefile.tests updated: 1.65 -> 1.65.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/test/Makefile.tests
diff -u llvm/test/Makefile.tests:1.65 llvm/test/Makefile.tests:1.65.2.1
--- llvm/test/Makefile.tests:1.65	Wed Jun 18 14:39:06 2003
+++ llvm/test/Makefile.tests	Thu Jun 26 16:35:31 2003
@@ -43,6 +43,7 @@
 LAS      = $(TOOLS)/as
 LGCCAS   = $(TOOLS)/gccas
 LGCCLD   = $(TOOLS)/gccld -L$(LLVMGCCDIR)/lib/gcc-lib/$(LLVMGCCARCH) -L$(LLVMGCCDIR)/lib
+LGCCLDPROG = $(TOOLS)/gccld
 LDIS     = $(TOOLS)/dis 
 LOPT     = $(TOOLS)/opt
 LLINK    = $(TOOLS)/link




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:41:04 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:41:04 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/InstCombine/2003-06-22-ConstantExprCrash.ll load.ll cast.ll getelementptr.ll mul.ll
Message-ID: <200306262137.h5QLbVI00340@choi.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/InstCombine:

2003-06-22-ConstantExprCrash.ll added (r1.1.2.1)
load.ll added (r1.1.2.1)
cast.ll updated: 1.6 -> 1.6.2.1
getelementptr.ll updated: 1.8 -> 1.8.2.1
mul.ll updated: 1.5 -> 1.5.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/InstCombine/2003-06-22-ConstantExprCrash.ll
diff -c /dev/null llvm/test/Regression/Transforms/InstCombine/2003-06-22-ConstantExprCrash.ll:1.1
*** /dev/null	Thu Jun 26 16:37:31 2003
--- llvm/test/Regression/Transforms/InstCombine/2003-06-22-ConstantExprCrash.ll	Sun Jun 22 15:45:39 2003
***************
*** 0 ****
--- 1,12 ----
+ ; This is a bug in the VMcode library, not instcombine, it's just convenient 
+ ; to expose it here.
+ 
+ ; RUN: as < %s | opt -instcombine -disable-output
+ 
+ %A = global int 1
+ %B = global int 2
+ 
+ bool %test() {
+ 	%C = setlt int* getelementptr (int* %A, long 1), getelementptr (int* %B, long 2)    ; Will get promoted to constantexpr
+ 	ret bool %C
+ }


Index: llvm/test/Regression/Transforms/InstCombine/load.ll
diff -c /dev/null llvm/test/Regression/Transforms/InstCombine/load.ll:1.1
*** /dev/null	Thu Jun 26 16:37:31 2003
--- llvm/test/Regression/Transforms/InstCombine/load.ll	Thu Jun 26 00:05:51 2003
***************
*** 0 ****
--- 1,29 ----
+ ; This test makes sure that these instructions are properly eliminated.
+ ;
+ 
+ ; RUN: if as < %s | opt -instcombine | dis | grep load
+ ; RUN: then exit 1
+ ; RUN: else exit 0
+ ; RUN: fi
+ 
+ %X = constant int 42
+ %Y = constant [2 x { int, float }] [ { int, float } { int 12, float 1.0 }, 
+                                      { int, float } { int 37, float 1.2312 } ]
+ int %test1() {
+ 	%B = load int* %X
+ 	ret int %B
+ }
+ 
+ float %test2() {
+ 	%A = getelementptr [2 x { int, float}]* %Y, long 0, long 1, ubyte 1
+ 	%B = load float* %A
+ 	ret float %B
+ }
+ 
+ int %test3() {
+ 	%A = getelementptr [2 x { int, float}]* %Y, long 0, long 0, ubyte 0
+ 	%B = load int* %A
+ 	ret int %B
+ }
+ 
+ 


Index: llvm/test/Regression/Transforms/InstCombine/cast.ll
diff -u llvm/test/Regression/Transforms/InstCombine/cast.ll:1.6 llvm/test/Regression/Transforms/InstCombine/cast.ll:1.6.2.1
--- llvm/test/Regression/Transforms/InstCombine/cast.ll:1.6	Mon Sep 23 18:39:17 2002
+++ llvm/test/Regression/Transforms/InstCombine/cast.ll	Thu Jun 26 16:35:51 2003
@@ -68,3 +68,22 @@
 	%c2 = cast uint %c1 to short
 	ret short %c2
 }
+
+bool %test11(ubyte %A, ubyte %B) {
+        %C = sub ubyte %A, %B
+        %D = cast ubyte %C to bool    ; == setne A, B
+        ret bool %D
+}
+
+bool %test12(ubyte %A) {
+        %B = add ubyte %A, 255
+        %C = cast ubyte %B to bool    ; === A != 1
+        ret bool %C
+}
+
+bool %test13(ubyte %A, ubyte %b) {
+        %B = add ubyte %A, %b
+        %C = cast ubyte %B to bool    ; === A != -b
+        ret bool %C
+}
+


Index: llvm/test/Regression/Transforms/InstCombine/getelementptr.ll
diff -u llvm/test/Regression/Transforms/InstCombine/getelementptr.ll:1.8 llvm/test/Regression/Transforms/InstCombine/getelementptr.ll:1.8.2.1
--- llvm/test/Regression/Transforms/InstCombine/getelementptr.ll:1.8	Wed Mar  5 16:32:48 2003
+++ llvm/test/Regression/Transforms/InstCombine/getelementptr.ll	Thu Jun 26 16:35:51 2003
@@ -48,3 +48,9 @@
 	%B = getelementptr int* %A, long %D
 	ret int* %B
 }
+
+sbyte* %foo8([10 x int]* %X) {
+	%A = getelementptr [10 x int]* %X, long 0, long 0   ;; Fold into the cast.
+	%B = cast int* %A to sbyte*
+	ret sbyte * %B
+}


Index: llvm/test/Regression/Transforms/InstCombine/mul.ll
diff -u llvm/test/Regression/Transforms/InstCombine/mul.ll:1.5 llvm/test/Regression/Transforms/InstCombine/mul.ll:1.5.2.1
--- llvm/test/Regression/Transforms/InstCombine/mul.ll:1.5	Mon Mar 10 16:43:56 2003
+++ llvm/test/Regression/Transforms/InstCombine/mul.ll	Thu Jun 26 16:35:51 2003
@@ -33,8 +33,18 @@
 	ret int %B
 }
 
-int %test6(ubyte %A) {
+ubyte %test6(ubyte %A) {
 	%B = mul ubyte %A, 8
 	%C = mul ubyte %B, 13
 	ret ubyte %C
+}
+
+int %test7(int %i) {
+        %tmp = mul int %i, -1   ; %tmp = sub 0, %i
+        ret int %tmp
+}
+
+ulong %test8(ulong %i) {
+	%j = mul ulong %i, 18446744073709551615 ; tmp = sub 0, %i
+	ret ulong %j
 }




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:41:11 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:41:11 2003
Subject: [llvm-commits] CVS: llvm/lib/Reoptimizer/Inst/rtl/pprtl.cpp pprtl.h
Message-ID: <200306262137.h5QLbWF00348@choi.cs.uiuc.edu>


Changes in directory llvm/lib/Reoptimizer/Inst/rtl:

pprtl.cpp updated: 1.2 -> 1.2.4.1
pprtl.h updated: 1.3 -> 1.3.4.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/lib/Reoptimizer/Inst/rtl/pprtl.cpp
diff -u llvm/lib/Reoptimizer/Inst/rtl/pprtl.cpp:1.2 llvm/lib/Reoptimizer/Inst/rtl/pprtl.cpp:1.2.4.1
--- llvm/lib/Reoptimizer/Inst/rtl/pprtl.cpp:1.2	Thu May 22 22:26:13 2003
+++ llvm/lib/Reoptimizer/Inst/rtl/pprtl.cpp	Thu Jun 26 16:35:21 2003
@@ -57,7 +57,7 @@
     printf("pp_L1_cache_miss_start invoked: retVal addr is 0x%lx\n", (unsigned long) retVal);
     start_papi_sample(events, 1);
     *retVal = PAPI_L1_ICM;
-    printf("pp_L1_cache_miss_start returning!\n");
+    printf("pp_L1_cache_miss_start returning value %x!\n", *retVal);
 }
 
 void pp_L1_cache_miss_end(int* retVal, int* start)


Index: llvm/lib/Reoptimizer/Inst/rtl/pprtl.h
diff -u llvm/lib/Reoptimizer/Inst/rtl/pprtl.h:1.3 llvm/lib/Reoptimizer/Inst/rtl/pprtl.h:1.3.4.1
--- llvm/lib/Reoptimizer/Inst/rtl/pprtl.h:1.3	Thu May 22 22:26:13 2003
+++ llvm/lib/Reoptimizer/Inst/rtl/pprtl.h	Thu Jun 26 16:35:22 2003
@@ -16,11 +16,22 @@
     void     pp_L1_cache_miss_end(int* retVal, int* start);
     unsigned pp_counterPrim(void);
 
-    // significant functions
+    // significant functions and rtl functions
+
+    void phase2();
+    void registerIntervalInst(unsigned siteID,
+                              void* startFunc,
+                              void* endFunc,
+                              unsigned paramSize,
+                              void* retVal);
 
     void     pp_regionPair(void*, void*);
     void     pp_metricSpec(void*);
 
+    // Interval declaration sigfuns: the placeholder variable is to connect the
+    // two sigfuns via uses.
+    void     pp_declareIntervalStart(double* placeHolder, unsigned intervalID);
+    void     pp_declareIntervalEnd(unsigned intervalID, double* placeHolder);
 
 #ifdef _BUILDING_RUNTIME_LIBRARY_
 }




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:41:18 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:41:18 2003
Subject: [llvm-commits] CVS: llvm/tools/gccld/gccld.cpp
Message-ID: <200306262137.h5QLbf900484@choi.cs.uiuc.edu>


Changes in directory llvm/tools/gccld:

gccld.cpp updated: 1.33 -> 1.33.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/tools/gccld/gccld.cpp
diff -u llvm/tools/gccld/gccld.cpp:1.33 llvm/tools/gccld/gccld.cpp:1.33.2.1
--- llvm/tools/gccld/gccld.cpp:1.33	Thu May 29 10:13:15 2003
+++ llvm/tools/gccld/gccld.cpp	Thu Jun 26 16:36:00 2003
@@ -396,6 +396,19 @@
     Passes.add(createInternalizePass());
   }
 
+  // Remove unused arguments from functions...
+  //
+  Passes.add(createDeadArgEliminationPass());
+
+  // The FuncResolve pass may leave cruft around if functions were prototyped
+  // differently than they were defined.  Remove this cruft.
+  //
+  Passes.add(createInstructionCombiningPass());
+
+  // Delete basic blocks, which optimization passes may have killed...
+  //
+  Passes.add(createCFGSimplificationPass());
+
   // Now that we have optimized the program, discard unreachable functions...
   //
   Passes.add(createGlobalDCEPass());




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:41:25 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:41:25 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/LevelRaise/2003-06-25-ExprAnalysis.ll
Message-ID: <200306262137.h5QLbWK00354@choi.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/LevelRaise:

2003-06-25-ExprAnalysis.ll added (r1.1.2.1)

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/LevelRaise/2003-06-25-ExprAnalysis.ll
diff -c /dev/null llvm/test/Regression/Transforms/LevelRaise/2003-06-25-ExprAnalysis.ll:1.1
*** /dev/null	Thu Jun 26 16:37:32 2003
--- llvm/test/Regression/Transforms/LevelRaise/2003-06-25-ExprAnalysis.ll	Wed Jun 25 12:08:34 2003
***************
*** 0 ****
--- 1,27 ----
+ ; The expr analysis routines were being too aggressive across cast instructions!
+ 
+ ; RUN: if as < %s | opt -raise | dis | grep 4294967295
+ ; RUN: then exit 1
+ ; RUN: else exit 0
+ ; RUN: fi
+ 
+ target endian = big
+ target pointersize = 64
+ 	%struct..istack_struct = type { %struct..istack_struct*, %struct..istk_entry*, uint }
+ 	%struct..istk_entry = type { double, int, int, double, double, sbyte* }
+ 
+ implementation   ; Functions:
+ bool %Intersection(%struct..istack_struct* %tmp.0, uint %tmp.12) {              ; No predecessors!
+         %tmp.8 = getelementptr %struct..istack_struct* %tmp.0, long 0, ubyte 1          ; <%struct..istk_entry**> [#uses=1]
+         %tmp.9 = load %struct..istk_entry** %tmp.8              ; <%struct..istk_entry*> [#uses=1]
+         %dec = sub uint %tmp.12, 1              ; <uint> [#uses=1]
+         %tmp.13 = cast uint %dec to ulong               ; <ulong> [#uses=1]
+         %tmp.14 = mul ulong %tmp.13, 40         ; <ulong> [#uses=1]
+         %tmp.16 = cast %struct..istk_entry* %tmp.9 to long              ; <long> [#uses=1]
+         %tmp.17 = cast ulong %tmp.14 to long            ; <long> [#uses=1]
+         %tmp.18 = add long %tmp.16, %tmp.17             ; <long> [#uses=1]
+         %tmp.19 = cast long %tmp.18 to %struct..istk_entry*             ; <%struct..istk_entry*> [#uses=1]
+         %tmp.21 = setne %struct..istk_entry* %tmp.19, null              ; <bool> [#uses=1]
+         ret bool %tmp.21
+ }
+ 




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:41:32 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:41:32 2003
Subject: [llvm-commits] CVS: llvm/lib/Support/Timer.cpp
Message-ID: <200306262137.h5QLbX800360@choi.cs.uiuc.edu>


Changes in directory llvm/lib/Support:

Timer.cpp updated: 1.18.2.1 -> 1.18.2.2

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/lib/Support/Timer.cpp
diff -u llvm/lib/Support/Timer.cpp:1.18.2.1 llvm/lib/Support/Timer.cpp:1.18.2.2
--- llvm/lib/Support/Timer.cpp:1.18.2.1	Mon Jun 23 14:07:06 2003
+++ llvm/lib/Support/Timer.cpp	Thu Jun 26 16:35:23 2003
@@ -12,6 +12,7 @@
 #include "Config/sys/resource.h"
 #include "Config/sys/time.h"
 #include "Config/unistd.h"
+#include "Config/malloc.h"
 #include "Config/stdio.h"
 #include <iostream>
 #include <algorithm>
@@ -82,12 +83,9 @@
   if (TrackSpace) {
     struct mallinfo MI = mallinfo();
     return MI.uordblks/*+MI.hblkhd*/;
-  } else {
-    return 0;
   }
-#else
-    return 0;
-#endif
+#endif // __FreeBSD__
+  return 0;
 }
 
 struct TimeRecord {




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:41:39 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:41:39 2003
Subject: [llvm-commits] CVS: llvm/lib/VMCore/AsmWriter.cpp ConstantHandling.cpp Constants.cpp PassManagerT.h Type.cpp Value.cpp
Message-ID: <200306262137.h5QLbed00478@choi.cs.uiuc.edu>


Changes in directory llvm/lib/VMCore:

AsmWriter.cpp updated: 1.88 -> 1.88.2.1
ConstantHandling.cpp updated: 1.31 -> 1.31.2.1
Constants.cpp updated: 1.44 -> 1.44.2.1
PassManagerT.h updated: 1.38.2.1 -> 1.38.2.2
Type.cpp updated: 1.47 -> 1.47.2.1
Value.cpp updated: 1.31 -> 1.31.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/lib/VMCore/AsmWriter.cpp
diff -u llvm/lib/VMCore/AsmWriter.cpp:1.88 llvm/lib/VMCore/AsmWriter.cpp:1.88.2.1
--- llvm/lib/VMCore/AsmWriter.cpp:1.88	Sat May 31 22:45:51 2003
+++ llvm/lib/VMCore/AsmWriter.cpp	Thu Jun 26 16:35:29 2003
@@ -234,7 +234,7 @@
     //
     if ((StrVal[0] >= '0' && StrVal[0] <= '9') ||
         ((StrVal[0] == '-' || StrVal[0] == '+') &&
-         (StrVal[0] >= '0' && StrVal[0] <= '9')))
+         (StrVal[1] >= '0' && StrVal[1] <= '9')))
       // Reparse stringized version!
       if (atof(StrVal.c_str()) == CFP->getValue()) {
         Out << StrVal; return;


Index: llvm/lib/VMCore/ConstantHandling.cpp
diff -u llvm/lib/VMCore/ConstantHandling.cpp:1.31 llvm/lib/VMCore/ConstantHandling.cpp:1.31.2.1
--- llvm/lib/VMCore/ConstantHandling.cpp:1.31	Tue May 27 14:16:07 2003
+++ llvm/lib/VMCore/ConstantHandling.cpp	Thu Jun 26 16:35:29 2003
@@ -141,12 +141,30 @@
 
   // FIXME: Implement folding of GEP constant exprs the same as instcombine does
 
-  // Implement folding of:
-  //    int* getelementptr ([2 x int]* cast ([3 x int]* %X to [2 x int]*),
-  //                        long 0, long 0)
-  // To: int* getelementptr ([3 x int]* %X, long 0, long 0)
-  //
-  if (const ConstantExpr *CE = dyn_cast<ConstantExpr>(C))
+  if (const ConstantExpr *CE = dyn_cast<ConstantExpr>(C)) {
+    // Implement folding of:
+    //    void ()** getelementptr (%struct..TorRec* getelementptr
+    //              ([N x %struct..TorRec]* %llvm.global_dtors, long 0, long 0),
+    //                 long 0, ubyte 1)
+    // Into:
+    //    %struct..TorRec* getelementptr ([N x %struct..TorRec]*
+    //                      %llvm.global_dtors, long 0, long 0, ubyte 1)
+    //
+    if (CE->getOpcode() == Instruction::GetElementPtr)
+      if (IdxList[0] == Constant::getNullValue(Type::LongTy)) {
+        std::vector<Constant*> NewIndices;
+        NewIndices.reserve(IdxList.size() + CE->getNumOperands());
+        for (unsigned i = 1, e = CE->getNumOperands(); i != e; ++i)
+          NewIndices.push_back(cast<Constant>(CE->getOperand(i)));
+        NewIndices.insert(NewIndices.end(), IdxList.begin()+1, IdxList.end());
+        return ConstantExpr::getGetElementPtr(CE->getOperand(0), NewIndices);
+      }
+
+    // Implement folding of:
+    //    int* getelementptr ([2 x int]* cast ([3 x int]* %X to [2 x int]*),
+    //                        long 0, long 0)
+    // To: int* getelementptr ([3 x int]* %X, long 0, long 0)
+    //
     if (CE->getOpcode() == Instruction::Cast && IdxList.size() > 1 &&
         IdxList[0]->isNullValue())
       if (const PointerType *SPT = 
@@ -157,6 +175,7 @@
             if (CAT->getElementType() == SAT->getElementType())
               return ConstantExpr::getGetElementPtr(
                       (Constant*)CE->getOperand(0), IdxList);
+  }
   return 0;
 }
 


Index: llvm/lib/VMCore/Constants.cpp
diff -u llvm/lib/VMCore/Constants.cpp:1.44 llvm/lib/VMCore/Constants.cpp:1.44.2.1
--- llvm/lib/VMCore/Constants.cpp:1.44	Mon Jun  2 12:42:47 2003
+++ llvm/lib/VMCore/Constants.cpp	Thu Jun 26 16:35:29 2003
@@ -225,7 +225,9 @@
          "Invalid initializer vector for constant structure");
   Operands.reserve(V.size());
   for (unsigned i = 0, e = V.size(); i != e; ++i) {
-    assert(V[i]->getType() == ETypes[i] &&
+    assert((V[i]->getType() == ETypes[i] ||
+            (ETypes[i]->isAbstract() &&
+             ETypes[i]->getPrimitiveID()==V[i]->getType()->getPrimitiveID())) &&
            "Initializer for struct element doesn't match struct element type!");
     Operands.push_back(Use(V[i], this));
   }
@@ -241,8 +243,14 @@
   Operands.push_back(Use(C, this));
 }
 
+static bool isSetCC(unsigned Opcode) {
+  return Opcode == Instruction::SetEQ || Opcode == Instruction::SetNE ||
+         Opcode == Instruction::SetLT || Opcode == Instruction::SetGT ||
+         Opcode == Instruction::SetLE || Opcode == Instruction::SetGE;
+}
+
 ConstantExpr::ConstantExpr(unsigned Opcode, Constant *C1, Constant *C2)
-  : Constant(C1->getType()), iType(Opcode) {
+  : Constant(isSetCC(Opcode) ? Type::BoolTy : C1->getType()), iType(Opcode) {
   Operands.push_back(Use(C1, this));
   Operands.push_back(Use(C2, this));
 }
@@ -563,9 +571,11 @@
   std::vector<Constant*> C;
   for (unsigned i = 0, e = getNumOperands(); i != e; ++i)
     C.push_back(cast<Constant>(getOperand(i)));
-  replaceAllUsesWith(ConstantArray::get(cast<ArrayType>(NewTy),
-                                        C));
-  destroyConstant();    // This constant is now dead, destroy it.
+  Constant *New = ConstantArray::get(cast<ArrayType>(NewTy), C);
+  if (New != this) {
+    replaceAllUsesWith(New);
+    destroyConstant();    // This constant is now dead, destroy it.
+  }
 }
 
 
@@ -633,9 +643,11 @@
   std::vector<Constant*> C;
   for (unsigned i = 0, e = getNumOperands(); i != e; ++i)
     C.push_back(cast<Constant>(getOperand(i)));
-  replaceAllUsesWith(ConstantStruct::get(cast<StructType>(NewTy),
-                                         C));
-  destroyConstant();    // This constant is now dead, destroy it.
+  Constant *New = ConstantStruct::get(cast<StructType>(NewTy), C);
+  if (New != this) {
+    replaceAllUsesWith(New);
+    destroyConstant();    // This constant is now dead, destroy it.
+  }
 }
 
 
@@ -672,10 +684,13 @@
   if (OldTy == NewTy) return;
 
   // Make everyone now use a constant of the new type...
-  replaceAllUsesWith(ConstantPointerNull::get(cast<PointerType>(NewTy)));
+  Constant *New = ConstantPointerNull::get(cast<PointerType>(NewTy));
+  if (New != this) {
+    replaceAllUsesWith(New);
     
-  // This constant is now dead, destroy it.
-  destroyConstant();
+    // This constant is now dead, destroy it.
+    destroyConstant();
+  }
 }
 
 
@@ -806,13 +821,14 @@
   // ::get methods intuit the type of the result based on the types of the
   // operands.  The operand types may not have had their types resolved yet.
   //
+  Constant *New;
   if (getOpcode() == Instruction::Cast) {
-    replaceAllUsesWith(getCast(getOperand(0), NewTy));
+    New = getCast(getOperand(0), NewTy);
   } else if (getOpcode() >= Instruction::BinaryOpsBegin &&
              getOpcode() < Instruction::BinaryOpsEnd) {
-    replaceAllUsesWith(get(getOpcode(), getOperand(0), getOperand(0)));
+    New = get(getOpcode(), getOperand(0), getOperand(0));
   } else if (getOpcode() == Instruction::Shl || getOpcode() ==Instruction::Shr){
-    replaceAllUsesWith(getShift(getOpcode(), getOperand(0), getOperand(0)));
+    New = getShift(getOpcode(), getOperand(0), getOperand(0));
   } else {
     assert(getOpcode() == Instruction::GetElementPtr);
 
@@ -820,10 +836,12 @@
     std::vector<Constant*> C;
     for (unsigned i = 1, e = getNumOperands(); i != e; ++i)
       C.push_back(cast<Constant>(getOperand(i)));
-    replaceAllUsesWith(ConstantExpr::getGetElementPtr(getOperand(0),
-                                                      C));
+    New = ConstantExpr::getGetElementPtr(getOperand(0), C);
+  }
+  if (New != this) {
+    replaceAllUsesWith(New);
+    destroyConstant();    // This constant is now dead, destroy it.
   }
-  destroyConstant();    // This constant is now dead, destroy it.
 }
 
 


Index: llvm/lib/VMCore/PassManagerT.h
diff -u llvm/lib/VMCore/PassManagerT.h:1.38.2.1 llvm/lib/VMCore/PassManagerT.h:1.38.2.2
--- llvm/lib/VMCore/PassManagerT.h:1.38.2.1	Mon Jun 23 14:06:45 2003
+++ llvm/lib/VMCore/PassManagerT.h	Thu Jun 26 16:35:29 2003
@@ -130,12 +130,6 @@
   typedef typename Traits::BatcherClass BatcherClass;
   typedef typename Traits::ParentClass   ParentClass;
 
-  //
-  // JTC:
-  //	GCC 3.3 does not permit friend declarations without the class keyword.
-  //	Hence, I changed the following two lines to allow the code to compiler
-  //	under the new compiler.
-  //
   friend class PassManagerTraits<UnitType>::PassClass;
   friend class PassManagerTraits<UnitType>::SubPassClass;  
   friend class Traits;


Index: llvm/lib/VMCore/Type.cpp
diff -u llvm/lib/VMCore/Type.cpp:1.47 llvm/lib/VMCore/Type.cpp:1.47.2.1
--- llvm/lib/VMCore/Type.cpp:1.47	Thu May 22 16:31:52 2003
+++ llvm/lib/VMCore/Type.cpp	Thu Jun 26 16:35:29 2003
@@ -216,11 +216,11 @@
 FunctionType::FunctionType(const Type *Result,
                            const std::vector<const Type*> &Params, 
                            bool IsVarArgs) : DerivedType(FunctionTyID), 
-    ResultType(PATypeHandle<Type>(Result, this)),
+    ResultType(PATypeHandle(Result, this)),
     isVarArgs(IsVarArgs) {
   ParamTys.reserve(Params.size());
   for (unsigned i = 0; i < Params.size(); ++i)
-    ParamTys.push_back(PATypeHandle<Type>(Params[i], this));
+    ParamTys.push_back(PATypeHandle(Params[i], this));
 
   setDerivedTypeProperties();
 }
@@ -230,7 +230,7 @@
   ETypes.reserve(Types.size());
   for (unsigned i = 0; i < Types.size(); ++i) {
     assert(Types[i] != Type::VoidTy && "Void type in method prototype!!");
-    ETypes.push_back(PATypeHandle<Type>(Types[i], this));
+    ETypes.push_back(PATypeHandle(Types[i], this));
   }
   setDerivedTypeProperties();
 }
@@ -427,20 +427,20 @@
 //
 template<class ValType, class TypeClass>
 class TypeMap : public AbstractTypeUser {
-  typedef std::map<ValType, PATypeHandle<TypeClass> > MapTy;
+  typedef std::map<ValType, PATypeHandle> MapTy;
   MapTy Map;
 public:
   ~TypeMap() { print("ON EXIT"); }
 
   inline TypeClass *get(const ValType &V) {
-    typename std::map<ValType, PATypeHandle<TypeClass> >::iterator I
+    typename std::map<ValType, PATypeHandle>::iterator I
       = Map.find(V);
     // TODO: FIXME: When Types are not CONST.
     return (I != Map.end()) ? (TypeClass*)I->second.get() : 0;
   }
 
   inline void add(const ValType &V, TypeClass *T) {
-    Map.insert(std::make_pair(V, PATypeHandle<TypeClass>(T, this)));
+    Map.insert(std::make_pair(V, PATypeHandle(T, this)));
     print("add");
   }
 
@@ -521,7 +521,7 @@
 
     TypeMap<ValType, TypeClass> &Table = MyTable;     // Copy MyTable reference
     ValType Tmp(*(ValType*)this);                     // Copy this.
-    PATypeHandle<TypeClass> OldType(Table.get(*(ValType*)this), this);
+    PATypeHandle OldType(Table.get(*(ValType*)this), this);
     Table.remove(*(ValType*)this);                    // Destroy's this!
 
     // Refine temporary to new state...
@@ -546,8 +546,8 @@
 // FunctionValType - Define a class to hold the key that goes into the TypeMap
 //
 class FunctionValType : public ValTypeBase<FunctionValType, FunctionType> {
-  PATypeHandle<Type> RetTy;
-  std::vector<PATypeHandle<Type> > ArgTypes;
+  PATypeHandle RetTy;
+  std::vector<PATypeHandle> ArgTypes;
   bool isVarArg;
 public:
   FunctionValType(const Type *ret, const std::vector<const Type*> &args,
@@ -555,7 +555,7 @@
     : ValTypeBase<FunctionValType, FunctionType>(Tab), RetTy(ret, this),
       isVarArg(IVA) {
     for (unsigned i = 0; i < args.size(); ++i)
-      ArgTypes.push_back(PATypeHandle<Type>(args[i], this));
+      ArgTypes.push_back(PATypeHandle(args[i], this));
   }
 
   // We *MUST* have an explicit copy ctor so that the TypeHandles think that
@@ -566,7 +566,7 @@
       isVarArg(MVT.isVarArg) {
     ArgTypes.reserve(MVT.ArgTypes.size());
     for (unsigned i = 0; i < MVT.ArgTypes.size(); ++i)
-      ArgTypes.push_back(PATypeHandle<Type>(MVT.ArgTypes[i], this));
+      ArgTypes.push_back(PATypeHandle(MVT.ArgTypes[i], this));
   }
 
   // Subclass should override this... to update self as usual
@@ -615,7 +615,7 @@
 // Array Type Factory...
 //
 class ArrayValType : public ValTypeBase<ArrayValType, ArrayType> {
-  PATypeHandle<Type> ValTy;
+  PATypeHandle ValTy;
   unsigned Size;
 public:
   ArrayValType(const Type *val, int sz, TypeMap<ArrayValType, ArrayType> &Tab)
@@ -671,14 +671,14 @@
 // StructValType - Define a class to hold the key that goes into the TypeMap
 //
 class StructValType : public ValTypeBase<StructValType, StructType> {
-  std::vector<PATypeHandle<Type> > ElTypes;
+  std::vector<PATypeHandle> ElTypes;
 public:
   StructValType(const std::vector<const Type*> &args,
 		TypeMap<StructValType, StructType> &Tab)
     : ValTypeBase<StructValType, StructType>(Tab) {
     ElTypes.reserve(args.size());
     for (unsigned i = 0, e = args.size(); i != e; ++i)
-      ElTypes.push_back(PATypeHandle<Type>(args[i], this));
+      ElTypes.push_back(PATypeHandle(args[i], this));
   }
 
   // We *MUST* have an explicit copy ctor so that the TypeHandles think that
@@ -688,7 +688,7 @@
     : ValTypeBase<StructValType, StructType>(SVT){
     ElTypes.reserve(SVT.ElTypes.size());
     for (unsigned i = 0, e = SVT.ElTypes.size(); i != e; ++i)
-      ElTypes.push_back(PATypeHandle<Type>(SVT.ElTypes[i], this));
+      ElTypes.push_back(PATypeHandle(SVT.ElTypes[i], this));
   }
 
   // Subclass should override this... to update self as usual
@@ -731,7 +731,7 @@
 // PointerValType - Define a class to hold the key that goes into the TypeMap
 //
 class PointerValType : public ValTypeBase<PointerValType, PointerType> {
-  PATypeHandle<Type> ValTy;
+  PATypeHandle ValTy;
 public:
   PointerValType(const Type *val, TypeMap<PointerValType, PointerType> &Tab)
     : ValTypeBase<PointerValType, PointerType>(Tab), ValTy(val, this) {}


Index: llvm/lib/VMCore/Value.cpp
diff -u llvm/lib/VMCore/Value.cpp:1.31 llvm/lib/VMCore/Value.cpp:1.31.2.1
--- llvm/lib/VMCore/Value.cpp:1.31	Thu Feb 13 13:46:22 2003
+++ llvm/lib/VMCore/Value.cpp	Thu Jun 26 16:35:29 2003
@@ -40,7 +40,7 @@
                 << **I << "\n";
   }
 #endif
-  assert(Uses.begin() == Uses.end());
+  assert(Uses.begin() == Uses.end() &&"Uses remain when a value is destroyed!");
 
   // There should be no uses of this object anymore, remove it.
   LeakDetector::removeGarbageObject(this);




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:41:46 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:41:46 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Support/CallSite.h
Message-ID: <200306262137.h5QLbXf00369@choi.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Support:

CallSite.h updated: 1.1 -> 1.1.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.
Kept includes of "Config/assert.h" to avoid implicit dependencies on header
files.


---
Diffs of the changes:

Index: llvm/include/llvm/Support/CallSite.h
diff -u llvm/include/llvm/Support/CallSite.h:1.1 llvm/include/llvm/Support/CallSite.h:1.1.2.1
--- llvm/include/llvm/Support/CallSite.h:1.1	Mon Feb 24 14:35:45 2003
+++ llvm/include/llvm/Support/CallSite.h	Thu Jun 26 16:34:53 2003
@@ -16,8 +16,30 @@
 class CallSite {
   Instruction *I;
 public:
+  CallSite() : I(0) {}
   CallSite(CallInst *CI) : I((Instruction*)CI) {}
   CallSite(InvokeInst *II) : I((Instruction*)II) {}
+  CallSite(const CallSite &CS) : I(CS.I) {}
+  CallSite &operator=(const CallSite &CS) { I = CS.I; return *this; }
+
+  /// CallSite::get - This static method is sort of like a constructor.  It will
+  /// create an appropriate call site for a Call or Invoke instruction, but it
+  /// can also create a null initialized CallSite object for something which is
+  /// NOT a call site.
+  ///
+  static CallSite get(Value *V) {
+    if (Instruction *I = dyn_cast<Instruction>(V)) {
+      if (I->getOpcode() == Instruction::Call)
+        return CallSite((CallInst*)I);
+      else if (I->getOpcode() == Instruction::Invoke)
+        return CallSite((InvokeInst*)I);
+    }
+    return CallSite();
+  }
+
+  /// getInstruction - Return the instruction this call site corresponds to
+  ///
+  Instruction *getInstruction() const { return I; }
 
   /// getCalledValue - Return the pointer to function that is being called...
   ///
@@ -30,6 +52,12 @@
     return dyn_cast<Function>(getCalledValue());
   }
 
+  /// setCalledFunction - Set the callee to the specied value...
+  ///
+  void setCalledFunction(Value *V) {
+    I->setOperand(0, V);
+  }
+
   /// arg_iterator - The type of iterator to use when looping over actual
   /// arguments at this call site...
   typedef User::op_iterator arg_iterator;
@@ -43,7 +71,7 @@
     else
       return I->op_begin()+3; // Skip Function, BB, BB
   }
-  arg_iterator arg_end() const { return I->op_begin(); }
+  arg_iterator arg_end() const { return I->op_end(); }
 };
 
 #endif




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:41:54 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:41:54 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/Mem2Reg/2003-06-26-IterativePromote.ll
Message-ID: <200306262137.h5QLbX300368@choi.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/Mem2Reg:

2003-06-26-IterativePromote.ll added (r1.1.2.1)

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/Mem2Reg/2003-06-26-IterativePromote.ll
diff -c /dev/null llvm/test/Regression/Transforms/Mem2Reg/2003-06-26-IterativePromote.ll:1.1
*** /dev/null	Thu Jun 26 16:37:33 2003
--- llvm/test/Regression/Transforms/Mem2Reg/2003-06-26-IterativePromote.ll	Wed Jun 25 09:55:30 2003
***************
*** 0 ****
--- 1,19 ----
+ ; Promoting some values allows promotion of other values.
+ ; RUN: if as < %s | opt -mem2reg | dis | grep alloca
+ ; RUN: then exit 1
+ ; RUN: else exit 0
+ ; RUN: fi
+ 
+ int %test2() {
+         %result = alloca int             ; ty=int*
+         %a = alloca int          ; ty=int*
+         %p = alloca int*                 ; ty=int**
+         store int 0, int* %a
+         store int* %a, int** %p
+         %tmp.0 = load int** %p           ; ty=int*
+         %tmp.1 = load int* %tmp.0                ; ty=int
+         store int %tmp.1, int* %result
+         %tmp.2 = load int* %result               ; ty=int
+         ret int %tmp.2
+ }
+ 




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:42:01 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:42:01 2003
Subject: [llvm-commits] CVS: llvm/tools/gccas/gccas.cpp
Message-ID: <200306262137.h5QLbd900462@choi.cs.uiuc.edu>


Changes in directory llvm/tools/gccas:

gccas.cpp updated: 1.65 -> 1.65.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/tools/gccas/gccas.cpp
diff -u llvm/tools/gccas/gccas.cpp:1.65 llvm/tools/gccas/gccas.cpp:1.65.2.1
--- llvm/tools/gccas/gccas.cpp:1.65	Fri May 30 14:24:06 2003
+++ llvm/tools/gccas/gccas.cpp	Thu Jun 26 16:35:59 2003
@@ -60,18 +60,16 @@
 
 
 void AddConfiguredTransformationPasses(PassManager &PM) {
-  if (Verify) PM.add(createVerifierPass());
-
+  PM.add(createVerifierPass());                  // Verify that input is correct
   addPass(PM, createFunctionResolvingPass());    // Resolve (...) functions
   addPass(PM, createGlobalDCEPass());            // Kill unused uinit g-vars
   addPass(PM, createDeadTypeEliminationPass());  // Eliminate dead types
   addPass(PM, createConstantMergePass());        // Merge dup global constants
-  addPass(PM, createCFGSimplificationPass());    // Merge & remove BBs
-  addPass(PM, createVerifierPass());             // Verify that input is correct
-  addPass(PM, createDeadInstEliminationPass());  // Remove Dead code/vars
   addPass(PM, createRaiseAllocationsPass());     // call %malloc -> malloc inst
   addPass(PM, createInstructionCombiningPass()); // Cleanup code for raise
   addPass(PM, createRaisePointerReferencesPass());// Recover type information
+  addPass(PM, createTailDuplicationPass());      // Simplify cfg by copying code
+  addPass(PM, createCFGSimplificationPass());    // Merge & remove BBs
   addPass(PM, createInstructionCombiningPass()); // Combine silly seq's
   addPass(PM, createScalarReplAggregatesPass()); // Break up aggregate allocas
   addPass(PM, createPromoteMemoryToRegister());  // Promote alloca's to regs
@@ -88,7 +86,7 @@
   // Run instcombine after redundancy elimination to exploit opportunities
   // opened up by them.
   addPass(PM, createInstructionCombiningPass());
-  addPass(PM, createAggressiveDCEPass());        // SSA based 'Agressive DCE'
+  addPass(PM, createAggressiveDCEPass());        // SSA based 'Aggressive DCE'
   addPass(PM, createCFGSimplificationPass());    // Merge & remove BBs
 }
 




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:42:08 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:42:08 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/SCCP/2003-06-24-OverdefinedPHIValue.ll
Message-ID: <200306262137.h5QLbYB00378@choi.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/SCCP:

2003-06-24-OverdefinedPHIValue.ll added (r1.1.2.1)

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/SCCP/2003-06-24-OverdefinedPHIValue.ll
diff -c /dev/null llvm/test/Regression/Transforms/SCCP/2003-06-24-OverdefinedPHIValue.ll:1.1
*** /dev/null	Thu Jun 26 16:37:34 2003
--- llvm/test/Regression/Transforms/SCCP/2003-06-24-OverdefinedPHIValue.ll	Tue Jun 24 15:29:36 2003
***************
*** 0 ****
--- 1,36 ----
+ ; RUN: if as < %s | opt -sccp -simplifycfg | dis | grep then:
+ ; RUN: then exit 1
+ ; RUN: else exit 0
+ ; RUN: fi
+ 
+ void %cprop_test11(int* %data.1) {
+ entry:		; No predecessors!
+ 	%tmp.1 = load int* %data.1		; <int> [#uses=3]
+ 	%tmp.41 = setgt int %tmp.1, 1		; <bool> [#uses=1]
+ 	br bool %tmp.41, label %no_exit, label %loopexit
+ 
+ no_exit:		; preds = %entry, %then, %endif
+ 	%j.0 = phi int [ %j.0, %endif ], [ %i.0, %then ], [ 1, %entry ]		; <int> [#uses=3]
+ 	%i.0 = phi int [ %inc, %endif ], [ %inc1, %then ], [ 1, %entry ]		; <int> [#uses=4]
+ 	%tmp.8.not = cast int %j.0 to bool		; <bool> [#uses=1]
+ 	br bool %tmp.8.not, label %endif, label %then
+ 
+ then:		; preds = %no_exit
+ 	%inc1 = add int %i.0, 1		; <int> [#uses=3]
+ 	%tmp.42 = setlt int %inc1, %tmp.1		; <bool> [#uses=1]
+ 	br bool %tmp.42, label %no_exit, label %loopexit
+ 
+ endif:		; preds = %no_exit
+ 	%inc = add int %i.0, 1		; <int> [#uses=3]
+ 	%tmp.4 = setlt int %inc, %tmp.1		; <bool> [#uses=1]
+ 	br bool %tmp.4, label %no_exit, label %loopexit
+ 
+ loopexit:		; preds = %entry, %endif, %then
+ 	%j.1 = phi int [ 1, %entry ], [ %j.0, %endif ], [ %i.0, %then ]		; <int> [#uses=1]
+ 	%i.1 = phi int [ 1, %entry ], [ %inc, %endif ], [ %inc1, %then ]		; <int> [#uses=1]
+ 	%tmp.17 = getelementptr int* %data.1, long 1		; <int*> [#uses=1]
+ 	store int %j.1, int* %tmp.17
+ 	%tmp.23 = getelementptr int* %data.1, long 2		; <int*> [#uses=1]
+ 	store int %i.1, int* %tmp.23
+ 	ret void
+ }




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:42:15 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:42:15 2003
Subject: [llvm-commits] CVS: llvm/lib/Target/X86/InstSelectSimple.cpp Printer.cpp X86InstrInfo.def X86TargetMachine.cpp X86TargetMachine.h
Message-ID: <200306262137.h5QLbYC00392@choi.cs.uiuc.edu>


Changes in directory llvm/lib/Target/X86:

InstSelectSimple.cpp updated: 1.106 -> 1.106.2.1
Printer.cpp updated: 1.35 -> 1.35.2.1
X86InstrInfo.def updated: 1.57 -> 1.57.2.1
X86TargetMachine.cpp updated: 1.17 -> 1.17.2.1
X86TargetMachine.h updated: 1.9 -> 1.9.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/lib/Target/X86/InstSelectSimple.cpp
diff -u llvm/lib/Target/X86/InstSelectSimple.cpp:1.106 llvm/lib/Target/X86/InstSelectSimple.cpp:1.106.2.1
--- llvm/lib/Target/X86/InstSelectSimple.cpp:1.106	Thu Jun  5 14:30:30 2003
+++ llvm/lib/Target/X86/InstSelectSimple.cpp	Thu Jun 26 16:35:24 2003
@@ -1069,8 +1069,6 @@
 /// registers op0Reg and op1Reg, and put the result in DestReg.  The type of the
 /// result should be given as DestTy.
 ///
-/// FIXME: doMultiply should use one of the two address IMUL instructions!
-///
 void ISel::doMultiply(MachineBasicBlock *MBB, MachineBasicBlock::iterator &MBBI,
                       unsigned DestReg, const Type *DestTy,
                       unsigned op0Reg, unsigned op1Reg) {
@@ -1079,28 +1077,20 @@
   case cFP:              // Floating point multiply
     BMI(BB, MBBI, X86::FpMUL, 2, DestReg).addReg(op0Reg).addReg(op1Reg);
     return;
+  case cInt:
+  case cShort:
+    BMI(BB, MBBI, Class == cInt ? X86::IMULr32 : X86::IMULr16, 2, DestReg)
+      .addReg(op0Reg).addReg(op1Reg);
+    return;
+  case cByte:
+    // Must use the MUL instruction, which forces use of AL...
+    BMI(MBB, MBBI, X86::MOVrr8, 1, X86::AL).addReg(op0Reg);
+    BMI(MBB, MBBI, X86::MULr8, 1).addReg(op1Reg);
+    BMI(MBB, MBBI, X86::MOVrr8, 1, DestReg).addReg(X86::AL);
+    return;
   default:
   case cLong: assert(0 && "doMultiply cannot operate on LONG values!");
-  case cByte:
-  case cShort:
-  case cInt:          // Small integerals, handled below...
-    break;
   }
- 
-  static const unsigned Regs[]     ={ X86::AL    , X86::AX     , X86::EAX     };
-  static const unsigned MulOpcode[]={ X86::MULr8 , X86::MULr16 , X86::MULr32  };
-  static const unsigned MovOpcode[]={ X86::MOVrr8, X86::MOVrr16, X86::MOVrr32 };
-  unsigned Reg     = Regs[Class];
-
-  // Emit a MOV to put the first operand into the appropriately-sized
-  // subreg of EAX.
-  BMI(MBB, MBBI, MovOpcode[Class], 1, Reg).addReg(op0Reg);
-  
-  // Emit the appropriate multiply instruction.
-  BMI(MBB, MBBI, MulOpcode[Class], 1).addReg(op1Reg);
-
-  // Emit another MOV to put the result into the destination register.
-  BMI(MBB, MBBI, MovOpcode[Class], 1, DestReg).addReg(Reg);
 }
 
 /// visitMul - Multiplies are not simple binary operators because they must deal
@@ -1126,16 +1116,16 @@
     BuildMI(BB, X86::MOVrr32, 1, OverflowReg).addReg(X86::EDX); // AL*BL >> 32
 
     MachineBasicBlock::iterator MBBI = BB->end();
-    unsigned AHBLReg = makeAnotherReg(Type::UIntTy);
-    doMultiply(BB, MBBI, AHBLReg, Type::UIntTy, Op0Reg+1, Op1Reg); // AH*BL
+    unsigned AHBLReg = makeAnotherReg(Type::UIntTy);   // AH*BL
+    BMI(BB, MBBI, X86::IMULr32, 2, AHBLReg).addReg(Op0Reg+1).addReg(Op1Reg);
 
     unsigned AHBLplusOverflowReg = makeAnotherReg(Type::UIntTy);
     BuildMI(BB, X86::ADDrr32, 2,                         // AH*BL+(AL*BL >> 32)
 	    AHBLplusOverflowReg).addReg(AHBLReg).addReg(OverflowReg);
     
     MBBI = BB->end();
-    unsigned ALBHReg = makeAnotherReg(Type::UIntTy);
-    doMultiply(BB, MBBI, ALBHReg, Type::UIntTy, Op0Reg, Op1Reg+1); // AL*BH
+    unsigned ALBHReg = makeAnotherReg(Type::UIntTy); // AL*BH
+    BMI(BB, MBBI, X86::IMULr32, 2, ALBHReg).addReg(Op0Reg).addReg(Op1Reg+1);
     
     BuildMI(BB, X86::ADDrr32, 2,               // AL*BH + AH*BL + (AL*BL >> 32)
 	    DestReg+1).addReg(AHBLplusOverflowReg).addReg(ALBHReg);
@@ -1189,7 +1179,7 @@
 
   static const unsigned Regs[]     ={ X86::AL    , X86::AX     , X86::EAX     };
   static const unsigned MovOpcode[]={ X86::MOVrr8, X86::MOVrr16, X86::MOVrr32 };
-  static const unsigned ExtOpcode[]={ X86::CBW   , X86::CWD    , X86::CDQ     };
+  static const unsigned SarOpcode[]={ X86::SARir8, X86::SARir16, X86::SARir32 };
   static const unsigned ClrOpcode[]={ X86::XORrr8, X86::XORrr16, X86::XORrr32 };
   static const unsigned ExtRegs[]  ={ X86::AH    , X86::DX     , X86::EDX     };
 
@@ -1207,7 +1197,9 @@
 
   if (isSigned) {
     // Emit a sign extension instruction...
-    BuildMI(BB, ExtOpcode[Class], 0);
+    unsigned ShiftResult = makeAnotherReg(I.getType());
+    BuildMI(BB, SarOpcode[Class], 2, ShiftResult).addReg(Op0Reg).addZImm(31);
+    BuildMI(BB, MovOpcode[Class], 1, ExtReg).addReg(ShiftResult);
   } else {
     // If unsigned, emit a zeroing instruction... (reg = xor reg, reg)
     BuildMI(BB, ClrOpcode[Class], 2, ExtReg).addReg(ExtReg).addReg(ExtReg);
@@ -1580,9 +1572,26 @@
 /// visitCastInst - Here we have various kinds of copying with or without
 /// sign extension going on.
 void ISel::visitCastInst(CastInst &CI) {
+  Value *Op = CI.getOperand(0);
+  // If this is a cast from a 32-bit integer to a Long type, and the only uses
+  // of the case are GEP instructions, then the cast does not need to be
+  // generated explicitly, it will be folded into the GEP.
+  if (CI.getType() == Type::LongTy &&
+      (Op->getType() == Type::IntTy || Op->getType() == Type::UIntTy)) {
+    bool AllUsesAreGEPs = true;
+    for (Value::use_iterator I = CI.use_begin(), E = CI.use_end(); I != E; ++I)
+      if (!isa<GetElementPtrInst>(*I)) {
+        AllUsesAreGEPs = false;
+        break;
+      }        
+
+    // No need to codegen this cast if all users are getelementptr instrs...
+    if (AllUsesAreGEPs) return;
+  }
+
   unsigned DestReg = getReg(CI);
   MachineBasicBlock::iterator MI = BB->end();
-  emitCastOperation(BB, MI, CI.getOperand(0), CI.getType(), DestReg);
+  emitCastOperation(BB, MI, Op, CI.getType(), DestReg);
 }
 
 /// emitCastOperation - Common code shared between visitCastInst and
@@ -1936,6 +1945,13 @@
       // indices, we may not know its actual value at code-generation
       // time.
       assert(idx->getType() == Type::LongTy && "Bad GEP array index!");
+
+      // Most GEP instructions use a [cast (int/uint) to LongTy] as their
+      // operand on X86.  Handle this case directly now...
+      if (CastInst *CI = dyn_cast<CastInst>(idx))
+        if (CI->getOperand(0)->getType() == Type::IntTy ||
+            CI->getOperand(0)->getType() == Type::UIntTy)
+          idx = CI->getOperand(0);
 
       // We want to add BaseReg to(idxReg * sizeof ElementType). First, we
       // must find the size of the pointed-to type (Not coincidentally, the next


Index: llvm/lib/Target/X86/Printer.cpp
diff -u llvm/lib/Target/X86/Printer.cpp:1.35 llvm/lib/Target/X86/Printer.cpp:1.35.2.1
--- llvm/lib/Target/X86/Printer.cpp:1.35	Mon May 26 19:03:17 2003
+++ llvm/lib/Target/X86/Printer.cpp	Thu Jun 26 16:35:24 2003
@@ -14,20 +14,41 @@
 #include "llvm/CodeGen/MachineConstantPool.h"
 #include "llvm/CodeGen/MachineInstr.h"
 #include "Support/Statistic.h"
+#include "Support/hash_map"
+#include "llvm/Type.h"
+#include "llvm/Constants.h"
+#include "llvm/Assembly/Writer.h"
+#include "llvm/DerivedTypes.h"
+#include "llvm/SlotCalculator.h"
+#include "Support/StringExtras.h"
+#include "llvm/Module.h"
 
 namespace {
+
+
+  std::set<const Value*> MangledGlobals;
   struct Printer : public MachineFunctionPass {
+
     std::ostream &O;
     unsigned ConstIdx;
     Printer(std::ostream &o) : O(o), ConstIdx(0) {}
+    const TargetData *TD;
 
     virtual const char *getPassName() const {
       return "X86 Assembly Printer";
     }
 
-    void printConstantPool(MachineConstantPool *MCP, const TargetData &TD);
-    bool runOnMachineFunction(MachineFunction &F);
+    void printConstantPool(MachineConstantPool *MCP);
+    bool runOnMachineFunction(MachineFunction &F);    
+    std::string ConstantExprToString(const ConstantExpr* CE);
+    std::string valToExprString(const Value* V);
+    bool doInitialization(Module &M);
+    bool doFinalization(Module &M);
+    void PrintZeroBytesToPad(int numBytes);
+    void printConstantValueOnly(const Constant* CV, int numPadBytesAfter = 0);
+    void printSingleConstantValue(const Constant* CV);
   };
+  std::map<const Value *, unsigned> NumberForBB;
 }
 
 /// createX86CodePrinterPass - Print out the specified machine code function to
@@ -38,44 +59,377 @@
   return new Printer(O);
 }
 
+// We dont want identifier names with ., space, -  in them. 
+// So we replace them with _
+static std::string makeNameProper(std::string x) {
+  std::string tmp;
+  for (std::string::iterator sI = x.begin(), sEnd = x.end(); sI != sEnd; sI++)
+    switch (*sI) {
+    case '.': tmp += "d_"; break;
+    case ' ': tmp += "s_"; break;
+    case '-': tmp += "D_"; break;
+    default:  tmp += *sI;
+    }
+
+  return tmp;
+}
+
+std::string getValueName(const Value *V) {
+  if (V->hasName()) { // Print out the label if it exists...
+    
+    // Name mangling occurs as follows:
+    // - If V is not a global, mangling always occurs.
+    // - Otherwise, mangling occurs when any of the following are true:
+    //   1) V has internal linkage
+    //   2) V's name would collide if it is not mangled.
+    //
+    
+    if(const GlobalValue* gv = dyn_cast<GlobalValue>(V)) {
+      if(!gv->hasInternalLinkage() && !MangledGlobals.count(gv)) {
+        // No internal linkage, name will not collide -> no mangling.
+        return makeNameProper(gv->getName());
+      }
+    }
+    
+    // Non-global, or global with internal linkage / colliding name -> mangle.
+    return "l" + utostr(V->getType()->getUniqueID()) + "_" +
+      makeNameProper(V->getName());      
+  }
+
+  static int Count = 0;
+  Count++;
+  return "ltmp_" + itostr(Count) + "_" + utostr(V->getType()->getUniqueID());
+}
+
+
+// valToExprString - Helper function for ConstantExprToString().
+// Appends result to argument string S.
+// 
+std::string Printer::valToExprString(const Value* V) {
+  std::string S;
+  bool failed = false;
+  if (const Constant* CV = dyn_cast<Constant>(V)) { // symbolic or known
+    if (const ConstantBool *CB = dyn_cast<ConstantBool>(CV))
+      S += std::string(CB == ConstantBool::True ? "1" : "0");
+    else if (const ConstantSInt *CI = dyn_cast<ConstantSInt>(CV))
+      S += itostr(CI->getValue());
+    else if (const ConstantUInt *CI = dyn_cast<ConstantUInt>(CV))
+      S += utostr(CI->getValue());
+    else if (const ConstantFP *CFP = dyn_cast<ConstantFP>(CV))
+      S += ftostr(CFP->getValue());
+    else if (isa<ConstantPointerNull>(CV))
+      S += "0";
+    else if (const ConstantPointerRef *CPR = dyn_cast<ConstantPointerRef>(CV))
+      S += valToExprString(CPR->getValue());
+    else if (const ConstantExpr *CE = dyn_cast<ConstantExpr>(CV))
+      S += ConstantExprToString(CE);
+    else
+      failed = true;
+  } else if (const GlobalValue* GV = dyn_cast<GlobalValue>(V)) {
+    S += getValueName(GV);
+  }
+  else
+    failed = true;
+
+  if (failed) {
+    assert(0 && "Cannot convert value to string");
+    S += "<illegal-value>";
+  }
+  return S;
+}
+
+// ConstantExprToString() - Convert a ConstantExpr to an asm expression
+// and return this as a string.
+std::string Printer::ConstantExprToString(const ConstantExpr* CE) {
+  std::string S;
+  switch(CE->getOpcode()) {
+  case Instruction::GetElementPtr:
+    { // generate a symbolic expression for the byte address
+      const Value* ptrVal = CE->getOperand(0);
+      std::vector<Value*> idxVec(CE->op_begin()+1, CE->op_end());
+      S += "(" + valToExprString(ptrVal) + ") + ("
+	+ utostr(TD->getIndexedOffset(ptrVal->getType(),idxVec)) + ")";
+      break;
+    }
+
+  case Instruction::Cast:
+    // Support only non-converting casts for now, i.e., a no-op.
+    // This assertion is not a complete check.
+    assert(TD->getTypeSize(CE->getType()) ==
+	   TD->getTypeSize(CE->getOperand(0)->getType()));
+    S += "(" + valToExprString(CE->getOperand(0)) + ")";
+    break;
+
+  case Instruction::Add:
+    S += "(" + valToExprString(CE->getOperand(0)) + ") + ("
+      + valToExprString(CE->getOperand(1)) + ")";
+    break;
+
+  default:
+    assert(0 && "Unsupported operator in ConstantExprToString()");
+    break;
+  }
+
+  return S;
+}
+
+// Print a single constant value.
+void
+Printer::printSingleConstantValue(const Constant* CV)
+{
+  assert(CV->getType() != Type::VoidTy &&
+         CV->getType() != Type::TypeTy &&
+         CV->getType() != Type::LabelTy &&
+         "Unexpected type for Constant");
+  
+  assert((!isa<ConstantArray>(CV) && ! isa<ConstantStruct>(CV))
+         && "Aggregate types should be handled outside this function");
+
+  const Type *type = CV->getType();
+  O << "\t";
+  switch(type->getPrimitiveID())
+    {
+    case Type::BoolTyID: case Type::UByteTyID: case Type::SByteTyID:
+      O << ".byte";
+      break;
+    case Type::UShortTyID: case Type::ShortTyID:
+      O << ".word";
+      break;
+    case Type::UIntTyID: case Type::IntTyID: case Type::PointerTyID:
+      O << ".long";
+      break;
+    case Type::ULongTyID: case Type::LongTyID:
+      O << ".quad";
+      break;
+    case Type::FloatTyID:
+      O << ".long";
+      break;
+    case Type::DoubleTyID:
+      O << ".quad";
+      break;
+    case Type::ArrayTyID:
+      if ((cast<ArrayType>(type)->getElementType() == Type::UByteTy) ||
+	  (cast<ArrayType>(type)->getElementType() == Type::SByteTy))
+	O << ".string";
+      else
+	assert (0 && "Can't handle printing this type of array");
+      break;
+    default:
+      assert (0 && "Can't handle printing this type of thing");
+      break;
+    }
+  O << "\t";
+  
+  if (type->isPrimitiveType())
+    {
+      if (type->isFloatingPoint()) {
+	// FP Constants are printed as integer constants to avoid losing
+	// precision...
+	double Val = cast<ConstantFP>(CV)->getValue();
+	if (type == Type::FloatTy) {
+	  float FVal = (float)Val;
+	  char *ProxyPtr = (char*)&FVal;        // Abide by C TBAA rules
+	  O << *(unsigned int*)ProxyPtr;            
+	} else if (type == Type::DoubleTy) {
+	  char *ProxyPtr = (char*)&Val;         // Abide by C TBAA rules
+	  O << *(uint64_t*)ProxyPtr;            
+	} else {
+	  assert(0 && "Unknown floating point type!");
+	}
+        
+	O << "\t# " << type->getDescription() << " value: " << Val << "\n";
+      } else {
+	WriteAsOperand(O, CV, false, false) << "\n";
+      }
+    }
+  else if (const ConstantPointerRef* CPR = dyn_cast<ConstantPointerRef>(CV))
+    {
+      // This is a constant address for a global variable or method.
+      // Use the name of the variable or method as the address value.
+      O << getValueName(CPR->getValue()) << "\n";
+    }
+  else if (isa<ConstantPointerNull>(CV))
+    {
+      // Null pointer value
+      O << "0\n";
+    }
+  else if (const ConstantExpr* CE = dyn_cast<ConstantExpr>(CV))
+    {
+      // Constant expression built from operators, constants, and
+      // symbolic addrs
+      O << ConstantExprToString(CE) << "\n";
+    }
+  else
+    {
+      assert(0 && "Unknown elementary type for constant");
+    }
+}
+
+// Can we treat the specified array as a string?  Only if it is an array of
+// ubytes or non-negative sbytes.
+//
+static bool isStringCompatible(const ConstantArray *CVA) {
+  const Type *ETy = cast<ArrayType>(CVA->getType())->getElementType();
+  if (ETy == Type::UByteTy) return true;
+  if (ETy != Type::SByteTy) return false;
+
+  for (unsigned i = 0; i < CVA->getNumOperands(); ++i)
+    if (cast<ConstantSInt>(CVA->getOperand(i))->getValue() < 0)
+      return false;
+
+  return true;
+}
+
+// toOctal - Convert the low order bits of X into an octal letter
+static inline char toOctal(int X) {
+  return (X&7)+'0';
+}
+
+// getAsCString - Return the specified array as a C compatible string, only if
+// the predicate isStringCompatible is true.
+//
+static std::string getAsCString(const ConstantArray *CVA) {
+  assert(isStringCompatible(CVA) && "Array is not string compatible!");
+
+  std::string Result;
+  const Type *ETy = cast<ArrayType>(CVA->getType())->getElementType();
+  Result = "\"";
+  for (unsigned i = 0; i < CVA->getNumOperands(); ++i) {
+    unsigned char C = (ETy == Type::SByteTy) ?
+      (unsigned char)cast<ConstantSInt>(CVA->getOperand(i))->getValue() :
+      (unsigned char)cast<ConstantUInt>(CVA->getOperand(i))->getValue();
+
+    if (C == '"') {
+      Result += "\\\"";
+    } else if (C == '\\') {
+      Result += "\\\\";
+    } else if (isprint(C)) {
+      Result += C;
+    } else {
+      switch(C) {
+      case '\a': Result += "\\a"; break;
+      case '\b': Result += "\\b"; break;
+      case '\f': Result += "\\f"; break;
+      case '\n': Result += "\\n"; break;
+      case '\r': Result += "\\r"; break;
+      case '\t': Result += "\\t"; break;
+      case '\v': Result += "\\v"; break;
+      default:
+        Result += '\\';
+        Result += toOctal(C >> 6);
+        Result += toOctal(C >> 3);
+        Result += toOctal(C >> 0);
+        break;
+      }
+    }
+  }
+  Result += "\"";
+  return Result;
+}
+
+// Print a constant value or values (it may be an aggregate).
+// Uses printSingleConstantValue() to print each individual value.
+void
+Printer::printConstantValueOnly(const Constant* CV,
+				int numPadBytesAfter /* = 0 */)
+{
+  const ConstantArray *CVA = dyn_cast<ConstantArray>(CV);
+
+  if (CVA && isStringCompatible(CVA))
+    { // print the string alone and return
+      O << "\t" << ".string" << "\t" << getAsCString(CVA) << "\n";
+    }
+  else if (CVA)
+    { // Not a string.  Print the values in successive locations
+      const std::vector<Use> &constValues = CVA->getValues();
+      for (unsigned i=0; i < constValues.size(); i++)
+        printConstantValueOnly(cast<Constant>(constValues[i].get()));
+    }
+  else if (const ConstantStruct *CVS = dyn_cast<ConstantStruct>(CV))
+    { // Print the fields in successive locations. Pad to align if needed!
+      const StructLayout *cvsLayout =
+        TD->getStructLayout(CVS->getType());
+      const std::vector<Use>& constValues = CVS->getValues();
+      unsigned sizeSoFar = 0;
+      for (unsigned i=0, N = constValues.size(); i < N; i++)
+        {
+          const Constant* field = cast<Constant>(constValues[i].get());
+
+          // Check if padding is needed and insert one or more 0s.
+          unsigned fieldSize = TD->getTypeSize(field->getType());
+          int padSize = ((i == N-1? cvsLayout->StructSize
+			  : cvsLayout->MemberOffsets[i+1])
+                         - cvsLayout->MemberOffsets[i]) - fieldSize;
+          sizeSoFar += (fieldSize + padSize);
+
+          // Now print the actual field value
+          printConstantValueOnly(field, padSize);
+        }
+      assert(sizeSoFar == cvsLayout->StructSize &&
+             "Layout of constant struct may be incorrect!");
+    }
+  else
+    printSingleConstantValue(CV);
+
+  if (numPadBytesAfter) {
+    unsigned numBytes = numPadBytesAfter;
+    for ( ; numBytes >= 8; numBytes -= 8)
+      printSingleConstantValue(Constant::getNullValue(Type::ULongTy));
+    if (numBytes >= 4)
+      {
+	printSingleConstantValue(Constant::getNullValue(Type::UIntTy));
+	numBytes -= 4;
+      }
+    while (numBytes--)
+      printSingleConstantValue(Constant::getNullValue(Type::UByteTy));
+  }
+}
 
 // printConstantPool - Print out any constants which have been spilled to
 // memory...
-void Printer::printConstantPool(MachineConstantPool *MCP, const TargetData &TD){
+void Printer::printConstantPool(MachineConstantPool *MCP){
   const std::vector<Constant*> &CP = MCP->getConstants();
   if (CP.empty()) return;
 
   for (unsigned i = 0, e = CP.size(); i != e; ++i) {
     O << "\t.section .rodata\n";
-    O << "\t.align " << (unsigned)TD.getTypeAlignment(CP[i]->getType()) << "\n";
-    O << ".CPI" << i+ConstIdx << ":\t\t\t\t\t;" << *CP[i] << "\n";
-    O << "\t*Constant output not implemented yet!*\n\n";
+    O << "\t.align " << (unsigned)TD->getTypeAlignment(CP[i]->getType()) << "\n";
+    O << ".CPI" << i+ConstIdx << ":\t\t\t\t\t#" << *CP[i] << "\n";
+    printConstantValueOnly (CP[i]);
   }
   ConstIdx += CP.size();  // Don't recycle constant pool index numbers
 }
 
-/// runOnFunction - This uses the X86InstructionInfo::print method
+/// runOnMachineFunction - This uses the X86InstructionInfo::print method
 /// to print assembly for each instruction.
 bool Printer::runOnMachineFunction(MachineFunction &MF) {
   static unsigned BBNumber = 0;
   const TargetMachine &TM = MF.getTarget();
   const TargetInstrInfo &TII = TM.getInstrInfo();
+  TD = &TM.getTargetData();
 
   // Print out constants referenced by the function
-  printConstantPool(MF.getConstantPool(), TM.getTargetData());
+  printConstantPool(MF.getConstantPool());
 
   // Print out labels for the function.
   O << "\t.text\n";
   O << "\t.align 16\n";
-  O << "\t.globl\t" << MF.getFunction()->getName() << "\n";
-  O << "\t.type\t" << MF.getFunction()->getName() << ", @function\n";
-  O << MF.getFunction()->getName() << ":\n";
+  O << "\t.globl\t" << getValueName(MF.getFunction()) << "\n";
+  O << "\t.type\t" << getValueName(MF.getFunction()) << ", @function\n";
+  O << getValueName(MF.getFunction()) << ":\n";
+
+  NumberForBB.clear();
+  for (MachineFunction::const_iterator I = MF.begin(), E = MF.end();
+       I != E; ++I) {
+    NumberForBB[I->getBasicBlock()] = BBNumber++;
+  }
 
   // Print out code for the function.
   for (MachineFunction::const_iterator I = MF.begin(), E = MF.end();
        I != E; ++I) {
     // Print a label for the basic block.
-    O << ".BB" << BBNumber++ << ":\n";
+    O << ".BB" << NumberForBB[I->getBasicBlock()] << ":\t# "
+      << I->getBasicBlock()->getName() << "\n";
     for (MachineBasicBlock::const_iterator II = I->begin(), E = I->end();
 	 II != E; ++II) {
       // Print the assembly for the instruction.
@@ -90,20 +444,20 @@
 
 static bool isScale(const MachineOperand &MO) {
   return MO.isImmediate() &&
-           (MO.getImmedValue() == 1 || MO.getImmedValue() == 2 ||
-            MO.getImmedValue() == 4 || MO.getImmedValue() == 8);
+    (MO.getImmedValue() == 1 || MO.getImmedValue() == 2 ||
+     MO.getImmedValue() == 4 || MO.getImmedValue() == 8);
 }
 
 static bool isMem(const MachineInstr *MI, unsigned Op) {
   if (MI->getOperand(Op).isFrameIndex()) return true;
   if (MI->getOperand(Op).isConstantPoolIndex()) return true;
   return Op+4 <= MI->getNumOperands() &&
-         MI->getOperand(Op  ).isRegister() &&isScale(MI->getOperand(Op+1)) &&
-         MI->getOperand(Op+2).isRegister() &&MI->getOperand(Op+3).isImmediate();
+    MI->getOperand(Op  ).isRegister() &&isScale(MI->getOperand(Op+1)) &&
+    MI->getOperand(Op+2).isRegister() &&MI->getOperand(Op+3).isImmediate();
 }
 
 static void printOp(std::ostream &O, const MachineOperand &MO,
-                    const MRegisterInfo &RI) {
+                    const MRegisterInfo &RI, bool elideOffsetKeyword = false) {
   switch (MO.getType()) {
   case MachineOperand::MO_VirtualRegister:
     if (Value *V = MO.getVRegValueOrNull()) {
@@ -123,29 +477,30 @@
     O << (int)MO.getImmedValue();
     return;
   case MachineOperand::MO_PCRelativeDisp:
-    O << "<" << MO.getVRegValue()->getName() << ">";
+    O << ".BB" << NumberForBB[MO.getVRegValue()] << " # PC rel: "
+      << MO.getVRegValue()->getName();
     return;
   case MachineOperand::MO_GlobalAddress:
-    O << "<" << MO.getGlobal()->getName() << ">";
+    if (!elideOffsetKeyword) O << "OFFSET "; O << getValueName(MO.getGlobal());
     return;
   case MachineOperand::MO_ExternalSymbol:
-    O << "<" << MO.getSymbolName() << ">";
+    O << MO.getSymbolName();
     return;
   default:
-    O << "<unknown op ty>"; return;    
+    O << "<unknown operand type>"; return;    
   }
 }
 
 static const std::string sizePtr(const TargetInstrDescriptor &Desc) {
   switch (Desc.TSFlags & X86II::ArgMask) {
-    default: assert(0 && "Unknown arg size!");
-    case X86II::Arg8:   return "BYTE PTR"; 
-    case X86II::Arg16:  return "WORD PTR"; 
-    case X86II::Arg32:  return "DWORD PTR"; 
-    case X86II::Arg64:  return "QWORD PTR"; 
-    case X86II::ArgF32:  return "DWORD PTR"; 
-    case X86II::ArgF64:  return "QWORD PTR"; 
-    case X86II::ArgF80:  return "XWORD PTR"; 
+  default: assert(0 && "Unknown arg size!");
+  case X86II::Arg8:   return "BYTE PTR"; 
+  case X86II::Arg16:  return "WORD PTR"; 
+  case X86II::Arg32:  return "DWORD PTR"; 
+  case X86II::Arg64:  return "QWORD PTR"; 
+  case X86II::ArgF32:  return "DWORD PTR"; 
+  case X86II::ArgF64:  return "QWORD PTR"; 
+  case X86II::ArgF80:  return "XWORD PTR"; 
   }
 }
 
@@ -208,6 +563,10 @@
 
   switch (Desc.TSFlags & X86II::FormMask) {
   case X86II::Pseudo:
+    // Print pseudo-instructions as comments; either they should have been
+    // turned into real instructions by now, or they don't need to be
+    // seen by the assembler (e.g., IMPLICIT_USEs.)
+    O << "# ";
     if (Opcode == X86::PHI) {
       printOp(O, MI->getOperand(0), RI);
       O << " = phi ";
@@ -256,7 +615,7 @@
     O << getName(MI->getOpcode()) << " ";
 
     if (MI->getNumOperands() == 1) {
-      printOp(O, MI->getOperand(0), RI);
+      printOp(O, MI->getOperand(0), RI, true); // Don't print "OFFSET"...
     }
     O << "\n";
     return;
@@ -409,7 +768,7 @@
             MI->getOperand(1).isRegister() || MI->getOperand(1).isImmediate())&&
            "Bad MRMSxR format!");
     assert((MI->getNumOperands() < 3 ||
-        (MI->getOperand(1).isRegister() && MI->getOperand(2).isImmediate())) &&
+	    (MI->getOperand(1).isRegister() && MI->getOperand(2).isImmediate())) &&
            "Bad MRMSxR format!");
 
     if (MI->getNumOperands() > 1 && MI->getOperand(1).isRegister() && 
@@ -456,4 +815,59 @@
   default:
     O << "\tUNKNOWN FORM:\t\t-"; MI->print(O, TM); break;
   }
+}
+
+bool Printer::doInitialization(Module &M)
+{
+  // Tell gas we are outputting Intel syntax (not AT&T syntax) assembly,
+  // with no % decorations on register names.
+  O << "\t.intel_syntax noprefix\n";
+
+  // Ripped from CWriter:
+  // Calculate which global values have names that will collide when we throw
+  // away type information.
+  {  // Scope to delete the FoundNames set when we are done with it...
+    std::set<std::string> FoundNames;
+    for (Module::iterator I = M.begin(), E = M.end(); I != E; ++I)
+      if (I->hasName())                      // If the global has a name...
+        if (FoundNames.count(I->getName()))  // And the name is already used
+          MangledGlobals.insert(I);          // Mangle the name
+        else
+          FoundNames.insert(I->getName());   // Otherwise, keep track of name
+
+    for (Module::giterator I = M.gbegin(), E = M.gend(); I != E; ++I)
+      if (I->hasName())                      // If the global has a name...
+        if (FoundNames.count(I->getName()))  // And the name is already used
+          MangledGlobals.insert(I);          // Mangle the name
+        else
+          FoundNames.insert(I->getName());   // Otherwise, keep track of name
+  }
+
+  return false; // success
+}
+
+bool Printer::doFinalization(Module &M)
+{
+  // Print out module-level global variables here.
+  for (Module::const_giterator I = M.gbegin(), E = M.gend(); I != E; ++I) {
+    std::string name(getValueName(I));
+    if (I->hasInitializer()) {
+      Constant *C = I->getInitializer();
+      O << "\t.data\n";
+      O << "\t.globl " << name << "\n";
+      O << "\t.type " << name << ", at object\n";
+      O << "\t.size " << name << ","
+	<< (unsigned)TD->getTypeSize(I->getType()) << "\n";
+      O << "\t.align " << (unsigned)TD->getTypeAlignment(C->getType()) << "\n";
+      O << name << ":\t\t\t\t\t#" << *C << "\n";
+      printConstantValueOnly (C);
+    } else {
+      O << "\t.globl " << name << "\n";
+      O << "\t.comm " << name << ", "
+        << (unsigned)TD->getTypeSize(I->getType()) << ", "
+        << (unsigned)TD->getTypeAlignment(I->getType()) << "\n";
+    }
+  }
+  MangledGlobals.clear();
+  return false; // success
 }


Index: llvm/lib/Target/X86/X86InstrInfo.def
diff -u llvm/lib/Target/X86/X86InstrInfo.def:1.57 llvm/lib/Target/X86/X86InstrInfo.def:1.57.2.1
--- llvm/lib/Target/X86/X86InstrInfo.def:1.57	Thu Jun  5 14:30:30 2003
+++ llvm/lib/Target/X86/X86InstrInfo.def	Thu Jun 26 16:35:24 2003
@@ -48,7 +48,10 @@
 
 // Arguments to be passed into the I macro
 //  #1: Enum name - This ends up being the opcode symbol in the X86 namespace
-//  #2: Opcode name, as used by the gnu assembler
+//  #2: Opcode name, as used by the Intel assembler. This should be in
+//      ALL CAPS for pseudo-instructions (which the assembler should never
+//      see), or all lowercase for real instructions (which the assembler
+//      should see).
 //  #3: The base opcode for the instruction
 //  #4: Instruction Flags - This should be a field or'd together that contains
 //      constants from the TargetInstrInfo.h file.
@@ -60,7 +63,7 @@
 //
 
 // The first instruction must always be the PHI instruction:
-I(PHI         , "phi",      0,             0, X86II::Pseudo              , NoIR, NoIR)
+I(PHI         , "PHI",      0,             0, X86II::Pseudo              , NoIR, NoIR)
 
 // The second instruction must always be the noop instruction:
 I(NOOP        , "nop",   0x90,             0, X86II::RawFrm | X86II::Void, NoIR, NoIR)             // nop
@@ -72,18 +75,18 @@
 // instruction turns into a noop, but if the frame pointer is retained, this
 // turns into a 'sub ESP, <amount>'.
 //
-I(ADJCALLSTACKDOWN, "adjcallstackdown", 0, 0, X86II::Pseudo,               NoIR, NoIR)
+I(ADJCALLSTACKDOWN, "ADJCALLSTACKDOWN", 0, 0, X86II::Pseudo,               NoIR, NoIR)
 
 // This instruction is used to mark readjustment of the stack after a function
 // call.  If the frame pointer is retained, this becomes a 'add ESP, <amount>'
 // instruction after the call.
-I(ADJCALLSTACKUP  , "adjcallstackup"  , 0, 0, X86II::Pseudo,               NoIR, NoIR)
+I(ADJCALLSTACKUP  , "ADJCALLSTACKUP"  , 0, 0, X86II::Pseudo,               NoIR, NoIR)
 
 // This pseudo-instruction is used to record implicit uses of physical registers
 // at the end of the function.  This ensures that bad things aren't done to
 // registes that are live on exit from the function (for example, EAX).
 //
-I(IMPLICIT_USE, "implicit_use", 0, 0, X86II::Pseudo, NoIR, NoIR)
+I(IMPLICIT_USE, "IMPLICIT_USE", 0, 0, X86II::Pseudo, NoIR, NoIR)
 
 
 // Flow control instructions
@@ -163,22 +166,24 @@
 
 I(MULr8       , "mul",  0xF6,             0, X86II::MRMS4r | X86II::Void, O_AL, O_AX)       // AX   = AL*R8
 I(MULr16      , "mul",  0xF7,             0, X86II::MRMS4r | X86II::Void |                  // DX:AX= AX*R16
-                                              X86II::OpSize, O_AX, T_AXDX)
+                                             X86II::OpSize, O_AX, T_AXDX)
 I(MULr32      , "mul",  0xF7,             0, X86II::MRMS4r | X86II::Void, O_EAX, T_EAXEDX)  // ED:EA= EA*R32
 
+I(IMULr16     , "imul", 0xAF, M_2_ADDR_FLAG, X86II::MRMSrcReg | X86II::TB |                 // R16 *= R16
+                                             X86II::OpSize, NoIR, NoIR)
+I(IMULr32     , "imul", 0xAF, M_2_ADDR_FLAG, X86II::MRMSrcReg | X86II::TB, NoIR, NoIR)      // R32 *= R32
+
 // unsigned division/remainder
 I(DIVr8       , "div",  0xF6,             0, X86II::MRMS6r | X86II::Void, O_AX, O_AX)       // AX/r8= AL&AH
 I(DIVr16      , "div",  0xF7,             0, X86II::MRMS6r | X86II::Void |                  // ED:EA/r16=AX&DX
-                                              X86II::OpSize, T_AXDX, T_AXDX)
-I(DIVr32      , "div",  0xF7,             0, X86II::MRMS6r | X86II::Void, T_EAXEDX,
-                                                                           T_EAXEDX)         // ED:EA/r32=EA&ED
+                                             X86II::OpSize, T_AXDX, T_AXDX)
+I(DIVr32      , "div",  0xF7,             0, X86II::MRMS6r | X86II::Void, T_EAXEDX,T_EAXEDX)// ED:EA/r32=EA&ED
 
 // signed division/remainder
 I(IDIVr8      , "idiv", 0xF6,             0, X86II::MRMS7r | X86II::Void, O_AX, O_AX)       // AX/r8= AL&AH
 I(IDIVr16     , "idiv", 0xF7,             0, X86II::MRMS7r | X86II::Void |                  // DA/r16=AX&DX
-                                              X86II::OpSize, T_AXDX, T_AXDX)
-I(IDIVr32     , "idiv", 0xF7,             0, X86II::MRMS7r | X86II::Void, T_EAXEDX,
-                                                                           T_EAXEDX)         // DA/r32=EAX&DX
+                                             X86II::OpSize, T_AXDX, T_AXDX)
+I(IDIVr32     , "idiv", 0xF7,             0, X86II::MRMS7r | X86II::Void, T_EAXEDX,T_EAXEDX)// DA/r32=EAX&DX
 
 // Logical operators
 I(ANDrr8      , "and",   0x20, M_2_ADDR_FLAG, X86II::MRMDestReg, NoIR, NoIR)                 // R8  &= R8
@@ -260,12 +265,12 @@
 
 
 // Integer comparisons
-I(CMPrr8      , "cmpb",  0x38,             0, X86II::Void | X86II::MRMDestReg                , NoIR, NoIR) // compare R8,R8
-I(CMPrr16     , "cmpw",  0x39,             0, X86II::Void | X86II::MRMDestReg | X86II::OpSize, NoIR, NoIR) // compare R16,R16
-I(CMPrr32     , "cmpl",  0x39,             0, X86II::Void | X86II::MRMDestReg                , NoIR, NoIR) // compare R32,R32
-I(CMPri8      , "cmpb",  0x80,             0, X86II::Void | X86II::MRMS7r     | X86II::Arg8  , NoIR, NoIR) // compare R8, imm8
-I(CMPri16     , "cmpw",  0x81,             0, X86II::Void | X86II::MRMS7r     | X86II::Arg16 | X86II::OpSize, NoIR, NoIR) // compare R8, imm8
-I(CMPri32     , "cmpl",  0x81,             0, X86II::Void | X86II::MRMS7r     | X86II::Arg32 , NoIR, NoIR) // compare R8, imm8
+I(CMPrr8      , "cmp",  0x38,             0, X86II::Void | X86II::MRMDestReg                , NoIR, NoIR) // compare R8,R8
+I(CMPrr16     , "cmp",  0x39,             0, X86II::Void | X86II::MRMDestReg | X86II::OpSize, NoIR, NoIR) // compare R16,R16
+I(CMPrr32     , "cmp",  0x39,             0, X86II::Void | X86II::MRMDestReg                , NoIR, NoIR) // compare R32,R32
+I(CMPri8      , "cmp",  0x80,             0, X86II::Void | X86II::MRMS7r     | X86II::Arg8  , NoIR, NoIR) // compare R8, imm8
+I(CMPri16     , "cmp",  0x81,             0, X86II::Void | X86II::MRMS7r     | X86II::Arg16 | X86II::OpSize, NoIR, NoIR) // compare R8, imm8
+I(CMPri32     , "cmp",  0x81,             0, X86II::Void | X86II::MRMS7r     | X86II::Arg32 , NoIR, NoIR) // compare R8, imm8
 
 // Sign extenders (first 3 are good for DIV/IDIV; the others are more general)
 I(CBW         , "cbw",   0x98,             0, X86II::Void | X86II::RawFrm | X86II::OpSize, O_AL, O_AH)     // AX = signext(AL)
@@ -299,60 +304,60 @@
 I(FpSETRESULT , "FSETRESULT",0, M_PSEUDO_FLAG | M_TERMINATOR_FLAG,         X86II::Void | X86II::Pseudo | X86II::SpecialFP, NoIR, NoIR)  // ST(0) = FPR
 
 // Floating point loads & stores...            PREFIX      ARGTYPE         ENCODING        FP INST TYPE      REF   MOD
-I(FLDr32      , "fld32",   0xD9,            0,                           X86II::ArgF32 | X86II::MRMS0m | X86II::ZeroArgFP, NoIR, NoIR)   // load float
-I(FLDr64      , "fld64",   0xDD,            0,                           X86II::ArgF64 | X86II::MRMS0m | X86II::ZeroArgFP, NoIR, NoIR)   // load double
-I(FLDr80      , "fld80",   0xDB,            0,                           X86II::ArgF80 | X86II::MRMS5m | X86II::ZeroArgFP, NoIR, NoIR)   // load extended
+I(FLDr32      , "fld",   0xD9,            0,                           X86II::ArgF32 | X86II::MRMS0m | X86II::ZeroArgFP, NoIR, NoIR)   // load float
+I(FLDr64      , "fld",   0xDD,            0,                           X86II::ArgF64 | X86II::MRMS0m | X86II::ZeroArgFP, NoIR, NoIR)   // load double
+I(FLDr80      , "fld",   0xDB,            0,                           X86II::ArgF80 | X86II::MRMS5m | X86II::ZeroArgFP, NoIR, NoIR)   // load extended
 I(FLDrr       , "fld"  ,   0xC0,            0, X86II::D9 |               X86II::ArgF80 | X86II::AddRegFrm                , NoIR, NoIR)   // push(ST(i))
-I(FILDr16     , "fild16",  0xDF,            0,                           X86II::Arg16  | X86II::MRMS0m | X86II::ZeroArgFP, NoIR, NoIR)   // load signed short
-I(FILDr32     , "fild32",  0xDB,            0,                           X86II::Arg32  | X86II::MRMS0m | X86II::ZeroArgFP, NoIR, NoIR)   // load signed int
-I(FILDr64     , "fild64",  0xDF,            0,                           X86II::Arg64  | X86II::MRMS5m | X86II::ZeroArgFP, NoIR, NoIR)   // load signed long
+I(FILDr16     , "fild",  0xDF,            0,                           X86II::Arg16  | X86II::MRMS0m | X86II::ZeroArgFP, NoIR, NoIR)   // load signed short
+I(FILDr32     , "fild",  0xDB,            0,                           X86II::Arg32  | X86II::MRMS0m | X86II::ZeroArgFP, NoIR, NoIR)   // load signed int
+I(FILDr64     , "fild",  0xDF,            0,                           X86II::Arg64  | X86II::MRMS5m | X86II::ZeroArgFP, NoIR, NoIR)   // load signed long
 
 
-I(FSTr32      , "fst32",   0xD9,            0,             X86II::Void | X86II::ArgF32 | X86II::MRMS2m | X86II::OneArgFP , NoIR, NoIR)   // store float
-I(FSTr64      , "fst64",   0xDD,            0,             X86II::Void | X86II::ArgF64 | X86II::MRMS2m | X86II::OneArgFP , NoIR, NoIR)   // store double
-I(FSTPr32     , "fst32p",  0xD9,            0,             X86II::Void | X86II::ArgF32 | X86II::MRMS3m                   , NoIR, NoIR)   // store float, pop
-I(FSTPr64     , "fst64p",  0xDD,            0,             X86II::Void | X86II::ArgF64 | X86II::MRMS3m                   , NoIR, NoIR)   // store double, pop
-I(FSTPr80     , "fst80p",  0xDB,            0,             X86II::Void | X86II::ArgF80 | X86II::MRMS7m | X86II::OneArgFP , NoIR, NoIR)   // store extended, pop
+I(FSTr32      , "fst",   0xD9,            0,             X86II::Void | X86II::ArgF32 | X86II::MRMS2m | X86II::OneArgFP , NoIR, NoIR)   // store float
+I(FSTr64      , "fst",   0xDD,            0,             X86II::Void | X86II::ArgF64 | X86II::MRMS2m | X86II::OneArgFP , NoIR, NoIR)   // store double
+I(FSTPr32     , "fstp",  0xD9,            0,             X86II::Void | X86II::ArgF32 | X86II::MRMS3m                   , NoIR, NoIR)   // store float, pop
+I(FSTPr64     , "fstp",  0xDD,            0,             X86II::Void | X86II::ArgF64 | X86II::MRMS3m                   , NoIR, NoIR)   // store double, pop
+I(FSTPr80     , "fstp",  0xDB,            0,             X86II::Void | X86II::ArgF80 | X86II::MRMS7m | X86II::OneArgFP , NoIR, NoIR)   // store extended, pop
 I(FSTrr       , "fst"  ,   0xD0,            0, X86II::DD | X86II::Void | X86II::ArgF80 | X86II::AddRegFrm                , NoIR, NoIR)   // ST(i) = ST(0)
 I(FSTPrr      , "fstp" ,   0xD8,            0, X86II::DD | X86II::Void | X86II::ArgF80 | X86II::AddRegFrm                , NoIR, NoIR)   // ST(i) = ST(0), pop
 
-I(FISTr16     , "fist16",  0xDF,            0,             X86II::Void | X86II::Arg16  | X86II::MRMS2m | X86II::OneArgFP , NoIR, NoIR)   // store signed short
-I(FISTr32     , "fist32",  0xDB,            0,             X86II::Void | X86II::Arg32  | X86II::MRMS2m | X86II::OneArgFP , NoIR, NoIR)   // store signed int
-I(FISTPr16    , "fist16p", 0xDF,            0,             X86II::Void | X86II::Arg16  | X86II::MRMS3m                   , NoIR, NoIR)   // store short, pop
-I(FISTPr32    , "fist32p", 0xDB,            0,             X86II::Void | X86II::Arg32  | X86II::MRMS3m                   , NoIR, NoIR)   // store int, pop
-I(FISTPr64    , "fist64p", 0xDF,            0,             X86II::Void | X86II::Arg64  | X86II::MRMS7m | X86II::OneArgFP , NoIR, NoIR)   // store long, pop
+I(FISTr16     , "fist",  0xDF,            0,             X86II::Void | X86II::Arg16  | X86II::MRMS2m | X86II::OneArgFP , NoIR, NoIR)   // store signed short
+I(FISTr32     , "fist",  0xDB,            0,             X86II::Void | X86II::Arg32  | X86II::MRMS2m | X86II::OneArgFP , NoIR, NoIR)   // store signed int
+I(FISTPr16    , "fistp", 0xDF,            0,             X86II::Void | X86II::Arg16  | X86II::MRMS3m                   , NoIR, NoIR)   // store short, pop
+I(FISTPr32    , "fistp", 0xDB,            0,             X86II::Void | X86II::Arg32  | X86II::MRMS3m                   , NoIR, NoIR)   // store int, pop
+I(FISTPr64    , "fistp", 0xDF,            0,             X86II::Void | X86II::Arg64  | X86II::MRMS7m | X86II::OneArgFP , NoIR, NoIR)   // store long, pop
 
 
 I(FXCH        , "fxch" , 0xC8,              0, X86II::D9 | X86II::Void | X86II::ArgF80 | X86II::AddRegFrm                , O_ST0, O_ST0) // fxch ST(i), ST(0)
 
 // Floating point constant loads...
-I(FLD0        , "fld0" , 0xEE,              0, X86II::D9 |               X86II::ArgF80 | X86II::RawFrm | X86II::ZeroArgFP, NoIR, NoIR)   // load +0.0
+I(FLD0        , "fldz" , 0xEE,              0, X86II::D9 |               X86II::ArgF80 | X86II::RawFrm | X86II::ZeroArgFP, NoIR, NoIR)   // load +0.0
 I(FLD1        , "fld1" , 0xE8,              0, X86II::D9 |               X86II::ArgF80 | X86II::RawFrm | X86II::ZeroArgFP, NoIR, NoIR)   // load +1.0
 
 // Binary arithmetic operations...
-I(FADDST0r    , "fadd_0",   0xC0,           0, X86II::D8 |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(0) = ST(0) + ST(i)
-I(FADDrST0    , "fadd_i",   0xC0,           0, X86II::DC |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(i) = ST(i) + ST(0)
-I(FADDPrST0   , "faddp_i",  0xC0,           0, X86II::DE |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(i) = ST(i) + ST(0), pop
-
-I(FSUBRST0r   , "fsubr_0" , 0xE8,           0, X86II::D8 |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(0) = ST(i) - ST(0)
-I(FSUBrST0    , "fsub_i"  , 0xE8,           0, X86II::DC |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(i) = ST(i) - ST(0)
-I(FSUBPrST0   , "fsubp_i" , 0xE8,           0, X86II::DE |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(i) = ST(i) - ST(0), pop
-
-I(FSUBST0r    , "fsub_0"  , 0xE0,           0, X86II::D8 |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(0) = ST(0) - ST(i)
-I(FSUBRrST0   , "fsubr_i" , 0xE0,           0, X86II::DC |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(i) = ST(0) - ST(i)
-I(FSUBRPrST0  , "fsubrp_i", 0xE0,           0, X86II::DE |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(i) = ST(0) - ST(i), pop
-
-I(FMULST0r    , "fmul_0",   0xC8,           0, X86II::D8 |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(0) = ST(0) * ST(i)
-I(FMULrST0    , "fmul_i",   0xC8,           0, X86II::DC |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(i) = ST(i) * ST(0)
-I(FMULPrST0   , "fmulp_i",  0xC8,           0, X86II::DE |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(i) = ST(i) * ST(0), pop
-
-I(FDIVRST0r   , "fdivr_0" , 0xF8,           0, X86II::D8 |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(0) = ST(i) / ST(0)
-I(FDIVrST0    , "fdiv_i"  , 0xF8,           0, X86II::DC |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(i) = ST(i) / ST(0)
-I(FDIVPrST0   , "fdivp_i" , 0xF8,           0, X86II::DE |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(i) = ST(i) / ST(0), pop
-
-I(FDIVST0r    , "fdiv_0"  , 0xF0,           0, X86II::D8 |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(0) = ST(0) / ST(i)
-I(FDIVRrST0   , "fdivr_i" , 0xF0,           0, X86II::DC |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(i) = ST(0) / ST(i)
-I(FDIVRPrST0  , "fdivrp_i", 0xF0,           0, X86II::DE |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(i) = ST(0) / ST(i), pop
+I(FADDST0r    , "fadd",   0xC0,           0, X86II::D8 |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(0) = ST(0) + ST(i)
+I(FADDrST0    , "fadd",   0xC0,           0, X86II::DC |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(i) = ST(i) + ST(0)
+I(FADDPrST0   , "faddp",  0xC0,           0, X86II::DE |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(i) = ST(i) + ST(0), pop
+
+I(FSUBRST0r   , "fsubr" , 0xE8,           0, X86II::D8 |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(0) = ST(i) - ST(0)
+I(FSUBrST0    , "fsub"  , 0xE8,           0, X86II::DC |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(i) = ST(i) - ST(0)
+I(FSUBPrST0   , "fsubp" , 0xE8,           0, X86II::DE |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(i) = ST(i) - ST(0), pop
+
+I(FSUBST0r    , "fsub"  , 0xE0,           0, X86II::D8 |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(0) = ST(0) - ST(i)
+I(FSUBRrST0   , "fsubr" , 0xE0,           0, X86II::DC |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(i) = ST(0) - ST(i)
+I(FSUBRPrST0  , "fsubrp", 0xE0,           0, X86II::DE |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(i) = ST(0) - ST(i), pop
+
+I(FMULST0r    , "fmul",   0xC8,           0, X86II::D8 |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(0) = ST(0) * ST(i)
+I(FMULrST0    , "fmul",   0xC8,           0, X86II::DC |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(i) = ST(i) * ST(0)
+I(FMULPrST0   , "fmulp",  0xC8,           0, X86II::DE |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(i) = ST(i) * ST(0), pop
+
+I(FDIVRST0r   , "fdivr" , 0xF8,           0, X86II::D8 |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(0) = ST(i) / ST(0)
+I(FDIVrST0    , "fdiv"  , 0xF8,           0, X86II::DC |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(i) = ST(i) / ST(0)
+I(FDIVPrST0   , "fdivp" , 0xF8,           0, X86II::DE |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(i) = ST(i) / ST(0), pop
+
+I(FDIVST0r    , "fdiv"  , 0xF0,           0, X86II::D8 |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(0) = ST(0) / ST(i)
+I(FDIVRrST0   , "fdivr" , 0xF0,           0, X86II::DC |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(i) = ST(0) / ST(i)
+I(FDIVRPrST0  , "fdivrp", 0xF0,           0, X86II::DE |               X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // ST(i) = ST(0) / ST(i), pop
 
 // Floating point compares
 I(FUCOMr      , "fucom"   , 0xE0,           0, X86II::DD | X86II::Void | X86II::ArgF80 | X86II::AddRegFrm               , NoIR, NoIR)   // FPSW = compare ST(0) with ST(i)


Index: llvm/lib/Target/X86/X86TargetMachine.cpp
diff -u llvm/lib/Target/X86/X86TargetMachine.cpp:1.17 llvm/lib/Target/X86/X86TargetMachine.cpp:1.17.2.1
--- llvm/lib/Target/X86/X86TargetMachine.cpp:1.17	Sat Apr 26 15:11:04 2003
+++ llvm/lib/Target/X86/X86TargetMachine.cpp	Thu Jun 26 16:35:24 2003
@@ -41,6 +41,19 @@
   FrameInfo(TargetFrameInfo::StackGrowsDown, 8/*16 for SSE*/, 4) {
 }
 
+// llc backend for x86
+bool X86TargetMachine::addPassesToEmitAssembly(PassManager &PM,
+					       std::ostream &Out) {
+  PM.add(createLowerSwitchPass());
+  PM.add(createSimpleX86InstructionSelector(*this));
+  PM.add(createLocalRegisterAllocator());
+  PM.add(createX86FloatingPointStackifierPass());
+  PM.add(createPrologEpilogCodeInserter());
+  PM.add(createX86PeepholeOptimizerPass());
+  PM.add(createX86CodePrinterPass(Out));
+  return false; // success!
+}
+
 /// addPassesToJITCompile - Add passes to the specified pass manager to
 /// implement a fast dynamic compiler for this target.  Return true if this is
 /// not supported for this target.


Index: llvm/lib/Target/X86/X86TargetMachine.h
diff -u llvm/lib/Target/X86/X86TargetMachine.h:1.9 llvm/lib/Target/X86/X86TargetMachine.h:1.9.2.1
--- llvm/lib/Target/X86/X86TargetMachine.h:1.9	Sat Dec 28 21:13:05 2002
+++ llvm/lib/Target/X86/X86TargetMachine.h	Thu Jun 26 16:35:24 2003
@@ -42,6 +42,8 @@
   ///
   virtual bool addPassesToEmitMachineCode(PassManager &PM,
                                           MachineCodeEmitter &MCE);
+  
+  virtual bool addPassesToEmitAssembly(PassManager &PM, std::ostream &Out);
 };
 
 #endif




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:42:24 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:42:24 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Scalar/TailDuplication.cpp ADCE.cpp GCSE.cpp InstructionCombining.cpp Mem2Reg.cpp SCCP.cpp
Message-ID: <200306262137.h5QLbc800444@choi.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Scalar:

TailDuplication.cpp added (r1.4.2.1)
ADCE.cpp updated: 1.53 -> 1.53.2.1
GCSE.cpp updated: 1.27 -> 1.27.2.1
InstructionCombining.cpp updated: 1.89 -> 1.89.2.1
Mem2Reg.cpp updated: 1.3 -> 1.3.2.1
SCCP.cpp updated: 1.69 -> 1.69.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/lib/Transforms/Scalar/TailDuplication.cpp
diff -c /dev/null llvm/lib/Transforms/Scalar/TailDuplication.cpp:1.1
*** /dev/null	Thu Jun 26 16:37:38 2003
--- llvm/lib/Transforms/Scalar/TailDuplication.cpp	Sun Jun 22 15:10:28 2003
***************
*** 0 ****
--- 1,324 ----
+ //===- TailDuplication.cpp - Simplify CFG through tail duplication --------===//
+ //
+ // This pass performs a limited form of tail duplication, intended to simplify
+ // CFGs by removing some unconditional branches.  This pass is necessary to
+ // straighten out loops created by the C front-end, but also is capable of
+ // making other code nicer.  After this pass is run, the CFG simplify pass
+ // should be run to clean up the mess.
+ //
+ // This pass could be enhanced in the future to use profile information to be
+ // more aggressive.
+ //
+ //===----------------------------------------------------------------------===//
+ 
+ #include "llvm/Transforms/Scalar.h"
+ #include "llvm/Function.h"
+ #include "llvm/iPHINode.h"
+ #include "llvm/iTerminators.h"
+ #include "llvm/Pass.h"
+ #include "llvm/Type.h"
+ #include "llvm/Support/CFG.h"
+ #include "llvm/Transforms/Utils/Local.h"
+ #include "Support/Statistic.h"
+ 
+ namespace {
+   Statistic<> NumEliminated("tailduplicate",
+                             "Number of unconditional branches eliminated");
+   Statistic<> NumPHINodes("tailduplicate", "Number of phi nodes inserted");
+ 
+   class TailDup : public FunctionPass {
+     bool runOnFunction(Function &F);
+   private:
+     inline bool shouldEliminateUnconditionalBranch(TerminatorInst *TI);
+     inline void eliminateUnconditionalBranch(BranchInst *BI);
+     inline void InsertPHINodesIfNecessary(Instruction *OrigInst, Value *NewInst,
+                                           BasicBlock *NewBlock);
+     inline Value *GetValueInBlock(BasicBlock *BB, Value *OrigVal,
+                                   std::map<BasicBlock*, Value*> &ValueMap,
+                                   std::map<BasicBlock*, Value*> &OutValueMap);
+     inline Value *GetValueOutBlock(BasicBlock *BB, Value *OrigVal,
+                                    std::map<BasicBlock*, Value*> &ValueMap,
+                                    std::map<BasicBlock*, Value*> &OutValueMap);
+   };
+   RegisterOpt<TailDup> X("tailduplicate", "Tail Duplication");
+ }
+ 
+ Pass *createTailDuplicationPass() { return new TailDup(); }
+ 
+ /// runOnFunction - Top level algorithm - Loop over each unconditional branch in
+ /// the function, eliminating it if it looks attractive enough.
+ ///
+ bool TailDup::runOnFunction(Function &F) {
+   bool Changed = false;
+   for (Function::iterator I = F.begin(), E = F.end(); I != E; )
+     if (shouldEliminateUnconditionalBranch(I->getTerminator())) {
+       eliminateUnconditionalBranch(cast<BranchInst>(I->getTerminator()));
+       Changed = true;
+     } else {
+       ++I;
+     }
+   return Changed;
+ }
+ 
+ /// shouldEliminateUnconditionalBranch - Return true if this branch looks
+ /// attractive to eliminate.  We eliminate the branch if the destination basic
+ /// block has <= 5 instructions in it, not counting PHI nodes.  In practice,
+ /// since one of these is a terminator instruction, this means that we will add
+ /// up to 4 instructions to the new block.
+ ///
+ /// We don't count PHI nodes in the count since they will be removed when the
+ /// contents of the block are copied over.
+ ///
+ bool TailDup::shouldEliminateUnconditionalBranch(TerminatorInst *TI) {
+   BranchInst *BI = dyn_cast<BranchInst>(TI);
+   if (!BI || !BI->isUnconditional()) return false;  // Not an uncond branch!
+ 
+   BasicBlock *Dest = BI->getSuccessor(0);
+   if (Dest == BI->getParent()) return false;        // Do not loop infinitely!
+ 
+   // Do not bother working on dead blocks...
+   pred_iterator PI = pred_begin(Dest), PE = pred_end(Dest);
+   if (PI == PE && Dest != Dest->getParent()->begin())
+     return false;   // It's just a dead block, ignore it...
+ 
+   // Also, do not bother with blocks with only a single predecessor: simplify
+   // CFG will fold these two blocks together!
+   ++PI;
+   if (PI == PE) return false;  // Exactly one predecessor!
+ 
+   BasicBlock::iterator I = Dest->begin();
+   while (isa<PHINode>(*I)) ++I;
+ 
+   for (unsigned Size = 0; I != Dest->end(); ++Size, ++I)
+     if (Size == 6) return false;  // The block is too large...
+   return true;  
+ }
+ 
+ 
+ /// eliminateUnconditionalBranch - Clone the instructions from the destination
+ /// block into the source block, eliminating the specified unconditional branch.
+ /// If the destination block defines values used by successors of the dest
+ /// block, we may need to insert PHI nodes.
+ ///
+ void TailDup::eliminateUnconditionalBranch(BranchInst *Branch) {
+   BasicBlock *SourceBlock = Branch->getParent();
+   BasicBlock *DestBlock = Branch->getSuccessor(0);
+   assert(SourceBlock != DestBlock && "Our predicate is broken!");
+ 
+   DEBUG(std::cerr << "TailDuplication[" << SourceBlock->getParent()->getName()
+                   << "]: Eliminating branch: " << *Branch);
+ 
+   // We are going to have to map operands from the original block B to the new
+   // copy of the block B'.  If there are PHI nodes in the DestBlock, these PHI
+   // nodes also define part of this mapping.  Loop over these PHI nodes, adding
+   // them to our mapping.
+   std::map<Value*, Value*> ValueMapping;
+ 
+   BasicBlock::iterator BI = DestBlock->begin();
+   bool HadPHINodes = isa<PHINode>(BI);
+   for (; PHINode *PN = dyn_cast<PHINode>(BI); ++BI)
+     ValueMapping[PN] = PN->getIncomingValueForBlock(SourceBlock);
+ 
+   // Clone the non-phi instructions of the dest block into the source block,
+   // keeping track of the mapping...
+   //
+   for (; BI != DestBlock->end(); ++BI) {
+     Instruction *New = BI->clone();
+     New->setName(BI->getName());
+     SourceBlock->getInstList().push_back(New);
+     ValueMapping[BI] = New;
+   }
+ 
+   // Now that we have built the mapping information and cloned all of the
+   // instructions (giving us a new terminator, among other things), walk the new
+   // instructions, rewriting references of old instructions to use new
+   // instructions.
+   //
+   BI = Branch; ++BI;  // Get an iterator to the first new instruction
+   for (; BI != SourceBlock->end(); ++BI)
+     for (unsigned i = 0, e = BI->getNumOperands(); i != e; ++i)
+       if (Value *Remapped = ValueMapping[BI->getOperand(i)])
+         BI->setOperand(i, Remapped);
+ 
+   // Next we check to see if any of the successors of DestBlock had PHI nodes.
+   // If so, we need to add entries to the PHI nodes for SourceBlock now.
+   for (succ_iterator SI = succ_begin(DestBlock), SE = succ_end(DestBlock);
+        SI != SE; ++SI) {
+     BasicBlock *Succ = *SI;
+     for (BasicBlock::iterator PNI = Succ->begin();
+          PHINode *PN = dyn_cast<PHINode>(PNI); ++PNI) {
+       // Ok, we have a PHI node.  Figure out what the incoming value was for the
+       // DestBlock.
+       Value *IV = PN->getIncomingValueForBlock(DestBlock);
+       
+       // Remap the value if necessary...
+       if (Value *MappedIV = ValueMapping[IV])
+         IV = MappedIV;
+       PN->addIncoming(IV, SourceBlock);
+     }
+   }
+   
+   // Now that all of the instructions are correctly copied into the SourceBlock,
+   // we have one more minor problem: the successors of the original DestBB may
+   // use the values computed in DestBB either directly (if DestBB dominated the
+   // block), or through a PHI node.  In either case, we need to insert PHI nodes
+   // into any successors of DestBB (which are now our successors) for each value
+   // that is computed in DestBB, but is used outside of it.  All of these uses
+   // we have to rewrite with the new PHI node.
+   //
+   if (succ_begin(SourceBlock) != succ_end(SourceBlock)) // Avoid wasting time...
+     for (BI = DestBlock->begin(); BI != DestBlock->end(); ++BI)
+       if (BI->getType() != Type::VoidTy)
+         InsertPHINodesIfNecessary(BI, ValueMapping[BI], SourceBlock);
+ 
+   // Final step: now that we have finished everything up, walk the cloned
+   // instructions one last time, constant propagating and DCE'ing them, because
+   // they may not be needed anymore.
+   //
+   BI = Branch; ++BI;  // Get an iterator to the first new instruction
+   if (HadPHINodes)
+     while (BI != SourceBlock->end())
+       if (!dceInstruction(BI) && !doConstantPropagation(BI))
+         ++BI;
+ 
+   DestBlock->removePredecessor(SourceBlock); // Remove entries in PHI nodes...
+   SourceBlock->getInstList().erase(Branch);  // Destroy the uncond branch...
+   
+   ++NumEliminated;  // We just killed a branch!
+ }
+ 
+ /// InsertPHINodesIfNecessary - So at this point, we cloned the OrigInst
+ /// instruction into the NewBlock with the value of NewInst.  If OrigInst was
+ /// used outside of its defining basic block, we need to insert a PHI nodes into
+ /// the successors.
+ ///
+ void TailDup::InsertPHINodesIfNecessary(Instruction *OrigInst, Value *NewInst,
+                                         BasicBlock *NewBlock) {
+   // Loop over all of the uses of OrigInst, rewriting them to be newly inserted
+   // PHI nodes, unless they are in the same basic block as OrigInst.
+   BasicBlock *OrigBlock = OrigInst->getParent();
+   std::vector<Instruction*> Users;
+   Users.reserve(OrigInst->use_size());
+   for (Value::use_iterator I = OrigInst->use_begin(), E = OrigInst->use_end();
+        I != E; ++I) {
+     Instruction *In = cast<Instruction>(*I);
+     if (In->getParent() != OrigBlock)  // Don't modify uses in the orig block!
+       Users.push_back(In);
+   }
+ 
+   // The common case is that the instruction is only used within the block that
+   // defines it.  If we have this case, quick exit.
+   //
+   if (Users.empty()) return; 
+ 
+   // Otherwise, we have a more complex case, handle it now.  This requires the
+   // construction of a mapping between a basic block and the value to use when
+   // in the scope of that basic block.  This map will map to the original and
+   // new values when in the original or new block, but will map to inserted PHI
+   // nodes when in other blocks.
+   //
+   std::map<BasicBlock*, Value*> ValueMap;
+   std::map<BasicBlock*, Value*> OutValueMap;   // The outgoing value map
+   OutValueMap[OrigBlock] = OrigInst;
+   OutValueMap[NewBlock ] = NewInst;    // Seed the initial values...
+ 
+   DEBUG(std::cerr << "  ** Inserting PHI nodes for " << OrigInst);
+   while (!Users.empty()) {
+     Instruction *User = Users.back(); Users.pop_back();
+ 
+     if (PHINode *PN = dyn_cast<PHINode>(User)) {
+       // PHI nodes must be handled specially here, because their operands are
+       // actually defined in predecessor basic blocks, NOT in the block that the
+       // PHI node lives in.  Note that we have already added entries to PHI nods
+       // which are in blocks that are immediate successors of OrigBlock, so
+       // don't modify them again.
+       for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i)
+         if (PN->getIncomingValue(i) == OrigInst &&
+             PN->getIncomingBlock(i) != OrigBlock) {
+           Value *V = GetValueOutBlock(PN->getIncomingBlock(i), OrigInst,
+                                       ValueMap, OutValueMap);
+           PN->setIncomingValue(i, V);
+         }
+       
+     } else {
+       // Any other user of the instruction can just replace any uses with the
+       // new value defined in the block it resides in.
+       Value *V = GetValueInBlock(User->getParent(), OrigInst, ValueMap,
+                                  OutValueMap);
+       User->replaceUsesOfWith(OrigInst, V);
+     }
+   }
+ }
+ 
+ /// GetValueInBlock - This is a recursive method which inserts PHI nodes into
+ /// the function until there is a value available in basic block BB.
+ ///
+ Value *TailDup::GetValueInBlock(BasicBlock *BB, Value *OrigVal,
+                                 std::map<BasicBlock*, Value*> &ValueMap,
+                                 std::map<BasicBlock*, Value*> &OutValueMap) {
+   Value*& BBVal = ValueMap[BB];
+   if (BBVal) return BBVal;       // Value already computed for this block?
+ 
+   assert(pred_begin(BB) != pred_end(BB) &&
+          "Propagating PHI nodes to unreachable blocks?");
+ 
+   // If there is no value already available in this basic block, we need to
+   // either reuse a value from an incoming, dominating, basic block, or we need
+   // to create a new PHI node to merge in different incoming values.  Because we
+   // don't know if we're part of a loop at this point or not, we create a PHI
+   // node, even if we will ultimately eliminate it.
+   PHINode *PN = new PHINode(OrigVal->getType(), OrigVal->getName()+".pn",
+                             BB->begin());
+   BBVal = PN;   // Insert this into the BBVal slot in case of cycles...
+ 
+   Value*& BBOutVal = OutValueMap[BB];
+   if (BBOutVal == 0) BBOutVal = PN;
+ 
+   // Now that we have created the PHI node, loop over all of the predecessors of
+   // this block, computing an incoming value for the predecessor.
+   std::vector<BasicBlock*> Preds(pred_begin(BB), pred_end(BB));
+   for (unsigned i = 0, e = Preds.size(); i != e; ++i)
+     PN->addIncoming(GetValueOutBlock(Preds[i], OrigVal, ValueMap, OutValueMap),
+                     Preds[i]);
+ 
+   // The PHI node is complete.  In many cases, however the PHI node was
+   // ultimately unnecessary: we could have just reused a dominating incoming
+   // value.  If this is the case, nuke the PHI node and replace the map entry
+   // with the dominating value.
+   //
+   assert(PN->getNumIncomingValues() > 0 && "No predecessors?");
+ 
+   // Check to see if all of the elements in the PHI node are either the PHI node
+   // itself or ONE particular value.
+   unsigned i = 0;
+   Value *ReplVal = PN->getIncomingValue(i);
+   for (; ReplVal == PN && i != PN->getNumIncomingValues(); ++i)
+     ReplVal = PN->getIncomingValue(i);  // Skip values equal to the PN
+ 
+   for (; i != PN->getNumIncomingValues(); ++i)
+     if (PN->getIncomingValue(i) != PN && PN->getIncomingValue(i) != ReplVal) {
+       ReplVal = 0;
+       break;
+     }
+ 
+   // Found a value to replace the PHI node with?
+   if (ReplVal) {
+     PN->replaceAllUsesWith(ReplVal);
+     BBVal = ReplVal;
+     if (BBOutVal == PN) BBOutVal = ReplVal;
+     BB->getInstList().erase(PN);   // Erase the PHI node...
+   } else {
+     ++NumPHINodes;
+   }
+ 
+   return BBVal;
+ }
+ 
+ Value *TailDup::GetValueOutBlock(BasicBlock *BB, Value *OrigVal,
+                                  std::map<BasicBlock*, Value*> &ValueMap,
+                                  std::map<BasicBlock*, Value*> &OutValueMap) {
+   Value*& BBVal = OutValueMap[BB];
+   if (BBVal) return BBVal;       // Value already computed for this block?
+ 
+   return BBVal = GetValueInBlock(BB, OrigVal, ValueMap, OutValueMap);
+ }


Index: llvm/lib/Transforms/Scalar/ADCE.cpp
diff -u llvm/lib/Transforms/Scalar/ADCE.cpp:1.53 llvm/lib/Transforms/Scalar/ADCE.cpp:1.53.2.1
--- llvm/lib/Transforms/Scalar/ADCE.cpp:1.53	Thu May 22 17:00:06 2003
+++ llvm/lib/Transforms/Scalar/ADCE.cpp	Thu Jun 26 16:35:28 2003
@@ -73,6 +73,8 @@
   // instructions that are dead according to LiveSet.
   bool dropReferencesOfDeadInstructionsInLiveBlock(BasicBlock *BB);
 
+  TerminatorInst *convertToUnconditionalBranch(TerminatorInst *TI);
+
   inline void markInstructionLive(Instruction *I) {
     if (LiveSet.count(I)) return;
     DEBUG(std::cerr << "Insn Live: " << I);
@@ -105,8 +107,11 @@
              bind_obj(this, &ADCE::markTerminatorLive));
   }
   
-  // If this basic block is live, then the terminator must be as well!
-  markTerminatorLive(BB);
+  // If this basic block is live, and it ends in an unconditional branch, then
+  // the branch is alive as well...
+  if (BranchInst *BI = dyn_cast<BranchInst>(BB->getTerminator()))
+    if (BI->isUnconditional())
+      markTerminatorLive(BB);
 }
 
 // dropReferencesOfDeadInstructionsInLiveBlock - Loop over all of the
@@ -136,6 +141,24 @@
 }
 
 
+/// convertToUnconditionalBranch - Transform this conditional terminator
+/// instruction into an unconditional branch because we don't care which of the
+/// successors it goes to.  This eliminate a use of the condition as well.
+///
+TerminatorInst *ADCE::convertToUnconditionalBranch(TerminatorInst *TI) {
+  BranchInst *NB = new BranchInst(TI->getSuccessor(0), TI);
+  BasicBlock *BB = TI->getParent();
+
+  // Remove entries from PHI nodes to avoid confusing ourself later...
+  for (unsigned i = 1, e = TI->getNumSuccessors(); i != e; ++i)
+    TI->getSuccessor(i)->removePredecessor(BB);
+  
+  // Delete the old branch itself...
+  BB->getInstList().erase(TI);
+  return NB;
+}
+
+
 // doADCE() - Run the Aggressive Dead Code Elimination algorithm, returning
 // true if the function was modified.
 //
@@ -166,6 +189,16 @@
     }
   }
 
+  // Check to ensure we have an exit node for this CFG.  If we don't, we won't
+  // have any post-dominance information, thus we cannot perform our
+  // transformations safely.
+  //
+  PostDominatorTree &DT = getAnalysis<PostDominatorTree>();
+  if (DT[&Func->getEntryNode()] == 0) {
+    WorkList.clear();
+    return MadeChanges;
+  }
+
   DEBUG(std::cerr << "Processing work list\n");
 
   // AliveBlocks - Set of basic blocks that we know have instructions that are
@@ -208,27 +241,38 @@
 
   DEBUG(
     std::cerr << "Current Function: X = Live\n";
-    for (Function::iterator I = Func->begin(), E = Func->end(); I != E; ++I)
+    for (Function::iterator I = Func->begin(), E = Func->end(); I != E; ++I){
+      std::cerr << I->getName() << ":\t"
+                << (AliveBlocks.count(I) ? "LIVE\n" : "DEAD\n");
       for (BasicBlock::iterator BI = I->begin(), BE = I->end(); BI != BE; ++BI){
         if (LiveSet.count(BI)) std::cerr << "X ";
         std::cerr << *BI;
       }
-    );
+    });
 
   // Find the first postdominator of the entry node that is alive.  Make it the
   // new entry node...
   //
-  PostDominatorTree &DT = getAnalysis<PostDominatorTree>();
-
-
   if (AliveBlocks.size() == Func->size()) {  // No dead blocks?
-    for (Function::iterator I = Func->begin(), E = Func->end(); I != E; ++I)
+    for (Function::iterator I = Func->begin(), E = Func->end(); I != E; ++I) {
       // Loop over all of the instructions in the function, telling dead
       // instructions to drop their references.  This is so that the next sweep
       // over the program can safely delete dead instructions without other dead
       // instructions still refering to them.
       //
       dropReferencesOfDeadInstructionsInLiveBlock(I);
+
+      // Check to make sure the terminator instruction is live.  If it isn't,
+      // this means that the condition that it branches on (we know it is not an
+      // unconditional branch), is not needed to make the decision of where to
+      // go to, because all outgoing edges go to the same place.  We must remove
+      // the use of the condition (because it's probably dead), so we convert
+      // the terminator to a conditional branch.
+      //
+      TerminatorInst *TI = I->getTerminator();
+      if (!LiveSet.count(TI))
+        convertToUnconditionalBranch(TI);
+    }
     
   } else {                                   // If there are some blocks dead...
     // If the entry node is dead, insert a new entry node to eliminate the entry
@@ -239,6 +283,7 @@
       NewEntry->getInstList().push_back(new BranchInst(&Func->front()));
       Func->getBasicBlockList().push_front(NewEntry);
       AliveBlocks.insert(NewEntry);    // This block is always alive!
+      LiveSet.insert(NewEntry->getTerminator());  // The branch is live
     }
     
     // Loop over all of the alive blocks in the function.  If any successor
@@ -251,6 +296,14 @@
         BasicBlock *BB = I;
         TerminatorInst *TI = BB->getTerminator();
       
+        // If the terminator instruction is alive, but the block it is contained
+        // in IS alive, this means that this terminator is a conditional branch
+        // on a condition that doesn't matter.  Make it an unconditional branch
+        // to ONE of the successors.  This has the side effect of dropping a use
+        // of the conditional value, which may also be dead.
+        if (!LiveSet.count(TI))
+          TI = convertToUnconditionalBranch(TI);
+
         // Loop over all of the successors, looking for ones that are not alive.
         // We cannot save the number of successors in the terminator instruction
         // here because we may remove them if we don't have a postdominator...


Index: llvm/lib/Transforms/Scalar/GCSE.cpp
diff -u llvm/lib/Transforms/Scalar/GCSE.cpp:1.27 llvm/lib/Transforms/Scalar/GCSE.cpp:1.27.2.1
--- llvm/lib/Transforms/Scalar/GCSE.cpp:1.27	Fri Jan 31 22:50:59 2003
+++ llvm/lib/Transforms/Scalar/GCSE.cpp	Thu Jun 26 16:35:28 2003
@@ -120,6 +120,7 @@
 
           // Erase the instruction from the program.
           I->getParent()->getInstList().erase(I);
+          WorkList.erase(I);
         }
       
       return true;


Index: llvm/lib/Transforms/Scalar/InstructionCombining.cpp
diff -u llvm/lib/Transforms/Scalar/InstructionCombining.cpp:1.89 llvm/lib/Transforms/Scalar/InstructionCombining.cpp:1.89.2.1
--- llvm/lib/Transforms/Scalar/InstructionCombining.cpp:1.89	Thu Jun  5 15:12:51 2003
+++ llvm/lib/Transforms/Scalar/InstructionCombining.cpp	Thu Jun 26 16:35:28 2003
@@ -22,8 +22,10 @@
 #include "llvm/Constants.h"
 #include "llvm/ConstantHandling.h"
 #include "llvm/DerivedTypes.h"
+#include "llvm/GlobalVariable.h"
 #include "llvm/Support/InstIterator.h"
 #include "llvm/Support/InstVisitor.h"
+#include "llvm/Support/CallSite.h"
 #include "Support/Statistic.h"
 #include <algorithm>
 
@@ -73,14 +75,20 @@
     Instruction *visitSetCondInst(BinaryOperator &I);
     Instruction *visitShiftInst(ShiftInst &I);
     Instruction *visitCastInst(CastInst &CI);
+    Instruction *visitCallInst(CallInst &CI);
+    Instruction *visitInvokeInst(InvokeInst &II);
     Instruction *visitPHINode(PHINode &PN);
     Instruction *visitGetElementPtrInst(GetElementPtrInst &GEP);
     Instruction *visitAllocationInst(AllocationInst &AI);
+    Instruction *visitLoadInst(LoadInst &LI);
     Instruction *visitBranchInst(BranchInst &BI);
 
     // visitInstruction - Specify what to return for unhandled instructions...
     Instruction *visitInstruction(Instruction &I) { return 0; }
 
+  private:
+    bool transformConstExprCastCall(CallSite CS);
+
     // InsertNewInstBefore - insert an instruction New before instruction Old
     // in the program.  Add the new instruction to the worklist.
     //
@@ -107,7 +115,6 @@
     // SimplifyCommutative - This performs a few simplifications for commutative
     // operators...
     bool SimplifyCommutative(BinaryOperator &I);
-
   };
 
   RegisterOpt<InstCombiner> X("instcombine", "Combine redundant instructions");
@@ -356,10 +363,11 @@
   if (Constant *Op1 = dyn_cast<Constant>(I.getOperand(1))) {
     if (ConstantInt *CI = dyn_cast<ConstantInt>(Op1)) {
       const Type *Ty = CI->getType();
-      uint64_t Val = Ty->isSigned() ?
-                          (uint64_t)cast<ConstantSInt>(CI)->getValue() : 
-                                    cast<ConstantUInt>(CI)->getValue();
+      int64_t Val = Ty->isSigned() ?        cast<ConstantSInt>(CI)->getValue() :
+                                   (int64_t)cast<ConstantUInt>(CI)->getValue();
       switch (Val) {
+      case -1:                               // X * -1 -> -X
+        return BinaryOperator::createNeg(Op0, I.getName());
       case 0:
         return ReplaceInstUsesWith(I, Op1);  // Eliminate 'mul double %X, 0'
       case 1:
@@ -893,15 +901,17 @@
 // CastInst simplification
 //
 Instruction *InstCombiner::visitCastInst(CastInst &CI) {
+  Value *Src = CI.getOperand(0);
+
   // If the user is casting a value to the same type, eliminate this cast
   // instruction...
-  if (CI.getType() == CI.getOperand(0)->getType())
-    return ReplaceInstUsesWith(CI, CI.getOperand(0));
+  if (CI.getType() == Src->getType())
+    return ReplaceInstUsesWith(CI, Src);
 
   // If casting the result of another cast instruction, try to eliminate this
   // one!
   //
-  if (CastInst *CSrc = dyn_cast<CastInst>(CI.getOperand(0))) {
+  if (CastInst *CSrc = dyn_cast<CastInst>(Src)) {
     if (isEliminableCastOfCast(CI, CSrc)) {
       // This instruction now refers directly to the cast's src operand.  This
       // has a good chance of making CSrc dead.
@@ -925,9 +935,192 @@
     }
   }
 
+  // If casting the result of a getelementptr instruction with no offset, turn
+  // this into a cast of the original pointer!
+  //
+  if (GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(Src)) {
+    bool AllZeroOperands = true;
+    for (unsigned i = 1, e = GEP->getNumOperands(); i != e; ++i)
+      if (!isa<Constant>(GEP->getOperand(i)) ||
+          !cast<Constant>(GEP->getOperand(i))->isNullValue()) {
+        AllZeroOperands = false;
+        break;
+      }
+    if (AllZeroOperands) {
+      CI.setOperand(0, GEP->getOperand(0));
+      return &CI;
+    }
+  }
+
+  // If this is a cast to bool (which is effectively a "!=0" test), then we can
+  // perform a few optimizations...
+  //
+  if (CI.getType() == Type::BoolTy) {
+    if (BinaryOperator *BO = dyn_cast<BinaryOperator>(Src)) {
+      Value *Op0 = BO->getOperand(0), *Op1 = BO->getOperand(1);
+
+      // Replace (cast (sub A, B) to bool) with (setne A, B)
+      if (BO->getOpcode() == Instruction::Sub)
+        return new SetCondInst(Instruction::SetNE, Op0, Op1);
+
+      // Replace (cast (add A, B) to bool) with (setne A, -B) if B is
+      // efficiently invertible, or if the add has just this one use.
+      if (BO->getOpcode() == Instruction::Add)
+        if (Value *NegVal = dyn_castNegVal(Op1))
+          return new SetCondInst(Instruction::SetNE, Op0, NegVal);
+        else if (Value *NegVal = dyn_castNegVal(Op0))
+          return new SetCondInst(Instruction::SetNE, NegVal, Op1);
+        else if (BO->use_size() == 1) {
+          Instruction *Neg = BinaryOperator::createNeg(Op1, BO->getName());
+          BO->setName("");
+          InsertNewInstBefore(Neg, CI);
+          return new SetCondInst(Instruction::SetNE, Op0, Neg);
+        }
+    }
+  }
+
   return 0;
 }
 
+// CallInst simplification
+//
+Instruction *InstCombiner::visitCallInst(CallInst &CI) {
+  if (transformConstExprCastCall(&CI)) return 0;
+  return 0;
+}
+
+// InvokeInst simplification
+//
+Instruction *InstCombiner::visitInvokeInst(InvokeInst &II) {
+  if (transformConstExprCastCall(&II)) return 0;
+  return 0;
+}
+
+// getPromotedType - Return the specified type promoted as it would be to pass
+// though a va_arg area...
+static const Type *getPromotedType(const Type *Ty) {
+  switch (Ty->getPrimitiveID()) {
+  case Type::SByteTyID:
+  case Type::ShortTyID:  return Type::IntTy;
+  case Type::UByteTyID:
+  case Type::UShortTyID: return Type::UIntTy;
+  case Type::FloatTyID:  return Type::DoubleTy;
+  default:               return Ty;
+  }
+}
+
+// transformConstExprCastCall - If the callee is a constexpr cast of a function,
+// attempt to move the cast to the arguments of the call/invoke.
+//
+bool InstCombiner::transformConstExprCastCall(CallSite CS) {
+  if (!isa<ConstantExpr>(CS.getCalledValue())) return false;
+  ConstantExpr *CE = cast<ConstantExpr>(CS.getCalledValue());
+  if (CE->getOpcode() != Instruction::Cast ||
+      !isa<ConstantPointerRef>(CE->getOperand(0)))
+    return false;
+  ConstantPointerRef *CPR = cast<ConstantPointerRef>(CE->getOperand(0));
+  if (!isa<Function>(CPR->getValue())) return false;
+  Function *Callee = cast<Function>(CPR->getValue());
+  Instruction *Caller = CS.getInstruction();
+
+  // Okay, this is a cast from a function to a different type.  Unless doing so
+  // would cause a type conversion of one of our arguments, change this call to
+  // be a direct call with arguments casted to the appropriate types.
+  //
+  const FunctionType *FT = Callee->getFunctionType();
+  const Type *OldRetTy = Caller->getType();
+
+  if (Callee->isExternal() &&
+      !OldRetTy->isLosslesslyConvertibleTo(FT->getReturnType()))
+    return false;   // Cannot transform this return value...
+
+  unsigned NumActualArgs = unsigned(CS.arg_end()-CS.arg_begin());
+  unsigned NumCommonArgs = std::min(FT->getNumParams(), NumActualArgs);
+                                    
+  CallSite::arg_iterator AI = CS.arg_begin();
+  for (unsigned i = 0, e = NumCommonArgs; i != e; ++i, ++AI) {
+    const Type *ParamTy = FT->getParamType(i);
+    bool isConvertible = (*AI)->getType()->isLosslesslyConvertibleTo(ParamTy);
+    if (Callee->isExternal() && !isConvertible) return false;    
+  }
+
+  if (FT->getNumParams() < NumActualArgs && !FT->isVarArg() &&
+      Callee->isExternal())
+    return false;   // Do not delete arguments unless we have a function body...
+
+  // Okay, we decided that this is a safe thing to do: go ahead and start
+  // inserting cast instructions as necessary...
+  std::vector<Value*> Args;
+  Args.reserve(NumActualArgs);
+
+  AI = CS.arg_begin();
+  for (unsigned i = 0; i != NumCommonArgs; ++i, ++AI) {
+    const Type *ParamTy = FT->getParamType(i);
+    if ((*AI)->getType() == ParamTy) {
+      Args.push_back(*AI);
+    } else {
+      Instruction *Cast = new CastInst(*AI, ParamTy, "tmp");
+      InsertNewInstBefore(Cast, *Caller);
+      Args.push_back(Cast);
+    }
+  }
+
+  // If the function takes more arguments than the call was taking, add them
+  // now...
+  for (unsigned i = NumCommonArgs; i != FT->getNumParams(); ++i)
+    Args.push_back(Constant::getNullValue(FT->getParamType(i)));
+
+  // If we are removing arguments to the function, emit an obnoxious warning...
+  if (FT->getNumParams() < NumActualArgs)
+    if (!FT->isVarArg()) {
+      std::cerr << "WARNING: While resolving call to function '"
+                << Callee->getName() << "' arguments were dropped!\n";
+    } else {
+      // Add all of the arguments in their promoted form to the arg list...
+      for (unsigned i = FT->getNumParams(); i != NumActualArgs; ++i, ++AI) {
+        const Type *PTy = getPromotedType((*AI)->getType());
+        if (PTy != (*AI)->getType()) {
+          // Must promote to pass through va_arg area!
+          Instruction *Cast = new CastInst(*AI, PTy, "tmp");
+          InsertNewInstBefore(Cast, *Caller);
+          Args.push_back(Cast);
+        } else {
+          Args.push_back(*AI);
+        }
+      }
+    }
+
+  if (FT->getReturnType() == Type::VoidTy)
+    Caller->setName("");   // Void type should not have a name...
+
+  Instruction *NC;
+  if (InvokeInst *II = dyn_cast<InvokeInst>(Caller)) {
+    NC = new InvokeInst(Callee, II->getNormalDest(), II->getExceptionalDest(),
+                        Args, Caller->getName(), Caller);
+  } else {
+    NC = new CallInst(Callee, Args, Caller->getName(), Caller);
+  }
+
+  // Insert a cast of the return type as necessary...
+  Value *NV = NC;
+  if (Caller->getType() != NV->getType() && !Caller->use_empty()) {
+    if (NV->getType() != Type::VoidTy) {
+      NV = NC = new CastInst(NC, Caller->getType(), "tmp");
+      InsertNewInstBefore(NC, *Caller);
+      AddUsesToWorkList(*Caller);
+    } else {
+      NV = Constant::getNullValue(Caller->getType());
+    }
+  }
+
+  if (Caller->getType() != Type::VoidTy && !Caller->use_empty())
+    Caller->replaceAllUsesWith(NV);
+  Caller->getParent()->getInstList().erase(Caller);
+  removeFromWorkList(Caller);
+  return true;
+}
+
+
 
 // PHINode simplification
 //
@@ -1069,6 +1262,52 @@
     }
   return 0;
 }
+
+/// GetGEPGlobalInitializer - Given a constant, and a getelementptr
+/// constantexpr, return the constant value being addressed by the constant
+/// expression, or null if something is funny.
+///
+static Constant *GetGEPGlobalInitializer(Constant *C, ConstantExpr *CE) {
+  if (CE->getOperand(1) != Constant::getNullValue(Type::LongTy))
+    return 0;  // Do not allow stepping over the value!
+
+  // Loop over all of the operands, tracking down which value we are
+  // addressing...
+  for (unsigned i = 2, e = CE->getNumOperands(); i != e; ++i)
+    if (ConstantUInt *CU = dyn_cast<ConstantUInt>(CE->getOperand(i))) {
+      ConstantStruct *CS = cast<ConstantStruct>(C);
+      if (CU->getValue() >= CS->getValues().size()) return 0;
+      C = cast<Constant>(CS->getValues()[CU->getValue()]);
+    } else if (ConstantSInt *CS = dyn_cast<ConstantSInt>(CE->getOperand(i))) {
+      ConstantArray *CA = cast<ConstantArray>(C);
+      if ((uint64_t)CS->getValue() >= CA->getValues().size()) return 0;
+      C = cast<Constant>(CA->getValues()[CS->getValue()]);
+    } else 
+      return 0;
+  return C;
+}
+
+Instruction *InstCombiner::visitLoadInst(LoadInst &LI) {
+  Value *Op = LI.getOperand(0);
+  if (ConstantPointerRef *CPR = dyn_cast<ConstantPointerRef>(Op))
+    Op = CPR->getValue();
+
+  // Instcombine load (constant global) into the value loaded...
+  if (GlobalVariable *GV = dyn_cast<GlobalVariable>(Op))
+    if (GV->isConstant())
+      return ReplaceInstUsesWith(LI, GV->getInitializer());
+
+  // Instcombine load (constantexpr_GEP global, 0, ...) into the value loaded...
+  if (ConstantExpr *CE = dyn_cast<ConstantExpr>(Op))
+    if (CE->getOpcode() == Instruction::GetElementPtr)
+      if (ConstantPointerRef *G=dyn_cast<ConstantPointerRef>(CE->getOperand(0)))
+        if (GlobalVariable *GV = dyn_cast<GlobalVariable>(G->getValue()))
+          if (GV->isConstant())
+            if (Constant *V = GetGEPGlobalInitializer(GV->getInitializer(), CE))
+              return ReplaceInstUsesWith(LI, V);
+  return 0;
+}
+
 
 Instruction *InstCombiner::visitBranchInst(BranchInst &BI) {
   // Change br (not X), label True, label False to: br X, label False, True


Index: llvm/lib/Transforms/Scalar/Mem2Reg.cpp
diff -u llvm/lib/Transforms/Scalar/Mem2Reg.cpp:1.3 llvm/lib/Transforms/Scalar/Mem2Reg.cpp:1.3.2.1
--- llvm/lib/Transforms/Scalar/Mem2Reg.cpp:1.3	Wed Apr 23 11:37:42 2003
+++ llvm/lib/Transforms/Scalar/Mem2Reg.cpp	Thu Jun 26 16:35:28 2003
@@ -40,19 +40,26 @@
 
   BasicBlock &BB = F.getEntryNode();  // Get the entry node for the function
 
-  // Find allocas that are safe to promote, by looking at all instructions in
-  // the entry node
-  for (BasicBlock::iterator I = BB.begin(), E = --BB.end(); I != E; ++I)
-    if (AllocaInst *AI = dyn_cast<AllocaInst>(I))       // Is it an alloca?
-      if (isAllocaPromotable(AI, TD))
-        Allocas.push_back(AI);
+  bool Changed  = false;
+  
+  while (1) {
+    Allocas.clear();
+
+    // Find allocas that are safe to promote, by looking at all instructions in
+    // the entry node
+    for (BasicBlock::iterator I = BB.begin(), E = --BB.end(); I != E; ++I)
+      if (AllocaInst *AI = dyn_cast<AllocaInst>(I))       // Is it an alloca?
+        if (isAllocaPromotable(AI, TD))
+          Allocas.push_back(AI);
+
+    if (Allocas.empty()) break;
 
-  if (!Allocas.empty()) {
     PromoteMemToReg(Allocas, getAnalysis<DominanceFrontier>(), TD);
     NumPromoted += Allocas.size();
-    return true;
+    Changed = true;
   }
-  return false;
+
+  return Changed;
 }
 
 // createPromoteMemoryToRegister - Provide an entry point to create this pass.


Index: llvm/lib/Transforms/Scalar/SCCP.cpp
diff -u llvm/lib/Transforms/Scalar/SCCP.cpp:1.69 llvm/lib/Transforms/Scalar/SCCP.cpp:1.69.2.1
--- llvm/lib/Transforms/Scalar/SCCP.cpp:1.69	Tue May 20 16:01:16 2003
+++ llvm/lib/Transforms/Scalar/SCCP.cpp	Thu Jun 26 16:35:28 2003
@@ -420,12 +420,13 @@
   for (unsigned i = 0, e = PN.getNumIncomingValues(); i != e; ++i) {
     InstVal &IV = getValueState(PN.getIncomingValue(i));
     if (IV.isUndefined()) continue;  // Doesn't influence PHI node.
-    if (IV.isOverdefined()) {   // PHI node becomes overdefined!
-      markOverdefined(&PN);
-      return;
-    }
     
     if (isEdgeFeasible(PN.getIncomingBlock(i), PN.getParent())) {
+      if (IV.isOverdefined()) {   // PHI node becomes overdefined!
+        markOverdefined(&PN);
+        return;
+      }
+
       if (OperandVal == 0) {   // Grab the first value...
         OperandVal = IV.getConstant();
       } else {                // Another value is being merged in!




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:42:32 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:42:32 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Transforms/IPO.h Scalar.h
Message-ID: <200306262137.h5QLbZ400400@choi.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Transforms:

IPO.h updated: 1.16 -> 1.16.2.1
Scalar.h updated: 1.17 -> 1.17.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.
Kept includes of "Config/assert.h" to avoid implicit dependencies on header
files.


---
Diffs of the changes:

Index: llvm/include/llvm/Transforms/IPO.h
diff -u llvm/include/llvm/Transforms/IPO.h:1.16 llvm/include/llvm/Transforms/IPO.h:1.16.2.1
--- llvm/include/llvm/Transforms/IPO.h:1.16	Mon Feb  3 13:08:33 2003
+++ llvm/include/llvm/Transforms/IPO.h	Thu Jun 26 16:34:55 2003
@@ -68,6 +68,15 @@
 //
 Pass *createInternalizePass();
 
+//===----------------------------------------------------------------------===//
+// createDeadArgEliminationPass - This pass removes arguments from functions
+// which are not used by the body of the function.  If
+// DeleteFromExternalFunctions is true, the pass will modify functions that have
+// external linkage, which is not usually safe (this is used by bugpoint to
+// reduce testcases).
+//
+Pass *createDeadArgEliminationPass(bool DeleteFromExternalFunctions=false);
+
 
 //===----------------------------------------------------------------------===//
 // These passes are wrappers that can do a few simple structure mutation


Index: llvm/include/llvm/Transforms/Scalar.h
diff -u llvm/include/llvm/Transforms/Scalar.h:1.17 llvm/include/llvm/Transforms/Scalar.h:1.17.2.1
--- llvm/include/llvm/Transforms/Scalar.h:1.17	Tue May 27 10:52:45 2003
+++ llvm/include/llvm/Transforms/Scalar.h	Thu Jun 26 16:34:55 2003
@@ -177,10 +177,19 @@
 
 //===----------------------------------------------------------------------===//
 //
+// TailDuplication - Eliminate unconditional branches through controlled code
+// duplication, creating simpler CFG structures.
+//
+Pass *createTailDuplicationPass();
+
+
+//===----------------------------------------------------------------------===//
+//
 // CFG Simplification - Merge basic blocks, eliminate unreachable blocks,
 // simplify terminator instructions, etc...
 //
 Pass *createCFGSimplificationPass();
+
 
 //===----------------------------------------------------------------------===//
 //




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:42:40 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:42:40 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/ExprTypeConvert.cpp TransformInternals.cpp
Message-ID: <200306262137.h5QLba500408@choi.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms:

ExprTypeConvert.cpp updated: 1.73 -> 1.73.2.1
TransformInternals.cpp updated: 1.35 -> 1.35.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/lib/Transforms/ExprTypeConvert.cpp
diff -u llvm/lib/Transforms/ExprTypeConvert.cpp:1.73 llvm/lib/Transforms/ExprTypeConvert.cpp:1.73.2.1
--- llvm/lib/Transforms/ExprTypeConvert.cpp:1.73	Thu May 29 10:12:27 2003
+++ llvm/lib/Transforms/ExprTypeConvert.cpp	Thu Jun 26 16:35:25 2003
@@ -49,6 +49,7 @@
 
   // Get information about the base datatype being allocated, before & after
   int ReqTypeSize = TD.getTypeSize(Ty);
+  if (ReqTypeSize == 0) return false;
   unsigned OldTypeSize = TD.getTypeSize(MI->getType()->getElementType());
 
   // Must have a scale or offset to analyze it...


Index: llvm/lib/Transforms/TransformInternals.cpp
diff -u llvm/lib/Transforms/TransformInternals.cpp:1.35 llvm/lib/Transforms/TransformInternals.cpp:1.35.2.1
--- llvm/lib/Transforms/TransformInternals.cpp:1.35	Sat Jun  7 16:45:42 2003
+++ llvm/lib/Transforms/TransformInternals.cpp	Thu Jun 26 16:35:25 2003
@@ -119,6 +119,7 @@
       if (!ElTy->isSized() || (isa<PointerType>(CompTy) && !Indices.empty()))
         return 0; // Type is unreasonable... escape!
       unsigned ElSize = TD.getTypeSize(ElTy);
+      if (ElSize == 0) return 0;   // Avoid division by zero...
       int64_t ElSizeS = ElSize;
 
       // See if the user is indexing into a different cell of this array...




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:42:48 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:42:48 2003
Subject: [llvm-commits] CVS: llvm/tools/bugpoint/CrashDebugger.cpp ExtractFunction.cpp Makefile SystemUtils.cpp
Message-ID: <200306262137.h5QLbcM00456@choi.cs.uiuc.edu>


Changes in directory llvm/tools/bugpoint:

CrashDebugger.cpp updated: 1.10 -> 1.10.2.1
ExtractFunction.cpp updated: 1.9 -> 1.9.2.1
Makefile updated: 1.3.2.1 -> 1.3.2.2
SystemUtils.cpp updated: 1.3.2.1 -> 1.3.2.2

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/tools/bugpoint/CrashDebugger.cpp
diff -u llvm/tools/bugpoint/CrashDebugger.cpp:1.10 llvm/tools/bugpoint/CrashDebugger.cpp:1.10.2.1
--- llvm/tools/bugpoint/CrashDebugger.cpp:1.10	Sun Jun  1 23:54:29 2003
+++ llvm/tools/bugpoint/CrashDebugger.cpp	Thu Jun 26 16:35:58 2003
@@ -360,20 +360,18 @@
   } while (Simplification);
 
   // Try to clean up the testcase by running funcresolve and globaldce...
-  if (AnyReduction) {
-    std::cout << "\n*** Attempting to perform final cleanups: ";
-    Module *M = performFinalCleanups();
-    std::swap(Program, M);
+  std::cout << "\n*** Attempting to perform final cleanups: ";
+  Module *M = performFinalCleanups();
+  std::swap(Program, M);
             
-    // Find out if the pass still crashes on the cleaned up program...
-    if (runPasses(PassesToRun)) {
-      // Yup, it does, keep the reduced version...
-      delete M;
-      AnyReduction = true;
-    } else {
-      delete Program;   // Otherwise, restore the original module...
-      Program = M;
-    }
+  // Find out if the pass still crashes on the cleaned up program...
+  if (runPasses(PassesToRun)) {
+    // Yup, it does, keep the reduced version...
+    delete M;
+    AnyReduction = true;
+  } else {
+    delete Program;   // Otherwise, restore the original module...
+    Program = M;
   }
 
   if (AnyReduction)


Index: llvm/tools/bugpoint/ExtractFunction.cpp
diff -u llvm/tools/bugpoint/ExtractFunction.cpp:1.9 llvm/tools/bugpoint/ExtractFunction.cpp:1.9.2.1
--- llvm/tools/bugpoint/ExtractFunction.cpp:1.9	Wed May 21 15:38:59 2003
+++ llvm/tools/bugpoint/ExtractFunction.cpp	Thu Jun 26 16:35:58 2003
@@ -97,6 +97,7 @@
   CleanupPasses.add(createFunctionResolvingPass());
   CleanupPasses.add(createGlobalDCEPass());
   CleanupPasses.add(createDeadTypeEliminationPass());
+  CleanupPasses.add(createDeadArgEliminationPass(true));
   CleanupPasses.add(createVerifierPass());
   CleanupPasses.run(*M);
   return M;


Index: llvm/tools/bugpoint/Makefile
diff -u llvm/tools/bugpoint/Makefile:1.3.2.1 llvm/tools/bugpoint/Makefile:1.3.2.2
--- llvm/tools/bugpoint/Makefile:1.3.2.1	Mon Jun 23 14:04:49 2003
+++ llvm/tools/bugpoint/Makefile	Thu Jun 26 16:35:58 2003
@@ -8,4 +8,6 @@
 USEDLIBS = ipo scalaropts analysis $(OPTLIBS) $(ANALIBS) \
            transformutils asmparser bcreader bcwriter vmcore support
 
+TOOLLINKOPTS = $(PLATFORMLIBDL)
+
 include $(LEVEL)/Makefile.common


Index: llvm/tools/bugpoint/SystemUtils.cpp
diff -u llvm/tools/bugpoint/SystemUtils.cpp:1.3.2.1 llvm/tools/bugpoint/SystemUtils.cpp:1.3.2.2
--- llvm/tools/bugpoint/SystemUtils.cpp:1.3.2.1	Mon Jun 23 14:04:49 2003
+++ llvm/tools/bugpoint/SystemUtils.cpp	Thu Jun 26 16:35:58 2003
@@ -11,12 +11,12 @@
 #include <iostream>
 #include <cstdlib>
 #include "Config/alloca.h"
-#include <sys/types.h>
-#include <sys/stat.h>
+#include "Config/sys/types.h"
+#include "Config/sys/stat.h"
 #include "Config/fcntl.h"
-#include <sys/wait.h>
+#include "Config/sys/wait.h"
 #include "Config/unistd.h"
-#include <errno.h>
+#include "Config/errno.h"
 
 /// removeFile - Delete the specified file
 ///
@@ -33,7 +33,7 @@
     return FilenameBase;    // Couldn't open the file? Use it!
 
   // Create a pattern for mkstemp...
-  char *FNBuffer = (char*)alloca(FilenameBase.size()+8);
+  char *FNBuffer = new char[FilenameBase.size()+8];
   strcpy(FNBuffer, FilenameBase.c_str());
   strcpy(FNBuffer+FilenameBase.size(), "-XXXXXX");
 
@@ -48,7 +48,9 @@
   // We don't need to hold the temp file descriptor... we will trust that noone
   // will overwrite/delete the file while we are working on it...
   close(TempFD);
-  return FNBuffer;
+  std::string Result(FNBuffer);
+  delete[] FNBuffer;
+  return Result;
 }
 
 /// isExecutableFile - This function returns true if the filename specified




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:42:56 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:42:56 2003
Subject: [llvm-commits] CVS: llvm/lib/Bytecode/Reader/ReadConst.cpp ReadInst.cpp ReaderInternals.h
Message-ID: <200306262137.h5QLbOJ32630@choi.cs.uiuc.edu>


Changes in directory llvm/lib/Bytecode/Reader:

ReadConst.cpp updated: 1.46 -> 1.46.2.1
ReadInst.cpp updated: 1.42 -> 1.42.2.1
ReaderInternals.h updated: 1.37 -> 1.37.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/lib/Bytecode/Reader/ReadConst.cpp
diff -u llvm/lib/Bytecode/Reader/ReadConst.cpp:1.46 llvm/lib/Bytecode/Reader/ReadConst.cpp:1.46.2.1
--- llvm/lib/Bytecode/Reader/ReadConst.cpp:1.46	Thu May 22 13:08:30 2003
+++ llvm/lib/Bytecode/Reader/ReadConst.cpp	Thu Jun 26 16:35:14 2003
@@ -139,7 +139,7 @@
 
   // Insert a bunch of opaque types to be resolved later...
   for (unsigned i = 0; i < NumEntries; ++i)
-    Tab.push_back(PATypeHandle<Type>(OpaqueType::get(), this));
+    Tab.push_back(PATypeHandle(OpaqueType::get(), this));
 
   // Loop through reading all of the types.  Forward types will make use of the
   // opaque types just inserted.


Index: llvm/lib/Bytecode/Reader/ReadInst.cpp
diff -u llvm/lib/Bytecode/Reader/ReadInst.cpp:1.42 llvm/lib/Bytecode/Reader/ReadInst.cpp:1.42.2.1
--- llvm/lib/Bytecode/Reader/ReadInst.cpp:1.42	Thu May 22 13:08:30 2003
+++ llvm/lib/Bytecode/Reader/ReadInst.cpp	Thu Jun 26 16:35:14 2003
@@ -302,16 +302,19 @@
       if (Raw.NumOperands < 3) return true;
 
       Normal = cast<BasicBlock>(getValue(Type::LabelTy, Raw.Arg2));
-      Except = cast<BasicBlock>(getValue(Type::LabelTy, args[0]));
+      if (Raw.NumOperands ==  3)
+        Except = cast<BasicBlock>(getValue(Type::LabelTy, Raw.Arg3));
+      else {
+        Except = cast<BasicBlock>(getValue(Type::LabelTy, args[0]));
 
-      FunctionType::ParamTypes::const_iterator It = PL.begin();
-      for (unsigned i = 1; i < args.size(); i++) {
-	if (It == PL.end()) return true;
-	// TODO: Check getValue for null!
-	Params.push_back(getValue(*It++, args[i]));
+        FunctionType::ParamTypes::const_iterator It = PL.begin();
+        for (unsigned i = 1; i < args.size(); i++) {
+          if (It == PL.end()) return true;
+          // TODO: Check getValue for null!
+          Params.push_back(getValue(*It++, args[i]));
+        }
+        if (It != PL.end()) return true;
       }
-
-      if (It != PL.end()) return true;
     } else {
       if (args.size() < 4) return true;
 
@@ -326,7 +329,8 @@
       }
     }
 
-    delete Raw.VarArgs;
+    if (Raw.NumOperands > 3)
+      delete Raw.VarArgs;
     Res = new InvokeInst(M, Normal, Except, Params);
     return false;
   }


Index: llvm/lib/Bytecode/Reader/ReaderInternals.h
diff -u llvm/lib/Bytecode/Reader/ReaderInternals.h:1.37 llvm/lib/Bytecode/Reader/ReaderInternals.h:1.37.2.1
--- llvm/lib/Bytecode/Reader/ReaderInternals.h:1.37	Thu May 22 13:26:48 2003
+++ llvm/lib/Bytecode/Reader/ReaderInternals.h	Thu Jun 26 16:35:14 2003
@@ -104,7 +104,7 @@
   // TypesLoaded - This vector mirrors the Values[TypeTyID] plane.  It is used
   // to deal with forward references to types.
   //
-  typedef std::vector<PATypeHandle<Type> > TypeValuesListTy;
+  typedef std::vector<PATypeHandle> TypeValuesListTy;
   TypeValuesListTy ModuleTypeValues;
   TypeValuesListTy FunctionTypeValues;
 




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:43:04 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:43:04 2003
Subject: [llvm-commits] CVS: llvm/lib/Analysis/DataStructure/BottomUpClosure.cpp DataStructure.cpp DataStructureAA.cpp Local.cpp Printer.cpp TopDownClosure.cpp
Message-ID: <200306262137.h5QLbo100642@choi.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis/DataStructure:

BottomUpClosure.cpp updated: 1.52 -> 1.52.2.1
DataStructure.cpp updated: 1.98 -> 1.98.2.1
DataStructureAA.cpp updated: 1.3 -> 1.3.2.1
Local.cpp updated: 1.52 -> 1.52.2.1
Printer.cpp updated: 1.51 -> 1.51.2.1
TopDownClosure.cpp updated: 1.40 -> 1.40.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/lib/Analysis/DataStructure/BottomUpClosure.cpp
diff -u llvm/lib/Analysis/DataStructure/BottomUpClosure.cpp:1.52 llvm/lib/Analysis/DataStructure/BottomUpClosure.cpp:1.52.2.1
--- llvm/lib/Analysis/DataStructure/BottomUpClosure.cpp:1.52	Wed Feb  5 15:59:56 2003
+++ llvm/lib/Analysis/DataStructure/BottomUpClosure.cpp	Thu Jun 26 16:35:10 2003
@@ -33,7 +33,7 @@
 // if the call sites are not external.
 //
 static inline bool isCompleteNode(DSNode *N) {
-  if (N->NodeType & DSNode::Incomplete) return false;
+  if (N->isIncomplete()) return false;
   const std::vector<GlobalValue*> &Callees = N->getGlobals();
   for (unsigned i = 0, e = Callees.size(); i != e; ++i)
     if (Callees[i]->isExternal())


Index: llvm/lib/Analysis/DataStructure/DataStructure.cpp
diff -u llvm/lib/Analysis/DataStructure/DataStructure.cpp:1.98 llvm/lib/Analysis/DataStructure/DataStructure.cpp:1.98.2.1
--- llvm/lib/Analysis/DataStructure/DataStructure.cpp:1.98	Tue May 20 13:45:31 2003
+++ llvm/lib/Analysis/DataStructure/DataStructure.cpp	Thu Jun 26 16:35:10 2003
@@ -49,8 +49,8 @@
 // DSNode Implementation
 //===----------------------------------------------------------------------===//
 
-DSNode::DSNode(unsigned NT, const Type *T, DSGraph *G)
-  : NumReferrers(0), Size(0), ParentGraph(G), Ty(Type::VoidTy), NodeType(NT) {
+DSNode::DSNode(const Type *T, DSGraph *G)
+  : NumReferrers(0), Size(0), ParentGraph(G), Ty(Type::VoidTy), NodeType(0) {
   // Add the type entry if it is specified...
   if (T) mergeTypeInfo(T, 0);
   G->getNodes().push_back(this);
@@ -107,14 +107,13 @@
 /// single byte with a single TypeEntry of "void".
 ///
 void DSNode::foldNodeCompletely() {
-  assert(!hasNoReferrers() &&
-         "Why would we collapse a node with no referrers?");
   if (isNodeCompletelyFolded()) return;  // If this node is already folded...
 
   ++NumFolds;
 
   // Create the node we are going to forward to...
-  DSNode *DestNode = new DSNode(NodeType|DSNode::Array, 0, ParentGraph);
+  DSNode *DestNode = new DSNode(0, ParentGraph);
+  DestNode->NodeType = NodeType|DSNode::Array;
   DestNode->Ty = Type::VoidTy;
   DestNode->Size = 1;
   DestNode->Globals.swap(Globals);
@@ -448,7 +447,7 @@
   // Merge the type entries of the two nodes together...
   if (NH.getNode()->Ty != Type::VoidTy)
     CurNodeH.getNode()->mergeTypeInfo(NH.getNode()->Ty, NOffset);
-  assert((CurNodeH.getNode()->NodeType & DSNode::DEAD) == 0);
+  assert(!CurNodeH.getNode()->isDeadNode());
 
   // If we are merging a node with a completely folded node, then both nodes are
   // now completely folded.
@@ -473,12 +472,14 @@
 
   DSNode *N = NH.getNode();
   if (CurNodeH.getNode() == N || N == 0) return;
-  assert((CurNodeH.getNode()->NodeType & DSNode::DEAD) == 0);
+  assert(!CurNodeH.getNode()->isDeadNode());
 
-  // Start forwarding to the new node!
+  // Merge the NodeType information...
   CurNodeH.getNode()->NodeType |= N->NodeType;
+
+  // Start forwarding to the new node!
   N->forwardNode(CurNodeH.getNode(), NOffset);
-  assert((CurNodeH.getNode()->NodeType & DSNode::DEAD) == 0);
+  assert(!CurNodeH.getNode()->isDeadNode());
 
   // Make all of the outgoing links of N now be outgoing links of CurNodeH.
   //
@@ -524,8 +525,7 @@
   if (N == 0 || (N == this && NH.getOffset() == Offset))
     return;  // Noop
 
-  assert((N->NodeType & DSNode::DEAD) == 0);
-  assert((NodeType & DSNode::DEAD) == 0);
+  assert(!N->isDeadNode() && !isDeadNode());
   assert(!hasNoReferrers() && "Should not try to fold a useless node!");
 
   if (N == this) {
@@ -632,13 +632,13 @@
   Nodes.reserve(FN+G.Nodes.size());
 
   // Remove alloca or mod/ref bits as specified...
-  unsigned clearBits = (CloneFlags & StripAllocaBit ? DSNode::AllocaNode : 0)
-    | (CloneFlags & StripModRefBits ? (DSNode::Modified | DSNode::Read) : 0);
-  clearBits |= DSNode::DEAD;  // Clear dead flag...
+  unsigned BitsToClear =((CloneFlags & StripAllocaBit) ? DSNode::AllocaNode : 0)
+    | ((CloneFlags & StripModRefBits) ? (DSNode::Modified | DSNode::Read) : 0);
+  BitsToClear |= DSNode::DEAD;  // Clear dead flag...
   for (unsigned i = 0, e = G.Nodes.size(); i != e; ++i) {
     DSNode *Old = G.Nodes[i];
     DSNode *New = new DSNode(*Old, this);
-    New->NodeType &= ~clearBits;
+    New->maskNodeTypes(~BitsToClear);
     OldNodeMap[Old] = New;
   }
 
@@ -745,16 +745,16 @@
 
 // markIncompleteNodes - Mark the specified node as having contents that are not
 // known with the current analysis we have performed.  Because a node makes all
-// of the nodes it can reach imcomplete if the node itself is incomplete, we
+// of the nodes it can reach incomplete if the node itself is incomplete, we
 // must recursively traverse the data structure graph, marking all reachable
 // nodes as incomplete.
 //
 static void markIncompleteNode(DSNode *N) {
   // Stop recursion if no node, or if node already marked...
-  if (N == 0 || (N->NodeType & DSNode::Incomplete)) return;
+  if (N == 0 || (N->isIncomplete())) return;
 
   // Actually mark the node
-  N->NodeType |= DSNode::Incomplete;
+  N->setIncompleteMarker();
 
   // Recusively process children...
   for (unsigned i = 0, e = N->getSize(); i < e; i += DS::PointerSize)
@@ -800,14 +800,15 @@
   // Mark all global nodes as incomplete...
   if ((Flags & DSGraph::IgnoreGlobals) == 0)
     for (unsigned i = 0, e = Nodes.size(); i != e; ++i)
-      if (Nodes[i]->NodeType & DSNode::GlobalNode && Nodes[i]->getNumLinks())
+      if (Nodes[i]->isGlobalNode() && Nodes[i]->getNumLinks())
         markIncompleteNode(Nodes[i]);
 }
 
 static inline void killIfUselessEdge(DSNodeHandle &Edge) {
   if (DSNode *N = Edge.getNode())  // Is there an edge?
     if (N->getNumReferrers() == 1)  // Does it point to a lonely node?
-      if ((N->NodeType & ~DSNode::Incomplete) == 0 && // No interesting info?
+      // No interesting info?
+      if ((N->getNodeFlags() & ~DSNode::Incomplete) == 0 &&
           N->getType() == Type::VoidTy && !N->isNodeCompletelyFolded())
         Edge.setNode(0);  // Kill the edge!
 }
@@ -837,7 +838,7 @@
     // If the Callee is a useless edge, this must be an unreachable call site,
     // eliminate it.
     if (CS.isIndirectCall() && CS.getCalleeNode()->getNumReferrers() == 1 &&
-        CS.getCalleeNode()->NodeType == 0) {  // No useful info?
+        CS.getCalleeNode()->getNodeFlags() == 0) {  // No useful info?
       std::cerr << "WARNING: Useless call site found??\n";
       CS.swap(Calls.back());
       Calls.pop_back();
@@ -916,8 +917,7 @@
 
   for (unsigned i = 0; i != Nodes.size(); ++i) {
     DSNode *Node = Nodes[i];
-    if (!(Node->NodeType & ~(DSNode::Composition | DSNode::Array |
-                             DSNode::DEAD))) {
+    if (!Node->isIncomplete() && !Node->isModified() && !Node->isRead()) {
       // This is a useless node if it has no mod/ref info (checked above),
       // outgoing edges (which it cannot, as it is not modified in this
       // context), and it has no incoming edges.  If it is a global node it may
@@ -925,7 +925,7 @@
       // scalar map, so we check those now.
       //
       if (Node->getNumReferrers() == Node->getGlobals().size()) {
-        std::vector<GlobalValue*> &Globals = Node->getGlobals();
+        const std::vector<GlobalValue*> &Globals = Node->getGlobals();
 
         // Loop through and make sure all of the globals are referring directly
         // to the node...
@@ -934,20 +934,16 @@
           assert(N == Node && "ScalarMap doesn't match globals list!");
         }
 
-        // Make sure numreferrers still agrees, if so, the node is truely dead.
+        // Make sure NumReferrers still agrees, if so, the node is truly dead.
         if (Node->getNumReferrers() == Globals.size()) {
           for (unsigned j = 0, e = Globals.size(); j != e; ++j)
             ScalarMap.erase(Globals[j]);
-
-          Globals.clear();
-          assert(Node->hasNoReferrers() && "Shouldn't have refs now!");
-          
-          Node->NodeType = DSNode::DEAD;
+          Node->makeNodeDead();
         }
       }
     }
 
-    if ((Node->NodeType & ~DSNode::DEAD) == 0 && Node->hasNoReferrers()) {
+    if (Node->getNodeFlags() == 0 && Node->hasNoReferrers()) {
       // This node is dead!
       delete Node;                        // Free memory...
       Nodes[i--] = Nodes.back();
@@ -1057,7 +1053,7 @@
       // these, prune the scalar pointing to it.
       //
       DSNode *N = I->second.getNode();
-      if (N->NodeType == DSNode::UnknownNode && !isa<Argument>(I->first)) {
+      if (N->isUnknownNode() && !isa<Argument>(I->first)) {
         ScalarMap.erase(I++);
       } else {
         I->second.getNode()->markReachableNodes(Alive);
@@ -1130,7 +1126,7 @@
         GlobalsGraph->Nodes.push_back(N);        // Move node to globals graph
         N->setParentGraph(GlobalsGraph);
       } else {                                 // Otherwise, delete the node
-        assert(((N->NodeType & DSNode::GlobalNode) == 0 ||
+        assert((!N->isGlobalNode() ||
                 (Flags & DSGraph::RemoveUnreachableGlobals))
                && "Killing a global?");
         //std::cerr << "[" << i+1 << "/" << DeadNodes.size()
@@ -1191,7 +1187,7 @@
     assert(I->second.getNode() && "Null node in scalarmap!");
     AssertNodeInGraph(I->second.getNode());
     if (GlobalValue *GV = dyn_cast<GlobalValue>(I->first)) {
-      assert((I->second.getNode()->NodeType & DSNode::GlobalNode) &&
+      assert(I->second.getNode()->isGlobalNode() &&
              "Global points to node, but node isn't global?");
       AssertNodeContainsGlobal(I->second.getNode(), GV);
     }


Index: llvm/lib/Analysis/DataStructure/DataStructureAA.cpp
diff -u llvm/lib/Analysis/DataStructure/DataStructureAA.cpp:1.3 llvm/lib/Analysis/DataStructure/DataStructureAA.cpp:1.3.2.1
--- llvm/lib/Analysis/DataStructure/DataStructureAA.cpp:1.3	Wed Feb 26 13:29:36 2003
+++ llvm/lib/Analysis/DataStructure/DataStructureAA.cpp	Thu Jun 26 16:35:10 2003
@@ -68,15 +68,12 @@
 // alias - This is the only method here that does anything interesting...
 AliasAnalysis::AliasResult DSAA::alias(const Value *V1, unsigned V1Size,
                                        const Value *V2, unsigned V2Size) {
-  // FIXME: This should handle the Size argument as well!
+  if (V1 == V2) return MustAlias;
+
   const Function *F1 = getValueFunction(V1);
   const Function *F2 = getValueFunction(V2);
   assert((!F1 || !F2 || F1 == F2) && "Alias query for 2 different functions?");
   
-
-  // FIXME: This can return must alias if querying a DSNode for a global value
-  // where the node has only the G composition bit set, and only one entry in
-  // the globals list...
   if (F2) F1 = F2;
   if (F1) {
     // Get the graph for a function...
@@ -88,17 +85,33 @@
       hash_map<Value*, DSNodeHandle>::iterator J = GSM.find((Value*)V2);
       if (J != GSM.end()) {
         assert(J->second.getNode() && "Scalar map points to null node?");
-        if (I->second.getNode() != J->second.getNode()) {
-          // Return noalias if one of the nodes is complete...
-          if ((~I->second.getNode()->NodeType | ~J->second.getNode()->NodeType)
-              & DSNode::Incomplete)
-            return NoAlias;
-          // both are incomplete, they may alias...
-        } else {
-          // Both point to the same node, see if they point to different
-          // offsets...  FIXME: This needs to know the size of the alias query
-          if (I->second.getOffset() != J->second.getOffset())
-            return NoAlias;
+
+        DSNode  *N1 = I->second.getNode(),  *N2 = J->second.getNode();
+        unsigned O1 = I->second.getOffset(), O2 = J->second.getOffset();
+        
+        // We can only make a judgement of one of the nodes is complete...
+        if (!N1->isIncomplete() || !N2->isIncomplete()) {
+          if (N1 != N2)
+            return NoAlias;   // Completely different nodes.
+
+          // Both point to the same node and same offset, and there is only one
+          // physical memory object represented in the node, return must alias.
+          //if (O1 == O2 && !N1->isMultiObject())
+          //  return MustAlias; // Exactly the same object & offset
+
+          // See if they point to different offsets...  if so, we may be able to
+          // determine that they do not alias...
+          if (O1 != O2) {
+            if (O2 < O1) {    // Ensure that O1 <= O2
+              std::swap(V1, V2);
+              std::swap(O1, O2);
+              std::swap(V1Size, V2Size);
+            }
+
+            // FIXME: This is not correct because we do not handle array
+            // indexing correctly with this check!
+            //if (O1+V1Size <= O2) return NoAlias;
+          }
         }
       }
     }


Index: llvm/lib/Analysis/DataStructure/Local.cpp
diff -u llvm/lib/Analysis/DataStructure/Local.cpp:1.52 llvm/lib/Analysis/DataStructure/Local.cpp:1.52.2.1
--- llvm/lib/Analysis/DataStructure/Local.cpp:1.52	Mon Mar  3 11:13:31 2003
+++ llvm/lib/Analysis/DataStructure/Local.cpp	Thu Jun 26 16:35:10 2003
@@ -86,9 +86,9 @@
   private:
     // Visitor functions, used to handle each instruction type we encounter...
     friend class InstVisitor<GraphBuilder>;
-    void visitMallocInst(MallocInst &MI) { handleAlloc(MI, DSNode::HeapNode); }
-    void visitAllocaInst(AllocaInst &AI) { handleAlloc(AI, DSNode::AllocaNode);}
-    void handleAlloc(AllocationInst &AI, DSNode::NodeTy NT);
+    void visitMallocInst(MallocInst &MI) { handleAlloc(MI, true); }
+    void visitAllocaInst(AllocaInst &AI) { handleAlloc(AI, false); }
+    void handleAlloc(AllocationInst &AI, bool isHeap);
 
     void visitPHINode(PHINode &PN);
 
@@ -108,10 +108,13 @@
     /// createNode - Create a new DSNode, ensuring that it is properly added to
     /// the graph.
     ///
-    DSNode *createNode(DSNode::NodeTy NodeType, const Type *Ty = 0) {
-      DSNode *N = new DSNode(NodeType, Ty, &G);   // Create the node
-      if (DisableFieldSensitivity)
+    DSNode *createNode(const Type *Ty = 0) {
+      DSNode *N = new DSNode(Ty, &G);   // Create the node
+      if (DisableFieldSensitivity) {
         N->foldNodeCompletely();
+        if (DSNode *FN = N->getForwardNode())
+          N = FN;
+      }
       return N;
     }
 
@@ -191,7 +194,7 @@
         NH = I->second;
       } else {
         // This returns a conservative unknown node for any unhandled ConstExpr
-        return NH = createNode(DSNode::UnknownNode);
+        return NH = createNode()->setUnknownNodeMarker();
       }
       if (NH.getNode() == 0) {  // (getelementptr null, X) returns null
         ScalarMap.erase(V);
@@ -201,7 +204,7 @@
 
     } else if (ConstantIntegral *CI = dyn_cast<ConstantIntegral>(C)) {
       // Random constants are unknown mem
-      return NH = createNode(DSNode::UnknownNode);
+      return NH = createNode()->setUnknownNodeMarker();
     } else {
       assert(0 && "Unknown constant type!");
     }
@@ -210,11 +213,11 @@
   DSNode *N;
   if (GlobalValue *GV = dyn_cast<GlobalValue>(V)) {
     // Create a new global node for this global variable...
-    N = createNode(DSNode::GlobalNode, GV->getType()->getElementType());
+    N = createNode(GV->getType()->getElementType());
     N->addGlobal(GV);
   } else {
     // Otherwise just create a shadow node
-    N = createNode(DSNode::ShadowNode);
+    N = createNode();
   }
 
   NH.setNode(N);      // Remember that we are pointing to it...
@@ -234,7 +237,7 @@
   DSNodeHandle &Link = Node.getLink(LinkNo);
   if (!Link.getNode()) {
     // If the link hasn't been created yet, make and return a new shadow node
-    Link = createNode(DSNode::ShadowNode);
+    Link = createNode();
   }
   return Link;
 }
@@ -260,8 +263,13 @@
 /// Alloca & Malloc instruction implementation - Simply create a new memory
 /// object, pointing the scalar to it.
 ///
-void GraphBuilder::handleAlloc(AllocationInst &AI, DSNode::NodeTy NodeType) {
-  setDestTo(AI, createNode(NodeType));
+void GraphBuilder::handleAlloc(AllocationInst &AI, bool isHeap) {
+  DSNode *N = createNode();
+  if (isHeap)
+    N->setHeapNodeMarker();
+  else
+    N->setAllocaNodeMarker();
+  setDestTo(AI, N);
 }
 
 // PHINode - Make the scalar for the PHI node point to all of the things the
@@ -365,7 +373,7 @@
   if (Ptr.getNode() == 0) return;
 
   // Make that the node is read from...
-  Ptr.getNode()->NodeType |= DSNode::Read;
+  Ptr.getNode()->setReadMarker();
 
   // Ensure a typerecord exists...
   Ptr.getNode()->mergeTypeInfo(LI.getType(), Ptr.getOffset(), false);
@@ -380,7 +388,7 @@
   if (Dest.getNode() == 0) return;
 
   // Mark that the node is written to...
-  Dest.getNode()->NodeType |= DSNode::Modified;
+  Dest.getNode()->setModifiedMarker();
 
   // Ensure a typerecord exists...
   Dest.getNode()->mergeTypeInfo(StoredTy, Dest.getOffset());
@@ -423,8 +431,9 @@
 
 void GraphBuilder::visitFreeInst(FreeInst &FI) {
   // Mark that the node is written to...
-  getValueDest(*FI.getOperand(0)).getNode()->NodeType
-    |= DSNode::Modified | DSNode::HeapNode;
+  DSNode *N = getValueDest(*FI.getOperand(0)).getNode();
+  N->setModifiedMarker();
+  N->setHeapNodeMarker();
 }
 
 /// Handle casts...
@@ -438,7 +447,7 @@
       // to track the fact that the node points to SOMETHING, just something we
       // don't know about.  Make an "Unknown" node.
       //
-      setDestTo(CI, createNode(DSNode::UnknownNode));
+      setDestTo(CI, createNode()->setUnknownNodeMarker());
     }
 }
 
@@ -455,7 +464,7 @@
       CurNode.mergeWith(getValueDest(**I));
 
   if (CurNode.getNode())
-    CurNode.getNode()->NodeType |= DSNode::UnknownNode;
+    CurNode.getNode()->setUnknownNodeMarker();
 }
 
 


Index: llvm/lib/Analysis/DataStructure/Printer.cpp
diff -u llvm/lib/Analysis/DataStructure/Printer.cpp:1.51 llvm/lib/Analysis/DataStructure/Printer.cpp:1.51.2.1
--- llvm/lib/Analysis/DataStructure/Printer.cpp:1.51	Fri Feb 14 14:25:47 2003
+++ llvm/lib/Analysis/DataStructure/Printer.cpp	Thu Jun 26 16:35:10 2003
@@ -38,16 +38,18 @@
     if (N->isArray())
       OS << " array";
   }
-  if (N->NodeType) {
+  if (unsigned NodeType = N->getNodeFlags()) {
     OS << ": ";
-    if (N->NodeType & DSNode::AllocaNode ) OS << "S";
-    if (N->NodeType & DSNode::HeapNode   ) OS << "H";
-    if (N->NodeType & DSNode::GlobalNode ) OS << "G";
-    if (N->NodeType & DSNode::UnknownNode) OS << "U";
-    if (N->NodeType & DSNode::Incomplete ) OS << "I";
-    if (N->NodeType & DSNode::Modified   ) OS << "M";
-    if (N->NodeType & DSNode::Read       ) OS << "R";
-    if (N->NodeType & DSNode::DEAD       ) OS << "<dead>";
+    if (NodeType & DSNode::AllocaNode ) OS << "S";
+    if (NodeType & DSNode::HeapNode   ) OS << "H";
+    if (NodeType & DSNode::GlobalNode ) OS << "G";
+    if (NodeType & DSNode::UnknownNode) OS << "U";
+    if (NodeType & DSNode::Incomplete ) OS << "I";
+    if (NodeType & DSNode::Modified   ) OS << "M";
+    if (NodeType & DSNode::Read       ) OS << "R";
+#ifndef NDEBUG
+    if (NodeType & DSNode::DEAD       ) OS << "<dead>";
+#endif
     OS << "\n";
   }
 


Index: llvm/lib/Analysis/DataStructure/TopDownClosure.cpp
diff -u llvm/lib/Analysis/DataStructure/TopDownClosure.cpp:1.40 llvm/lib/Analysis/DataStructure/TopDownClosure.cpp:1.40.2.1
--- llvm/lib/Analysis/DataStructure/TopDownClosure.cpp:1.40	Tue Feb 11 17:11:44 2003
+++ llvm/lib/Analysis/DataStructure/TopDownClosure.cpp	Thu Jun 26 16:35:10 2003
@@ -45,6 +45,7 @@
 // has no way to extend the lifetime of the pass, which screws up ds-aa.
 //
 void TDDataStructures::releaseMyMemory() {
+  return;
   for (hash_map<const Function*, DSGraph*>::iterator I = DSInfo.begin(),
          E = DSInfo.end(); I != E; ++I)
     delete I->second;




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:43:13 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:43:13 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/C++Frontend/ConstructorDestructorAttributes.c
Message-ID: <200306262137.h5QLbO432618@choi.cs.uiuc.edu>


Changes in directory llvm/test/Regression/C++Frontend:

ConstructorDestructorAttributes.c added (r1.1.2.1)

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/test/Regression/C++Frontend/ConstructorDestructorAttributes.c
diff -c /dev/null llvm/test/Regression/C++Frontend/ConstructorDestructorAttributes.c:1.1
*** /dev/null	Thu Jun 26 16:37:24 2003
--- llvm/test/Regression/C++Frontend/ConstructorDestructorAttributes.c	Thu Jun 26 09:56:36 2003
***************
*** 0 ****
--- 1,14 ----
+ #include <stdio.h>
+ 
+ void ctor() __attribute__((constructor));
+ 
+ void ctor() {
+    printf("Create!\n");
+ }
+ void dtor() __attribute__((destructor));
+ 
+ void dtor() {
+    printf("Create!\n");
+ }
+ 
+ int main() { return 0; }




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:43:21 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:43:21 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/Makefile
Message-ID: <200306262137.h5QLbOo32608@choi.cs.uiuc.edu>


Changes in directory llvm/tools/lli:

Makefile updated: 1.24.2.1 -> 1.24.2.2

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/tools/lli/Makefile
diff -u llvm/tools/lli/Makefile:1.24.2.1 llvm/tools/lli/Makefile:1.24.2.2
--- llvm/tools/lli/Makefile:1.24.2.1	Mon Jun 23 14:04:51 2003
+++ llvm/tools/lli/Makefile	Thu Jun 26 16:36:14 2003
@@ -2,8 +2,8 @@
 TOOLNAME = lli
 PARALLEL_DIRS = Interpreter JIT
 
-# FIXME: This enables testing the Sparc JIT on x86.
-# Perhaps in the future this should be a ./configure option.
+# Get the config name...
+include $(LEVEL)/Makefile.config
 
 # Generic JIT libraries
 JITLIBS   = lli-jit codegen
@@ -11,15 +11,24 @@
 
 # What the X86 JIT requires
 JITLIBS  += x86
-ARCHLIBS += 
+# X86 doesn't require any ARCHLIBS
+
+
 
 # What the Sparc JIT requires
+ifeq ($(OS),SunOS)
 JITLIBS  += sparc 
-ARCHLIBS  = sched livevar instrument.a profpaths transformutils.a \
+ARCHLIBS  = sched livevar instrument.a profpaths \
            bcwriter transforms.a ipo.a ipa.a datastructure.a regalloc \
            mapping select postopts.a preopts
+endif
+
+USEDLIBS = lli-interpreter $(JITLIBS) $(ARCHLIBS) scalaropts analysis.a \
+           transformutils.a bcreader vmcore support.a target.a
 
-USEDLIBS = lli-interpreter $(JITLIBS) bcreader vmcore scalaropts \
-           analysis.a support.a target.a $(ARCHLIBS)
+# Have gcc tell the linker to export symbols from the program so that
+# dynamically loaded modules can be linked against them.
+#
+TOOLLINKOPTS = $(PLATFORMLIBDL)
 
 include $(LEVEL)/Makefile.common




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:43:29 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:43:29 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/BugPoint/remove_arguments_test.ll Makefile
Message-ID: <200306262137.h5QLbN732602@choi.cs.uiuc.edu>


Changes in directory llvm/test/Regression/BugPoint:

remove_arguments_test.ll added (r1.1.2.1)
Makefile updated: 1.2 -> 1.2.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/test/Regression/BugPoint/remove_arguments_test.ll
diff -c /dev/null llvm/test/Regression/BugPoint/remove_arguments_test.ll:1.1
*** /dev/null	Thu Jun 26 16:37:23 2003
--- llvm/test/Regression/BugPoint/remove_arguments_test.ll	Tue Jun 24 18:40:27 2003
***************
*** 0 ****
--- 1,9 ----
+ ; RUN: bugpoint %s  -bugpoint-crashcalls
+ 
+ ; Test to make sure that arguments are removed from the function if they are unnecessary.
+ 
+ declare int %test2()
+ int %test(int %A, int %B, float %C) {
+ 	call int %test2()
+ 	ret int %0
+ }


Index: llvm/test/Regression/BugPoint/Makefile
diff -u llvm/test/Regression/BugPoint/Makefile:1.2 llvm/test/Regression/BugPoint/Makefile:1.2.2.1
--- llvm/test/Regression/BugPoint/Makefile:1.2	Mon Dec 23 17:47:39 2002
+++ llvm/test/Regression/BugPoint/Makefile	Thu Jun 26 16:35:43 2003
@@ -12,8 +12,8 @@
 
 Output/%.ll.out: %.ll Output/.dir $(LAS) $(LBUGPOINT)
 	-$(TESTRUNR) $<
-	-# Remove temporary files generated...
-	@rm -f bugpoint-$<-*
+	-@# Remove temporary files generated...
+	- at rm -f bugpoint-*
 
 clean::
 	rm -f *.bc




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:43:38 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:43:38 2003
Subject: [llvm-commits] CVS: llvm/www/docs/GettingStarted.html
Message-ID: <200306262137.h5QLbNH32593@choi.cs.uiuc.edu>


Changes in directory llvm/www/docs:

GettingStarted.html updated: 1.12.2.3 -> 1.12.2.4

---
Log message:

Merged with mainline.


---
Diffs of the changes:

Index: llvm/www/docs/GettingStarted.html
diff -u llvm/www/docs/GettingStarted.html:1.12.2.3 llvm/www/docs/GettingStarted.html:1.12.2.4
--- llvm/www/docs/GettingStarted.html:1.12.2.3	Thu Jun 26 11:15:54 2003
+++ llvm/www/docs/GettingStarted.html	Thu Jun 26 16:33:12 2003
@@ -54,7 +54,7 @@
     <!--=====================================================================-->
 
     <p>The <a href"starting">next section</a> of this guide is meant to get
-    you up and running with LLVM, and to give you some basic information about
+    you up and running with LLVM and to give you some basic information about
     the LLVM environment.  The <a href"#quickstart">first subsection</a> gives
     a short summary for those who are already familiar with the system and
     want to get started as quickly as possible.
@@ -262,7 +262,7 @@
     This environment variable is used to locate "system" libraries like
     "<tt>-lc</tt>" and "<tt>-lm</tt>" when linking.  This variable should be set
     to the absolute path for the bytecode-libs subdirectory of the C front-end
-    install.  For example, <tt>LLVM_LIB_SEARCH_PATH</tt> is set to
+    install.  For example, one might set <tt>LLVM_LIB_SEARCH_PATH</tt> to
     <tt>/home/vadve/lattner/local/x86/llvm-gcc/bytecode-libs</tt> for the X86
     version of the C front-end on our research machines.<p>
 
@@ -305,8 +305,8 @@
     included in the CVS tree you just checked out.
 
     <p>The other <a href="#tools">LLVM tools</a> are part of the LLVM
-    source base and are built when compiling LLVM.  They will be built into the
-    <tt><i>LLVM_OBJ_DIR</i>/tools/Debug</tt> directory.</p>
+    source base and built when compiling LLVM.  They will be built into the
+    <tt><i>OBJ_ROOT</i>/tools/Debug</tt> directory.</p>
 
     <!------------------------------------------------------------------------->
     <h3><a name="compile">Compiling the source code</a></h3>
@@ -314,7 +314,9 @@
 
     <p>Every directory in the LLVM source tree includes a <tt>Makefile</tt> to
     build it and any subdirectories that it contains.  These makefiles require
-    that you use <tt>gmake</tt>, instead of <tt>make</tt> to build them, but can
+    that you use GNU Make (sometimes called <tt>gmake</tt>) instead of
+    <tt>make</tt> to
+    build them, but can
     otherwise be used freely.  To build the entire LLVM system, just enter the
     top level <tt>llvm</tt> directory and type <tt>gmake</tt>.  A few minutes
     later you will hopefully have a freshly compiled toolchain waiting for you
@@ -322,8 +324,8 @@
     libraries that
     were compiled, look in <tt><i>OBJ_ROOT</i>/llvm/lib/Debug</tt>.</p>
 
-    If you get an error talking about a <tt>/localhome</tt> directory, follow
-    the instructions in the section about <a href="#environment">Setting Up Your
+    If you get an error about a <tt>/localhome</tt> directory, follow the
+    instructions in the section about <a href="#environment">Setting Up Your
     Environment.</a>
 
 
@@ -344,7 +346,7 @@
     <h3><a name="cvsdir"><tt>CVS</tt> directories</a></h3>
     <!------------------------------------------------------------------------->
 
-    Every directory checked out of CVS will contain a <tt>CVS</tt> directory,
+    Every directory checked out of CVS will contain a <tt>CVS</tt> directory;
     for the most part these can just be ignored.
 
 
@@ -361,7 +363,7 @@
     directory contains automatically generated dependance files which are used
     during compilation to make sure that source files get rebuilt if a header
     file they use is modified. The <tt>Debug</tt> directory holds the object
-    files, library files and executables that are used for building a debug
+    files, library files, and executables that are used for building a debug
     enabled build.  The <tt>Release</tt> directory is created to hold the same
     files when the <tt>ENABLE_OPTIMIZED=1</tt> flag is passed to <tt>gmake</tt>,
     causing an optimized build to be performed.<p>




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:43:46 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:43:46 2003
Subject: [llvm-commits] CVS: llvm/lib/AsmParser/llvmAsmParser.y
Message-ID: <200306262137.h5QLbMd32583@choi.cs.uiuc.edu>


Changes in directory llvm/lib/AsmParser:

llvmAsmParser.y updated: 1.113 -> 1.113.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/lib/AsmParser/llvmAsmParser.y
diff -u llvm/lib/AsmParser/llvmAsmParser.y:1.113 llvm/lib/AsmParser/llvmAsmParser.y:1.113.2.1
--- llvm/lib/AsmParser/llvmAsmParser.y:1.113	Wed May 21 12:48:56 2003
+++ llvm/lib/AsmParser/llvmAsmParser.y	Thu Jun 26 16:35:12 2003
@@ -816,7 +816,7 @@
   | UpRTypesV '(' ArgTypeListI ')' {           // Function derived type?
     std::vector<const Type*> Params;
     mapto($3->begin(), $3->end(), std::back_inserter(Params), 
-	  std::mem_fun_ref(&PATypeHandle<Type>::get));
+	  std::mem_fun_ref(&PATypeHandle::get));
     bool isVarArg = Params.size() && Params.back() == Type::VoidTy;
     if (isVarArg) Params.pop_back();
 
@@ -831,7 +831,7 @@
   | '{' TypeListI '}' {                        // Structure type?
     std::vector<const Type*> Elements;
     mapto($2->begin(), $2->end(), std::back_inserter(Elements), 
-	std::mem_fun_ref(&PATypeHandle<Type>::get));
+	std::mem_fun_ref(&PATypeHandle::get));
 
     $$ = new PATypeHolder(HandleUpRefs(StructType::get(Elements)));
     delete $2;




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:43:54 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:43:54 2003
Subject: [llvm-commits] CVS: llvm/tools/llc/Makefile llc.cpp
Message-ID: <200306262137.h5QLbMO32577@choi.cs.uiuc.edu>


Changes in directory llvm/tools/llc:

Makefile updated: 1.35.2.1 -> 1.35.2.2
llc.cpp updated: 1.73 -> 1.73.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/tools/llc/Makefile
diff -u llvm/tools/llc/Makefile:1.35.2.1 llvm/tools/llc/Makefile:1.35.2.2
--- llvm/tools/llc/Makefile:1.35.2.1	Mon Jun 23 14:04:50 2003
+++ llvm/tools/llc/Makefile	Thu Jun 26 16:36:12 2003
@@ -2,6 +2,7 @@
 TOOLNAME = llc
 USEDLIBS = mapping \
            sparc \
+           x86 \
            regalloc \
            sched \
            select \
@@ -11,13 +12,14 @@
            target.a \
            livevar \
            transforms.a \
-	   scalaropts.a \
+           scalaropts.a \
            analysis.a \
            transformutils.a \
            bcreader \
-	   bcwriter \
+           bcwriter \
            vmcore \
            support
+TOOLLINKOPTS = $(PLATFORMLIBDL)
 
 include $(LEVEL)/Makefile.common
 


Index: llvm/tools/llc/llc.cpp
diff -u llvm/tools/llc/llc.cpp:1.73 llvm/tools/llc/llc.cpp:1.73.2.1
--- llvm/tools/llc/llc.cpp:1.73	Thu Jun  5 15:51:10 2003
+++ llvm/tools/llc/llc.cpp	Thu Jun 26 16:36:12 2003
@@ -1,6 +1,6 @@
-//===-- llc.cpp - Implement the LLVM Compiler -----------------------------===//
+//===-- llc.cpp - Implement the LLVM Native Code Generator ----------------===//
 //
-// This is the llc compiler driver.
+// This is the llc code generator.
 //
 //===----------------------------------------------------------------------===//
 
@@ -8,31 +8,14 @@
 #include "llvm/Target/TargetMachineImpls.h"
 #include "llvm/Target/TargetMachine.h"
 #include "llvm/Transforms/Scalar.h"
-#include "llvm/Assembly/PrintModulePass.h"
 #include "llvm/Module.h"
 #include "llvm/PassManager.h"
 #include "llvm/Pass.h"
-#include "llvm/Support/PassNameParser.h"
 #include "Support/CommandLine.h"
 #include "Support/Signals.h"
 #include <memory>
 #include <fstream>
 
-//------------------------------------------------------------------------------
-// Option declarations for LLC.
-//------------------------------------------------------------------------------
-
-// Make all registered optimization passes available to llc.  These passes
-// will all be run before the simplification and lowering steps used by the
-// back-end code generator, and will be run in the order specified on the
-// command line. The OptimizationList is automatically populated with
-// registered Passes by the PassNameParser.
-//
-static cl::list<const PassInfo*, bool,
-                FilteredPassNameParser<PassInfo::Optimization> >
-OptimizationList(cl::desc("Optimizations available:"));
-
-
 // General options for llc.  Other pass-specific options are specified
 // within the corresponding llc passes, and target-specific options
 // and back-end code generation options are specified with the target machine.
@@ -45,13 +28,14 @@
 
 static cl::opt<bool> Force("f", cl::desc("Overwrite output files"));
 
-static cl::opt<bool>
-DisableStrip("disable-strip",
-          cl::desc("Do not strip the LLVM bytecode included in the executable"));
-
-static cl::opt<bool>
-DumpAsm("d", cl::desc("Print bytecode before native code generation"),
-        cl::Hidden);
+enum ArchName { noarch, x86, Sparc };
+
+static cl::opt<ArchName>
+Arch("march", cl::desc("Architecture to generate assembly for:"), cl::Prefix,
+     cl::values(clEnumVal(x86, "  IA-32 (Pentium and above)"),
+		clEnumValN(Sparc, "sparc", "  SPARC V9"),
+		0),
+     cl::init(noarch));
 
 // GetFileNameRoot - Helper function to get the basename of a filename...
 static inline std::string
@@ -69,73 +53,84 @@
 }
 
 
-//===---------------------------------------------------------------------===//
-// Function main()
-// 
-// Entry point for the llc compiler.
-//===---------------------------------------------------------------------===//
-
-int
-main(int argc, char **argv)
-{
+// main - Entry point for the llc compiler.
+//
+int main(int argc, char **argv) {
   cl::ParseCommandLineOptions(argc, argv, " llvm system compiler\n");
   
-  // Allocate a target... in the future this will be controllable on the
-  // command line.
-  std::auto_ptr<TargetMachine> target(allocateSparcTargetMachine());
-  assert(target.get() && "Could not allocate target machine!");
+  // Load the module to be compiled...
+  std::auto_ptr<Module> M(ParseBytecodeFile(InputFilename));
+  if (M.get() == 0) {
+    std::cerr << argv[0] << ": bytecode didn't read correctly.\n";
+    return 1;
+  }
+  Module &mod = *M.get();
 
+  // Allocate target machine.  First, check whether the user has
+  // explicitly specified an architecture to compile for.
+  unsigned Config = (mod.isLittleEndian()   ? TM::LittleEndian : TM::BigEndian)|
+                    (mod.has32BitPointers() ? TM::PtrSize32    : TM::PtrSize64);
+  TargetMachine* (*TargetMachineAllocator)(unsigned) = 0;
+  switch (Arch) {
+  case x86:
+    TargetMachineAllocator = allocateX86TargetMachine;
+    break;
+  case Sparc:
+    TargetMachineAllocator = allocateSparcTargetMachine;
+    break;
+  default:
+    // Decide what the default target machine should be, by looking at
+    // the module. This heuristic (ILP32, LE -> IA32; LP64, BE ->
+    // SPARCV9) is kind of gross, but it will work until we have more
+    // sophisticated target information to work from.
+    if (mod.isLittleEndian() && mod.has32BitPointers()) { 
+      TargetMachineAllocator = allocateX86TargetMachine;
+    } else if (mod.isBigEndian() && mod.has64BitPointers()) {
+      TargetMachineAllocator = allocateSparcTargetMachine;
+    } else {
+      assert(0 && "You must specify -march; I could not guess the default");
+    } 
+    break;
+  }
+  std::auto_ptr<TargetMachine> target((*TargetMachineAllocator)(Config));
+  assert(target.get() && "Could not allocate target machine!");
   TargetMachine &Target = *target.get();
   const TargetData &TD = Target.getTargetData();
 
-  // Load the module to be compiled...
-  std::auto_ptr<Module> M(ParseBytecodeFile(InputFilename));
-  if (M.get() == 0)
-    {
-      std::cerr << argv[0] << ": bytecode didn't read correctly.\n";
-      return 1;
-    }
-
   // Build up all of the passes that we want to do to the module...
   PassManager Passes;
 
   Passes.add(new TargetData("llc", TD.isLittleEndian(), TD.getPointerSize(),
                             TD.getPointerAlignment(), TD.getDoubleAlignment()));
 
-  // Create a new optimization pass for each one specified on the command line
-  // Deal specially with tracing passes, which must be run differently than opt.
-  // 
-  for (unsigned i = 0; i < OptimizationList.size(); ++i) {
-    const PassInfo *Opt = OptimizationList[i];
-    
-    // handle other passes as normal optimization passes
-    if (Opt->getNormalCtor())
-      Passes.add(Opt->getNormalCtor()());
-    else if (Opt->getTargetCtor())
-      Passes.add(Opt->getTargetCtor()(Target));
-    else
-      std::cerr << argv[0] << ": cannot create pass: "
-                << Opt->getPassName() << "\n";
-  }
-
-  // Replace malloc and free instructions with library calls.
-  // Do this after tracing until lli implements these lib calls.
-  // For now, it will emulate malloc and free internally.
-  // FIXME: This is sparc specific!
-  Passes.add(createLowerAllocationsPass());
-
-  // If LLVM dumping after transformations is requested, add it to the pipeline
-  if (DumpAsm)
-    Passes.add(new PrintFunctionPass("Code after xformations: \n", &std::cerr));
-
-  // Strip all of the symbols from the bytecode so that it will be smaller...
-  if (!DisableStrip)
-    Passes.add(createSymbolStrippingPass());
-
   // Figure out where we are going to send the output...
   std::ostream *Out = 0;
-  if (OutputFilename != "")
-    {   // Specified an output filename?
+  if (OutputFilename != "") {
+    if (OutputFilename != "-") {
+      // Specified an output filename?
+      if (!Force && std::ifstream(OutputFilename.c_str())) {
+	// If force is not specified, make sure not to overwrite a file!
+	std::cerr << argv[0] << ": error opening '" << OutputFilename
+		  << "': file exists!\n"
+		  << "Use -f command line argument to force output\n";
+	return 1;
+      }
+      Out = new std::ofstream(OutputFilename.c_str());
+
+      // Make sure that the Out file gets unlink'd from the disk if we get a
+      // SIGINT
+      RemoveFileOnSignal(OutputFilename);
+    } else {
+      Out = &std::cout;
+    }
+  } else {
+    if (InputFilename == "-") {
+      OutputFilename = "-";
+      Out = &std::cout;
+    } else {
+      OutputFilename = GetFileNameRoot(InputFilename); 
+      OutputFilename += ".s";
+      
       if (!Force && std::ifstream(OutputFilename.c_str())) {
         // If force is not specified, make sure not to overwrite a file!
         std::cerr << argv[0] << ": error opening '" << OutputFilename
@@ -143,52 +138,28 @@
                   << "Use -f command line argument to force output\n";
         return 1;
       }
+      
       Out = new std::ofstream(OutputFilename.c_str());
-
+      if (!Out->good()) {
+        std::cerr << argv[0] << ": error opening " << OutputFilename << "!\n";
+        delete Out;
+        return 1;
+      }
+      
       // Make sure that the Out file gets unlink'd from the disk if we get a
       // SIGINT
       RemoveFileOnSignal(OutputFilename);
     }
-  else
-    {
-      if (InputFilename == "-")
-        {
-          OutputFilename = "-";
-          Out = &std::cout;
-        }
-      else
-        {
-          std::string OutputFilename = GetFileNameRoot(InputFilename); 
-          OutputFilename += ".s";
-
-          if (!Force && std::ifstream(OutputFilename.c_str()))
-            {
-              // If force is not specified, make sure not to overwrite a file!
-              std::cerr << argv[0] << ": error opening '" << OutputFilename
-                        << "': file exists!\n"
-                        << "Use -f command line argument to force output\n";
-              return 1;
-            }
-
-          Out = new std::ofstream(OutputFilename.c_str());
-          if (!Out->good())
-            {
-              std::cerr << argv[0] << ": error opening " << OutputFilename
-                        << "!\n";
-              delete Out;
-              return 1;
-            }
-
-          // Make sure that the Out file gets unlink'd from the disk if we get a
-          // SIGINT
-          RemoveFileOnSignal(OutputFilename);
-        }
-    }
+  }
 
-  // Ask the target to add backend passes as neccesary
+  // Ask the target to add backend passes as necessary
   if (Target.addPassesToEmitAssembly(Passes, *Out)) {
     std::cerr << argv[0] << ": target '" << Target.getName()
-              << " does not support static compilation!\n";
+              << "' does not support static compilation!\n";
+    if (Out != &std::cout) delete Out;
+    // And the Out file is empty and useless, so remove it now.
+    std::remove(OutputFilename.c_str());
+    return 1;
   } else {
     // Run our queue of passes all at once now, efficiently.
     Passes.run(*M.get());




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:44:03 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:44:03 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Analysis/AliasSetTracker.h CallGraph.h ConstantsScanner.h DSGraph.h DSGraphTraits.h DSNode.h DSSupport.h DataStructure.h DependenceGraph.h InstForest.h IntervalIterator.h MemoryDepAnalysis.h PgmDependenceGraph.h
Message-ID: <200306262137.h5QLbMC32564@choi.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Analysis:

AliasSetTracker.h updated: 1.5.2.1 -> 1.5.2.2
CallGraph.h updated: 1.27.2.1 -> 1.27.2.2
ConstantsScanner.h updated: 1.10.2.1 -> 1.10.2.2
DSGraph.h updated: 1.48.2.1 -> 1.48.2.2
DSGraphTraits.h updated: 1.14.2.1 -> 1.14.2.2
DSNode.h updated: 1.24.2.1 -> 1.24.2.2
DSSupport.h updated: 1.16.2.1 -> 1.16.2.2
DataStructure.h updated: 1.63.2.1 -> 1.63.2.2
DependenceGraph.h updated: 1.4.2.1 -> 1.4.2.2
InstForest.h updated: 1.17.2.1 -> 1.17.2.2
IntervalIterator.h updated: 1.11.2.1 -> 1.11.2.2
MemoryDepAnalysis.h updated: 1.2.2.1 -> 1.2.2.2
PgmDependenceGraph.h updated: 1.2.2.1 -> 1.2.2.2

---
Log message:

Merged with mainline on Thursday, June 26, 2003.
Kept includes of "Config/assert.h" to avoid implicit dependencies on header
files.


---
Diffs of the changes:

Index: llvm/include/llvm/Analysis/AliasSetTracker.h
diff -u llvm/include/llvm/Analysis/AliasSetTracker.h:1.5.2.1 llvm/include/llvm/Analysis/AliasSetTracker.h:1.5.2.2
--- llvm/include/llvm/Analysis/AliasSetTracker.h:1.5.2.1	Mon Jun 23 13:51:46 2003
+++ llvm/include/llvm/Analysis/AliasSetTracker.h	Thu Jun 26 16:34:42 2003
@@ -11,7 +11,6 @@
 #define LLVM_ANALYSIS_ALIASSETTRACKER_H
 
 #include "Config/assert.h"
-
 #include "llvm/Support/CallSite.h"
 #include "Support/iterator"
 #include "Support/hash_map"


Index: llvm/include/llvm/Analysis/CallGraph.h
diff -u llvm/include/llvm/Analysis/CallGraph.h:1.27.2.1 llvm/include/llvm/Analysis/CallGraph.h:1.27.2.2
--- llvm/include/llvm/Analysis/CallGraph.h:1.27.2.1	Mon Jun 23 13:51:47 2003
+++ llvm/include/llvm/Analysis/CallGraph.h	Thu Jun 26 16:34:42 2003
@@ -42,7 +42,6 @@
 #define LLVM_ANALYSIS_CALLGRAPH_H
 
 #include "Config/assert.h"
-
 #include "Support/GraphTraits.h"
 #include "Support/STLExtras.h"
 #include "llvm/Pass.h"


Index: llvm/include/llvm/Analysis/ConstantsScanner.h
diff -u llvm/include/llvm/Analysis/ConstantsScanner.h:1.10.2.1 llvm/include/llvm/Analysis/ConstantsScanner.h:1.10.2.2
--- llvm/include/llvm/Analysis/ConstantsScanner.h:1.10.2.1	Mon Jun 23 13:51:47 2003
+++ llvm/include/llvm/Analysis/ConstantsScanner.h	Thu Jun 26 16:34:42 2003
@@ -10,7 +10,6 @@
 #define LLVM_ANALYSIS_CONSTANTSSCANNER_H
 
 #include "Config/assert.h"
-
 #include "llvm/Support/InstIterator.h"
 #include "llvm/Instruction.h"
 #include "Support/iterator"


Index: llvm/include/llvm/Analysis/DSGraph.h
diff -u llvm/include/llvm/Analysis/DSGraph.h:1.48.2.1 llvm/include/llvm/Analysis/DSGraph.h:1.48.2.2
--- llvm/include/llvm/Analysis/DSGraph.h:1.48.2.1	Mon Jun 23 13:51:47 2003
+++ llvm/include/llvm/Analysis/DSGraph.h	Thu Jun 26 16:34:42 2003
@@ -8,7 +8,6 @@
 #define LLVM_ANALYSIS_DSGRAPH_H
 
 #include "Config/assert.h"
-
 #include "llvm/Analysis/DSNode.h"
 
 //===----------------------------------------------------------------------===//
@@ -137,9 +136,9 @@
   /// maskNodeTypes - Apply a mask to all of the node types in the graph.  This
   /// is useful for clearing out markers like Incomplete.
   ///
-  void maskNodeTypes(unsigned char Mask) {
+  void maskNodeTypes(unsigned Mask) {
     for (unsigned i = 0, e = Nodes.size(); i != e; ++i)
-      Nodes[i]->NodeType &= Mask;
+      Nodes[i]->maskNodeTypes(Mask);
   }
   void maskIncompleteMarkers() { maskNodeTypes(~DSNode::Incomplete); }
 


Index: llvm/include/llvm/Analysis/DSGraphTraits.h
diff -u llvm/include/llvm/Analysis/DSGraphTraits.h:1.14.2.1 llvm/include/llvm/Analysis/DSGraphTraits.h:1.14.2.2
--- llvm/include/llvm/Analysis/DSGraphTraits.h:1.14.2.1	Mon Jun 23 13:51:47 2003
+++ llvm/include/llvm/Analysis/DSGraphTraits.h	Thu Jun 26 16:34:42 2003
@@ -10,7 +10,6 @@
 #define LLVM_ANALYSIS_DSGRAPHTRAITS_H
 
 #include "Config/assert.h"
-
 #include "llvm/Analysis/DSGraph.h"
 #include "Support/GraphTraits.h"
 #include "Support/iterator"
@@ -28,7 +27,7 @@
   DSNodeIterator(NodeTy *N, bool) : Node(N) {         // Create end iterator
     Offset = N->getNumLinks() << DS::PointerShift;
     if (Offset == 0 && Node->getForwardNode() &&
-        (Node->NodeType & DSNode::DEAD))        // Model Forward link
+        Node->isDeadNode())        // Model Forward link
       Offset += DS::PointerSize;
   }
 public:
@@ -47,7 +46,7 @@
   }
   
   pointer operator*() const {
-    if (Node->NodeType & DSNode::DEAD)
+    if (Node->isDeadNode())
       return Node->getForwardNode();
     else
       return Node->getLink(Offset).getNode();


Index: llvm/include/llvm/Analysis/DSNode.h
diff -u llvm/include/llvm/Analysis/DSNode.h:1.24.2.1 llvm/include/llvm/Analysis/DSNode.h:1.24.2.2
--- llvm/include/llvm/Analysis/DSNode.h:1.24.2.1	Mon Jun 23 13:51:47 2003
+++ llvm/include/llvm/Analysis/DSNode.h	Thu Jun 26 16:34:42 2003
@@ -8,8 +8,8 @@
 #define LLVM_ANALYSIS_DSNODE_H
 
 #include "Config/assert.h"
-
 #include "llvm/Analysis/DSSupport.h"
+
 template<typename BaseType>
 class DSNodeIterator;          // Data structure graph traversal iterator
 
@@ -69,24 +69,27 @@
     GlobalNode  = 1 << 2,   // This node was allocated by a global var decl
     UnknownNode = 1 << 3,   // This node points to unknown allocated memory 
     Incomplete  = 1 << 4,   // This node may not be complete
+
     Modified    = 1 << 5,   // This node is modified in this context
     Read        = 1 << 6,   // This node is read in this context
+
     Array       = 1 << 7,   // This node is treated like an array
-#if 1
+    //#ifndef NDEBUG
     DEAD        = 1 << 8,   // This node is dead and should not be pointed to
-#endif
+    //#endif
 
     Composition = AllocaNode | HeapNode | GlobalNode | UnknownNode,
   };
   
   /// NodeType - A union of the above bits.  "Shadow" nodes do not add any flags
   /// to the nodes in the data structure graph, so it is possible to have nodes
-  /// with a value of 0 for their NodeType.  Scalar and Alloca markers go away
-  /// when function graphs are inlined.
+  /// with a value of 0 for their NodeType.
   ///
+private:
   unsigned short NodeType;
+public:
 
-  DSNode(unsigned NodeTy, const Type *T, DSGraph *G);
+  DSNode(const Type *T, DSGraph *G);
   DSNode(const DSNode &, DSGraph *G);
 
   ~DSNode() {
@@ -122,14 +125,6 @@
   /// return the number of nodes forwarding over the node!
   unsigned getNumReferrers() const { return NumReferrers; }
 
-  /// isModified - Return true if this node may be modified in this context
-  ///
-  bool isModified() const { return (NodeType & Modified) != 0; }
-
-  /// isRead - Return true if this node may be read in this context
-  ///
-  bool isRead() const { return (NodeType & Read) != 0; }
-
   DSGraph *getParentGraph() const { return ParentGraph; }
   void setParentGraph(DSGraph *G) { ParentGraph = G; }
 
@@ -234,7 +229,42 @@
   ///
   void addGlobal(GlobalValue *GV);
   const std::vector<GlobalValue*> &getGlobals() const { return Globals; }
-  std::vector<GlobalValue*> &getGlobals() { return Globals; }
+
+  /// maskNodeTypes - Apply a mask to the node types bitfield.
+  ///
+  void maskNodeTypes(unsigned Mask) {
+    NodeType &= Mask;
+  }
+
+  /// getNodeFlags - Return all of the flags set on the node.  If the DEAD flag
+  /// is set, hide it from the caller.
+  unsigned getNodeFlags() const { return NodeType & ~DEAD; }
+
+  bool isAllocaNode()  const { return NodeType & AllocaNode; }
+  bool isHeapNode()    const { return NodeType & HeapNode; }
+  bool isGlobalNode()  const { return NodeType & GlobalNode; }
+  bool isUnknownNode() const { return NodeType & UnknownNode; }
+
+  bool isModified() const   { return NodeType & Modified; }
+  bool isRead() const       { return NodeType & Read; }
+
+  bool isIncomplete() const { return NodeType & Incomplete; }
+  bool isDeadNode() const   { return NodeType & DEAD; }
+
+  DSNode *setAllocaNodeMarker()  { NodeType |= AllocaNode;  return this; }
+  DSNode *setHeapNodeMarker()    { NodeType |= HeapNode;    return this; }
+  DSNode *setGlobalNodeMarker()  { NodeType |= GlobalNode;  return this; }
+  DSNode *setUnknownNodeMarker() { NodeType |= UnknownNode; return this; }
+
+  DSNode *setIncompleteMarker() { NodeType |= Incomplete; return this; }
+  DSNode *setModifiedMarker()   { NodeType |= Modified;   return this; }
+  DSNode *setReadMarker()       { NodeType |= Read;       return this; }
+
+  void makeNodeDead() {
+    Globals.clear();
+    assert(hasNoReferrers() && "Dead node shouldn't have refs!");
+    NodeType = DEAD;
+  }
 
   /// forwardNode - Mark this node as being obsolete, and all references to it
   /// should be forwarded to the specified node and offset.
@@ -295,7 +325,6 @@
     }
   }
   assert(!N || ((N->NodeType & DSNode::DEAD) == 0));
-
   assert((!N || Offset < N->Size || (N->Size == 0 && Offset == 0) ||
           !N->ForwardNH.isNull()) && "Node handle offset out of range!");
 }


Index: llvm/include/llvm/Analysis/DSSupport.h
diff -u llvm/include/llvm/Analysis/DSSupport.h:1.16.2.1 llvm/include/llvm/Analysis/DSSupport.h:1.16.2.2
--- llvm/include/llvm/Analysis/DSSupport.h:1.16.2.1	Mon Jun 23 13:51:47 2003
+++ llvm/include/llvm/Analysis/DSSupport.h	Thu Jun 26 16:34:42 2003
@@ -8,10 +8,10 @@
 #define LLVM_ANALYSIS_DSSUPPORT_H
 
 #include "Config/assert.h"
-
 #include <vector>
 #include <functional>
 #include <string>
+#include <assert.h>
 #include "Support/HashExtras.h"
 #include "Support/hash_set"
 


Index: llvm/include/llvm/Analysis/DataStructure.h
diff -u llvm/include/llvm/Analysis/DataStructure.h:1.63.2.1 llvm/include/llvm/Analysis/DataStructure.h:1.63.2.2
--- llvm/include/llvm/Analysis/DataStructure.h:1.63.2.1	Mon Jun 23 13:51:47 2003
+++ llvm/include/llvm/Analysis/DataStructure.h	Thu Jun 26 16:34:42 2003
@@ -8,7 +8,6 @@
 #define LLVM_ANALYSIS_DATA_STRUCTURE_H
 
 #include "Config/assert.h"
-
 #include "llvm/Pass.h"
 #include "Support/HashExtras.h"
 #include "Support/hash_set"


Index: llvm/include/llvm/Analysis/DependenceGraph.h
diff -u llvm/include/llvm/Analysis/DependenceGraph.h:1.4.2.1 llvm/include/llvm/Analysis/DependenceGraph.h:1.4.2.2
--- llvm/include/llvm/Analysis/DependenceGraph.h:1.4.2.1	Mon Jun 23 13:51:47 2003
+++ llvm/include/llvm/Analysis/DependenceGraph.h	Thu Jun 26 16:34:42 2003
@@ -14,7 +14,6 @@
 //
 //===----------------------------------------------------------------------===//
 
-
 #ifndef LLVM_ANALYSIS_DEPENDENCEGRAPH_H
 #define LLVM_ANALYSIS_DEPENDENCEGRAPH_H
 


Index: llvm/include/llvm/Analysis/InstForest.h
diff -u llvm/include/llvm/Analysis/InstForest.h:1.17.2.1 llvm/include/llvm/Analysis/InstForest.h:1.17.2.2
--- llvm/include/llvm/Analysis/InstForest.h:1.17.2.1	Mon Jun 23 13:51:47 2003
+++ llvm/include/llvm/Analysis/InstForest.h	Thu Jun 26 16:34:42 2003
@@ -14,8 +14,6 @@
 #ifndef LLVM_ANALYSIS_INSTFOREST_H
 #define LLVM_ANALYSIS_INSTFOREST_H
 
-#include "llvm/Instruction.h"
-#include "llvm/BasicBlock.h"
 #include "llvm/Function.h"
 #include "Support/Tree.h"
 #include <map>


Index: llvm/include/llvm/Analysis/IntervalIterator.h
diff -u llvm/include/llvm/Analysis/IntervalIterator.h:1.11.2.1 llvm/include/llvm/Analysis/IntervalIterator.h:1.11.2.2
--- llvm/include/llvm/Analysis/IntervalIterator.h:1.11.2.1	Mon Jun 23 13:51:47 2003
+++ llvm/include/llvm/Analysis/IntervalIterator.h	Thu Jun 26 16:34:42 2003
@@ -28,7 +28,6 @@
 
 #include "llvm/Analysis/IntervalPartition.h"
 #include "llvm/Function.h"
-#include "llvm/BasicBlock.h"
 #include "llvm/Support/CFG.h"
 #include <stack>
 #include <set>


Index: llvm/include/llvm/Analysis/MemoryDepAnalysis.h
diff -u llvm/include/llvm/Analysis/MemoryDepAnalysis.h:1.2.2.1 llvm/include/llvm/Analysis/MemoryDepAnalysis.h:1.2.2.2
--- llvm/include/llvm/Analysis/MemoryDepAnalysis.h:1.2.2.1	Mon Jun 23 13:51:47 2003
+++ llvm/include/llvm/Analysis/MemoryDepAnalysis.h	Thu Jun 26 16:34:42 2003
@@ -7,6 +7,7 @@
 //
 // The result of this pass is a DependenceGraph for each function
 // representing the memory-based data dependences between instructions.
+//
 //===----------------------------------------------------------------------===//
 
 #ifndef LLVM_ANALYSIS_MEMORYDEPANALYSIS_H
@@ -17,16 +18,11 @@
 #include "llvm/Analysis/DataStructure.h"
 #include "llvm/Pass.h"
 #include "Support/TarjanSCCIterator.h"
-#include "Support/NonCopyable.h"
 #include "Support/hash_map"
-
 #include "Config/assert.h"
 
 class Instruction;
-class Function;
-class DSGraph;
 class ModRefTable;
-
 
 ///---------------------------------------------------------------------------
 /// class MemoryDepGraph:


Index: llvm/include/llvm/Analysis/PgmDependenceGraph.h
diff -u llvm/include/llvm/Analysis/PgmDependenceGraph.h:1.2.2.1 llvm/include/llvm/Analysis/PgmDependenceGraph.h:1.2.2.2
--- llvm/include/llvm/Analysis/PgmDependenceGraph.h:1.2.2.1	Mon Jun 23 13:51:47 2003
+++ llvm/include/llvm/Analysis/PgmDependenceGraph.h	Thu Jun 26 16:34:42 2003
@@ -26,6 +26,7 @@
 //
 // class PgmDependenceGraph -- Interface to obtain PDGIterators for each
 //                          instruction.
+//
 //===----------------------------------------------------------------------===//
 
 #ifndef LLVM_ANALYSIS_PGMDEPENDENCEGRAPH_H
@@ -35,15 +36,10 @@
 #include "llvm/Analysis/MemoryDepAnalysis.h"
 /* #include "llvm/Analysis/PostDominators.h" -- see below */
 #include "llvm/Instruction.h"
-#include "llvm/Value.h"
 #include "llvm/Pass.h"
-#include "Support/NonCopyable.h"
 #include <iterator>
 
 #include "Config/assert.h"
-
-class Instruction;
-class Function;
 class DSGraph;
 class DependenceGraph;
 class PgmDependenceGraph;




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:44:11 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:44:11 2003
Subject: [llvm-commits] CVS: llvm/test/Programs/External/SPEC/CFP2000/Makefile
Message-ID: <200306262137.h5QLbLp32534@choi.cs.uiuc.edu>


Changes in directory llvm/test/Programs/External/SPEC/CFP2000:

Makefile updated: 1.1 -> 1.1.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/test/Programs/External/SPEC/CFP2000/Makefile
diff -u llvm/test/Programs/External/SPEC/CFP2000/Makefile:1.1 llvm/test/Programs/External/SPEC/CFP2000/Makefile:1.1.2.1
--- llvm/test/Programs/External/SPEC/CFP2000/Makefile:1.1	Sun Jun  1 21:45:11 2003
+++ llvm/test/Programs/External/SPEC/CFP2000/Makefile	Thu Jun 26 16:35:41 2003
@@ -1,3 +1,3 @@
 LEVEL = ../../../../..
-DIRS := $(sort $(filter-out CVS/, $(wildcard */)))
+PARALLEL_DIRS := $(filter-out %-disabled/, $(sort $(filter-out CVS/, $(wildcard */))))
 include ${LEVEL}/test/Makefile.tests




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:44:21 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:44:21 2003
Subject: [llvm-commits] CVS: llvm/www/docs/CommandLine.html LangRef.html OpenProjects.html
Message-ID: <200306262137.h5QLb2n32519@choi.cs.uiuc.edu>


Changes in directory llvm/www/docs:

CommandLine.html updated: 1.9 -> 1.9.2.1
LangRef.html updated: 1.23 -> 1.23.2.1
OpenProjects.html updated: 1.6 -> 1.6.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/www/docs/CommandLine.html
diff -u llvm/www/docs/CommandLine.html:1.9 llvm/www/docs/CommandLine.html:1.9.2.1
--- llvm/www/docs/CommandLine.html:1.9	Mon Jun  2 23:40:06 2003
+++ llvm/www/docs/CommandLine.html	Thu Jun 26 16:36:22 2003
@@ -94,8 +94,9 @@
 <li>Speed: The CommandLine library is very quick and uses little resources.  The
 parsing time of the library is directly proportional to the number of arguments
 parsed, not the the number of options recognized.  Additionally, command line
-argument values are captured transparently into user defined variables, which
-can be accessed like any other variable (and with the same performance).<p>
+argument values are captured transparently into user defined global variables,
+which can be accessed like any other variable (and with the same
+performance).<p>
 
 <li>Type Safe: As a user of CommandLine, you don't have to worry about
 remembering the type of arguments that you want (is it an int?  a string? a
@@ -183,24 +184,24 @@
 
 Now that you are ready to support command line arguments, we need to tell the
 system which ones we want, and what type of argument they are.  The CommandLine
-library uses a declarative syntax to model cammand line arguments with the
-variable declarations that capture the parsed values.  This means that for every
-command line option that you would like to support, there should be a variable
-declaration to capture the result.  For example, in a compiler, we would like to
-support the unix standard '<tt>-o &lt;filename&gt;</tt>' option to specify where
-to put the output.  With the CommandLine library, this is represented like
-this:<p>
+library uses a declarative syntax to model command line arguments with the
+global variable declarations that capture the parsed values.  This means that
+for every command line option that you would like to support, there should be a
+global variable declaration to capture the result.  For example, in a compiler,
+we would like to support the unix standard '<tt>-o &lt;filename&gt;</tt>' option
+to specify where to put the output.  With the CommandLine library, this is
+represented like this:<p>
 
 <pre><a name="value_desc_example">
 <a href="#cl::opt">cl::opt</a>&lt;string&gt; OutputFilename("<i>o</i>", <a href="#cl::desc">cl::desc</a>("<i>Specify output filename</i>"), <a href="#cl::value_desc">cl::value_desc</a>("<i>filename</i>"));
 </pre><p>
 
-This declares a variable "<tt>OutputFilename</tt>" that is used to capture the
-result of the "<tt>o</tt>" argument (first parameter).  We specify that this is
-a simple scalar option by using the "<tt><a href="#cl::opt">cl::opt</a></tt>"
-template (as opposed to the <a href="#list">"<tt>cl::list</tt>
-template</a>), and tell the CommandLine library that the data type that we are
-parsing is a string.<p>
+This declares a global variable "<tt>OutputFilename</tt>" that is used to
+capture the result of the "<tt>o</tt>" argument (first parameter).  We specify
+that this is a simple scalar option by using the "<tt><a
+href="#cl::opt">cl::opt</a></tt>" template (as opposed to the <a
+href="#list">"<tt>cl::list</tt> template</a>), and tell the CommandLine library
+that the data type that we are parsing is a string.<p>
 
 The second and third parameters (which are optional) are used to specify what to
 output for the "<tt>--help</tt>" option.  In this case, we get a line that looks
@@ -1495,7 +1496,7 @@
 <address><a href="mailto:sabre at nondot.org">Chris Lattner</a></address>
 <!-- Created: Tue Jan 23 15:19:28 CST 2001 -->
 <!-- hhmts start -->
-Last modified: Mon Jun  2 23:39:44 CDT 2003
+Last modified: Sat Jun 21 16:45:29 CDT 2003
 <!-- hhmts end -->
 </font>
 </body></html>


Index: llvm/www/docs/LangRef.html
diff -u llvm/www/docs/LangRef.html:1.23 llvm/www/docs/LangRef.html:1.23.2.1
--- llvm/www/docs/LangRef.html:1.23	Thu May  8 10:55:44 2003
+++ llvm/www/docs/LangRef.html	Thu Jun 26 16:36:22 2003
@@ -295,7 +295,7 @@
 <table border=1 cellspacing=0 cellpadding=4 align=center>
 <tr><td><a name="t_signed">signed</td>    <td><tt>sbyte, short, int, long, float, double</tt></td></tr>
 <tr><td><a name="t_unsigned">unsigned</td><td><tt>ubyte, ushort, uint, ulong</tt></td></tr>
-<tr><td><a name="t_integral">integer</td><td><tt>ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td></tr>
+<tr><td><a name="t_integer">integer</td><td><tt>ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td></tr>
 <tr><td><a name="t_integral">integral</td><td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td></tr>
 <tr><td><a name="t_floating">floating point</td><td><tt>float, double</tt></td></tr>
 <tr><td><a name="t_firstclass">first class</td><td><tt>bool, ubyte, sbyte, ushort, short,<br> uint, int, ulong, long, float, double, <a href="#t_pointer">pointer</a></tt></td></tr>
@@ -1193,9 +1193,10 @@
 <h5>Example:</h5>
 <pre>
   &lt;result&gt; = shr int 4, ubyte %var   <i>; yields {int}:result = 4 >> %var</i>
-  &lt;result&gt; = shr int 4, ubyte 1      <i>; yields {int}:result = 2</i>
+  &lt;result&gt; = shr uint 4, ubyte 1     <i>; yields {uint}:result = 2</i>
   &lt;result&gt; = shr int 4, ubyte 2      <i>; yields {int}:result = 1</i>
-  &lt;result&gt; = shr int 4, ubyte 3      <i>; yields {int}:result = 0</i>
+  &lt;result&gt; = shr sbyte 4, ubyte 3    <i>; yields {sbyte}:result = 0</i>
+  &lt;result&gt; = shr sbyte -2, ubyte 1   <i>; yields {sbyte}:result = -1</i>
 </pre>
 
 
@@ -1810,7 +1811,7 @@
 <address><a href="mailto:sabre at nondot.org">Chris Lattner</a></address>
 <!-- Created: Tue Jan 23 15:19:28 CST 2001 -->
 <!-- hhmts start -->
-Last modified: Thu May  8 10:48:46 CDT 2003
+Last modified: Wed Jun 18 16:29:55 CDT 2003
 <!-- hhmts end -->
 </font>
 </body></html>


Index: llvm/www/docs/OpenProjects.html
diff -u llvm/www/docs/OpenProjects.html:1.6 llvm/www/docs/OpenProjects.html:1.6.2.1
--- llvm/www/docs/OpenProjects.html:1.6	Tue May  6 15:47:11 2003
+++ llvm/www/docs/OpenProjects.html	Thu Jun 26 16:36:22 2003
@@ -130,7 +130,6 @@
 <li>Transform <tt>setjmp</tt> and <tt>longjmp</tt> calls to use the LLVM
     <tt>invoke</tt> mechanism.
 <li>Add support for a volatile attribute on loads and stores
-<li>Support for variable argument functions
 <li>Add a new conditional move instruction: <tt>X = select bool Cond, Y, Z</tt>
 <li>Add support for platform independent prefetch support.  The GCC <a
     href="http://gcc.gnu.org/projects/prefetch.html">prefetch project</a> page
@@ -192,7 +191,7 @@
 <li>Implement a flow-sensitive context-sensitive alias analysis algorithm<br>
   - Pick one of the somewhat efficient algorithms, but strive for maximum
     precision
-<li>Implement a flow-sensitive context-insensitive alias anlaysis algorithm<br>
+<li>Implement a flow-sensitive context-insensitive alias analysis algorithm<br>
   - Just an efficient local algorithm perhaps?
 
 <li>Implement an interface to update analyses in response to common code motion
@@ -200,15 +199,6 @@
 <li>Implement alias analysis based optimizations:
   <ul>
   <li>Dead store elimination
-  <li>Store+Reload or "store forwarding" elimination:<p>
-      Change:
-       <pre>
-        store int X, int* P
-        Y = load int* P</pre>
-      into:
-       <pre>
-        store int X, int *P
-        Y = X</pre>
   </ul>
 </ol>
 
@@ -290,6 +280,6 @@
 <address><a href="mailto:sabre at nondot.org">Chris Lattner</a></address>
 <!-- Created: Tue Aug  6 15:00:33 CDT 2002 -->
 <!-- hhmts start -->
-Last modified: Tue May  6 15:46:48 CDT 2003
+Last modified: Wed Jun 25 09:30:27 CDT 2003
 <!-- hhmts end -->
 </font></body></html>




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:44:29 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:44:29 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/IPO/DeadArgumentElimination.cpp FunctionResolution.cpp GlobalDCE.cpp Internalize.cpp PoolAllocate.cpp
Message-ID: <200306262137.h5QLbbw00422@choi.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/IPO:

DeadArgumentElimination.cpp added (r1.3.2.1)
FunctionResolution.cpp updated: 1.31 -> 1.31.2.1
GlobalDCE.cpp updated: 1.23 -> 1.23.2.1
Internalize.cpp updated: 1.16 -> 1.16.2.1
PoolAllocate.cpp updated: 1.8 -> 1.8.2.1

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/lib/Transforms/IPO/DeadArgumentElimination.cpp
diff -c /dev/null llvm/lib/Transforms/IPO/DeadArgumentElimination.cpp:1.1
*** /dev/null	Thu Jun 26 16:37:37 2003
--- llvm/lib/Transforms/IPO/DeadArgumentElimination.cpp	Tue Jun 17 17:21:05 2003
***************
*** 0 ****
--- 1,304 ----
+ //===-- DeadArgumentElimination.cpp - Eliminate dead arguments ------------===//
+ //
+ // This pass deletes dead arguments from internal functions.  Dead argument
+ // elimination removes arguments which are directly dead, as well as arguments
+ // only passed into function calls as dead arguments of other functions.
+ //
+ // This pass is often useful as a cleanup pass to run after aggressive
+ // interprocedural passes, which add possibly-dead arguments.
+ //
+ //===----------------------------------------------------------------------===//
+ 
+ #include "llvm/Transforms/IPO.h"
+ #include "llvm/Module.h"
+ #include "llvm/Pass.h"
+ #include "llvm/DerivedTypes.h"
+ #include "llvm/Constant.h"
+ #include "llvm/iOther.h"
+ #include "llvm/iTerminators.h"
+ #include "llvm/Support/CallSite.h"
+ #include "Support/Statistic.h"
+ #include "Support/iterator"
+ #include <set>
+ 
+ namespace {
+   Statistic<> NumArgumentsEliminated("deadargelim", "Number of args removed");
+ 
+   struct DAE : public Pass {
+     bool run(Module &M);
+   };
+   RegisterOpt<DAE> X("deadargelim", "Dead Argument Elimination");
+ }
+ 
+ // createDeadArgEliminationPass - This pass removes arguments from functions
+ // which are not used by the body of the function.
+ //
+ Pass *createDeadArgEliminationPass() { return new DAE(); }
+ 
+ 
+ // FunctionArgumentsIntrinsicallyAlive - Return true if the arguments of the
+ // specified function are intrinsically alive.
+ //
+ // We consider arguments of non-internal functions to be intrinsically alive as
+ // well as arguments to functions which have their "address taken".
+ //
+ static bool FunctionArgumentsIntrinsicallyAlive(const Function &F) {
+   if (!F.hasInternalLinkage()) return true;
+ 
+   for (Value::use_const_iterator I = F.use_begin(), E = F.use_end(); I!=E; ++I){
+     // If this use is anything other than a call site, the function is alive.
+     CallSite CS = CallSite::get(const_cast<User*>(*I));
+     if (!CS.getInstruction()) return true;  // Not a valid call site?
+ 
+     // If the function is PASSED IN as an argument, its address has been taken
+     for (CallSite::arg_iterator AI = CS.arg_begin(), E = CS.arg_end(); AI != E;
+          ++AI)
+       if (AI->get() == &F) return true;
+   }
+   return false;
+ }
+ 
+ namespace {
+   enum ArgumentLiveness { Alive, MaybeLive, Dead };
+ }
+ 
+ // getArgumentLiveness - Inspect an argument, determining if is known Alive
+ // (used in a computation), MaybeLive (only passed as an argument to a call), or
+ // Dead (not used).
+ static ArgumentLiveness getArgumentLiveness(const Argument &A) {
+   if (A.use_empty()) return Dead;  // First check, directly dead?
+ 
+   // Scan through all of the uses, looking for non-argument passing uses.
+   for (Value::use_const_iterator I = A.use_begin(), E = A.use_end(); I!=E;++I) {
+     CallSite CS = CallSite::get(const_cast<User*>(*I));
+     if (!CS.getInstruction()) {
+       // If its used by something that is not a call or invoke, it's alive!
+       return Alive;
+     }
+     // If it's an indirect call, mark it alive...
+     Function *Callee = CS.getCalledFunction();
+     if (!Callee) return Alive;
+ 
+     // FIXME: check to see if it's passed through a va_arg area
+   }
+ 
+   return MaybeLive;  // It must be used, but only as argument to a function
+ }
+ 
+ // isMaybeLiveArgumentNowAlive - Check to see if Arg is alive.  At this point,
+ // we know that the only uses of Arg are to be passed in as an argument to a
+ // function call.  Check to see if the formal argument passed in is in the
+ // LiveArguments set.  If so, return true.
+ //
+ static bool isMaybeLiveArgumentNowAlive(Argument *Arg,
+                                      const std::set<Argument*> &LiveArguments) {
+   for (Value::use_iterator I = Arg->use_begin(), E = Arg->use_end(); I!=E; ++I){
+     CallSite CS = CallSite::get(*I);
+ 
+     // We know that this can only be used for direct calls...
+     Function *Callee = cast<Function>(CS.getCalledValue());
+ 
+     // Loop over all of the arguments (because Arg may be passed into the call
+     // multiple times) and check to see if any are now alive...
+     CallSite::arg_iterator CSAI = CS.arg_begin();
+     for (Function::aiterator AI = Callee->abegin(), E = Callee->aend();
+          AI != E; ++AI, ++CSAI)
+       // If this is the argument we are looking for, check to see if it's alive
+       if (*CSAI == Arg && LiveArguments.count(AI))
+         return true;
+   }
+   return false;
+ }
+ 
+ // MarkArgumentLive - The MaybeLive argument 'Arg' is now known to be alive.
+ // Mark it live in the specified sets and recursively mark arguments in callers
+ // live that are needed to pass in a value.
+ //
+ static void MarkArgumentLive(Argument *Arg,
+                              std::set<Argument*> &MaybeLiveArguments,
+                              std::set<Argument*> &LiveArguments,
+                           const std::multimap<Function*, CallSite> &CallSites) {
+   DEBUG(std::cerr << "  MaybeLive argument now live: " << Arg->getName()<<"\n");
+   assert(MaybeLiveArguments.count(Arg) && !LiveArguments.count(Arg) &&
+          "Arg not MaybeLive?");
+   MaybeLiveArguments.erase(Arg);
+   LiveArguments.insert(Arg);
+   
+   // Loop over all of the call sites of the function, making any arguments
+   // passed in to provide a value for this argument live as necessary.
+   //
+   Function *Fn = Arg->getParent();
+   unsigned ArgNo = std::distance(Fn->abegin(), Function::aiterator(Arg));
+ 
+   std::multimap<Function*, CallSite>::const_iterator I =
+     CallSites.lower_bound(Fn);
+   for (; I != CallSites.end() && I->first == Fn; ++I) {
+     const CallSite &CS = I->second;
+     if (Argument *ActualArg = dyn_cast<Argument>(*(CS.arg_begin()+ArgNo)))
+       if (MaybeLiveArguments.count(ActualArg))
+         MarkArgumentLive(ActualArg, MaybeLiveArguments, LiveArguments,
+                          CallSites);
+   }
+ }
+ 
+ // RemoveDeadArgumentsFromFunction - We know that F has dead arguments, as
+ // specified by the DeadArguments list.  Transform the function and all of the
+ // callees of the function to not have these arguments.
+ //
+ static void RemoveDeadArgumentsFromFunction(Function *F,
+                                             std::set<Argument*> &DeadArguments){
+   // Start by computing a new prototype for the function, which is the same as
+   // the old function, but has fewer arguments.
+   const FunctionType *FTy = F->getFunctionType();
+   std::vector<const Type*> Params;
+ 
+   for (Function::aiterator I = F->abegin(), E = F->aend(); I != E; ++I)
+     if (!DeadArguments.count(I))
+       Params.push_back(I->getType());
+ 
+   FunctionType *NFTy = FunctionType::get(FTy->getReturnType(), Params,
+                                          FTy->isVarArg());
+   
+   // Create the new function body and insert it into the module...
+   Function *NF = new Function(NFTy, Function::InternalLinkage, F->getName());
+   F->getParent()->getFunctionList().insert(F, NF);
+ 
+   // Loop over all of the callers of the function, transforming the call sites
+   // to pass in a smaller number of arguments into the new function.
+   //
+   while (!F->use_empty()) {
+     CallSite CS = CallSite::get(F->use_back());
+     Instruction *Call = CS.getInstruction();
+     CS.setCalledFunction(NF);   // Reduce the uses count of F
+     
+     // Loop over the operands, deleting dead ones...
+     CallSite::arg_iterator AI = CS.arg_begin();
+     for (Function::aiterator I = F->abegin(), E = F->aend(); I != E; ++I)
+       if (DeadArguments.count(I)) {        // Remove operands for dead arguments
+         AI = Call->op_erase(AI);
+       }  else {
+         ++AI;  // Leave live operands alone...
+       }
+   }
+ 
+   // Since we have now created the new function, splice the body of the old
+   // function right into the new function, leaving the old rotting hulk of the
+   // function empty.
+   NF->getBasicBlockList().splice(NF->begin(), F->getBasicBlockList());
+ 
+   // Loop over the argument list, transfering uses of the old arguments over to
+   // the new arguments, also transfering over the names as well.  While we're at
+   // it, remove the dead arguments from the DeadArguments list.
+   //
+   for (Function::aiterator I = F->abegin(), E = F->aend(), I2 = NF->abegin();
+        I != E; ++I)
+     if (!DeadArguments.count(I)) {
+       // If this is a live argument, move the name and users over to the new
+       // version.
+       I->replaceAllUsesWith(I2);
+       I2->setName(I->getName());
+       ++I2;
+     } else {
+       // If this argument is dead, replace any uses of it with null constants
+       // (these are guaranteed to only be operands to call instructions which
+       // will later be simplified).
+       I->replaceAllUsesWith(Constant::getNullValue(I->getType()));
+       DeadArguments.erase(I);
+     }
+ 
+   // Now that the old function is dead, delete it.
+   F->getParent()->getFunctionList().erase(F);
+ }
+ 
+ bool DAE::run(Module &M) {
+   // First phase: loop through the module, determining which arguments are live.
+   // We assume all arguments are dead unless proven otherwise (allowing us to
+   // determing that dead arguments passed into recursive functions are dead).
+   //
+   std::set<Argument*> LiveArguments, MaybeLiveArguments, DeadArguments;
+   std::multimap<Function*, CallSite> CallSites;
+ 
+   DEBUG(std::cerr << "DAE - Determining liveness\n");
+   for (Module::iterator I = M.begin(), E = M.end(); I != E; ++I) {
+     Function &Fn = *I;
+     // If the function is intrinsically alive, just mark the arguments alive.
+     if (FunctionArgumentsIntrinsicallyAlive(Fn)) {
+       for (Function::aiterator AI = Fn.abegin(), E = Fn.aend(); AI != E; ++AI)
+         LiveArguments.insert(AI);
+       DEBUG(std::cerr << "  Args intrinsically live for fn: " << Fn.getName()
+                       << "\n");
+     } else {
+       DEBUG(std::cerr << "  Inspecting args for fn: " << Fn.getName() << "\n");
+ 
+       // If it is not intrinsically alive, we know that all users of the
+       // function are call sites.  Mark all of the arguments live which are
+       // directly used, and keep track of all of the call sites of this function
+       // if there are any arguments we assume that are dead.
+       //
+       bool AnyMaybeLiveArgs = false;
+       for (Function::aiterator AI = Fn.abegin(), E = Fn.aend(); AI != E; ++AI)
+         switch (getArgumentLiveness(*AI)) {
+         case Alive:
+           DEBUG(std::cerr << "    Arg live by use: " << AI->getName() << "\n");
+           LiveArguments.insert(AI);
+           break;
+         case Dead:
+           DEBUG(std::cerr << "    Arg definately dead: " <<AI->getName()<<"\n");
+           DeadArguments.insert(AI);
+           break;
+         case MaybeLive:
+           DEBUG(std::cerr << "    Arg only passed to calls: "
+                           << AI->getName() << "\n");
+           AnyMaybeLiveArgs = true;
+           MaybeLiveArguments.insert(AI);
+           break;
+         }
+ 
+       // If there are any "MaybeLive" arguments, we need to check callees of
+       // this function when/if they become alive.  Record which functions are
+       // callees...
+       if (AnyMaybeLiveArgs)
+         for (Value::use_iterator I = Fn.use_begin(), E = Fn.use_end();
+              I != E; ++I)
+           CallSites.insert(std::make_pair(&Fn, CallSite::get(*I)));
+     }
+   }
+ 
+   // Now we loop over all of the MaybeLive arguments, promoting them to be live
+   // arguments if one of the calls that uses the arguments to the calls they are
+   // passed into requires them to be live.  Of course this could make other
+   // arguments live, so process callers recursively.
+   //
+   // Because elements can be removed from the MaybeLiveArguments list, copy it
+   // to a temporary vector.
+   //
+   std::vector<Argument*> TmpArgList(MaybeLiveArguments.begin(),
+                                     MaybeLiveArguments.end());
+   for (unsigned i = 0, e = TmpArgList.size(); i != e; ++i) {
+     Argument *MLA = TmpArgList[i];
+     if (MaybeLiveArguments.count(MLA) &&
+         isMaybeLiveArgumentNowAlive(MLA, LiveArguments)) {
+       MarkArgumentLive(MLA, MaybeLiveArguments, LiveArguments, CallSites);
+     }
+   }
+ 
+   // Recover memory early...
+   CallSites.clear();
+ 
+   // At this point, we know that all arguments in DeadArguments and
+   // MaybeLiveArguments are dead.  If the two sets are empty, there is nothing
+   // to do.
+   if (MaybeLiveArguments.empty() && DeadArguments.empty())
+     return false;
+   
+   // Otherwise, compact into one set, and start eliminating the arguments from
+   // the functions.
+   DeadArguments.insert(MaybeLiveArguments.begin(), MaybeLiveArguments.end());
+   MaybeLiveArguments.clear();
+ 
+   NumArgumentsEliminated += DeadArguments.size();
+   while (!DeadArguments.empty())
+     RemoveDeadArgumentsFromFunction((*DeadArguments.begin())->getParent(),
+                                     DeadArguments);
+   return true;
+ }


Index: llvm/lib/Transforms/IPO/FunctionResolution.cpp
diff -u llvm/lib/Transforms/IPO/FunctionResolution.cpp:1.31 llvm/lib/Transforms/IPO/FunctionResolution.cpp:1.31.2.1
--- llvm/lib/Transforms/IPO/FunctionResolution.cpp:1.31	Sat May 31 16:57:06 2003
+++ llvm/lib/Transforms/IPO/FunctionResolution.cpp	Thu Jun 26 16:35:27 2003
@@ -36,94 +36,6 @@
   return new FunctionResolvingPass();
 }
 
-// ConvertCallTo - Convert a call to a varargs function with no arg types
-// specified to a concrete nonvarargs function.
-//
-static void ConvertCallTo(CallInst *CI, Function *Dest) {
-  const FunctionType::ParamTypes &ParamTys =
-    Dest->getFunctionType()->getParamTypes();
-  BasicBlock *BB = CI->getParent();
-
-  // Keep an iterator to where we want to insert cast instructions if the
-  // argument types don't agree.
-  //
-  unsigned NumArgsToCopy = CI->getNumOperands()-1;
-  if (NumArgsToCopy != ParamTys.size() &&
-      !(NumArgsToCopy > ParamTys.size() &&
-        Dest->getFunctionType()->isVarArg())) {
-    std::cerr << "WARNING: Call arguments do not match expected number of"
-              << " parameters.\n";
-    std::cerr << "WARNING: In function '"
-              << CI->getParent()->getParent()->getName() << "': call: " << *CI;
-    std::cerr << "Function resolved to: ";
-    WriteAsOperand(std::cerr, Dest);
-    std::cerr << "\n";
-    if (NumArgsToCopy > ParamTys.size())
-      NumArgsToCopy = ParamTys.size();
-  }
-
-  std::vector<Value*> Params;
-
-  // Convert all of the call arguments over... inserting cast instructions if
-  // the types are not compatible.
-  for (unsigned i = 1; i <= NumArgsToCopy; ++i) {
-    Value *V = CI->getOperand(i);
-
-    if (i-1 < ParamTys.size() && V->getType() != ParamTys[i-1]) {
-      // Must insert a cast...
-      V = new CastInst(V, ParamTys[i-1], "argcast", CI);
-    }
-
-    Params.push_back(V);
-  }
-  
-  // If the function takes extra parameters that are not being passed in, pass
-  // null values in now...
-  for (unsigned i = NumArgsToCopy; i < ParamTys.size(); ++i)
-    Params.push_back(Constant::getNullValue(ParamTys[i]));
-
-  // Replace the old call instruction with a new call instruction that calls
-  // the real function.
-  //
-  Instruction *NewCall = new CallInst(Dest, Params, "", CI);
-  std::string Name = CI->getName(); CI->setName("");
-
-  // Transfer the name over...
-  if (NewCall->getType() != Type::VoidTy)
-    NewCall->setName(Name);
-
-  // Replace uses of the old instruction with the appropriate values...
-  //
-  if (NewCall->getType() == CI->getType()) {
-    CI->replaceAllUsesWith(NewCall);
-    NewCall->setName(Name);
-
-  } else if (NewCall->getType() == Type::VoidTy) {
-    // Resolved function does not return a value but the prototype does.  This
-    // often occurs because undefined functions default to returning integers.
-    // Just replace uses of the call (which are broken anyway) with dummy
-    // values.
-    CI->replaceAllUsesWith(Constant::getNullValue(CI->getType()));
-  } else if (CI->getType() == Type::VoidTy) {
-    // If we are gaining a new return value, we don't have to do anything
-    // special here, because it will automatically be ignored.
-  } else {
-    // Insert a cast instruction to convert the return value of the function
-    // into it's new type.  Of course we only need to do this if the return
-    // value of the function is actually USED.
-    //
-    if (!CI->use_empty()) {
-      // Insert the new cast instruction...
-      CastInst *NewCast = new CastInst(NewCall, CI->getType(), Name, CI);
-      CI->replaceAllUsesWith(NewCast);
-    }
-  }
-
-  // The old instruction is no longer needed, destroy it!
-  BB->getInstList().erase(CI);
-}
-
-
 static bool ResolveFunctions(Module &M, std::vector<GlobalValue*> &Globals,
                              Function *Concrete) {
   bool Changed = false;
@@ -168,36 +80,12 @@
       // functions and that the Old function has no varargs fns specified.  In
       // otherwords it's just <retty> (...)
       //
-      for (unsigned i = 0; i < Old->use_size(); ) {
-        User *U = *(Old->use_begin()+i);
-        if (CastInst *CI = dyn_cast<CastInst>(U)) {
-          // Convert casts directly
-          assert(CI->getOperand(0) == Old);
-          CI->setOperand(0, Concrete);
-          Changed = true;
-          ++NumResolved;
-        } else if (CallInst *CI = dyn_cast<CallInst>(U)) {
-          // Can only fix up calls TO the argument, not args passed in.
-          if (CI->getCalledValue() == Old) {
-            ConvertCallTo(CI, Concrete);
-            Changed = true;
-            ++NumResolved;
-          } else {
-            ++i;
-          }
-        } else {
-          ++i;
-        }
-      }
-
-      // If there are any more uses that we could not resolve, force them to use
-      // a casted pointer now.
-      if (!Old->use_empty()) {
-        NumResolved += Old->use_size();
-        Constant *NewCPR = ConstantPointerRef::get(Concrete);
-        Old->replaceAllUsesWith(ConstantExpr::getCast(NewCPR, Old->getType()));
-        Changed = true;
-      }
+      Value *Replacement = Concrete;
+      if (Concrete->getType() != Old->getType())
+        Replacement = ConstantExpr::getCast(ConstantPointerRef::get(Concrete),
+                                            Old->getType());
+      NumResolved += Old->use_size();
+      Old->replaceAllUsesWith(Replacement);
 
       // Since there are no uses of Old anymore, remove it from the module.
       M.getFunctionList().erase(Old);


Index: llvm/lib/Transforms/IPO/GlobalDCE.cpp
diff -u llvm/lib/Transforms/IPO/GlobalDCE.cpp:1.23 llvm/lib/Transforms/IPO/GlobalDCE.cpp:1.23.2.1
--- llvm/lib/Transforms/IPO/GlobalDCE.cpp:1.23	Sun Oct 13 12:12:47 2002
+++ llvm/lib/Transforms/IPO/GlobalDCE.cpp	Thu Jun 26 16:35:27 2003
@@ -46,12 +46,18 @@
     // Nothing to do if no unreachable functions have been found...
     if (FunctionsToDelete.empty()) return false;
     
-    // Unreachables functions have been found and should have no references to
+    // Unreachable functions have been found and should have no references to
     // them, delete them now.
     //
     for (std::vector<CallGraphNode*>::iterator I = FunctionsToDelete.begin(),
            E = FunctionsToDelete.end(); I != E; ++I)
       delete CallGraph.removeFunctionFromModule(*I);
+
+    // Walk the function list, removing prototypes for functions which are not
+    // used.
+    for (Module::iterator I = M.begin(), E = M.end(); I != E; ++I)
+      if (I->use_size() == 0 && I->isExternal())
+        delete CallGraph.removeFunctionFromModule(I);
     
     return true;
   }


Index: llvm/lib/Transforms/IPO/Internalize.cpp
diff -u llvm/lib/Transforms/IPO/Internalize.cpp:1.16 llvm/lib/Transforms/IPO/Internalize.cpp:1.16.2.1
--- llvm/lib/Transforms/IPO/Internalize.cpp:1.16	Thu May 22 15:27:13 2003
+++ llvm/lib/Transforms/IPO/Internalize.cpp	Thu Jun 26 16:35:27 2003
@@ -87,6 +87,14 @@
       for (Module::giterator I = M.gbegin(), E = M.gend(); I != E; ++I)
         if (!I->isExternal() && !I->hasInternalLinkage() &&
             !ExternalNames.count(I->getName())) {
+          // Special case handling of the global ctor and dtor list.  When we
+          // internalize it, we mark it constant, which allows elimination of
+          // the list if it's empty.
+          //
+          if (I->hasAppendingLinkage() && (I->getName() == "llvm.global_ctors"||
+                                           I->getName() == "llvm.global_dtors"))
+            I->setConstant(true);
+
           I->setLinkage(GlobalValue::InternalLinkage);
           Changed = true;
           ++NumGlobals;


Index: llvm/lib/Transforms/IPO/PoolAllocate.cpp
diff -u llvm/lib/Transforms/IPO/PoolAllocate.cpp:1.8 llvm/lib/Transforms/IPO/PoolAllocate.cpp:1.8.2.1
--- llvm/lib/Transforms/IPO/PoolAllocate.cpp:1.8	Sat Jun  7 15:29:58 2003
+++ llvm/lib/Transforms/IPO/PoolAllocate.cpp	Thu Jun 26 16:35:27 2003
@@ -82,10 +82,10 @@
 	   CSE = callSites.end(); CSI != CSE ; ++CSI) {
       if (CSI->isIndirectCall()) {
 	DSNode *DSN = CSI->getCalleeNode();
-	if (DSN->NodeType == DSNode::Incomplete) 
+	if (DSN->isIncomplete())
 	  std::cerr << "Incomplete node " << CSI->getCallInst();
-	// assert(DSN->NodeType == DSNode::GlobalNode);
-	std::vector<GlobalValue*> &Callees = DSN->getGlobals();
+	// assert(DSN->isGlobalNode());
+	const std::vector<GlobalValue*> &Callees = DSN->getGlobals();
 	if (Callees.size() > 0) {
 	  Function *firstCalledF = dyn_cast<Function>(*Callees.begin());
 	  FuncECs.addElement(firstCalledF);
@@ -93,7 +93,7 @@
 				 (&CSI->getCallInst(),
 				  firstCalledF));
 	  if (Callees.size() > 1) {
-	    for (std::vector<GlobalValue*>::iterator CalleesI = 
+	    for (std::vector<GlobalValue*>::const_iterator CalleesI = 
 		   Callees.begin()+1, CalleesE = Callees.end(); 
 		 CalleesI != CalleesE; ++CalleesI) {
 	      Function *calledF = dyn_cast<Function>(*CalleesI);
@@ -229,13 +229,13 @@
   // Mark globals and incomplete nodes as live... (this handles arguments)
   if (F.getName() != "main")
     for (unsigned i = 0, e = Nodes.size(); i != e; ++i)
-      if (Nodes[i]->NodeType & (DSNode::GlobalNode | DSNode::Incomplete) &&
-          Nodes[i]->NodeType & (DSNode::HeapNode))
+      if ((Nodes[i]->isGlobalNode() || Nodes[i]->isIncomplete()) &&
+          Nodes[i]->isHeapNode())
         Nodes[i]->markReachableNodes(MarkedNodes);
 
   // Marked the returned node as alive...
   if (DSNode *RetNode = G.getRetNode().getNode())
-    if (RetNode->NodeType & DSNode::HeapNode)
+    if (RetNode->isHeapNode())
       RetNode->markReachableNodes(MarkedNodes);
 
   if (MarkedNodes.empty())   // We don't need to clone the function if there
@@ -411,7 +411,7 @@
   // ones to the NodesToPA vector.
   std::vector<DSNode*> NodesToPA;
   for (unsigned i = 0, e = Nodes.size(); i != e; ++i)
-    if (Nodes[i]->NodeType & DSNode::HeapNode &&   // Pick nodes with heap elems
+    if (Nodes[i]->isHeapNode() &&   // Pick nodes with heap elems
         !MarkedNodes.count(Nodes[i]))              // Can't be marked
       NodesToPA.push_back(Nodes[i]);
   
@@ -577,7 +577,7 @@
   if (!DSN) {
     return 0;
   }
-  std::vector<GlobalValue*> &Callees = DSN->getGlobals();
+  const std::vector<GlobalValue*> &Callees = DSN->getGlobals();
   if (Callees.size() > 0) {
     Function *calledF = dyn_cast<Function>(*Callees.begin());
     assert(PAInfo.FuncECs.findClass(calledF) && "should exist in some eq. class");




From criswell at choi.cs.uiuc.edu  Thu Jun 26 16:44:38 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Thu Jun 26 16:44:38 2003
Subject: [llvm-commits] CVS: llvm/tools/analyze/Makefile
Message-ID: <200306262137.h5QLbbT00428@choi.cs.uiuc.edu>


Changes in directory llvm/tools/analyze:

Makefile updated: 1.18.2.1 -> 1.18.2.2

---
Log message:

Merged with mainline on Thursday, June 26, 2003.


---
Diffs of the changes:

Index: llvm/tools/analyze/Makefile
diff -u llvm/tools/analyze/Makefile:1.18.2.1 llvm/tools/analyze/Makefile:1.18.2.2
--- llvm/tools/analyze/Makefile:1.18.2.1	Mon Jun 23 14:04:46 2003
+++ llvm/tools/analyze/Makefile	Thu Jun 26 16:35:57 2003
@@ -2,6 +2,7 @@
 TOOLNAME = analyze
 USEDLIBS = asmparser bcreader scalaropts.a transforms.a analysis ipa \
            datastructure target.a transformutils.a scalaropts.a vmcore support
+TOOLLINKOPTS = $(PLATFORMLIBDL)
 
 include $(LEVEL)/Makefile.common
 




From gaeke at cs.uiuc.edu  Thu Jun 26 19:01:01 2003
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Thu Jun 26 19:01:01 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Target/TargetInstrInfo.h
Message-ID: <200306270000.TAA32730@zion.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Target:

TargetInstrInfo.h updated: 1.45 -> 1.46

---
Log message:

Nice tasty llc fixes. These should fix LLC for x86 for everything in
SingleSource except oopack and Oscar.  (Sorry, Oscar.)

include/llvm/Target/TargetInstrInfo.h: Remove virtual print method. Add
 accessors for ImplicitUses/Defs.
lib/Target/TargetInstrInfo.cpp: Remove virtual print method. If you
 really wanted this, just use MI->print(O, TM); instead...
lib/Target/X86:
FloatingPoint.cpp: ...like this.
X86InstrInfo.h: Remove virtual print method. Define the PrintImplUses
 target-specific flag bit.
X86InstrInfo.def: Add the PrintImplUses flag to all the instructions
 which implicitly use CL, because the assembler needs to see the CL in
 order to generate the right instruction.
Printer.cpp: Ditch fnIndex at Chris's request. Now we use CurrentFnName
  to name constants in the constant pool for each function instead. This
  avoids keeping state between runOnMachineFunction() invocations, which
  is a no-no. Having MangledGlobals be global is a bogon I'd like to get
  rid of too, but making it a static member of Printer causes link errors
  (why???).
 Make NumberForBB into a member of Printer instead of a global, too.
 Make printOp and printMemReference into methods of Printer.
 X86InstrInfo::print is now Printer::printMachineInstruction, because
  TargetInstrInfo::print is history. (Because of this, we have to qualify
  the names of some TargetInstrInfo methods we call.)
 Print out the ImplicitUses field of any instruction we print that has
  the PrintImplUses bit set. 


---
Diffs of the changes:

Index: llvm/include/llvm/Target/TargetInstrInfo.h
diff -u llvm/include/llvm/Target/TargetInstrInfo.h:1.45 llvm/include/llvm/Target/TargetInstrInfo.h:1.46
--- llvm/include/llvm/Target/TargetInstrInfo.h:1.45	Wed Jun 11 09:01:36 2003
+++ llvm/include/llvm/Target/TargetInstrInfo.h	Thu Jun 26 19:00:48 2003
@@ -109,13 +109,6 @@
     return desc[opCode];
   }
 
-  /// print - Print out the specified machine instruction in the appropriate
-  /// target specific assembly language.  If this method is not overridden, the
-  /// default implementation uses the crummy machine independant printer.
-  ///
-  virtual void print(const MachineInstr *MI, std::ostream &O,
-                     const TargetMachine &TM) const;
-
   const char *getName(MachineOpCode opCode) const {
     return get(opCode).Name;
   }
@@ -135,7 +128,15 @@
   InstrSchedClass getSchedClass(MachineOpCode opCode) const {
     return get(opCode).schedClass;
   }
-  
+
+  const unsigned *getImplicitUses(MachineOpCode opCode) const {
+    return get(opCode).ImplicitUses;
+  }
+
+  const unsigned *getImplicitDefs(MachineOpCode opCode) const {
+    return get(opCode).ImplicitDefs;
+  }
+
   //
   // Query instruction class flags according to the machine-independent
   // flags listed above.




From gaeke at cs.uiuc.edu  Thu Jun 26 19:01:18 2003
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Thu Jun 26 19:01:18 2003
Subject: [llvm-commits] CVS: llvm/lib/Target/X86/FloatingPoint.cpp Printer.cpp X86InstrInfo.def X86InstrInfo.h
Message-ID: <200306270000.TAA32723@zion.cs.uiuc.edu>


Changes in directory llvm/lib/Target/X86:

FloatingPoint.cpp updated: 1.2 -> 1.3
Printer.cpp updated: 1.40 -> 1.41
X86InstrInfo.def updated: 1.60 -> 1.61
X86InstrInfo.h updated: 1.23 -> 1.24

---
Log message:

Nice tasty llc fixes. These should fix LLC for x86 for everything in
SingleSource except oopack and Oscar.  (Sorry, Oscar.)

include/llvm/Target/TargetInstrInfo.h: Remove virtual print method. Add
 accessors for ImplicitUses/Defs.
lib/Target/TargetInstrInfo.cpp: Remove virtual print method. If you
 really wanted this, just use MI->print(O, TM); instead...
lib/Target/X86:
FloatingPoint.cpp: ...like this.
X86InstrInfo.h: Remove virtual print method. Define the PrintImplUses
 target-specific flag bit.
X86InstrInfo.def: Add the PrintImplUses flag to all the instructions
 which implicitly use CL, because the assembler needs to see the CL in
 order to generate the right instruction.
Printer.cpp: Ditch fnIndex at Chris's request. Now we use CurrentFnName
  to name constants in the constant pool for each function instead. This
  avoids keeping state between runOnMachineFunction() invocations, which
  is a no-no. Having MangledGlobals be global is a bogon I'd like to get
  rid of too, but making it a static member of Printer causes link errors
  (why???).
 Make NumberForBB into a member of Printer instead of a global, too.
 Make printOp and printMemReference into methods of Printer.
 X86InstrInfo::print is now Printer::printMachineInstruction, because
  TargetInstrInfo::print is history. (Because of this, we have to qualify
  the names of some TargetInstrInfo methods we call.)
 Print out the ImplicitUses field of any instruction we print that has
  the PrintImplUses bit set. 


---
Diffs of the changes:

Index: llvm/lib/Target/X86/FloatingPoint.cpp
diff -u llvm/lib/Target/X86/FloatingPoint.cpp:1.2 llvm/lib/Target/X86/FloatingPoint.cpp:1.3
--- llvm/lib/Target/X86/FloatingPoint.cpp:1.2	Tue Jan 14 15:59:16 2003
+++ llvm/lib/Target/X86/FloatingPoint.cpp	Thu Jun 26 19:00:45 2003
@@ -196,8 +196,8 @@
 	    MachineBasicBlock::iterator Start = I;
 	    // Rewind to first instruction newly inserted.
 	    while (Start != BB.begin() && *(Start-1) != PrevMI) --Start;
-	    std::cerr << "Inserted instructions:\n";
-	    do TII.print(*Start, std::cerr << "\t", MF.getTarget());
+	    std::cerr << "Inserted instructions:\n\t";
+	    (*Start)->print(std::cerr, MF.getTarget());
 	    while (++Start != I+1);
 	  }
 	  dumpStack();


Index: llvm/lib/Target/X86/Printer.cpp
diff -u llvm/lib/Target/X86/Printer.cpp:1.40 llvm/lib/Target/X86/Printer.cpp:1.41
--- llvm/lib/Target/X86/Printer.cpp:1.40	Thu Jun 26 13:02:30 2003
+++ llvm/lib/Target/X86/Printer.cpp	Thu Jun 26 19:00:46 2003
@@ -24,16 +24,25 @@
 #include "llvm/Module.h"
 
 namespace {
-  unsigned fnIndex;
-  std::set<const Value*> MangledGlobals;
+  std::set<const Value *> MangledGlobals;
   struct Printer : public MachineFunctionPass {
     std::ostream &O;
+    typedef std::map<const Value *, unsigned> ValueMapTy;
+    ValueMapTy NumberForBB;
     Printer(std::ostream &o) : O(o) {}
     const TargetData *TD;
+    std::string CurrentFnName;
     virtual const char *getPassName() const {
       return "X86 Assembly Printer";
     }
 
+    void printMachineInstruction(const MachineInstr *MI, std::ostream &O,
+				 const TargetMachine &TM) const;
+    void printOp(std::ostream &O, const MachineOperand &MO,
+		 const MRegisterInfo &RI, bool elideOffsetKeyword = false) const;
+    void printMemReference(std::ostream &O, const MachineInstr *MI,
+			   unsigned Op,
+			   const MRegisterInfo &RI) const;
     void printConstantPool(MachineConstantPool *MCP);
     bool runOnMachineFunction(MachineFunction &F);    
     std::string ConstantExprToString(const ConstantExpr* CE);
@@ -44,8 +53,7 @@
     void printConstantValueOnly(const Constant* CV, int numPadBytesAfter = 0);
     void printSingleConstantValue(const Constant* CV);
   };
-  std::map<const Value *, unsigned> NumberForBB;
-}
+} // end of anonymous namespace
 
 /// createX86CodePrinterPass - Print out the specified machine code function to
 /// the specified stream.  This function should work regardless of whether or
@@ -55,8 +63,8 @@
   return new Printer(O);
 }
 
-// We dont want identifier names with ., space, -  in them. 
-// So we replace them with _
+// We don't want identifier names with ., space, or - in them, 
+// so we replace them with underscores.
 static std::string makeNameProper(std::string x) {
   std::string tmp;
   for (std::string::iterator sI = x.begin(), sEnd = x.end(); sI != sEnd; sI++)
@@ -66,38 +74,32 @@
     case '-': tmp += "D_"; break;
     default:  tmp += *sI;
     }
-
   return tmp;
 }
 
-std::string getValueName(const Value *V) {
+static std::string getValueName(const Value *V) {
   if (V->hasName()) { // Print out the label if it exists...
-    
     // Name mangling occurs as follows:
     // - If V is not a global, mangling always occurs.
     // - Otherwise, mangling occurs when any of the following are true:
     //   1) V has internal linkage
     //   2) V's name would collide if it is not mangled.
     //
-    
     if(const GlobalValue* gv = dyn_cast<GlobalValue>(V)) {
       if(!gv->hasInternalLinkage() && !MangledGlobals.count(gv)) {
         // No internal linkage, name will not collide -> no mangling.
         return makeNameProper(gv->getName());
       }
     }
-    
     // Non-global, or global with internal linkage / colliding name -> mangle.
     return "l" + utostr(V->getType()->getUniqueID()) + "_" +
       makeNameProper(V->getName());      
   }
-
   static int Count = 0;
   Count++;
   return "ltmp_" + itostr(Count) + "_" + utostr(V->getType()->getUniqueID());
 }
 
-
 // valToExprString - Helper function for ConstantExprToString().
 // Appends result to argument string S.
 // 
@@ -391,7 +393,8 @@
     O << "\t.section .rodata\n";
     O << "\t.align " << (unsigned)TD->getTypeAlignment(CP[i]->getType())
       << "\n";
-    O << ".CPI" << fnIndex << "_" << i << ":\t\t\t\t\t#" << *CP[i] << "\n";
+    O << ".CPI" << CurrentFnName << "_" << i << ":\t\t\t\t\t#"
+      << *CP[i] << "\n";
     printConstantValueOnly (CP[i]);
   }
 }
@@ -404,16 +407,21 @@
   const TargetInstrInfo &TII = TM.getInstrInfo();
   TD = &TM.getTargetData();
 
+  // What's my mangled name?
+  CurrentFnName = getValueName(MF.getFunction());
+
   // Print out constants referenced by the function
   printConstantPool(MF.getConstantPool());
 
   // Print out labels for the function.
   O << "\t.text\n";
   O << "\t.align 16\n";
-  O << "\t.globl\t" << getValueName(MF.getFunction()) << "\n";
-  O << "\t.type\t" << getValueName(MF.getFunction()) << ", @function\n";
-  O << getValueName(MF.getFunction()) << ":\n";
+  O << "\t.globl\t" << CurrentFnName << "\n";
+  O << "\t.type\t" << CurrentFnName << ", @function\n";
+  O << CurrentFnName << ":\n";
 
+  // Number each basic block so that we can consistently refer to them
+  // in PC-relative references.
   NumberForBB.clear();
   for (MachineFunction::const_iterator I = MF.begin(), E = MF.end();
        I != E; ++I) {
@@ -430,11 +438,10 @@
 	 II != E; ++II) {
       // Print the assembly for the instruction.
       O << "\t";
-      TII.print(*II, O, TM);
+      printMachineInstruction(*II, O, TM);
     }
   }
 
-  fnIndex++;
   // We didn't modify anything.
   return false;
 }
@@ -453,8 +460,9 @@
     MI->getOperand(Op+2).isRegister() &&MI->getOperand(Op+3).isImmediate();
 }
 
-static void printOp(std::ostream &O, const MachineOperand &MO,
-                    const MRegisterInfo &RI, bool elideOffsetKeyword = false) {
+void Printer::printOp(std::ostream &O, const MachineOperand &MO,
+		      const MRegisterInfo &RI,
+		      bool elideOffsetKeyword /* = false */) const {
   switch (MO.getType()) {
   case MachineOperand::MO_VirtualRegister:
     if (Value *V = MO.getVRegValueOrNull()) {
@@ -474,8 +482,12 @@
     O << (int)MO.getImmedValue();
     return;
   case MachineOperand::MO_PCRelativeDisp:
-    O << ".BB" << NumberForBB[MO.getVRegValue()] << " # PC rel: "
-      << MO.getVRegValue()->getName();
+    {
+      ValueMapTy::const_iterator i = NumberForBB.find(MO.getVRegValue());
+      assert (i != NumberForBB.end()
+	      && "Could not find a BB I previously put in the NumberForBB map!");
+      O << ".BB" << i->second << " # PC rel: " << MO.getVRegValue()->getName();
+    }
     return;
   case MachineOperand::MO_GlobalAddress:
     if (!elideOffsetKeyword) O << "OFFSET "; O << getValueName(MO.getGlobal());
@@ -501,8 +513,9 @@
   }
 }
 
-static void printMemReference(std::ostream &O, const MachineInstr *MI,
-                              unsigned Op, const MRegisterInfo &RI) {
+void Printer::printMemReference(std::ostream &O, const MachineInstr *MI,
+				unsigned Op,
+				const MRegisterInfo &RI) const {
   assert(isMem(MI, Op) && "Invalid memory reference!");
 
   if (MI->getOperand(Op).isFrameIndex()) {
@@ -512,7 +525,7 @@
     O << "]";
     return;
   } else if (MI->getOperand(Op).isConstantPoolIndex()) {
-    O << "[.CPI" << fnIndex << "_"
+    O << "[.CPI" << CurrentFnName << "_"
       << MI->getOperand(Op).getConstantPoolIndex();
     if (MI->getOperand(Op+3).getImmedValue())
       O << " + " << MI->getOperand(Op+3).getImmedValue();
@@ -553,11 +566,14 @@
   O << "]";
 }
 
-// print - Print out an x86 instruction in intel syntax
-void X86InstrInfo::print(const MachineInstr *MI, std::ostream &O,
-                         const TargetMachine &TM) const {
+/// printMachineInstruction -- Print out an x86 instruction in intel syntax
+///
+void Printer::printMachineInstruction(const MachineInstr *MI, std::ostream &O,
+				      const TargetMachine &TM) const {
   unsigned Opcode = MI->getOpcode();
-  const TargetInstrDescriptor &Desc = get(Opcode);
+  const TargetInstrInfo &TII = TM.getInstrInfo();
+  const TargetInstrDescriptor &Desc = TII.get(Opcode);
+  const MRegisterInfo &RI = *TM.getRegisterInfo();
 
   switch (Desc.TSFlags & X86II::FormMask) {
   case X86II::Pseudo:
@@ -584,7 +600,7 @@
 	O << " = ";
 	++i;
       }
-      O << getName(MI->getOpcode());
+      O << TII.getName(MI->getOpcode());
 
       for (unsigned e = MI->getNumOperands(); i != e; ++i) {
 	O << " ";
@@ -610,7 +626,7 @@
 	     MI->getOperand(0).isGlobalAddress() ||
 	     MI->getOperand(0).isExternalSymbol())) &&
            "Illegal raw instruction!");
-    O << getName(MI->getOpcode()) << " ";
+    O << TII.getName(MI->getOpcode()) << " ";
 
     if (MI->getNumOperands() == 1) {
       printOp(O, MI->getOperand(0), RI, true); // Don't print "OFFSET"...
@@ -639,7 +655,7 @@
 
     unsigned Reg = MI->getOperand(0).getReg();
     
-    O << getName(MI->getOpCode()) << " ";
+    O << TII.getName(MI->getOpCode()) << " ";
     printOp(O, MI->getOperand(0), RI);
     if (MI->getNumOperands() == 2 &&
 	(!MI->getOperand(1).isRegister() ||
@@ -666,7 +682,7 @@
     // 4 Operands: This form is for instructions which are 3 operands forms, but
     // have a constant argument as well.
     //
-    bool isTwoAddr = isTwoAddrInstr(Opcode);
+    bool isTwoAddr = TII.isTwoAddrInstr(Opcode);
     assert(MI->getOperand(0).isRegister() &&
            (MI->getNumOperands() == 2 ||
 	    (isTwoAddr && MI->getOperand(1).isRegister() &&
@@ -675,7 +691,7 @@
 	      (MI->getNumOperands() == 4 && MI->getOperand(3).isImmediate()))))
            && "Bad format for MRMDestReg!");
 
-    O << getName(MI->getOpCode()) << " ";
+    O << TII.getName(MI->getOpCode()) << " ";
     printOp(O, MI->getOperand(0), RI);
     O << ", ";
     printOp(O, MI->getOperand(1+isTwoAddr), RI);
@@ -694,7 +710,7 @@
     assert(isMem(MI, 0) && MI->getNumOperands() == 4+1 &&
            MI->getOperand(4).isRegister() && "Bad format for MRMDestMem!");
 
-    O << getName(MI->getOpCode()) << " " << sizePtr(Desc) << " ";
+    O << TII.getName(MI->getOpCode()) << " " << sizePtr(Desc) << " ";
     printMemReference(O, MI, 0, RI);
     O << ", ";
     printOp(O, MI->getOperand(4), RI);
@@ -721,7 +737,7 @@
         MI->getOperand(0).getReg() != MI->getOperand(1).getReg())
       O << "**";
 
-    O << getName(MI->getOpCode()) << " ";
+    O << TII.getName(MI->getOpCode()) << " ";
     printOp(O, MI->getOperand(0), RI);
     O << ", ";
     printOp(O, MI->getOperand(MI->getNumOperands()-1), RI);
@@ -742,7 +758,7 @@
         MI->getOperand(0).getReg() != MI->getOperand(1).getReg())
       O << "**";
 
-    O << getName(MI->getOpCode()) << " ";
+    O << TII.getName(MI->getOpCode()) << " ";
     printOp(O, MI->getOperand(0), RI);
     O << ", " << sizePtr(Desc) << " ";
     printMemReference(O, MI, MI->getNumOperands()-4, RI);
@@ -773,12 +789,17 @@
         MI->getOperand(0).getReg() != MI->getOperand(1).getReg())
       O << "**";
 
-    O << getName(MI->getOpCode()) << " ";
+    O << TII.getName(MI->getOpCode()) << " ";
     printOp(O, MI->getOperand(0), RI);
     if (MI->getOperand(MI->getNumOperands()-1).isImmediate()) {
       O << ", ";
       printOp(O, MI->getOperand(MI->getNumOperands()-1), RI);
     }
+    if (Desc.TSFlags & X86II::PrintImplUses) {
+      for (const unsigned *p = Desc.ImplicitUses; *p; ++p) {
+	O << ", " << RI.get(*p).Name;
+      }
+    }
     O << "\n";
 
     return;
@@ -799,7 +820,7 @@
     assert((MI->getNumOperands() != 5 || MI->getOperand(4).isImmediate()) &&
            "Bad MRMSxM format!");
 
-    O << getName(MI->getOpCode()) << " ";
+    O << TII.getName(MI->getOpCode()) << " ";
     O << sizePtr(Desc) << " ";
     printMemReference(O, MI, 0, RI);
     if (MI->getNumOperands() == 5) {
@@ -821,9 +842,6 @@
   // with no % decorations on register names.
   O << "\t.intel_syntax noprefix\n";
 
-  // Start function index at 0
-  fnIndex = 0;
-
   // Ripped from CWriter:
   // Calculate which global values have names that will collide when we throw
   // away type information.
@@ -872,3 +890,5 @@
   MangledGlobals.clear();
   return false; // success
 }
+
+


Index: llvm/lib/Target/X86/X86InstrInfo.def
diff -u llvm/lib/Target/X86/X86InstrInfo.def:1.60 llvm/lib/Target/X86/X86InstrInfo.def:1.61
--- llvm/lib/Target/X86/X86InstrInfo.def:1.60	Sat Jun 21 12:13:35 2003
+++ llvm/lib/Target/X86/X86InstrInfo.def	Thu Jun 26 19:00:46 2003
@@ -220,29 +220,29 @@
 
 
 // Shift instructions
-I(SHLrr8      , "shl",  0xD2, M_2_ADDR_FLAG, X86II::MRMS4r, O_CL, NoIR)                                   // R8   <<= cl
-I(SHLrr16     , "shl",  0xD3, M_2_ADDR_FLAG, X86II::MRMS4r | X86II::OpSize, O_CL, NoIR)                   // R16  <<= cl
-I(SHLrr32     , "shl",  0xD3, M_2_ADDR_FLAG, X86II::MRMS4r, O_CL, NoIR)                                   // R32  <<= cl
+I(SHLrr8      , "shl",  0xD2, M_2_ADDR_FLAG, X86II::MRMS4r | X86II::PrintImplUses, O_CL, NoIR)            // R8   <<= cl
+I(SHLrr16     , "shl",  0xD3, M_2_ADDR_FLAG, X86II::MRMS4r | X86II::OpSize | X86II::PrintImplUses, O_CL, NoIR) // R16  <<= cl
+I(SHLrr32     , "shl",  0xD3, M_2_ADDR_FLAG, X86II::MRMS4r | X86II::PrintImplUses, O_CL, NoIR)            // R32  <<= cl
 I(SHLir8      , "shl",  0xC0, M_2_ADDR_FLAG, X86II::MRMS4r | X86II::Arg8, NoIR, NoIR)                     // R8   <<= imm8
 I(SHLir16     , "shl",  0xC1, M_2_ADDR_FLAG, X86II::MRMS4r | X86II::Arg8 | X86II::OpSize, NoIR, NoIR)     // R16  <<= imm8
 I(SHLir32     , "shl",  0xC1, M_2_ADDR_FLAG, X86II::MRMS4r | X86II::Arg8, NoIR, NoIR)                     // R32  <<= imm8
-I(SHRrr8      , "shr",  0xD2, M_2_ADDR_FLAG, X86II::MRMS5r, O_CL, NoIR)                                   // R8  >>>= cl
-I(SHRrr16     , "shr",  0xD3, M_2_ADDR_FLAG, X86II::MRMS5r | X86II::OpSize, O_CL, NoIR)                   // R16 >>>= cl
-I(SHRrr32     , "shr",  0xD3, M_2_ADDR_FLAG, X86II::MRMS5r, O_CL, NoIR)                                   // R32 >>>= cl
+I(SHRrr8      , "shr",  0xD2, M_2_ADDR_FLAG, X86II::MRMS5r | X86II::PrintImplUses, O_CL, NoIR)            // R8  >>>= cl
+I(SHRrr16     , "shr",  0xD3, M_2_ADDR_FLAG, X86II::MRMS5r | X86II::OpSize | X86II::PrintImplUses, O_CL, NoIR) // R16 >>>= cl
+I(SHRrr32     , "shr",  0xD3, M_2_ADDR_FLAG, X86II::MRMS5r | X86II::PrintImplUses, O_CL, NoIR)            // R32 >>>= cl
 I(SHRir8      , "shr",  0xC0, M_2_ADDR_FLAG, X86II::MRMS5r | X86II::Arg8, NoIR, NoIR)                     // R8  >>>= imm8
 I(SHRir16     , "shr",  0xC1, M_2_ADDR_FLAG, X86II::MRMS5r | X86II::Arg8 | X86II::OpSize, NoIR, NoIR)     // R16 >>>= imm8
 I(SHRir32     , "shr",  0xC1, M_2_ADDR_FLAG, X86II::MRMS5r | X86II::Arg8, NoIR, NoIR)                     // R32 >>>= imm8
-I(SARrr8      , "sar",  0xD2, M_2_ADDR_FLAG, X86II::MRMS7r, O_CL, NoIR)                                   // R8   >>= cl
-I(SARrr16     , "sar",  0xD3, M_2_ADDR_FLAG, X86II::MRMS7r | X86II::OpSize, O_CL, NoIR)                   // R16  >>= cl
-I(SARrr32     , "sar",  0xD3, M_2_ADDR_FLAG, X86II::MRMS7r, O_CL, NoIR)                                   // R32  >>= cl
+I(SARrr8      , "sar",  0xD2, M_2_ADDR_FLAG, X86II::MRMS7r | X86II::PrintImplUses, O_CL, NoIR)            // R8   >>= cl
+I(SARrr16     , "sar",  0xD3, M_2_ADDR_FLAG, X86II::MRMS7r | X86II::OpSize | X86II::PrintImplUses, O_CL, NoIR) // R16  >>= cl
+I(SARrr32     , "sar",  0xD3, M_2_ADDR_FLAG, X86II::MRMS7r | X86II::PrintImplUses, O_CL, NoIR)            // R32  >>= cl
 I(SARir8      , "sar",  0xC0, M_2_ADDR_FLAG, X86II::MRMS7r | X86II::Arg8, NoIR, NoIR)                     // R8   >>= imm8
 I(SARir16     , "sar",  0xC1, M_2_ADDR_FLAG, X86II::MRMS7r | X86II::Arg8 | X86II::OpSize, NoIR, NoIR)     // R16  >>= imm8
 I(SARir32     , "sar",  0xC1, M_2_ADDR_FLAG, X86II::MRMS7r | X86II::Arg8, NoIR, NoIR)                     // R32  >>= imm8
 
 I(SHLDir32    , "shld",  0xA4, M_2_ADDR_FLAG, X86II::MRMDestReg | X86II::TB | X86II::Arg8, NoIR, NoIR)    // R32 >>= R32,R32 imm8
-I(SHLDrr32    , "shld",  0xA5, M_2_ADDR_FLAG, X86II::MRMDestReg | X86II::TB, O_CL, NoIR)                  // R32 >>= R32,R32 cl
+I(SHLDrr32    , "shld",  0xA5, M_2_ADDR_FLAG, X86II::MRMDestReg | X86II::TB | X86II::PrintImplUses, O_CL, NoIR) // R32 >>= R32,R32 cl
 I(SHRDir32    , "shrd",  0xAC, M_2_ADDR_FLAG, X86II::MRMDestReg | X86II::TB | X86II::Arg8, NoIR, NoIR)    // R32 >>= R32,R32 imm8
-I(SHRDrr32    , "shrd",  0xAD, M_2_ADDR_FLAG, X86II::MRMDestReg | X86II::TB, O_CL, NoIR)                  // R32 >>= R32,R32 cl
+I(SHRDrr32    , "shrd",  0xAD, M_2_ADDR_FLAG, X86II::MRMDestReg | X86II::TB | X86II::PrintImplUses, O_CL, NoIR) // R32 >>= R32,R32 cl
 
 
 // Condition code ops, incl. set if equal/not equal/...


Index: llvm/lib/Target/X86/X86InstrInfo.h
diff -u llvm/lib/Target/X86/X86InstrInfo.h:1.23 llvm/lib/Target/X86/X86InstrInfo.h:1.24
--- llvm/lib/Target/X86/X86InstrInfo.h:1.23	Fri May 23 20:08:43 2003
+++ llvm/lib/Target/X86/X86InstrInfo.h	Thu Jun 26 19:00:46 2003
@@ -133,7 +133,10 @@
     // FPTypeMask - Mask for all of the FP types...
     FPTypeMask = 7 << 14,
 
-    // Bits 17 -> 31 are unused
+    // PrintImplUses - Print out implicit uses in the assembly output.
+    PrintImplUses = 1 << 17
+
+    // Bits 18 -> 31 are unused
   };
 }
 
@@ -159,11 +162,6 @@
   /// more than one way to `do nothing' but only one canonical way to slack off.
   ///
   bool isNOPinstr(const MachineInstr &MI) const;
-
-  /// print - Print out an x86 instruction in intel syntax
-  ///
-  virtual void print(const MachineInstr *MI, std::ostream &O,
-                     const TargetMachine &TM) const;
 
   // getBaseOpcodeFor - This function returns the "base" X86 opcode for the
   // specified opcode number.




From gaeke at cs.uiuc.edu  Thu Jun 26 19:01:27 2003
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Thu Jun 26 19:01:27 2003
Subject: [llvm-commits] CVS: llvm/lib/Target/TargetInstrInfo.cpp
Message-ID: <200306270000.TAA32710@zion.cs.uiuc.edu>


Changes in directory llvm/lib/Target:

TargetInstrInfo.cpp updated: 1.7 -> 1.8

---
Log message:

Nice tasty llc fixes. These should fix LLC for x86 for everything in
SingleSource except oopack and Oscar.  (Sorry, Oscar.)

include/llvm/Target/TargetInstrInfo.h: Remove virtual print method. Add
 accessors for ImplicitUses/Defs.
lib/Target/TargetInstrInfo.cpp: Remove virtual print method. If you
 really wanted this, just use MI->print(O, TM); instead...
lib/Target/X86:
FloatingPoint.cpp: ...like this.
X86InstrInfo.h: Remove virtual print method. Define the PrintImplUses
 target-specific flag bit.
X86InstrInfo.def: Add the PrintImplUses flag to all the instructions
 which implicitly use CL, because the assembler needs to see the CL in
 order to generate the right instruction.
Printer.cpp: Ditch fnIndex at Chris's request. Now we use CurrentFnName
  to name constants in the constant pool for each function instead. This
  avoids keeping state between runOnMachineFunction() invocations, which
  is a no-no. Having MangledGlobals be global is a bogon I'd like to get
  rid of too, but making it a static member of Printer causes link errors
  (why???).
 Make NumberForBB into a member of Printer instead of a global, too.
 Make printOp and printMemReference into methods of Printer.
 X86InstrInfo::print is now Printer::printMachineInstruction, because
  TargetInstrInfo::print is history. (Because of this, we have to qualify
  the names of some TargetInstrInfo methods we call.)
 Print out the ImplicitUses field of any instruction we print that has
  the PrintImplUses bit set. 


---
Diffs of the changes:

Index: llvm/lib/Target/TargetInstrInfo.cpp
diff -u llvm/lib/Target/TargetInstrInfo.cpp:1.7 llvm/lib/Target/TargetInstrInfo.cpp:1.8
--- llvm/lib/Target/TargetInstrInfo.cpp:1.7	Tue Jan 14 15:59:13 2003
+++ llvm/lib/Target/TargetInstrInfo.cpp	Thu Jun 26 19:00:41 2003
@@ -28,11 +28,6 @@
   TargetInstrDescriptors = NULL;	// reset global variable
 }
 
-void TargetInstrInfo::print(const MachineInstr *MI, std::ostream &O,
-			    const TargetMachine &TM) const {
-  MI->print(O, TM);
-}
-
 bool TargetInstrInfo::constantFitsInImmedField(MachineOpCode opCode,
 					       int64_t intValue) const {
   // First, check if opCode has an immed field.




From lattner at cs.uiuc.edu  Thu Jun 26 21:39:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Jun 26 21:39:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/CFrontend/2003-06-26-CFECrash.c
Message-ID: <200306270238.VAA24679@morpheus.cs.uiuc.edu>


Changes in directory llvm/test/Regression/CFrontend:

2003-06-26-CFECrash.c added (r1.1)

---
Log message:

New testcase that crashes the cfe


---
Diffs of the changes:

Index: llvm/test/Regression/CFrontend/2003-06-26-CFECrash.c
diff -c /dev/null llvm/test/Regression/CFrontend/2003-06-26-CFECrash.c:1.1
*** /dev/null	Thu Jun 26 21:38:04 2003
--- llvm/test/Regression/CFrontend/2003-06-26-CFECrash.c	Thu Jun 26 21:37:53 2003
***************
*** 0 ****
--- 1,17 ----
+ typedef struct min_info {
+   long offset;
+   unsigned file_attr;
+ } min_info;
+ 
+ typedef struct Globals {
+   char answerbuf;
+   min_info info[1];
+   min_info *pInfo;
+ } Uz_Globs;
+ 
+ extern Uz_Globs G;
+ 
+ int extract_or_test_files() {  
+   G.pInfo = G.info;
+ }
+ 




From lattner at cs.uiuc.edu  Thu Jun 26 23:21:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Jun 26 23:21:01 2003
Subject: [llvm-commits] CVS: llvm/www/www-index.html
Message-ID: <200306270420.XAA28492@tank.cs.uiuc.edu>


Changes in directory llvm/www:

www-index.html updated: 1.40 -> 1.41

---
Log message:

Add news entry so that people know that LLVM isn't stagnant.  :)


---
Diffs of the changes:

Index: llvm/www/www-index.html
diff -u llvm/www/www-index.html:1.40 llvm/www/www-index.html:1.41
--- llvm/www/www-index.html:1.40	Fri Jun 20 15:20:21 2003
+++ llvm/www/www-index.html	Thu Jun 26 23:20:30 2003
@@ -231,6 +231,11 @@
                 <td>
                 <table border="0">
                     <tr>
+                        <td valign="top"><b>Jun&nbsp;26,&nbsp;2003</b></td>
+                        <td valign="top">-</td>
+                        <td valign="top">Added a new <a href="http://mail.cs.uiuc.edu/pipermail/llvmdev/2003-June/000416.html">status update</a>.</td>
+                    </tr>
+                    <tr>
                         <td valign="top"><b>May&nbsp;21,&nbsp;2003</b></td>
                         <td valign="top">-</td>
                         <td valign="top">Added some new documents to the




From criswell at choi.cs.uiuc.edu  Fri Jun 27 11:59:01 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Fri Jun 27 11:59:01 2003
Subject: [llvm-commits] CVS: llvm/Makefile.common
Message-ID: <200306271658.h5RGwtQ21248@choi.cs.uiuc.edu>


Changes in directory llvm:

Makefile.common updated: 1.93 -> 1.94

---
Log message:

Added the OPTIONAL_DIRS variable for specifying optional directories to build.


---
Diffs of the changes:

Index: llvm/Makefile.common
diff -u llvm/Makefile.common:1.93 llvm/Makefile.common:1.94
--- llvm/Makefile.common:1.93	Fri Jun 20 16:24:54 2003
+++ llvm/Makefile.common	Fri Jun 27 11:58:44 2003
@@ -40,6 +40,7 @@
 #
 #===-----------------------------------------------------------------------====
 
+#
 # Configuration file to set paths specific to local installation of LLVM
 # 
 include $(LEVEL)/Makefile.config
@@ -340,6 +341,16 @@
 	$(VERB) cd $(@D); $(MAKE) $(subst $(@D)/.make,,$@)
 endif
 
+ifdef OPTIONAL_DIRS
+all install clean test ::
+	$(VERB) for dir in ${OPTIONAL_DIRS}; do \
+		if [ -d $$dir ]; \
+		then\
+			(cd $$dir; $(MAKE) $@) || exit 1; \
+		fi \
+	done
+endif
+
 #---------------------------------------------------------
 # Handle the LIBRARYNAME option - used when building libs...
 #---------------------------------------------------------
@@ -489,6 +500,7 @@
         $(addsuffix /.makeall, $(PARALLEL_DIRS))
 
 all::   $(TOOLEXENAMES)
+
 clean::
 	$(VERB) rm -f $(TOOLEXENAMES)
 




From criswell at choi.cs.uiuc.edu  Fri Jun 27 12:04:02 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Fri Jun 27 12:04:02 2003
Subject: [llvm-commits] CVS: llvm/utils/Burg/Makefile
Message-ID: <200306271703.h5RH3UZ21304@choi.cs.uiuc.edu>


Changes in directory llvm/utils/Burg:

Makefile updated: 1.14 -> 1.15

---
Log message:

Added code to the clean rule that deletes files placed in the source directory.
These files interfer with a second build attempt.


---
Diffs of the changes:

Index: llvm/utils/Burg/Makefile
diff -u llvm/utils/Burg/Makefile:1.14 llvm/utils/Burg/Makefile:1.15
--- llvm/utils/Burg/Makefile:1.14	Tue Jun 17 09:30:21 2003
+++ llvm/utils/Burg/Makefile	Fri Jun 27 12:03:20 2003
@@ -25,3 +25,4 @@
 	$(TOOLEXENAME_G) -I      sample.gr -o tmp && cmp tmp sample.c
 	$(TOOLEXENAME_G) -I -O0 <sample.gr   >tmp && cmp tmp sample.c
 	$(TOOLEXENAME_G) -I -=  <sample.gr   >tmp && cmp tmp sample.c
+	rm -f tmp sample.c




From criswell at choi.cs.uiuc.edu  Fri Jun 27 13:33:00 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Fri Jun 27 13:33:00 2003
Subject: [llvm-commits] CVS: llvm/include/Support/Annotation.h BitSetVector.h Casting.h CommandLine.h TarjanSCCIterator.h Timer.h Tree.h ilist
Message-ID: <200306271832.h5RIWe325062@choi.cs.uiuc.edu>


Changes in directory llvm/include/Support:

Annotation.h updated: 1.8.2.1 -> 1.8.2.2
BitSetVector.h updated: 1.5.2.2 -> 1.5.2.3
Casting.h updated: 1.5.2.1 -> 1.5.2.2
CommandLine.h updated: 1.12.2.2 -> 1.12.2.3
TarjanSCCIterator.h updated: 1.5.2.2 -> 1.5.2.3
Timer.h updated: 1.6.2.2 -> 1.6.2.3
Tree.h updated: 1.3.2.2 -> 1.3.2.3
ilist updated: 1.6.2.2 -> 1.6.2.3

---
Log message:

Removed the user of assert.h.  It is now globally included in
llvm/AbstractTypeUser.h.
Updated some of the files to use the header files from include/Config instead
of directly including the system header files.


---
Diffs of the changes:

Index: llvm/include/Support/Annotation.h
diff -u llvm/include/Support/Annotation.h:1.8.2.1 llvm/include/Support/Annotation.h:1.8.2.2
--- llvm/include/Support/Annotation.h:1.8.2.1	Mon Jun 23 13:50:38 2003
+++ llvm/include/Support/Annotation.h	Fri Jun 27 13:31:59 2003
@@ -16,7 +16,6 @@
 #define SUPPORT_ANNOTATION_H
 
 #include <string>
-#include "Config/assert.h"
 
 class AnnotationID;
 class Annotation;


Index: llvm/include/Support/BitSetVector.h
diff -u llvm/include/Support/BitSetVector.h:1.5.2.2 llvm/include/Support/BitSetVector.h:1.5.2.3
--- llvm/include/Support/BitSetVector.h:1.5.2.2	Thu Jun 26 16:34:32 2003
+++ llvm/include/Support/BitSetVector.h	Fri Jun 27 13:31:59 2003
@@ -24,7 +24,6 @@
 #include <vector>
 #include <functional>
 #include <iostream>
-#include "Config/assert.h"
 
 class BitSetVector {
   enum { BITSET_WORDSIZE = sizeof(long)*8 };


Index: llvm/include/Support/Casting.h
diff -u llvm/include/Support/Casting.h:1.5.2.1 llvm/include/Support/Casting.h:1.5.2.2
--- llvm/include/Support/Casting.h:1.5.2.1	Mon Jun 23 13:50:38 2003
+++ llvm/include/Support/Casting.h	Fri Jun 27 13:31:59 2003
@@ -8,8 +8,6 @@
 #ifndef SUPPORT_CASTING_H
 #define SUPPORT_CASTING_H
 
-#include "Config/assert.h"
-
 //===----------------------------------------------------------------------===//
 //                          isa<x> Support Templates
 //===----------------------------------------------------------------------===//


Index: llvm/include/Support/CommandLine.h
diff -u llvm/include/Support/CommandLine.h:1.12.2.2 llvm/include/Support/CommandLine.h:1.12.2.3
--- llvm/include/Support/CommandLine.h:1.12.2.2	Thu Jun 26 16:34:32 2003
+++ llvm/include/Support/CommandLine.h	Fri Jun 27 13:31:59 2003
@@ -19,8 +19,6 @@
 #include <cstdarg>
 #include "boost/type_traits/object_traits.hpp"
 
-#include "Config/assert.h"
-
 /// cl Namespace - This namespace contains all of the command line option
 /// processing machinery.  It is intentionally a short name to make qualified
 /// usage concise.


Index: llvm/include/Support/TarjanSCCIterator.h
diff -u llvm/include/Support/TarjanSCCIterator.h:1.5.2.2 llvm/include/Support/TarjanSCCIterator.h:1.5.2.3
--- llvm/include/Support/TarjanSCCIterator.h:1.5.2.2	Thu Jun 26 16:34:33 2003
+++ llvm/include/Support/TarjanSCCIterator.h	Fri Jun 27 13:31:59 2003
@@ -21,7 +21,6 @@
 #include <vector>
 #include <stack>
 #include <map>
-#include "Config/assert.h"
 
 //--------------------------------------------------------------------------
 // class SCC : A simple representation of an SCC in a generic Graph.


Index: llvm/include/Support/Timer.h
diff -u llvm/include/Support/Timer.h:1.6.2.2 llvm/include/Support/Timer.h:1.6.2.3
--- llvm/include/Support/Timer.h:1.6.2.2	Thu Jun 26 16:34:33 2003
+++ llvm/include/Support/Timer.h	Fri Jun 27 13:31:59 2003
@@ -29,8 +29,6 @@
 #include <vector>
 #include <iosfwd>
 
-#include "Config/assert.h"
-
 class TimerGroup;
 
 class Timer {


Index: llvm/include/Support/Tree.h
diff -u llvm/include/Support/Tree.h:1.3.2.2 llvm/include/Support/Tree.h:1.3.2.3
--- llvm/include/Support/Tree.h:1.3.2.2	Thu Jun 26 16:34:33 2003
+++ llvm/include/Support/Tree.h	Fri Jun 27 13:31:59 2003
@@ -9,7 +9,6 @@
 #define SUPPORT_TREE_H
 
 #include <vector>
-#include "Config/assert.h"
 
 template<class ConcreteTreeNode, class Payload>
 class Tree {


Index: llvm/include/Support/ilist
diff -u llvm/include/Support/ilist:1.6.2.2 llvm/include/Support/ilist:1.6.2.3
--- llvm/include/Support/ilist:1.6.2.2	Thu Jun 26 16:34:33 2003
+++ llvm/include/Support/ilist	Fri Jun 27 13:31:59 2003
@@ -31,7 +31,6 @@
 #ifndef SUPPORT_ILIST
 #define SUPPORT_ILIST
 
-#include "Config/assert.h"
 #include <algorithm>
 #include <Support/iterator>
 




From criswell at choi.cs.uiuc.edu  Fri Jun 27 13:33:10 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Fri Jun 27 13:33:10 2003
Subject: [llvm-commits] CVS: llvm/lib/Reoptimizer/BinInterface/analyze.cpp construct.cpp emit.cpp select.cpp sparcbin.cpp sparcdis.cpp
Message-ID: <200306271832.h5RIWMb25041@choi.cs.uiuc.edu>


Changes in directory llvm/lib/Reoptimizer/BinInterface:

analyze.cpp updated: 1.10 -> 1.10.2.1
construct.cpp updated: 1.5 -> 1.5.2.1
emit.cpp updated: 1.5 -> 1.5.2.1
select.cpp updated: 1.6 -> 1.6.2.1
sparcbin.cpp updated: 1.7.2.1 -> 1.7.2.2
sparcdis.cpp updated: 1.12 -> 1.12.2.1

---
Log message:

Removed the user of assert.h.  It is now globally included in
llvm/AbstractTypeUser.h.
Updated some of the files to use the header files from include/Config instead
of directly including the system header files.


---
Diffs of the changes:

Index: llvm/lib/Reoptimizer/BinInterface/analyze.cpp
diff -u llvm/lib/Reoptimizer/BinInterface/analyze.cpp:1.10 llvm/lib/Reoptimizer/BinInterface/analyze.cpp:1.10.2.1
--- llvm/lib/Reoptimizer/BinInterface/analyze.cpp:1.10	Thu Jun  5 01:06:03 2003
+++ llvm/lib/Reoptimizer/BinInterface/analyze.cpp	Fri Jun 27 13:32:11 2003
@@ -15,9 +15,8 @@
 #include "llvm/Reoptimizer/BinInterface/analyze.h"  // SPARC analysis library 
 #include "llvm/Reoptimizer/BinInterface/bitmath.h"  // binary math routines
 #include "llvm/Reoptimizer/BinInterface/sparc9.h"   // SPARC9 opcode definitions
-#include <stdio.h>
-#include <stdlib.h>
-#include <assert.h>
+#include "Config/stdio.h"
+#include "Config/stdlib.h"
 
 //*****************************************************************************
 // unsigned sparc_analyzebr(unsigned instr)


Index: llvm/lib/Reoptimizer/BinInterface/construct.cpp
diff -u llvm/lib/Reoptimizer/BinInterface/construct.cpp:1.5 llvm/lib/Reoptimizer/BinInterface/construct.cpp:1.5.2.1
--- llvm/lib/Reoptimizer/BinInterface/construct.cpp:1.5	Wed Jun  4 04:40:38 2003
+++ llvm/lib/Reoptimizer/BinInterface/construct.cpp	Fri Jun 27 13:32:11 2003
@@ -11,9 +11,8 @@
 #include "llvm/Reoptimizer/BinInterface/bitmath.h"  // binary math routines
 #include "llvm/Reoptimizer/BinInterface/sparc9.h"   // SPARC9 opcode definitions
 #include "llvm/Reoptimizer/BinInterface/regmask.h"  // register mask allocator
-#include <stdio.h>
-#include <stdlib.h>
-#include <assert.h>
+#include "Config/stdio.h"
+#include "Config/stdlib.h"
 #include <vector>
 
 using std::pair;            // use STL pair class


Index: llvm/lib/Reoptimizer/BinInterface/emit.cpp
diff -u llvm/lib/Reoptimizer/BinInterface/emit.cpp:1.5 llvm/lib/Reoptimizer/BinInterface/emit.cpp:1.5.2.1
--- llvm/lib/Reoptimizer/BinInterface/emit.cpp:1.5	Wed Jun  4 04:40:43 2003
+++ llvm/lib/Reoptimizer/BinInterface/emit.cpp	Fri Jun 27 13:32:11 2003
@@ -10,10 +10,9 @@
 #include "llvm/Reoptimizer/BinInterface/bitmath.h"  // binary math routines
 #include "llvm/Reoptimizer/BinInterface/sparc9.h"   // SPARC9 opcode definitions
 #include "llvm/Reoptimizer/BinInterface/regmask.h"  // Register mask allocator
-#include <stdio.h>
-#include <stdlib.h>
+#include "Config/stdio.h"
+#include "Config/stdlib.h"
 #include <vector>
-#include <assert.h>
 
 using std::pair;
 using std::vector;


Index: llvm/lib/Reoptimizer/BinInterface/select.cpp
diff -u llvm/lib/Reoptimizer/BinInterface/select.cpp:1.6 llvm/lib/Reoptimizer/BinInterface/select.cpp:1.6.2.1
--- llvm/lib/Reoptimizer/BinInterface/select.cpp:1.6	Wed Jun  4 04:40:48 2003
+++ llvm/lib/Reoptimizer/BinInterface/select.cpp	Fri Jun 27 13:32:11 2003
@@ -10,10 +10,9 @@
 #include "llvm/Reoptimizer/BinInterface/bitmath.h"  // binary math routines
 #include "llvm/Reoptimizer/BinInterface/sparc9.h"   // SPARC9 opcode definitions
 #include "llvm/Reoptimizer/BinInterface/regmask.h"  // Register mask allocator
-#include <stdio.h>
-#include <stdlib.h>
+#include "Config/stdio.h"
+#include "Config/stdlib.h"
 #include <vector>
-#include <assert.h>
 
 using std::pair;
 


Index: llvm/lib/Reoptimizer/BinInterface/sparcbin.cpp
diff -u llvm/lib/Reoptimizer/BinInterface/sparcbin.cpp:1.7.2.1 llvm/lib/Reoptimizer/BinInterface/sparcbin.cpp:1.7.2.2
--- llvm/lib/Reoptimizer/BinInterface/sparcbin.cpp:1.7.2.1	Thu Jun 26 16:35:18 2003
+++ llvm/lib/Reoptimizer/BinInterface/sparcbin.cpp	Fri Jun 27 13:32:12 2003
@@ -14,10 +14,9 @@
 #include "llvm/Reoptimizer/BinInterface/sparcdis.h"  // sparc disassembler
 #include "llvm/Reoptimizer/BinInterface/bitmath.h" // binary math routines
 #include "llvm/Reoptimizer/BinInterface/sparc9.h"  // SPARC 9 opcode definitions
-#include <stdio.h>
-#include <stdlib.h>
+#include "Config/stdio.h"
+#include "Config/stdlib.h"
 #include <vector>
-#include <assert.h>
 #include <new>
 
 using std::pair;


Index: llvm/lib/Reoptimizer/BinInterface/sparcdis.cpp
diff -u llvm/lib/Reoptimizer/BinInterface/sparcdis.cpp:1.12 llvm/lib/Reoptimizer/BinInterface/sparcdis.cpp:1.12.2.1
--- llvm/lib/Reoptimizer/BinInterface/sparcdis.cpp:1.12	Thu Jun  5 01:06:03 2003
+++ llvm/lib/Reoptimizer/BinInterface/sparcdis.cpp	Fri Jun 27 13:32:12 2003
@@ -19,9 +19,8 @@
 
 #include "llvm/Reoptimizer/BinInterface/bitmath.h"  // binary math routines
 #include "llvm/Reoptimizer/BinInterface/sparc9.h"   // SPARC9 opcode definitions
-#include <stdio.h>
-#include <stdlib.h>
-#include <assert.h>
+#include "Config/stdio.h"
+#include "Config/stdlib.h"
 
 void sparc_printop_rs1(unsigned instr, int labelrs1)
 {




From criswell at choi.cs.uiuc.edu  Fri Jun 27 13:33:19 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Fri Jun 27 13:33:19 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Target/MRegisterInfo.h TargetCacheInfo.h TargetInstrInfo.h TargetRegInfo.h TargetSchedInfo.h
Message-ID: <200306271832.h5RIWKA25023@choi.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Target:

MRegisterInfo.h updated: 1.16.2.1 -> 1.16.2.2
TargetCacheInfo.h updated: 1.9.2.1 -> 1.9.2.2
TargetInstrInfo.h updated: 1.45.2.1 -> 1.45.2.2
TargetRegInfo.h updated: 1.38.2.1 -> 1.38.2.2
TargetSchedInfo.h updated: 1.17.2.1 -> 1.17.2.2

---
Log message:

Removed the user of assert.h.  It is now globally included in
llvm/AbstractTypeUser.h.
Updated some of the files to use the header files from include/Config instead
of directly including the system header files.


---
Diffs of the changes:

Index: llvm/include/llvm/Target/MRegisterInfo.h
diff -u llvm/include/llvm/Target/MRegisterInfo.h:1.16.2.1 llvm/include/llvm/Target/MRegisterInfo.h:1.16.2.2
--- llvm/include/llvm/Target/MRegisterInfo.h:1.16.2.1	Mon Jun 23 13:51:59 2003
+++ llvm/include/llvm/Target/MRegisterInfo.h	Fri Jun 27 13:32:10 2003
@@ -10,7 +10,6 @@
 #define LLVM_TARGET_MREGISTERINFO_H
 
 #include "llvm/CodeGen/MachineBasicBlock.h"
-#include "Config/assert.h"
 
 class Type;
 class MachineFunction;


Index: llvm/include/llvm/Target/TargetCacheInfo.h
diff -u llvm/include/llvm/Target/TargetCacheInfo.h:1.9.2.1 llvm/include/llvm/Target/TargetCacheInfo.h:1.9.2.2
--- llvm/include/llvm/Target/TargetCacheInfo.h:1.9.2.1	Mon Jun 23 13:51:59 2003
+++ llvm/include/llvm/Target/TargetCacheInfo.h	Fri Jun 27 13:32:10 2003
@@ -8,7 +8,6 @@
 #define LLVM_TARGET_TARGETCACHEINFO_H
 
 #include "Support/DataTypes.h"
-#include "Config/assert.h"
 
 class TargetMachine;
 


Index: llvm/include/llvm/Target/TargetInstrInfo.h
diff -u llvm/include/llvm/Target/TargetInstrInfo.h:1.45.2.1 llvm/include/llvm/Target/TargetInstrInfo.h:1.45.2.2
--- llvm/include/llvm/Target/TargetInstrInfo.h:1.45.2.1	Mon Jun 23 13:51:59 2003
+++ llvm/include/llvm/Target/TargetInstrInfo.h	Fri Jun 27 13:32:10 2003
@@ -9,7 +9,6 @@
 
 #include "Support/DataTypes.h"
 #include <vector>
-#include "Config/assert.h"
 
 class MachineInstr;
 class TargetMachine;


Index: llvm/include/llvm/Target/TargetRegInfo.h
diff -u llvm/include/llvm/Target/TargetRegInfo.h:1.38.2.1 llvm/include/llvm/Target/TargetRegInfo.h:1.38.2.2
--- llvm/include/llvm/Target/TargetRegInfo.h:1.38.2.1	Mon Jun 23 13:51:59 2003
+++ llvm/include/llvm/Target/TargetRegInfo.h	Fri Jun 27 13:32:10 2003
@@ -10,7 +10,6 @@
 
 #include "Support/hash_map"
 #include <string>
-#include "Config/assert.h"
 
 class TargetMachine;
 class IGNode;


Index: llvm/include/llvm/Target/TargetSchedInfo.h
diff -u llvm/include/llvm/Target/TargetSchedInfo.h:1.17.2.1 llvm/include/llvm/Target/TargetSchedInfo.h:1.17.2.2
--- llvm/include/llvm/Target/TargetSchedInfo.h:1.17.2.1	Mon Jun 23 13:51:59 2003
+++ llvm/include/llvm/Target/TargetSchedInfo.h	Fri Jun 27 13:32:10 2003
@@ -10,7 +10,6 @@
 #include "llvm/Target/TargetInstrInfo.h"
 #include "Support/hash_map"
 #include <string>
-#include "Config/assert.h"
 
 typedef long long cycles_t; 
 static const cycles_t HUGE_LATENCY = ~((long long) 1 << (sizeof(cycles_t)-2));




From criswell at choi.cs.uiuc.edu  Fri Jun 27 13:33:29 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Fri Jun 27 13:33:29 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Support/CFG.h InstVisitor.h PassNameParser.h
Message-ID: <200306271832.h5RIWJi25007@choi.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Support:

CFG.h updated: 1.11.2.1 -> 1.11.2.2
InstVisitor.h updated: 1.17.2.1 -> 1.17.2.2
PassNameParser.h updated: 1.6.2.1 -> 1.6.2.2

---
Log message:

Removed the user of assert.h.  It is now globally included in
llvm/AbstractTypeUser.h.
Updated some of the files to use the header files from include/Config instead
of directly including the system header files.


---
Diffs of the changes:

Index: llvm/include/llvm/Support/CFG.h
diff -u llvm/include/llvm/Support/CFG.h:1.11.2.1 llvm/include/llvm/Support/CFG.h:1.11.2.2
--- llvm/include/llvm/Support/CFG.h:1.11.2.1	Mon Jun 23 13:51:57 2003
+++ llvm/include/llvm/Support/CFG.h	Fri Jun 27 13:32:09 2003
@@ -13,8 +13,6 @@
 #include "llvm/InstrTypes.h"
 #include "Support/iterator"
 
-#include "Config/assert.h"
-
 //===--------------------------------------------------------------------===//
 // BasicBlock pred_iterator definition
 //===--------------------------------------------------------------------===//


Index: llvm/include/llvm/Support/InstVisitor.h
diff -u llvm/include/llvm/Support/InstVisitor.h:1.17.2.1 llvm/include/llvm/Support/InstVisitor.h:1.17.2.2
--- llvm/include/llvm/Support/InstVisitor.h:1.17.2.1	Mon Jun 23 13:51:57 2003
+++ llvm/include/llvm/Support/InstVisitor.h	Fri Jun 27 13:32:09 2003
@@ -44,7 +44,6 @@
 #define LLVM_SUPPORT_INSTVISITOR_H
 
 #include "llvm/Instruction.h"
-#include "Config/assert.h"
 
 class Module;
 


Index: llvm/include/llvm/Support/PassNameParser.h
diff -u llvm/include/llvm/Support/PassNameParser.h:1.6.2.1 llvm/include/llvm/Support/PassNameParser.h:1.6.2.2
--- llvm/include/llvm/Support/PassNameParser.h:1.6.2.1	Mon Jun 23 13:51:57 2003
+++ llvm/include/llvm/Support/PassNameParser.h	Fri Jun 27 13:32:09 2003
@@ -20,7 +20,6 @@
 #include "llvm/Pass.h"
 #include <algorithm>
 #include <iostream>
-#include "Config/assert.h"
 
 //===----------------------------------------------------------------------===//
 // PassNameParser class - Make use of the pass registration mechanism to




From criswell at choi.cs.uiuc.edu  Fri Jun 27 13:33:38 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Fri Jun 27 13:33:38 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Reoptimizer/BinInterface/regmask.h sparcbin.h
Message-ID: <200306271832.h5RIWHr24993@choi.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Reoptimizer/BinInterface:

regmask.h updated: 1.2.2.1 -> 1.2.2.2
sparcbin.h updated: 1.3.2.1 -> 1.3.2.2

---
Log message:

Removed the user of assert.h.  It is now globally included in
llvm/AbstractTypeUser.h.
Updated some of the files to use the header files from include/Config instead
of directly including the system header files.


---
Diffs of the changes:

Index: llvm/include/llvm/Reoptimizer/BinInterface/regmask.h
diff -u llvm/include/llvm/Reoptimizer/BinInterface/regmask.h:1.2.2.1 llvm/include/llvm/Reoptimizer/BinInterface/regmask.h:1.2.2.2
--- llvm/include/llvm/Reoptimizer/BinInterface/regmask.h:1.2.2.1	Mon Jun 23 13:51:55 2003
+++ llvm/include/llvm/Reoptimizer/BinInterface/regmask.h	Fri Jun 27 13:32:07 2003
@@ -11,7 +11,6 @@
 #include "bitmath.h"
 #include <vector>
 #include <iostream>
-#include "Config/assert.h"
 
 #define VREG_ISMEM(x) (x >= 32)
 #define VREG_ISREG(x) (x <  32)


Index: llvm/include/llvm/Reoptimizer/BinInterface/sparcbin.h
diff -u llvm/include/llvm/Reoptimizer/BinInterface/sparcbin.h:1.3.2.1 llvm/include/llvm/Reoptimizer/BinInterface/sparcbin.h:1.3.2.2
--- llvm/include/llvm/Reoptimizer/BinInterface/sparcbin.h:1.3.2.1	Mon Jun 23 13:51:55 2003
+++ llvm/include/llvm/Reoptimizer/BinInterface/sparcbin.h	Fri Jun 27 13:32:07 2003
@@ -18,7 +18,6 @@
 #include "sparcpriv.h"
 #include <map>
 #include <vector>
-#include "Config/assert.h"
 
 using std::vector;
 




From criswell at choi.cs.uiuc.edu  Fri Jun 27 13:33:48 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Fri Jun 27 13:33:48 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Reoptimizer/TraceCache.h VirtualMem.h
Message-ID: <200306271832.h5RIWGj24983@choi.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Reoptimizer:

TraceCache.h updated: 1.10.2.1 -> 1.10.2.2
VirtualMem.h updated: 1.7.2.1 -> 1.7.2.2

---
Log message:

Removed the user of assert.h.  It is now globally included in
llvm/AbstractTypeUser.h.
Updated some of the files to use the header files from include/Config instead
of directly including the system header files.


---
Diffs of the changes:

Index: llvm/include/llvm/Reoptimizer/TraceCache.h
diff -u llvm/include/llvm/Reoptimizer/TraceCache.h:1.10.2.1 llvm/include/llvm/Reoptimizer/TraceCache.h:1.10.2.2
--- llvm/include/llvm/Reoptimizer/TraceCache.h:1.10.2.1	Mon Jun 23 13:51:52 2003
+++ llvm/include/llvm/Reoptimizer/TraceCache.h	Fri Jun 27 13:32:06 2003
@@ -11,7 +11,6 @@
 #include <vector>
 #include <list>
 #include <utility>
-#include "Config/assert.h"
 
 class VirtualMem;
 class MemoryManager;


Index: llvm/include/llvm/Reoptimizer/VirtualMem.h
diff -u llvm/include/llvm/Reoptimizer/VirtualMem.h:1.7.2.1 llvm/include/llvm/Reoptimizer/VirtualMem.h:1.7.2.2
--- llvm/include/llvm/Reoptimizer/VirtualMem.h:1.7.2.1	Mon Jun 23 13:51:52 2003
+++ llvm/include/llvm/Reoptimizer/VirtualMem.h	Fri Jun 27 13:32:06 2003
@@ -12,7 +12,6 @@
 #include "Config/unistd.h"
 
 #include "Config/stdio.h"
-#include "Config/assert.h"
 #include <iostream>
 
 #include <vector>




From criswell at choi.cs.uiuc.edu  Fri Jun 27 13:33:57 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Fri Jun 27 13:33:57 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/CodeGen/IGNode.h InstrForest.h InstrSelection.h LiveRange.h LiveVariables.h MachineCodeForInstruction.h MachineFrameInfo.h MachineInstr.h MachineInstrAnnot.h SSARegMap.h
Message-ID: <200306271832.h5RIWFm24973@choi.cs.uiuc.edu>


Changes in directory llvm/include/llvm/CodeGen:

IGNode.h updated: 1.12.2.1 -> 1.12.2.2
InstrForest.h updated: 1.26.2.2 -> 1.26.2.3
InstrSelection.h updated: 1.23.2.2 -> 1.23.2.3
LiveRange.h updated: 1.19.2.2 -> 1.19.2.3
LiveVariables.h updated: 1.5.2.1 -> 1.5.2.2
MachineCodeForInstruction.h updated: 1.9.2.1 -> 1.9.2.2
MachineFrameInfo.h updated: 1.5.2.1 -> 1.5.2.2
MachineInstr.h updated: 1.103.2.1 -> 1.103.2.2
MachineInstrAnnot.h updated: 1.9.2.1 -> 1.9.2.2
SSARegMap.h updated: 1.4.2.1 -> 1.4.2.2

---
Log message:

Removed the user of assert.h.  It is now globally included in
llvm/AbstractTypeUser.h.
Updated some of the files to use the header files from include/Config instead
of directly including the system header files.


---
Diffs of the changes:

Index: llvm/include/llvm/CodeGen/IGNode.h
diff -u llvm/include/llvm/CodeGen/IGNode.h:1.12.2.1 llvm/include/llvm/CodeGen/IGNode.h:1.12.2.2
--- llvm/include/llvm/CodeGen/IGNode.h:1.12.2.1	Mon Jun 23 13:51:50 2003
+++ llvm/include/llvm/CodeGen/IGNode.h	Fri Jun 27 13:32:05 2003
@@ -26,7 +26,6 @@
 #define IG_NODE_H
 
 #include "llvm/CodeGen/LiveRange.h"
-#include "Config/assert.h"
 class RegClass;
 
 //----------------------------------------------------------------------------


Index: llvm/include/llvm/CodeGen/InstrForest.h
diff -u llvm/include/llvm/CodeGen/InstrForest.h:1.26.2.2 llvm/include/llvm/CodeGen/InstrForest.h:1.26.2.3
--- llvm/include/llvm/CodeGen/InstrForest.h:1.26.2.2	Thu Jun 26 16:34:44 2003
+++ llvm/include/llvm/CodeGen/InstrForest.h	Fri Jun 27 13:32:05 2003
@@ -21,7 +21,6 @@
 
 #include "llvm/Instruction.h"
 #include "Support/HashExtras.h"
-#include "Config/assert.h"
 
 class Constant;
 class Function;


Index: llvm/include/llvm/CodeGen/InstrSelection.h
diff -u llvm/include/llvm/CodeGen/InstrSelection.h:1.23.2.2 llvm/include/llvm/CodeGen/InstrSelection.h:1.23.2.3
--- llvm/include/llvm/CodeGen/InstrSelection.h:1.23.2.2	Thu Jun 26 16:34:44 2003
+++ llvm/include/llvm/CodeGen/InstrSelection.h	Fri Jun 27 13:32:05 2003
@@ -8,7 +8,6 @@
 #define LLVM_CODEGEN_INSTR_SELECTION_H
 
 #include "llvm/Instruction.h"
-#include "Config/assert.h"
 
 class Function;
 class InstrForest;


Index: llvm/include/llvm/CodeGen/LiveRange.h
diff -u llvm/include/llvm/CodeGen/LiveRange.h:1.19.2.2 llvm/include/llvm/CodeGen/LiveRange.h:1.19.2.3
--- llvm/include/llvm/CodeGen/LiveRange.h:1.19.2.2	Thu Jun 26 16:34:44 2003
+++ llvm/include/llvm/CodeGen/LiveRange.h	Fri Jun 27 13:32:05 2003
@@ -14,8 +14,6 @@
 #include "llvm/CodeGen/ValueSet.h"
 #include "llvm/Value.h"
 
-#include "Config/assert.h"
-
 class RegClass;
 class IGNode;
 


Index: llvm/include/llvm/CodeGen/LiveVariables.h
diff -u llvm/include/llvm/CodeGen/LiveVariables.h:1.5.2.1 llvm/include/llvm/CodeGen/LiveVariables.h:1.5.2.2
--- llvm/include/llvm/CodeGen/LiveVariables.h:1.5.2.1	Mon Jun 23 13:51:50 2003
+++ llvm/include/llvm/CodeGen/LiveVariables.h	Fri Jun 27 13:32:05 2003
@@ -24,7 +24,6 @@
 
 #include "llvm/CodeGen/MachineFunctionPass.h"
 #include <map>
-#include "Config/assert.h"
 
 class MRegisterInfo;
 


Index: llvm/include/llvm/CodeGen/MachineCodeForInstruction.h
diff -u llvm/include/llvm/CodeGen/MachineCodeForInstruction.h:1.9.2.1 llvm/include/llvm/CodeGen/MachineCodeForInstruction.h:1.9.2.2
--- llvm/include/llvm/CodeGen/MachineCodeForInstruction.h:1.9.2.1	Mon Jun 23 13:51:50 2003
+++ llvm/include/llvm/CodeGen/MachineCodeForInstruction.h	Fri Jun 27 13:32:05 2003
@@ -20,7 +20,6 @@
 
 #include "Support/Annotation.h"
 #include <vector>
-#include "Config/assert.h"
 
 class MachineInstr;
 class Instruction;


Index: llvm/include/llvm/CodeGen/MachineFrameInfo.h
diff -u llvm/include/llvm/CodeGen/MachineFrameInfo.h:1.5.2.1 llvm/include/llvm/CodeGen/MachineFrameInfo.h:1.5.2.2
--- llvm/include/llvm/CodeGen/MachineFrameInfo.h:1.5.2.1	Mon Jun 23 13:51:50 2003
+++ llvm/include/llvm/CodeGen/MachineFrameInfo.h	Fri Jun 27 13:32:05 2003
@@ -35,7 +35,6 @@
 class TargetRegisterClass;
 class MachineFunction;
 #include <vector>
-#include "Config/assert.h"
 
 class MachineFrameInfo {
 


Index: llvm/include/llvm/CodeGen/MachineInstr.h
diff -u llvm/include/llvm/CodeGen/MachineInstr.h:1.103.2.1 llvm/include/llvm/CodeGen/MachineInstr.h:1.103.2.2
--- llvm/include/llvm/CodeGen/MachineInstr.h:1.103.2.1	Mon Jun 23 13:51:50 2003
+++ llvm/include/llvm/CodeGen/MachineInstr.h	Fri Jun 27 13:32:05 2003
@@ -13,7 +13,6 @@
 #include "Support/Annotation.h"
 #include "Support/iterator"
 #include <set>
-#include "Config/assert.h"
 
 class Value;
 class Function;


Index: llvm/include/llvm/CodeGen/MachineInstrAnnot.h
diff -u llvm/include/llvm/CodeGen/MachineInstrAnnot.h:1.9.2.1 llvm/include/llvm/CodeGen/MachineInstrAnnot.h:1.9.2.2
--- llvm/include/llvm/CodeGen/MachineInstrAnnot.h:1.9.2.1	Mon Jun 23 13:51:50 2003
+++ llvm/include/llvm/CodeGen/MachineInstrAnnot.h	Fri Jun 27 13:32:05 2003
@@ -10,8 +10,6 @@
 #include "llvm/CodeGen/MachineInstr.h"
 #include "llvm/Target/TargetRegInfo.h"
 
-#include "Config/assert.h"
-
 class Value;
 class TmpInstruction;
 class CallInst;


Index: llvm/include/llvm/CodeGen/SSARegMap.h
diff -u llvm/include/llvm/CodeGen/SSARegMap.h:1.4.2.1 llvm/include/llvm/CodeGen/SSARegMap.h:1.4.2.2
--- llvm/include/llvm/CodeGen/SSARegMap.h:1.4.2.1	Mon Jun 23 13:51:50 2003
+++ llvm/include/llvm/CodeGen/SSARegMap.h	Fri Jun 27 13:32:05 2003
@@ -12,8 +12,6 @@
 
 #include "llvm/Target/MRegisterInfo.h"
 
-#include "Config/assert.h"
-
 class TargetRegisterClass;
 
 class SSARegMap {




From criswell at choi.cs.uiuc.edu  Fri Jun 27 13:34:06 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Fri Jun 27 13:34:06 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Analysis/AliasSetTracker.h CallGraph.h ConstantsScanner.h DSGraph.h DSGraphTraits.h DSNode.h DSSupport.h DataStructure.h DependenceGraph.h Dominators.h IPModRef.h InstForest.h IntervalIterator.h MemoryDepAnalysis.h PgmDependenceGraph.h
Message-ID: <200306271832.h5RIWE924946@choi.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Analysis:

AliasSetTracker.h updated: 1.5.2.2 -> 1.5.2.3
CallGraph.h updated: 1.27.2.2 -> 1.27.2.3
ConstantsScanner.h updated: 1.10.2.2 -> 1.10.2.3
DSGraph.h updated: 1.48.2.2 -> 1.48.2.3
DSGraphTraits.h updated: 1.14.2.2 -> 1.14.2.3
DSNode.h updated: 1.24.2.2 -> 1.24.2.3
DSSupport.h updated: 1.16.2.2 -> 1.16.2.3
DataStructure.h updated: 1.63.2.2 -> 1.63.2.3
DependenceGraph.h updated: 1.4.2.2 -> 1.4.2.3
Dominators.h updated: 1.33.2.1 -> 1.33.2.2
IPModRef.h updated: 1.10.2.1 -> 1.10.2.2
InstForest.h updated: 1.17.2.2 -> 1.17.2.3
IntervalIterator.h updated: 1.11.2.2 -> 1.11.2.3
MemoryDepAnalysis.h updated: 1.2.2.2 -> 1.2.2.3
PgmDependenceGraph.h updated: 1.2.2.2 -> 1.2.2.3

---
Log message:

Removed the user of assert.h.  It is now globally included in
llvm/AbstractTypeUser.h.
Updated some of the files to use the header files from include/Config instead
of directly including the system header files.


---
Diffs of the changes:

Index: llvm/include/llvm/Analysis/AliasSetTracker.h
diff -u llvm/include/llvm/Analysis/AliasSetTracker.h:1.5.2.2 llvm/include/llvm/Analysis/AliasSetTracker.h:1.5.2.3
--- llvm/include/llvm/Analysis/AliasSetTracker.h:1.5.2.2	Thu Jun 26 16:34:42 2003
+++ llvm/include/llvm/Analysis/AliasSetTracker.h	Fri Jun 27 13:32:03 2003
@@ -10,7 +10,6 @@
 #ifndef LLVM_ANALYSIS_ALIASSETTRACKER_H
 #define LLVM_ANALYSIS_ALIASSETTRACKER_H
 
-#include "Config/assert.h"
 #include "llvm/Support/CallSite.h"
 #include "Support/iterator"
 #include "Support/hash_map"


Index: llvm/include/llvm/Analysis/CallGraph.h
diff -u llvm/include/llvm/Analysis/CallGraph.h:1.27.2.2 llvm/include/llvm/Analysis/CallGraph.h:1.27.2.3
--- llvm/include/llvm/Analysis/CallGraph.h:1.27.2.2	Thu Jun 26 16:34:42 2003
+++ llvm/include/llvm/Analysis/CallGraph.h	Fri Jun 27 13:32:03 2003
@@ -41,7 +41,6 @@
 #ifndef LLVM_ANALYSIS_CALLGRAPH_H
 #define LLVM_ANALYSIS_CALLGRAPH_H
 
-#include "Config/assert.h"
 #include "Support/GraphTraits.h"
 #include "Support/STLExtras.h"
 #include "llvm/Pass.h"


Index: llvm/include/llvm/Analysis/ConstantsScanner.h
diff -u llvm/include/llvm/Analysis/ConstantsScanner.h:1.10.2.2 llvm/include/llvm/Analysis/ConstantsScanner.h:1.10.2.3
--- llvm/include/llvm/Analysis/ConstantsScanner.h:1.10.2.2	Thu Jun 26 16:34:42 2003
+++ llvm/include/llvm/Analysis/ConstantsScanner.h	Fri Jun 27 13:32:03 2003
@@ -9,7 +9,6 @@
 #ifndef LLVM_ANALYSIS_CONSTANTSSCANNER_H
 #define LLVM_ANALYSIS_CONSTANTSSCANNER_H
 
-#include "Config/assert.h"
 #include "llvm/Support/InstIterator.h"
 #include "llvm/Instruction.h"
 #include "Support/iterator"


Index: llvm/include/llvm/Analysis/DSGraph.h
diff -u llvm/include/llvm/Analysis/DSGraph.h:1.48.2.2 llvm/include/llvm/Analysis/DSGraph.h:1.48.2.3
--- llvm/include/llvm/Analysis/DSGraph.h:1.48.2.2	Thu Jun 26 16:34:42 2003
+++ llvm/include/llvm/Analysis/DSGraph.h	Fri Jun 27 13:32:03 2003
@@ -7,7 +7,6 @@
 #ifndef LLVM_ANALYSIS_DSGRAPH_H
 #define LLVM_ANALYSIS_DSGRAPH_H
 
-#include "Config/assert.h"
 #include "llvm/Analysis/DSNode.h"
 
 //===----------------------------------------------------------------------===//


Index: llvm/include/llvm/Analysis/DSGraphTraits.h
diff -u llvm/include/llvm/Analysis/DSGraphTraits.h:1.14.2.2 llvm/include/llvm/Analysis/DSGraphTraits.h:1.14.2.3
--- llvm/include/llvm/Analysis/DSGraphTraits.h:1.14.2.2	Thu Jun 26 16:34:42 2003
+++ llvm/include/llvm/Analysis/DSGraphTraits.h	Fri Jun 27 13:32:03 2003
@@ -9,7 +9,6 @@
 #ifndef LLVM_ANALYSIS_DSGRAPHTRAITS_H
 #define LLVM_ANALYSIS_DSGRAPHTRAITS_H
 
-#include "Config/assert.h"
 #include "llvm/Analysis/DSGraph.h"
 #include "Support/GraphTraits.h"
 #include "Support/iterator"


Index: llvm/include/llvm/Analysis/DSNode.h
diff -u llvm/include/llvm/Analysis/DSNode.h:1.24.2.2 llvm/include/llvm/Analysis/DSNode.h:1.24.2.3
--- llvm/include/llvm/Analysis/DSNode.h:1.24.2.2	Thu Jun 26 16:34:42 2003
+++ llvm/include/llvm/Analysis/DSNode.h	Fri Jun 27 13:32:03 2003
@@ -7,7 +7,6 @@
 #ifndef LLVM_ANALYSIS_DSNODE_H
 #define LLVM_ANALYSIS_DSNODE_H
 
-#include "Config/assert.h"
 #include "llvm/Analysis/DSSupport.h"
 
 template<typename BaseType>


Index: llvm/include/llvm/Analysis/DSSupport.h
diff -u llvm/include/llvm/Analysis/DSSupport.h:1.16.2.2 llvm/include/llvm/Analysis/DSSupport.h:1.16.2.3
--- llvm/include/llvm/Analysis/DSSupport.h:1.16.2.2	Thu Jun 26 16:34:42 2003
+++ llvm/include/llvm/Analysis/DSSupport.h	Fri Jun 27 13:32:04 2003
@@ -7,11 +7,9 @@
 #ifndef LLVM_ANALYSIS_DSSUPPORT_H
 #define LLVM_ANALYSIS_DSSUPPORT_H
 
-#include "Config/assert.h"
 #include <vector>
 #include <functional>
 #include <string>
-#include <assert.h>
 #include "Support/HashExtras.h"
 #include "Support/hash_set"
 


Index: llvm/include/llvm/Analysis/DataStructure.h
diff -u llvm/include/llvm/Analysis/DataStructure.h:1.63.2.2 llvm/include/llvm/Analysis/DataStructure.h:1.63.2.3
--- llvm/include/llvm/Analysis/DataStructure.h:1.63.2.2	Thu Jun 26 16:34:42 2003
+++ llvm/include/llvm/Analysis/DataStructure.h	Fri Jun 27 13:32:04 2003
@@ -7,7 +7,6 @@
 #ifndef LLVM_ANALYSIS_DATA_STRUCTURE_H
 #define LLVM_ANALYSIS_DATA_STRUCTURE_H
 
-#include "Config/assert.h"
 #include "llvm/Pass.h"
 #include "Support/HashExtras.h"
 #include "Support/hash_set"


Index: llvm/include/llvm/Analysis/DependenceGraph.h
diff -u llvm/include/llvm/Analysis/DependenceGraph.h:1.4.2.2 llvm/include/llvm/Analysis/DependenceGraph.h:1.4.2.3
--- llvm/include/llvm/Analysis/DependenceGraph.h:1.4.2.2	Thu Jun 26 16:34:42 2003
+++ llvm/include/llvm/Analysis/DependenceGraph.h	Fri Jun 27 13:32:04 2003
@@ -22,8 +22,6 @@
 #include <vector>
 #include <utility>
 
-#include "Config/assert.h"
-
 class Instruction;
 class Function;
 class Dependence;


Index: llvm/include/llvm/Analysis/Dominators.h
diff -u llvm/include/llvm/Analysis/Dominators.h:1.33.2.1 llvm/include/llvm/Analysis/Dominators.h:1.33.2.2
--- llvm/include/llvm/Analysis/Dominators.h:1.33.2.1	Mon Jun 23 13:51:47 2003
+++ llvm/include/llvm/Analysis/Dominators.h	Fri Jun 27 13:32:04 2003
@@ -20,7 +20,7 @@
 
 #include "llvm/Pass.h"
 #include <set>
-#include "Config/assert.h"
+
 class Instruction;
 
 template <typename GraphType> struct GraphTraits;


Index: llvm/include/llvm/Analysis/IPModRef.h
diff -u llvm/include/llvm/Analysis/IPModRef.h:1.10.2.1 llvm/include/llvm/Analysis/IPModRef.h:1.10.2.2
--- llvm/include/llvm/Analysis/IPModRef.h:1.10.2.1	Mon Jun 23 13:51:47 2003
+++ llvm/include/llvm/Analysis/IPModRef.h	Fri Jun 27 13:32:04 2003
@@ -39,8 +39,6 @@
 #ifndef LLVM_ANALYSIS_IPMODREF_H
 #define LLVM_ANALYSIS_IPMODREF_H
 
-#include "Config/assert.h"
-
 #include "llvm/Pass.h"
 #include "Support/BitSetVector.h"
 #include "Support/hash_map"


Index: llvm/include/llvm/Analysis/InstForest.h
diff -u llvm/include/llvm/Analysis/InstForest.h:1.17.2.2 llvm/include/llvm/Analysis/InstForest.h:1.17.2.3
--- llvm/include/llvm/Analysis/InstForest.h:1.17.2.2	Thu Jun 26 16:34:42 2003
+++ llvm/include/llvm/Analysis/InstForest.h	Fri Jun 27 13:32:04 2003
@@ -17,7 +17,6 @@
 #include "llvm/Function.h"
 #include "Support/Tree.h"
 #include <map>
-#include "Config/assert.h"
 
 template<class Payload> class InstTreeNode;
 template<class Payload> class InstForest;


Index: llvm/include/llvm/Analysis/IntervalIterator.h
diff -u llvm/include/llvm/Analysis/IntervalIterator.h:1.11.2.2 llvm/include/llvm/Analysis/IntervalIterator.h:1.11.2.3
--- llvm/include/llvm/Analysis/IntervalIterator.h:1.11.2.2	Thu Jun 26 16:34:42 2003
+++ llvm/include/llvm/Analysis/IntervalIterator.h	Fri Jun 27 13:32:04 2003
@@ -32,7 +32,6 @@
 #include <stack>
 #include <set>
 #include <algorithm>
-#include "Config/assert.h"
 
 // getNodeHeader - Given a source graph node and the source graph, return the 
 // BasicBlock that is the header node.  This is the opposite of


Index: llvm/include/llvm/Analysis/MemoryDepAnalysis.h
diff -u llvm/include/llvm/Analysis/MemoryDepAnalysis.h:1.2.2.2 llvm/include/llvm/Analysis/MemoryDepAnalysis.h:1.2.2.3
--- llvm/include/llvm/Analysis/MemoryDepAnalysis.h:1.2.2.2	Thu Jun 26 16:34:42 2003
+++ llvm/include/llvm/Analysis/MemoryDepAnalysis.h	Fri Jun 27 13:32:04 2003
@@ -19,7 +19,6 @@
 #include "llvm/Pass.h"
 #include "Support/TarjanSCCIterator.h"
 #include "Support/hash_map"
-#include "Config/assert.h"
 
 class Instruction;
 class ModRefTable;


Index: llvm/include/llvm/Analysis/PgmDependenceGraph.h
diff -u llvm/include/llvm/Analysis/PgmDependenceGraph.h:1.2.2.2 llvm/include/llvm/Analysis/PgmDependenceGraph.h:1.2.2.3
--- llvm/include/llvm/Analysis/PgmDependenceGraph.h:1.2.2.2	Thu Jun 26 16:34:42 2003
+++ llvm/include/llvm/Analysis/PgmDependenceGraph.h	Fri Jun 27 13:32:04 2003
@@ -39,7 +39,6 @@
 #include "llvm/Pass.h"
 #include <iterator>
 
-#include "Config/assert.h"
 class DSGraph;
 class DependenceGraph;
 class PgmDependenceGraph;




From criswell at choi.cs.uiuc.edu  Fri Jun 27 13:34:16 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Fri Jun 27 13:34:16 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/AbstractTypeUser.h Constant.h ConstantHandling.h Constants.h DerivedTypes.h GlobalVariable.h Instruction.h Pass.h PassAnalysisSupport.h PassSupport.h SymbolTable.h User.h iOther.h iPHINode.h iTerminators.h
Message-ID: <200306271832.h5RIWCE24909@choi.cs.uiuc.edu>


Changes in directory llvm/include/llvm:

AbstractTypeUser.h updated: 1.6.2.2 -> 1.6.2.3
Constant.h updated: 1.7.2.1 -> 1.7.2.2
ConstantHandling.h updated: 1.26.2.1 -> 1.26.2.2
Constants.h updated: 1.26.2.2 -> 1.26.2.3
DerivedTypes.h updated: 1.29.2.2 -> 1.29.2.3
GlobalVariable.h updated: 1.20.2.2 -> 1.20.2.3
Instruction.h updated: 1.40.2.2 -> 1.40.2.3
Pass.h updated: 1.33.2.1 -> 1.33.2.2
PassAnalysisSupport.h updated: 1.12.2.1 -> 1.12.2.2
PassSupport.h updated: 1.12.2.1 -> 1.12.2.2
SymbolTable.h updated: 1.21.2.2 -> 1.21.2.3
User.h updated: 1.18.2.2 -> 1.18.2.3
iOther.h updated: 1.35.2.2 -> 1.35.2.3
iPHINode.h updated: 1.9.2.2 -> 1.9.2.3
iTerminators.h updated: 1.27.2.2 -> 1.27.2.3

---
Log message:

Removed the user of assert.h.  It is now globally included in
llvm/AbstractTypeUser.h.
Updated some of the files to use the header files from include/Config instead
of directly including the system header files.


---
Diffs of the changes:

Index: llvm/include/llvm/AbstractTypeUser.h
diff -u llvm/include/llvm/AbstractTypeUser.h:1.6.2.2 llvm/include/llvm/AbstractTypeUser.h:1.6.2.3
--- llvm/include/llvm/AbstractTypeUser.h:1.6.2.2	Thu Jun 26 16:34:37 2003
+++ llvm/include/llvm/AbstractTypeUser.h	Fri Jun 27 13:32:02 2003
@@ -21,7 +21,16 @@
 #ifndef LLVM_ABSTRACT_TYPE_USER_H
 #define LLVM_ABSTRACT_TYPE_USER_H
 
+//
+// This is the "master" include for assert.h
+// Whether this file needs it or not, it must always include assert.h for the
+// files which include llvm/AbstractTypeUser.h
+//
+// In this way, most every LLVM source file will have access to the assert()
+// macro without having to #include <assert.h> directly.
+//
 #include "Config/assert.h"
+
 class Type;
 class DerivedType;
 


Index: llvm/include/llvm/Constant.h
diff -u llvm/include/llvm/Constant.h:1.7.2.1 llvm/include/llvm/Constant.h:1.7.2.2
--- llvm/include/llvm/Constant.h:1.7.2.1	Mon Jun 23 13:51:40 2003
+++ llvm/include/llvm/Constant.h	Fri Jun 27 13:32:02 2003
@@ -7,7 +7,6 @@
 #ifndef LLVM_CONSTANT_H
 #define LLVM_CONSTANT_H
 
-#include "Config/assert.h"
 #include "llvm/User.h"
 
 class Constant : public User {


Index: llvm/include/llvm/ConstantHandling.h
diff -u llvm/include/llvm/ConstantHandling.h:1.26.2.1 llvm/include/llvm/ConstantHandling.h:1.26.2.2
--- llvm/include/llvm/ConstantHandling.h:1.26.2.1	Mon Jun 23 13:51:40 2003
+++ llvm/include/llvm/ConstantHandling.h	Fri Jun 27 13:32:02 2003
@@ -33,8 +33,6 @@
 #ifndef LLVM_CONSTANTHANDLING_H
 #define LLVM_CONSTANTHANDLING_H
 
-#include "Config/assert.h"
-
 #include "llvm/Constants.h"
 #include "llvm/Type.h"
 class PointerType;


Index: llvm/include/llvm/Constants.h
diff -u llvm/include/llvm/Constants.h:1.26.2.2 llvm/include/llvm/Constants.h:1.26.2.3
--- llvm/include/llvm/Constants.h:1.26.2.2	Thu Jun 26 16:34:37 2003
+++ llvm/include/llvm/Constants.h	Fri Jun 27 13:32:02 2003
@@ -8,7 +8,6 @@
 #ifndef LLVM_CONSTANTS_H
 #define LLVM_CONSTANTS_H
 
-#include "Config/assert.h"
 #include "llvm/Constant.h"
 #include "Support/DataTypes.h"
 


Index: llvm/include/llvm/DerivedTypes.h
diff -u llvm/include/llvm/DerivedTypes.h:1.29.2.2 llvm/include/llvm/DerivedTypes.h:1.29.2.3
--- llvm/include/llvm/DerivedTypes.h:1.29.2.2	Thu Jun 26 16:34:37 2003
+++ llvm/include/llvm/DerivedTypes.h	Fri Jun 27 13:32:02 2003
@@ -11,7 +11,6 @@
 #ifndef LLVM_DERIVED_TYPES_H
 #define LLVM_DERIVED_TYPES_H
 
-#include "Config/assert.h"
 #include "llvm/Type.h"
 
 class DerivedType : public Type {


Index: llvm/include/llvm/GlobalVariable.h
diff -u llvm/include/llvm/GlobalVariable.h:1.20.2.2 llvm/include/llvm/GlobalVariable.h:1.20.2.3
--- llvm/include/llvm/GlobalVariable.h:1.20.2.2	Thu Jun 26 16:34:37 2003
+++ llvm/include/llvm/GlobalVariable.h	Fri Jun 27 13:32:02 2003
@@ -13,7 +13,6 @@
 #ifndef LLVM_GLOBAL_VARIABLE_H
 #define LLVM_GLOBAL_VARIABLE_H
 
-#include "Config/assert.h"
 #include "llvm/GlobalValue.h"
 
 class Module;


Index: llvm/include/llvm/Instruction.h
diff -u llvm/include/llvm/Instruction.h:1.40.2.2 llvm/include/llvm/Instruction.h:1.40.2.3
--- llvm/include/llvm/Instruction.h:1.40.2.2	Thu Jun 26 16:34:37 2003
+++ llvm/include/llvm/Instruction.h	Fri Jun 27 13:32:02 2003
@@ -8,7 +8,6 @@
 #ifndef LLVM_INSTRUCTION_H
 #define LLVM_INSTRUCTION_H
 
-#include "Config/assert.h"
 #include "llvm/User.h"
 
 template<typename SC> struct ilist_traits;


Index: llvm/include/llvm/Pass.h
diff -u llvm/include/llvm/Pass.h:1.33.2.1 llvm/include/llvm/Pass.h:1.33.2.2
--- llvm/include/llvm/Pass.h:1.33.2.1	Mon Jun 23 13:51:40 2003
+++ llvm/include/llvm/Pass.h	Fri Jun 27 13:32:02 2003
@@ -22,8 +22,6 @@
 #ifndef LLVM_PASS_H
 #define LLVM_PASS_H
 
-#include "Config/assert.h"
-
 #include <vector>
 #include <map>
 #include <iosfwd>


Index: llvm/include/llvm/PassAnalysisSupport.h
diff -u llvm/include/llvm/PassAnalysisSupport.h:1.12.2.1 llvm/include/llvm/PassAnalysisSupport.h:1.12.2.2
--- llvm/include/llvm/PassAnalysisSupport.h:1.12.2.1	Mon Jun 23 13:51:40 2003
+++ llvm/include/llvm/PassAnalysisSupport.h	Fri Jun 27 13:32:02 2003
@@ -14,8 +14,6 @@
 
 // No need to include Pass.h, we are being included by it!
 
-#include "Config/assert.h"
-
 //===----------------------------------------------------------------------===//
 // AnalysisUsage - Represent the analysis usage information of a pass.  This
 // tracks analyses that the pass REQUIRES (must available when the pass runs),


Index: llvm/include/llvm/PassSupport.h
diff -u llvm/include/llvm/PassSupport.h:1.12.2.1 llvm/include/llvm/PassSupport.h:1.12.2.2
--- llvm/include/llvm/PassSupport.h:1.12.2.1	Mon Jun 23 13:51:40 2003
+++ llvm/include/llvm/PassSupport.h	Fri Jun 27 13:32:02 2003
@@ -14,8 +14,6 @@
 #ifndef LLVM_PASS_SUPPORT_H
 #define LLVM_PASS_SUPPORT_H
 
-#include "Config/assert.h"
-
 // No need to include Pass.h, we are being included by it!
 
 class TargetMachine;


Index: llvm/include/llvm/SymbolTable.h
diff -u llvm/include/llvm/SymbolTable.h:1.21.2.2 llvm/include/llvm/SymbolTable.h:1.21.2.3
--- llvm/include/llvm/SymbolTable.h:1.21.2.2	Thu Jun 26 16:34:37 2003
+++ llvm/include/llvm/SymbolTable.h	Fri Jun 27 13:32:02 2003
@@ -16,7 +16,6 @@
 #ifndef LLVM_SYMBOL_TABLE_H
 #define LLVM_SYMBOL_TABLE_H
 
-#include "Config/assert.h"
 #include "llvm/Value.h"
 #include <map>
 


Index: llvm/include/llvm/User.h
diff -u llvm/include/llvm/User.h:1.18.2.2 llvm/include/llvm/User.h:1.18.2.3
--- llvm/include/llvm/User.h:1.18.2.2	Thu Jun 26 16:34:37 2003
+++ llvm/include/llvm/User.h	Fri Jun 27 13:32:02 2003
@@ -12,7 +12,6 @@
 #ifndef LLVM_USER_H
 #define LLVM_USER_H
 
-#include "Config/assert.h"
 #include "llvm/Value.h"
 
 class User : public Value {


Index: llvm/include/llvm/iOther.h
diff -u llvm/include/llvm/iOther.h:1.35.2.2 llvm/include/llvm/iOther.h:1.35.2.3
--- llvm/include/llvm/iOther.h:1.35.2.2	Thu Jun 26 16:34:37 2003
+++ llvm/include/llvm/iOther.h	Fri Jun 27 13:32:02 2003
@@ -8,7 +8,6 @@
 #ifndef LLVM_IOTHER_H
 #define LLVM_IOTHER_H
 
-#include "Config/assert.h"
 #include "llvm/InstrTypes.h"
 
 //===----------------------------------------------------------------------===//


Index: llvm/include/llvm/iPHINode.h
diff -u llvm/include/llvm/iPHINode.h:1.9.2.2 llvm/include/llvm/iPHINode.h:1.9.2.3
--- llvm/include/llvm/iPHINode.h:1.9.2.2	Thu Jun 26 16:34:37 2003
+++ llvm/include/llvm/iPHINode.h	Fri Jun 27 13:32:02 2003
@@ -7,7 +7,6 @@
 #ifndef LLVM_IPHINODE_H
 #define LLVM_IPHINODE_H
 
-#include "Config/assert.h"
 #include "llvm/Instruction.h"
 class BasicBlock;
 


Index: llvm/include/llvm/iTerminators.h
diff -u llvm/include/llvm/iTerminators.h:1.27.2.2 llvm/include/llvm/iTerminators.h:1.27.2.3
--- llvm/include/llvm/iTerminators.h:1.27.2.2	Thu Jun 26 16:34:37 2003
+++ llvm/include/llvm/iTerminators.h	Fri Jun 27 13:32:02 2003
@@ -9,7 +9,6 @@
 #ifndef LLVM_ITERMINATORS_H
 #define LLVM_ITERMINATORS_H
 
-#include "Config/assert.h"
 #include "llvm/InstrTypes.h"
 
 //===---------------------------------------------------------------------------




From criswell at choi.cs.uiuc.edu  Fri Jun 27 13:36:01 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Fri Jun 27 13:36:01 2003
Subject: [llvm-commits] CVS: llvm/Makefile.common
Message-ID: <200306271835.h5RIZZY25508@choi.cs.uiuc.edu>


Changes in directory llvm:

Makefile.common updated: 1.88.2.6 -> 1.88.2.7

---
Log message:

Merged with HEAD on Jun 27, 2003 to get the OPTIONAL_DIRS functionality.


---
Diffs of the changes:

Index: llvm/Makefile.common
diff -u llvm/Makefile.common:1.88.2.6 llvm/Makefile.common:1.88.2.7
--- llvm/Makefile.common:1.88.2.6	Thu Jun 26 16:33:47 2003
+++ llvm/Makefile.common	Fri Jun 27 13:35:24 2003
@@ -50,6 +50,7 @@
 # 	LLVM_OBJ_ROOT = BUILD_OBJ_ROOT.
 #===-----------------------------------------------------------------------====
 
+#
 # Configuration file to set paths specific to local installation of LLVM
 # 
 include $(LEVEL)/Makefile.config
@@ -421,6 +422,17 @@
 	$(VERB) cd $(@D); $(MAKE) $(subst $(@D)/.make,,$@)
 endif
 
+# Handle directories that may or may not exist
+ifdef OPTIONAL_DIRS
+all install clean test ::
+	$(VERB) for dir in ${OPTIONAL_DIRS}; do \
+		if [ -d $$dir ]; \
+		then\
+			(cd $$dir; $(MAKE) $@) || exit 1; \
+		fi \
+	done
+endif
+
 ###########################################################################
 # Library Build Rules:
 #
@@ -606,6 +618,7 @@
         $(addsuffix /.makeall, $(PARALLEL_DIRS))
 
 all::   $(TOOLEXENAMES)
+
 clean::
 	$(VERB) $(RM) -f $(TOOLEXENAMES)
 




From criswell at choi.cs.uiuc.edu  Fri Jun 27 14:16:00 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Fri Jun 27 14:16:00 2003
Subject: [llvm-commits] CVS: llvm/lib/Reoptimizer/BinInterface/sparcdis.cpp
Message-ID: <200306271915.h5RJFhu28555@choi.cs.uiuc.edu>


Changes in directory llvm/lib/Reoptimizer/BinInterface:

sparcdis.cpp updated: 1.12.2.1 -> 1.12.2.2

---
Log message:

Added Config/assert.h since we're not grabbing it via other header files.


---
Diffs of the changes:

Index: llvm/lib/Reoptimizer/BinInterface/sparcdis.cpp
diff -u llvm/lib/Reoptimizer/BinInterface/sparcdis.cpp:1.12.2.1 llvm/lib/Reoptimizer/BinInterface/sparcdis.cpp:1.12.2.2
--- llvm/lib/Reoptimizer/BinInterface/sparcdis.cpp:1.12.2.1	Fri Jun 27 13:32:12 2003
+++ llvm/lib/Reoptimizer/BinInterface/sparcdis.cpp	Fri Jun 27 14:15:32 2003
@@ -21,6 +21,7 @@
 #include "llvm/Reoptimizer/BinInterface/sparc9.h"   // SPARC9 opcode definitions
 #include "Config/stdio.h"
 #include "Config/stdlib.h"
+#include "Config/assert.h"
 
 void sparc_printop_rs1(unsigned instr, int labelrs1)
 {




From criswell at choi.cs.uiuc.edu  Fri Jun 27 14:56:01 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Fri Jun 27 14:56:01 2003
Subject: [llvm-commits] CVS: llvm/include/Config/alloca.h
Message-ID: <200306271955.h5RJtRf25267@choi.cs.uiuc.edu>


Changes in directory llvm/include/Config:

alloca.h updated: 1.1.2.1 -> 1.1.2.2

---
Log message:

Merged in Brian's suggestion of including stdlib.h if alloca.h does not exist.
This will hopefully fix the build on FreeBSD.


---
Diffs of the changes:

Index: llvm/include/Config/alloca.h
diff -u llvm/include/Config/alloca.h:1.1.2.1 llvm/include/Config/alloca.h:1.1.2.2
--- llvm/include/Config/alloca.h:1.1.2.1	Mon Jun 23 13:45:36 2003
+++ llvm/include/Config/alloca.h	Fri Jun 27 14:55:17 2003
@@ -13,8 +13,10 @@
 #include "Config/config.h"
 
 /*
- * This is the suggested use by the Autoconf manual.
+ * This is a modified form of what is suggested by the Autoconf manual.
  *	1) The #pragma is indented so that pre-ANSI C compilers ignore it.
+ *	2) We attempt to include stdlib.h if we don't find alloca.h since
+ *	   on some systems (notably FreeBSD) alloca() requires stdlib.h.
  */
 #ifndef __GNUC__
 #	ifdef HAVE_ALLOCA_H
@@ -32,7 +34,11 @@
 #	ifdef HAVE_ALLOCA_H
 #		include <alloca.h>
 #	else
-#		error "The function alloca() is required but not found!"
+#		ifdef HAVE_STDLIB_H
+#			include <stdlib.h>
+#		else
+#			error "The function alloca() is required but not found!"
+#		endif
 #	endif
 #endif
 




From criswell at choi.cs.uiuc.edu  Fri Jun 27 16:45:01 2003
From: criswell at choi.cs.uiuc.edu (John Criswell)
Date: Fri Jun 27 16:45:01 2003
Subject: [llvm-commits] CVS: llvm/Makefile.config.in
Message-ID: <200306272144.h5RLirB24912@choi.cs.uiuc.edu>


Changes in directory llvm:

Makefile.config.in updated: 1.1.2.2 -> 1.1.2.3

---
Log message:

Commented out LIBS since it we don't currently use it and it causes the
tests to break.


---
Diffs of the changes:

Index: llvm/Makefile.config.in
diff -u llvm/Makefile.config.in:1.1.2.2 llvm/Makefile.config.in:1.1.2.3
--- llvm/Makefile.config.in:1.1.2.2	Tue Jun 24 16:19:14 2003
+++ llvm/Makefile.config.in	Fri Jun 27 16:44:42 2003
@@ -29,7 +29,7 @@
 CPPFLAGS=@DEFS@
 CCFLAGS=@DEFS@
 LDFLAGS=@LDFLAGS@
-LIBS=@LIBS@
+#LIBS=@LIBS@
 
 #
 # Libraries needed by tools




From lattner at cs.uiuc.edu  Sat Jun 28 10:48:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 10:48:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Support/CommandLine.cpp
Message-ID: <200306281547.KAA20143@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Support:

CommandLine.cpp updated: 1.29 -> 1.30

---
Log message:

Add support for 'unsigned' command line arguments



---
Diffs of the changes:

Index: llvm/lib/Support/CommandLine.cpp
diff -u llvm/lib/Support/CommandLine.cpp:1.29 llvm/lib/Support/CommandLine.cpp:1.30
--- llvm/lib/Support/CommandLine.cpp:1.29	Thu May 22 15:26:17 2003
+++ llvm/lib/Support/CommandLine.cpp	Sat Jun 28 10:47:20 2003
@@ -575,11 +575,22 @@
 //
 bool parser<int>::parse(Option &O, const char *ArgName,
                         const std::string &Arg, int &Value) {
-  const char *ArgStart = Arg.c_str();
   char *End;
-  Value = (int)strtol(ArgStart, &End, 0);
+  Value = (int)strtol(Arg.c_str(), &End, 0);
   if (*End != 0) 
     return O.error(": '" + Arg + "' value invalid for integer argument!");
+  return false;
+}
+
+// parser<unsigned> implementation
+//
+bool parser<unsigned>::parse(Option &O, const char *ArgName,
+                             const std::string &Arg, unsigned &Value) {
+  char *End;
+  long long int V = strtoll(Arg.c_str(), &End, 0);
+  Value = (unsigned)V;
+  if (*End != 0 || V < 0 || Value != V) 
+    return O.error(": '" + Arg + "' value invalid for uint argument!");
   return false;
 }
 




From lattner at cs.uiuc.edu  Sat Jun 28 10:48:14 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 10:48:14 2003
Subject: [llvm-commits] CVS: llvm/include/Support/CommandLine.h
Message-ID: <200306281547.KAA20136@apoc.cs.uiuc.edu>


Changes in directory llvm/include/Support:

CommandLine.h updated: 1.14 -> 1.15

---
Log message:

Add support for 'unsigned' command line arguments


---
Diffs of the changes:

Index: llvm/include/Support/CommandLine.h
diff -u llvm/include/Support/CommandLine.h:1.14 llvm/include/Support/CommandLine.h:1.15
--- llvm/include/Support/CommandLine.h:1.14	Sat Jun 21 22:08:05 2003
+++ llvm/include/Support/CommandLine.h	Sat Jun 28 10:47:18 2003
@@ -516,6 +516,21 @@
 
 
 //--------------------------------------------------
+// parser<unsigned>
+//
+template<>
+struct parser<unsigned> : public basic_parser<unsigned> {
+  
+  // parse - Return true on error.
+  bool parse(Option &O, const char *ArgName, const std::string &Arg,
+             unsigned &Val);
+
+  // getValueName - Overload in subclass to provide a better default value.
+  virtual const char *getValueName() const { return "uint"; }
+};
+
+
+//--------------------------------------------------
 // parser<double>
 //
 template<>




From lattner at cs.uiuc.edu  Sat Jun 28 10:58:02 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 10:58:02 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/IPO/FunctionInlining.cpp
Message-ID: <200306281557.KAA21806@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/IPO:

FunctionInlining.cpp updated: 1.45 -> 1.46

---
Log message:

Allow the inlining limit to be controlled from the command line!


---
Diffs of the changes:

Index: llvm/lib/Transforms/IPO/FunctionInlining.cpp
diff -u llvm/lib/Transforms/IPO/FunctionInlining.cpp:1.45 llvm/lib/Transforms/IPO/FunctionInlining.cpp:1.46
--- llvm/lib/Transforms/IPO/FunctionInlining.cpp:1.45	Thu May 29 10:11:31 2003
+++ llvm/lib/Transforms/IPO/FunctionInlining.cpp	Sat Jun 28 10:57:04 2003
@@ -11,10 +11,14 @@
 #include "llvm/iOther.h"
 #include "llvm/iMemory.h"
 #include "Support/Statistic.h"
+#include "Support/CommandLine.h"
 #include <set>
 
 namespace {
   Statistic<> NumInlined("inline", "Number of functions inlined");
+  cl::opt<unsigned>             // FIXME: 200 is VERY conservative
+  InlineLimit("inline-threshold", cl::Hidden, cl::init(200),
+              cl::desc("Control the amount of inlining to perform (default = 200)"));
 
   struct FunctionInlining : public Pass {
     virtual bool run(Module &M) {
@@ -55,7 +59,7 @@
   // inliner is.  If this value is negative after the final computation,
   // inlining is not performed.
   //
-  int InlineQuality = 200;            // FIXME: This is VERY conservative
+  int InlineQuality = InlineLimit;
 
   // If there is only one call of the function, and it has internal linkage,
   // make it almost guaranteed to be inlined.




From lattner at cs.uiuc.edu  Sat Jun 28 12:08:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 12:08:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/CBackend/2003-06-28-LinkOnceGlobalVars.llx
Message-ID: <200306281707.MAA24612@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/CBackend:

2003-06-28-LinkOnceGlobalVars.llx added (r1.1)

---
Log message:

New testcase


---
Diffs of the changes:

Index: llvm/test/Regression/CBackend/2003-06-28-LinkOnceGlobalVars.llx
diff -c /dev/null llvm/test/Regression/CBackend/2003-06-28-LinkOnceGlobalVars.llx:1.1
*** /dev/null	Sat Jun 28 12:07:34 2003
--- llvm/test/Regression/CBackend/2003-06-28-LinkOnceGlobalVars.llx	Sat Jun 28 12:07:23 2003
***************
*** 0 ****
--- 1,4 ----
+ ; RUN: as < %s | dis -c | grep common | grep X
+ 
+ %X = linkonce global int 5
+ 




From lattner at cs.uiuc.edu  Sat Jun 28 12:09:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 12:09:01 2003
Subject: [llvm-commits] CVS: llvm/lib/CWriter/Writer.cpp
Message-ID: <200306281708.MAA24626@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/CWriter:

Writer.cpp updated: 1.100 -> 1.101

---
Log message:

Fix bug: CBackend/2003-06-28-LinkOnceGlobalVars.llx


---
Diffs of the changes:

Index: llvm/lib/CWriter/Writer.cpp
diff -u llvm/lib/CWriter/Writer.cpp:1.100 llvm/lib/CWriter/Writer.cpp:1.101
--- llvm/lib/CWriter/Writer.cpp:1.100	Tue Jun 24 23:52:09 2003
+++ llvm/lib/CWriter/Writer.cpp	Sat Jun 28 12:08:36 2003
@@ -682,6 +682,8 @@
         if (I->hasInternalLinkage())
           Out << "static ";
         printType(Out, I->getType()->getElementType(), getValueName(I));
+        if (I->hasLinkOnceLinkage())
+          Out << " __attribute__((common))";
         if (!I->getInitializer()->isNullValue()) {
           Out << " = " ;
           writeOperand(I->getInitializer());




From lattner at cs.uiuc.edu  Sat Jun 28 12:16:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 12:16:01 2003
Subject: [llvm-commits] CVS: llvm/lib/CWriter/Writer.cpp
Message-ID: <200306281715.MAA24789@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/CWriter:

Writer.cpp updated: 1.101 -> 1.102

---
Log message:

Preserve compatibility with non-gcc compilers


---
Diffs of the changes:

Index: llvm/lib/CWriter/Writer.cpp
diff -u llvm/lib/CWriter/Writer.cpp:1.101 llvm/lib/CWriter/Writer.cpp:1.102
--- llvm/lib/CWriter/Writer.cpp:1.101	Sat Jun 28 12:08:36 2003
+++ llvm/lib/CWriter/Writer.cpp	Sat Jun 28 12:15:12 2003
@@ -572,9 +572,13 @@
   return Changed;
 }
 
-static void generateAllocaDecl(std::ostream& Out) {
-  // On SunOS, we need to insert the alloca macro & proto for the builtin.
-  Out << "#ifdef sun\n"
+// generateCompilerSpecificCode - This is where we add conditional compilation
+// directives to cater to specific compilers as need be.
+//
+static void generateCompilerSpecificCode(std::ostream& Out) {
+  // Alloca is hard to get, and we don't want to include stdlib.h here...
+  Out << "/* get a declaration for alloca */\n"
+      << "#ifdef sun\n"
       << "extern void *__builtin_alloca(unsigned long);\n"
       << "#define alloca(x) __builtin_alloca(x)\n"
       << "#else\n"
@@ -582,6 +586,12 @@
       << "#include <alloca.h>\n"
       << "#endif\n"
       << "#endif\n\n";
+
+  // We output GCC specific attributes to preserve 'linkonce'ness on globals.
+  // If we aren't being compiled with GCC, just drop these attributes.
+  Out << "#ifndef __GNUC__\n"
+      << "#define __attribute__(X)\n"
+      << "#endif\n";
 }
 
 void CWriter::printModule(Module *M) {
@@ -606,9 +616,9 @@
 
   // get declaration for alloca
   Out << "/* Provide Declarations */\n";
-  generateAllocaDecl(Out);
   Out << "#include <stdarg.h>\n";
   Out << "#include <setjmp.h>\n";
+  generateCompilerSpecificCode(Out);
   
   // Provide a definition for `bool' if not compiling with a C++ compiler.
   Out << "\n"




From lattner at cs.uiuc.edu  Sat Jun 28 12:53:03 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 12:53:03 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/CBackend/2003-06-28-InvokeSupport.ll
Message-ID: <200306281752.MAA26279@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/CBackend:

2003-06-28-InvokeSupport.ll added (r1.1)

---
Log message:

New testcase for invoke support


---
Diffs of the changes:

Index: llvm/test/Regression/CBackend/2003-06-28-InvokeSupport.ll
diff -c /dev/null llvm/test/Regression/CBackend/2003-06-28-InvokeSupport.ll:1.1
*** /dev/null	Sat Jun 28 12:52:49 2003
--- llvm/test/Regression/CBackend/2003-06-28-InvokeSupport.ll	Sat Jun 28 12:52:38 2003
***************
*** 0 ****
--- 1,12 ----
+ 
+ declare int %callee(int, int)
+ 
+ 
+ int %test(int %X) {
+ 	%A = invoke int %callee(int %X, int 5) to label %Ok except label %Threw
+ Ok:
+ 	%B = phi int [%A, %0], [-1, %Threw]
+ 	ret int %A
+ Threw:
+ 	br label %Ok
+ }




From lattner at cs.uiuc.edu  Sat Jun 28 12:54:02 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 12:54:02 2003
Subject: [llvm-commits] CVS: llvm/lib/CWriter/Writer.cpp
Message-ID: <200306281753.MAA26362@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/CWriter:

Writer.cpp updated: 1.102 -> 1.103

---
Log message:

Add support for the Invoke instruction!


---
Diffs of the changes:

Index: llvm/lib/CWriter/Writer.cpp
diff -u llvm/lib/CWriter/Writer.cpp:1.102 llvm/lib/CWriter/Writer.cpp:1.103
--- llvm/lib/CWriter/Writer.cpp:1.102	Sat Jun 28 12:15:12 2003
+++ llvm/lib/CWriter/Writer.cpp	Sat Jun 28 12:53:05 2003
@@ -17,6 +17,7 @@
 #include "llvm/Analysis/ConstantsScanner.h"
 #include "llvm/Support/InstVisitor.h"
 #include "llvm/Support/InstIterator.h"
+#include "llvm/Support/CallSite.h"
 #include "Support/StringExtras.h"
 #include "Support/STLExtras.h"
 #include <algorithm>
@@ -30,7 +31,7 @@
     const Module *TheModule;
     std::map<const Type *, std::string> TypeNames;
     std::set<const Value*> MangledGlobals;
-    bool needsMalloc;
+    bool needsMalloc, emittedInvoke;
 
     std::map<const ConstantFP *, unsigned> FPConstantMap;
   public:
@@ -117,12 +118,14 @@
     void visitReturnInst(ReturnInst &I);
     void visitBranchInst(BranchInst &I);
     void visitSwitchInst(SwitchInst &I);
+    void visitInvokeInst(InvokeInst &I);
 
     void visitPHINode(PHINode &I);
     void visitBinaryOperator(Instruction &I);
 
     void visitCastInst (CastInst &I);
     void visitCallInst (CallInst &I);
+    void visitCallSite (CallSite CS);
     void visitShiftInst(ShiftInst &I) { visitBinaryOperator(I); }
 
     void visitMallocInst(MallocInst &I);
@@ -589,7 +592,7 @@
 
   // We output GCC specific attributes to preserve 'linkonce'ness on globals.
   // If we aren't being compiled with GCC, just drop these attributes.
-  Out << "#ifndef __GNUC__\n"
+  Out << "#ifndef __GNUC__  /* Can only support \"linkonce\" vars with GCC */\n"
       << "#define __attribute__(X)\n"
       << "#endif\n";
 }
@@ -628,6 +631,11 @@
       << "typedef unsigned long long ConstantDoubleTy;\n"
       << "typedef unsigned int        ConstantFloatTy;\n"
     
+      << "\n\n/* Support for the invoke instruction */\n"
+      << "extern struct __llvm_jmpbuf_list_t {\n"
+      << "  jmp_buf buf; struct __llvm_jmpbuf_list_t *next;\n"
+      << "} *__llvm_jmpbuf_list;\n"
+
       << "\n\n/* Global Declarations */\n";
 
   // First output all the declarations for the program, because C requires
@@ -703,11 +711,20 @@
   }
 
   // Output all of the functions...
+  emittedInvoke = false;
   if (!M->empty()) {
     Out << "\n\n/* Function Bodies */\n";
     for (Module::iterator I = M->begin(), E = M->end(); I != E; ++I)
       printFunction(I);
   }
+
+  // If the program included an invoke instruction, we need to output the
+  // support code for it here!
+  if (emittedInvoke) {
+    Out << "\n/* More support for the invoke instruction */\n"
+        << "struct __llvm_jmpbuf_list_t *__llvm_jmpbuf_list "
+        << "__attribute__((common)) = 0;\n";
+  }
 }
 
 
@@ -982,6 +999,24 @@
   Out << "  }\n";
 }
 
+void CWriter::visitInvokeInst(InvokeInst &II) {
+  Out << "  {\n"
+      << "    struct __llvm_jmpbuf_list_t Entry;\n"
+      << "    Entry.next = __llvm_jmpbuf_list;\n"
+      << "    if (setjmp(Entry.buf)) {\n"
+      << "      __llvm_jmpbuf_list = Entry.next;\n";
+  printBranchToBlock(II.getParent(), II.getExceptionalDest(), 4);
+  Out << "    }\n"
+      << "    __llvm_jmpbuf_list = &Entry;\n"
+      << "    ";
+  visitCallSite(&II);
+  Out << ";\n"
+      << "    __llvm_jmpbuf_list = Entry.next;\n"
+      << "  }\n";
+  printBranchToBlock(II.getParent(), II.getNormalDest(), 0);
+  emittedInvoke = true;
+}
+
 
 static bool isGotoCodeNeccessary(BasicBlock *From, BasicBlock *To) {
   // If PHI nodes need copies, we need the copy code...
@@ -994,7 +1029,7 @@
 }
 
 void CWriter::printBranchToBlock(BasicBlock *CurBB, BasicBlock *Succ,
-                                           unsigned Indent) {
+                                 unsigned Indent) {
   for (BasicBlock::iterator I = Succ->begin();
        PHINode *PN = dyn_cast<PHINode>(I); ++I) {
     //  now we have to do the printing
@@ -1156,20 +1191,24 @@
         return;
       }
     }
+  visitCallSite(&I);
+}
 
-  const PointerType  *PTy   = cast<PointerType>(I.getCalledValue()->getType());
+void CWriter::visitCallSite(CallSite CS) {
+  const PointerType  *PTy   = cast<PointerType>(CS.getCalledValue()->getType());
   const FunctionType *FTy   = cast<FunctionType>(PTy->getElementType());
   const Type         *RetTy = FTy->getReturnType();
   
-  writeOperand(I.getOperand(0));
+  writeOperand(CS.getCalledValue());
   Out << "(";
 
-  if (I.getNumOperands() > 1) {
-    writeOperand(I.getOperand(1));
+  if (CS.arg_begin() != CS.arg_end()) {
+    CallSite::arg_iterator AI = CS.arg_begin(), AE = CS.arg_end();
+    writeOperand(*AI);
 
-    for (unsigned op = 2, Eop = I.getNumOperands(); op != Eop; ++op) {
+    for (++AI; AI != AE; ++AI) {
       Out << ", ";
-      writeOperand(I.getOperand(op));
+      writeOperand(*AI);
     }
   }
   Out << ")";




From lattner at cs.uiuc.edu  Sat Jun 28 14:30:02 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 14:30:02 2003
Subject: [llvm-commits] CVS: llvm/lib/CWriter/Writer.cpp
Message-ID: <200306281929.OAA30448@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/CWriter:

Writer.cpp updated: 1.103 -> 1.104

---
Log message:

jmp_buf is really a pointer type that is passed around...


---
Diffs of the changes:

Index: llvm/lib/CWriter/Writer.cpp
diff -u llvm/lib/CWriter/Writer.cpp:1.103 llvm/lib/CWriter/Writer.cpp:1.104
--- llvm/lib/CWriter/Writer.cpp:1.103	Sat Jun 28 12:53:05 2003
+++ llvm/lib/CWriter/Writer.cpp	Sat Jun 28 14:29:34 2003
@@ -1178,12 +1178,12 @@
         return;
         
       case LLVMIntrinsic::setjmp:
-        Out << "setjmp((jmp_buf)";
+        Out << "setjmp(*(jmp_buf*)";
         writeOperand(I.getOperand(1));
         Out << ")";
         return;
       case LLVMIntrinsic::longjmp:
-        Out << "longjmp((jmp_buf)";
+        Out << "longjmp(*(jmp_buf*)";
         writeOperand(I.getOperand(1));
         Out << ", ";
         writeOperand(I.getOperand(2));




From lattner at cs.uiuc.edu  Sat Jun 28 15:02:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 15:02:01 2003
Subject: [llvm-commits] CVS: llvm/lib/AsmParser/Lexer.l llvmAsmParser.y
Message-ID: <200306282001.PAA01743@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/AsmParser:

Lexer.l updated: 1.34 -> 1.35
llvmAsmParser.y updated: 1.114 -> 1.115

---
Log message:

Add support for a new zeroinitializer token which can be used to get rid of
huge arrays of zero initialized values


---
Diffs of the changes:

Index: llvm/lib/AsmParser/Lexer.l
diff -u llvm/lib/AsmParser/Lexer.l:1.34 llvm/lib/AsmParser/Lexer.l:1.35
--- llvm/lib/AsmParser/Lexer.l:1.34	Wed May  7 21:44:01 2003
+++ llvm/lib/AsmParser/Lexer.l	Sat Jun 28 15:01:34 2003
@@ -174,6 +174,7 @@
 uninitialized   { return EXTERNAL; }    /* Deprecated, turn into external */
 external        { return EXTERNAL; }
 implementation  { return IMPLEMENTATION; }
+zeroinitializer { return ZEROINITIALIZER; }
 \.\.\.          { return DOTDOTDOT; }
 null            { return NULL_TOK; }
 to              { return TO; }


Index: llvm/lib/AsmParser/llvmAsmParser.y
diff -u llvm/lib/AsmParser/llvmAsmParser.y:1.114 llvm/lib/AsmParser/llvmAsmParser.y:1.115
--- llvm/lib/AsmParser/llvmAsmParser.y:1.114	Wed Jun 18 14:22:32 2003
+++ llvm/lib/AsmParser/llvmAsmParser.y	Sat Jun 28 15:01:34 2003
@@ -702,7 +702,8 @@
 %type  <StrVal>  OptVAR_ID OptAssign FuncName
 
 
-%token IMPLEMENTATION TRUE FALSE BEGINTOK ENDTOK DECLARE GLOBAL CONSTANT
+%token IMPLEMENTATION ZEROINITIALIZER TRUE FALSE BEGINTOK ENDTOK
+%token  DECLARE GLOBAL CONSTANT
 %token TO EXCEPT DOTDOTDOT NULL_TOK CONST INTERNAL LINKONCE APPENDING
 %token OPAQUE NOT EXTERNAL TARGET ENDIAN POINTERSIZE LITTLE BIG
 
@@ -1038,6 +1039,10 @@
     if ($1->get() != $2->getType())
       ThrowException("Mismatched types for constant expression!");
     $$ = $2;
+    delete $1;
+  }
+  | Types ZEROINITIALIZER {
+    $$ = Constant::getNullValue($1->get());
     delete $1;
   };
 




From lattner at cs.uiuc.edu  Sat Jun 28 15:02:13 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 15:02:13 2003
Subject: [llvm-commits] CVS: llvm/test/Feature/globalvars.ll
Message-ID: <200306282001.PAA01730@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Feature:

globalvars.ll updated: 1.6 -> 1.7

---
Log message:

Add test for new zeroinitializer token


---
Diffs of the changes:

Index: llvm/test/Feature/globalvars.ll
diff -u llvm/test/Feature/globalvars.ll:1.6 llvm/test/Feature/globalvars.ll:1.7
--- llvm/test/Feature/globalvars.ll:1.6	Sun Oct  6 17:32:38 2002
+++ llvm/test/Feature/globalvars.ll	Sat Jun 28 15:01:03 2003
@@ -7,6 +7,9 @@
 
 %AString   = constant [4 x ubyte] c"test"
 
+%ZeroInit  = global { [100 x int ], [40 x float ] } { [100 x int] zeroinitializer,
+                                                      [40  x float] zeroinitializer }
+
 implementation
 
 int "foo"(int %blah)




From lattner at cs.uiuc.edu  Sat Jun 28 15:09:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 15:09:01 2003
Subject: [llvm-commits] CVS: llvm/lib/VMCore/AsmWriter.cpp
Message-ID: <200306282008.PAA02760@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/VMCore:

AsmWriter.cpp updated: 1.89 -> 1.90

---
Log message:

Avoid printing out huge structures or arrays if they are just filled with zeros


---
Diffs of the changes:

Index: llvm/lib/VMCore/AsmWriter.cpp
diff -u llvm/lib/VMCore/AsmWriter.cpp:1.89 llvm/lib/VMCore/AsmWriter.cpp:1.90
--- llvm/lib/VMCore/AsmWriter.cpp:1.89	Tue Jun 17 18:55:35 2003
+++ llvm/lib/VMCore/AsmWriter.cpp	Sat Jun 28 15:08:24 2003
@@ -253,6 +253,11 @@
     Out << "0x" << utohexstr(*(uint64_t*)Ptr);
 
   } else if (const ConstantArray *CA = dyn_cast<ConstantArray>(CV)) {
+    if (CA->getNumOperands() > 5 && CA->isNullValue()) {
+      Out << "zeroinitializer";
+      return;
+    }
+
     // As a special case, print the array as a string if it is an array of
     // ubytes or an array of sbytes with positive values.
     // 
@@ -300,6 +305,11 @@
       Out << " ]";
     }
   } else if (const ConstantStruct *CS = dyn_cast<ConstantStruct>(CV)) {
+    if (CS->getNumOperands() > 5 && CS->isNullValue()) {
+      Out << "zeroinitializer";
+      return;
+    }
+
     Out << "{";
     if (CS->getNumOperands()) {
       Out << " ";




From lattner at cs.uiuc.edu  Sat Jun 28 16:17:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 16:17:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Analysis/DSGraph/Makefile
Message-ID: <200306282116.QAA03302@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Analysis/DSGraph:

Makefile added (r1.1)

---
Log message:

New makefile for Dsgraph tests


---
Diffs of the changes:

Index: llvm/test/Regression/Analysis/DSGraph/Makefile
diff -c /dev/null llvm/test/Regression/Analysis/DSGraph/Makefile:1.1
*** /dev/null	Sat Jun 28 16:16:48 2003
--- llvm/test/Regression/Analysis/DSGraph/Makefile	Sat Jun 28 16:16:38 2003
***************
*** 0 ****
--- 1,10 ----
+ 
+ LEVEL = ../../../..
+ include $(LEVEL)/test/Makefile.tests
+ 
+ TESTS := $(wildcard *.ll)
+ 
+ all:: $(addprefix Output/, $(TESTS:%.ll=%.ll.out))
+ 
+ Output/%.ll.out: %.ll Output/.dir $(LOPT) $(LANALYZE)
+ 	-$(TESTRUNR) $<




From lattner at cs.uiuc.edu  Sat Jun 28 16:17:13 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 16:17:13 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Analysis/DSGraph/
Message-ID: <200306282116.QAA03280@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Analysis/DSGraph:


---
Log message:

Directory /home/vadve/vadve/Research/DynOpt/CVSRepository/llvm/test/Regression/Analysis/DSGraph added to the repository


---
Diffs of the changes:



From lattner at cs.uiuc.edu  Sat Jun 28 16:18:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 16:18:00 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Analysis/Makefile
Message-ID: <200306282117.QAA03314@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Analysis:

Makefile updated: 1.2 -> 1.3

---
Log message:

Add DSGraph to the tests


---
Diffs of the changes:

Index: llvm/test/Regression/Analysis/Makefile
diff -u llvm/test/Regression/Analysis/Makefile:1.2 llvm/test/Regression/Analysis/Makefile:1.3
--- llvm/test/Regression/Analysis/Makefile:1.2	Thu May 15 13:02:54 2003
+++ llvm/test/Regression/Analysis/Makefile	Sat Jun 28 16:16:27 2003
@@ -1,5 +1,7 @@
 #                   test/Regression/Analysis/Makefile
 
 LEVEL = ../../..
-DIRS  = Dominators LoopInfo
+DIRS  = Dominators \
+	DSGraph \
+        LoopInfo
 include $(LEVEL)/Makefile.common




From lattner at cs.uiuc.edu  Sat Jun 28 16:55:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 16:55:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Analysis/DSGraph/buglobals.ll
Message-ID: <200306282154.QAA04037@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Analysis/DSGraph:

buglobals.ll added (r1.1)

---
Log message:

New testcase


---
Diffs of the changes:

Index: llvm/test/Regression/Analysis/DSGraph/buglobals.ll
diff -c /dev/null llvm/test/Regression/Analysis/DSGraph/buglobals.ll:1.1
*** /dev/null	Sat Jun 28 16:54:35 2003
--- llvm/test/Regression/Analysis/DSGraph/buglobals.ll	Sat Jun 28 16:54:24 2003
***************
*** 0 ****
--- 1,21 ----
+ ; This tests to make sure that G ends up in the globals graph of the BU pass.
+ ; If it is not, then %G will get converted to a 'constant' from a 'global'
+ ;
+ ; RUN: as < %s | opt -ds-opt -globaldce | dis | grep %G
+ 
+ 
+ %G = internal global int 0		; <int*> [#uses=2]
+ 
+ implementation   ; Functions:
+ 
+ internal void %foo() {
+ 	%tmp.0 = load int* %G		; <int> [#uses=1]
+ 	%tmp.1 = add int %tmp.0, 1		; <int> [#uses=1]
+ 	store int %tmp.1, int* %G
+ 	ret void
+ }
+ 
+ int %main() {
+ 	call void %foo( )
+ 	ret int 0
+ }




From lattner at cs.uiuc.edu  Sat Jun 28 16:56:02 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 16:56:02 2003
Subject: [llvm-commits] CVS: llvm/lib/Analysis/DataStructure/DataStructureOpt.cpp
Message-ID: <200306282155.QAA04056@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis/DataStructure:

DataStructureOpt.cpp added (r1.1)

---
Log message:

New pass to perform DSA based optimizations.  Initially we just support turning
globals into constants if we can prove it's safe


---
Diffs of the changes:

Index: llvm/lib/Analysis/DataStructure/DataStructureOpt.cpp
diff -c /dev/null llvm/lib/Analysis/DataStructure/DataStructureOpt.cpp:1.1
*** /dev/null	Sat Jun 28 16:55:05 2003
--- llvm/lib/Analysis/DataStructure/DataStructureOpt.cpp	Sat Jun 28 16:54:55 2003
***************
*** 0 ****
--- 1,76 ----
+ //===- DataStructureOpt.cpp - Data Structure Analysis Based Optimizations -===//
+ //
+ // This pass uses DSA to a series of simple optimizations, like marking
+ // unwritten global variables 'constant'.
+ //
+ //===----------------------------------------------------------------------===//
+ 
+ #include "llvm/Analysis/DataStructure.h"
+ #include "llvm/Analysis/DSGraph.h"
+ #include "llvm/Module.h"
+ #include "llvm/Constant.h"
+ #include "Support/Statistic.h"
+ 
+ namespace {
+   Statistic<>
+   NumGlobalsConstanted("ds-opt", "Number of globals marked constant");
+ 
+   class DSOpt : public Pass {
+     TDDataStructures *TD;
+   public:
+     bool run(Module &M) {
+       TD = &getAnalysis<TDDataStructures>();
+       bool Changed = OptimizeGlobals(M);
+       return Changed;
+     }
+ 
+     virtual void getAnalysisUsage(AnalysisUsage &AU) const {
+       AU.addRequired<TDDataStructures>();      // Uses TD Datastructures
+       AU.addPreserved<LocalDataStructures>();  // Preserves local...
+       AU.addPreserved<TDDataStructures>();     // Preserves bu...
+       AU.addPreserved<BUDataStructures>();     // Preserves td...
+     }
+ 
+   private:
+     bool OptimizeGlobals(Module &M);
+   };
+ 
+   RegisterOpt<DSOpt> X("ds-opt", "DSA-based simple optimizations");
+ }
+ 
+ 
+ /// OptimizeGlobals - This method uses information taken from DSA to optimize
+ /// global variables.
+ ///
+ bool DSOpt::OptimizeGlobals(Module &M) {
+   DSGraph &GG = TD->getGlobalsGraph();
+   const DSGraph::ScalarMapTy &SM = GG.getScalarMap();
+   bool Changed = false;
+ 
+   for (Module::giterator I = M.gbegin(), E = M.gend(); I != E; ++I)
+     if (!I->isExternal()) { // Loop over all of the non-external globals...
+       // Look up the node corresponding to this global, if it exists.
+       DSNode *GNode = 0;
+       DSGraph::ScalarMapTy::const_iterator SMI = SM.find(I);
+       if (SMI != SM.end()) GNode = SMI->second.getNode();
+     
+       if (GNode == 0 && I->hasInternalLinkage()) {
+         // If there is no entry in the scalar map for this global, it was never
+         // referenced in the program.  If it has internal linkage, that means we
+         // can delete it.  We don't ACTUALLY want to delete the global, just
+         // remove anything that references the global: later passes will take
+         // care of nuking it.
+         I->replaceAllUsesWith(Constant::getNullValue((Type*)I->getType()));
+       } else if (GNode && GNode->isComplete()) {
+         // We expect that there will almost always be a node for this global.
+         // If there is, and the node doesn't have the M bit set, we can set the
+         // 'constant' bit on the global.
+         if (!GNode->isModified() && !I->isConstant()) {
+           I->setConstant(true);
+           ++NumGlobalsConstanted;
+           Changed = true;
+         }
+       }
+     }
+   return Changed;
+ }




From lattner at cs.uiuc.edu  Sat Jun 28 16:57:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 16:57:00 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Analysis/DSNode.h
Message-ID: <200306282156.QAA04150@apoc.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Analysis:

DSNode.h updated: 1.27 -> 1.28

---
Log message:

Add new 'isComplete' method


---
Diffs of the changes:

Index: llvm/include/llvm/Analysis/DSNode.h
diff -u llvm/include/llvm/Analysis/DSNode.h:1.27 llvm/include/llvm/Analysis/DSNode.h:1.28
--- llvm/include/llvm/Analysis/DSNode.h:1.27	Sat Jun 21 22:04:58 2003
+++ llvm/include/llvm/Analysis/DSNode.h	Sat Jun 28 16:56:42 2003
@@ -248,6 +248,7 @@
   bool isRead() const       { return NodeType & Read; }
 
   bool isIncomplete() const { return NodeType & Incomplete; }
+  bool isComplete() const   { return !isIncomplete(); }
   bool isDeadNode() const   { return NodeType & DEAD; }
 
   DSNode *setAllocaNodeMarker()  { NodeType |= AllocaNode;  return this; }




From lattner at cs.uiuc.edu  Sat Jun 28 16:57:12 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 16:57:12 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Analysis/DSGraph/constantize.ll
Message-ID: <200306282156.QAA04138@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Analysis/DSGraph:

constantize.ll added (r1.1)

---
Log message:

New testcase


---
Diffs of the changes:

Index: llvm/test/Regression/Analysis/DSGraph/constantize.ll
diff -c /dev/null llvm/test/Regression/Analysis/DSGraph/constantize.ll:1.1
*** /dev/null	Sat Jun 28 16:56:32 2003
--- llvm/test/Regression/Analysis/DSGraph/constantize.ll	Sat Jun 28 16:56:22 2003
***************
*** 0 ****
--- 1,13 ----
+ ; Make sure that the ds-opt pass is constantizing globals
+ ;
+ ; RUN: as < %s | opt -ds-opt | dis | grep %G | grep constant
+ 
+ 
+ %G = internal global int 0		; <int*> [#uses=2]
+ 
+ implementation   ; Functions:
+ 
+ int %main() {
+ 	%A = load int* %G
+ 	ret int %A
+ }




From lattner at cs.uiuc.edu  Sat Jun 28 16:58:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 16:58:01 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Analysis/DSGraph.h
Message-ID: <200306282157.QAA04164@apoc.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Analysis:

DSGraph.h updated: 1.50 -> 1.51

---
Log message:

Add new DSGraph::ScalarMapTy typedef to avoid pulling representation issues
into callers of getScalarMap


---
Diffs of the changes:

Index: llvm/include/llvm/Analysis/DSGraph.h
diff -u llvm/include/llvm/Analysis/DSGraph.h:1.50 llvm/include/llvm/Analysis/DSGraph.h:1.51
--- llvm/include/llvm/Analysis/DSGraph.h:1.50	Sat Jun 21 22:05:45 2003
+++ llvm/include/llvm/Analysis/DSGraph.h	Sat Jun 28 16:57:13 2003
@@ -12,14 +12,18 @@
 //===----------------------------------------------------------------------===//
 /// DSGraph - The graph that represents a function.
 ///
-class DSGraph {
+struct DSGraph {
+  // Public data-type declarations...
+  typedef hash_map<Value*, DSNodeHandle> ScalarMapTy;
+
+private:
   Function *Func;          // Func - The LLVM function this graph corresponds to
   DSGraph *GlobalsGraph;   // Pointer to the common graph of global objects
   bool PrintAuxCalls;      // Should this graph print the Aux calls vector?
 
   DSNodeHandle RetNode;    // The node that gets returned...
   std::vector<DSNode*> Nodes;
-  hash_map<Value*, DSNodeHandle> ScalarMap;
+  ScalarMapTy ScalarMap;
 
   // FunctionCalls - This vector maintains a single entry for each call
   // instruction in the current graph.  The first entry in the vector is the
@@ -80,8 +84,8 @@
   /// getScalarMap - Get a map that describes what the nodes the scalars in this
   /// function point to...
   ///
-  hash_map<Value*, DSNodeHandle> &getScalarMap() { return ScalarMap; }
-  const hash_map<Value*, DSNodeHandle> &getScalarMap() const {return ScalarMap;}
+  ScalarMapTy &getScalarMap() { return ScalarMap; }
+  const ScalarMapTy &getScalarMap() const {return ScalarMap;}
 
   /// getFunctionCalls - Return the list of call sites in the original local
   /// graph...
@@ -106,7 +110,7 @@
   DSNodeHandle &getNodeForValue(Value *V) { return ScalarMap[V]; }
 
   const DSNodeHandle &getNodeForValue(Value *V) const {
-    hash_map<Value*, DSNodeHandle>::const_iterator I = ScalarMap.find(V);
+    ScalarMapTy::const_iterator I = ScalarMap.find(V);
     assert(I != ScalarMap.end() &&
            "Use non-const lookup function if node may not be in the map");
     return I->second;
@@ -181,8 +185,7 @@
   // 'StripAllocaBit', Alloca markers are removed from the graph as the graph is
   // being cloned.
   //
-  DSNodeHandle cloneInto(const DSGraph &G,
-                         hash_map<Value*, DSNodeHandle> &OldValMap,
+  DSNodeHandle cloneInto(const DSGraph &G, ScalarMapTy &OldValMap,
                          hash_map<const DSNode*, DSNodeHandle> &OldNodeMap,
                          unsigned CloneFlags = 0);
 




From lattner at cs.uiuc.edu  Sat Jun 28 16:59:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 16:59:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Analysis/DataStructure/DataStructure.cpp DataStructureAA.cpp
Message-ID: <200306282158.QAA04181@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis/DataStructure:

DataStructure.cpp updated: 1.101 -> 1.102
DataStructureAA.cpp updated: 1.5 -> 1.6

---
Log message:

Avoid double negatives


---
Diffs of the changes:

Index: llvm/lib/Analysis/DataStructure/DataStructure.cpp
diff -u llvm/lib/Analysis/DataStructure/DataStructure.cpp:1.101 llvm/lib/Analysis/DataStructure/DataStructure.cpp:1.102
--- llvm/lib/Analysis/DataStructure/DataStructure.cpp:1.101	Sat Jun 21 22:03:52 2003
+++ llvm/lib/Analysis/DataStructure/DataStructure.cpp	Sat Jun 28 16:58:28 2003
@@ -751,7 +751,7 @@
 //
 static void markIncompleteNode(DSNode *N) {
   // Stop recursion if no node, or if node already marked...
-  if (N == 0 || (N->isIncomplete())) return;
+  if (N == 0 || N->isIncomplete()) return;
 
   // Actually mark the node
   N->setIncompleteMarker();
@@ -917,7 +917,7 @@
 
   for (unsigned i = 0; i != Nodes.size(); ++i) {
     DSNode *Node = Nodes[i];
-    if (!Node->isIncomplete() && !Node->isModified() && !Node->isRead()) {
+    if (Node->isComplete() && !Node->isModified() && !Node->isRead()) {
       // This is a useless node if it has no mod/ref info (checked above),
       // outgoing edges (which it cannot, as it is not modified in this
       // context), and it has no incoming edges.  If it is a global node it may


Index: llvm/lib/Analysis/DataStructure/DataStructureAA.cpp
diff -u llvm/lib/Analysis/DataStructure/DataStructureAA.cpp:1.5 llvm/lib/Analysis/DataStructure/DataStructureAA.cpp:1.6
--- llvm/lib/Analysis/DataStructure/DataStructureAA.cpp:1.5	Sat Jun 21 22:03:52 2003
+++ llvm/lib/Analysis/DataStructure/DataStructureAA.cpp	Sat Jun 28 16:58:28 2003
@@ -90,7 +90,7 @@
         unsigned O1 = I->second.getOffset(), O2 = J->second.getOffset();
         
         // We can only make a judgement of one of the nodes is complete...
-        if (!N1->isIncomplete() || !N2->isIncomplete()) {
+        if (N1->isComplete() || N2->isComplete()) {
           if (N1 != N2)
             return NoAlias;   // Completely different nodes.
 




From lattner at cs.uiuc.edu  Sat Jun 28 17:12:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 17:12:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Analysis/DataStructure/DataStructureOpt.cpp
Message-ID: <200306282211.RAA04640@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis/DataStructure:

DataStructureOpt.cpp updated: 1.1 -> 1.2

---
Log message:

Drop references to globals who do exist in the globals graph, but are never
read or written to.  Keep track of how many times this happens.  This should
be good for deleting things like references to type information in C++ classes



---
Diffs of the changes:

Index: llvm/lib/Analysis/DataStructure/DataStructureOpt.cpp
diff -u llvm/lib/Analysis/DataStructure/DataStructureOpt.cpp:1.1 llvm/lib/Analysis/DataStructure/DataStructureOpt.cpp:1.2
--- llvm/lib/Analysis/DataStructure/DataStructureOpt.cpp:1.1	Sat Jun 28 16:54:55 2003
+++ llvm/lib/Analysis/DataStructure/DataStructureOpt.cpp	Sat Jun 28 17:10:58 2003
@@ -14,6 +14,8 @@
 namespace {
   Statistic<>
   NumGlobalsConstanted("ds-opt", "Number of globals marked constant");
+  Statistic<>
+  NumGlobalsIsolated("ds-opt", "Number of globals with references dropped");
 
   class DSOpt : public Pass {
     TDDataStructures *TD;
@@ -60,8 +62,21 @@
         // can delete it.  We don't ACTUALLY want to delete the global, just
         // remove anything that references the global: later passes will take
         // care of nuking it.
-        I->replaceAllUsesWith(Constant::getNullValue((Type*)I->getType()));
+        if (!I->use_empty()) {
+          I->replaceAllUsesWith(Constant::getNullValue((Type*)I->getType()));
+          ++NumGlobalsIsolated;
+        }
       } else if (GNode && GNode->isComplete()) {
+
+        // If the node has not been read or written, and it is not externally
+        // visible, kill any references to it so it can be DCE'd.
+        if (!GNode->isModified() && !GNode->isRead() &&I->hasInternalLinkage()){
+          if (!I->use_empty()) {
+            I->replaceAllUsesWith(Constant::getNullValue((Type*)I->getType()));
+            ++NumGlobalsIsolated;
+          }
+        }
+
         // We expect that there will almost always be a node for this global.
         // If there is, and the node doesn't have the M bit set, we can set the
         // 'constant' bit on the global.




From lattner at cs.uiuc.edu  Sat Jun 28 17:16:02 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 17:16:02 2003
Subject: [llvm-commits] CVS: llvm/lib/Analysis/DataStructure/BottomUpClosure.cpp TopDownClosure.cpp
Message-ID: <200306282215.RAA05134@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis/DataStructure:

BottomUpClosure.cpp updated: 1.53 -> 1.54
TopDownClosure.cpp updated: 1.41 -> 1.42

---
Log message:

Propagate globals graph from the local to bu to td globals graphs.  This
fixes bug: DSGraph/buglobals.ll


---
Diffs of the changes:

Index: llvm/lib/Analysis/DataStructure/BottomUpClosure.cpp
diff -u llvm/lib/Analysis/DataStructure/BottomUpClosure.cpp:1.53 llvm/lib/Analysis/DataStructure/BottomUpClosure.cpp:1.54
--- llvm/lib/Analysis/DataStructure/BottomUpClosure.cpp:1.53	Thu Jun 19 16:15:10 2003
+++ llvm/lib/Analysis/DataStructure/BottomUpClosure.cpp	Sat Jun 28 17:14:55 2003
@@ -17,7 +17,7 @@
   Statistic<> MaxSCC("budatastructure", "Maximum SCC Size in Call Graph");
   
   RegisterAnalysis<BUDataStructures>
-  X("budatastructure", "Bottom-up Data Structure Analysis Closure");
+  X("budatastructure", "Bottom-up Data Structure Analysis");
 }
 
 using namespace DS;
@@ -120,7 +120,8 @@
 // program.
 //
 bool BUDataStructures::run(Module &M) {
-  GlobalsGraph = new DSGraph();
+  LocalDataStructures &LocalDSA = getAnalysis<LocalDataStructures>();
+  GlobalsGraph = new DSGraph(LocalDSA.getGlobalsGraph());
   GlobalsGraph->setPrintAuxCalls();
 
   Function *MainFunc = M.getMainFunction();


Index: llvm/lib/Analysis/DataStructure/TopDownClosure.cpp
diff -u llvm/lib/Analysis/DataStructure/TopDownClosure.cpp:1.41 llvm/lib/Analysis/DataStructure/TopDownClosure.cpp:1.42
--- llvm/lib/Analysis/DataStructure/TopDownClosure.cpp:1.41	Sat Jun 21 22:03:52 2003
+++ llvm/lib/Analysis/DataStructure/TopDownClosure.cpp	Sat Jun 28 17:14:55 2003
@@ -15,7 +15,7 @@
 
 namespace {
   RegisterAnalysis<TDDataStructures>   // Register the pass
-  Y("tddatastructure", "Top-down Data Structure Analysis Closure");
+  Y("tddatastructure", "Top-down Data Structure Analysis");
 }
 
 // run - Calculate the top down data structure graphs for each function in the
@@ -23,7 +23,7 @@
 //
 bool TDDataStructures::run(Module &M) {
   BUDataStructures &BU = getAnalysis<BUDataStructures>();
-  GlobalsGraph = new DSGraph();
+  GlobalsGraph = new DSGraph(BU.getGlobalsGraph());
 
   // Calculate top-down from main...
   if (Function *F = M.getMainFunction())




From lattner at cs.uiuc.edu  Sat Jun 28 17:18:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 17:18:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Analysis/DSGraph/Makefile
Message-ID: <200306282217.RAA05157@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Analysis/DSGraph:

Makefile updated: 1.1 -> 1.2

---
Log message:

Improve the make rule


---
Diffs of the changes:

Index: llvm/test/Regression/Analysis/DSGraph/Makefile
diff -u llvm/test/Regression/Analysis/DSGraph/Makefile:1.1 llvm/test/Regression/Analysis/DSGraph/Makefile:1.2
--- llvm/test/Regression/Analysis/DSGraph/Makefile:1.1	Sat Jun 28 16:16:38 2003
+++ llvm/test/Regression/Analysis/DSGraph/Makefile	Sat Jun 28 17:16:59 2003
@@ -6,5 +6,6 @@
 
 all:: $(addprefix Output/, $(TESTS:%.ll=%.ll.out))
 
-Output/%.ll.out: %.ll Output/.dir $(LOPT) $(LANALYZE)
+$(addprefix Output/, $(TESTS:%.ll=%.ll.out)):
+Output/%.ll.out: %.ll Output/.dir $(LOPT)
 	-$(TESTRUNR) $<




From lattner at cs.uiuc.edu  Sat Jun 28 17:20:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 17:20:01 2003
Subject: [llvm-commits] CVS: llvm/test/Scripts/
Message-ID: <200306282219.RAA05355@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Scripts:


---
Log message:

Directory /home/vadve/vadve/Research/DynOpt/CVSRepository/llvm/test/Scripts added to the repository


---
Diffs of the changes:



From lattner at cs.uiuc.edu  Sat Jun 28 17:21:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 17:21:00 2003
Subject: [llvm-commits] CVS: llvm/test/Scripts/README.txt
Message-ID: <200306282220.RAA05370@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Scripts:

README.txt added (r1.1)

---
Log message:

Describe directory


---
Diffs of the changes:

Index: llvm/test/Scripts/README.txt
diff -c /dev/null llvm/test/Scripts/README.txt:1.1
*** /dev/null	Sat Jun 28 17:20:29 2003
--- llvm/test/Scripts/README.txt	Sat Jun 28 17:20:19 2003
***************
*** 0 ****
--- 1,2 ----
+ This directory contains scripts which are used by the TestRunner style
+ tests, which allows them to be simpler and more direct.




From lattner at cs.uiuc.edu  Sat Jun 28 17:22:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 17:22:01 2003
Subject: [llvm-commits] CVS: llvm/test/Makefile.tests
Message-ID: <200306282221.RAA05383@apoc.cs.uiuc.edu>


Changes in directory llvm/test:

Makefile.tests updated: 1.66 -> 1.67

---
Log message:

Add the test/Scripts directory to the path used by TestRunner


---
Diffs of the changes:

Index: llvm/test/Makefile.tests
diff -u llvm/test/Makefile.tests:1.66 llvm/test/Makefile.tests:1.67
--- llvm/test/Makefile.tests:1.66	Thu Jun 26 00:28:00 2003
+++ llvm/test/Makefile.tests	Sat Jun 28 17:21:39 2003
@@ -54,7 +54,8 @@
 LCXXFLAGS += -O2 -Wall
 LLCFLAGS =
 FAILURE  = $(LEVEL)/test/Failure.sh
-TESTRUNR = PATH=$(LLVMTOOLCURRENT):$(PATH) $(LEVEL)/test/TestRunner.sh
+TESTRUNR = PATH=$(LLVMTOOLCURRENT):$(LEVEL)/test/Scripts:$(PATH) \
+               $(LEVEL)/test/TestRunner.sh
 
 # Native Tool Definitions
 NATGCC  = $(CC)




From lattner at cs.uiuc.edu  Sat Jun 28 17:29:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 17:29:01 2003
Subject: [llvm-commits] CVS: llvm/test/Scripts/grep-not
Message-ID: <200306282228.RAA05714@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Scripts:

grep-not added (r1.1)

---
Log message:

Initial checkin of grep-not script


---
Diffs of the changes:

Index: llvm/test/Scripts/grep-not
diff -c /dev/null llvm/test/Scripts/grep-not:1.1
*** /dev/null	Sat Jun 28 17:28:41 2003
--- llvm/test/Scripts/grep-not	Sat Jun 28 17:28:31 2003
***************
*** 0 ****
--- 1,19 ----
+ #!/bin/sh
+ #
+ # Program: grep-not
+ #
+ # Synopsis: Search the input to the program for a regular expression, working
+ #           just like grep.  The only difference is that we return success if
+ #           the pattern was not found, and failure if it was.  This is useful
+ #           for TestRunner tests which want to make sure something DOES NOT
+ #           occur in the output of a command.
+ #
+ # Syntax:   The same as grep
+ 
+ if grep $@
+ then exit 1
+ else exit 0
+ fi
+     
+ 
+ 




From lattner at cs.uiuc.edu  Sat Jun 28 17:32:02 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 17:32:02 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/InstCombine/mul.ll
Message-ID: <200306282231.RAA06402@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/InstCombine:

mul.ll updated: 1.6 -> 1.7

---
Log message:

Fix testcase


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/InstCombine/mul.ll
diff -u llvm/test/Regression/Transforms/InstCombine/mul.ll:1.6 llvm/test/Regression/Transforms/InstCombine/mul.ll:1.7
--- llvm/test/Regression/Transforms/InstCombine/mul.ll:1.6	Wed Jun 25 12:10:34 2003
+++ llvm/test/Regression/Transforms/InstCombine/mul.ll	Sat Jun 28 17:31:37 2003
@@ -1,10 +1,7 @@
 ; This test makes sure that mul instructions are properly eliminated.
 ;
 
-; RUN: if as < %s | opt -instcombine | dis | grep mul
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -instcombine | dis | grep-not mul
 
 implementation
 
@@ -35,7 +32,7 @@
 
 ubyte %test6(ubyte %A) {
 	%B = mul ubyte %A, 8
-	%C = mul ubyte %B, 13
+	%C = mul ubyte %B, 8
 	ret ubyte %C
 }
 




From lattner at cs.uiuc.edu  Sat Jun 28 17:36:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 17:36:01 2003
Subject: [llvm-commits] CVS: llvm/test/Makefile.tests
Message-ID: <200306282235.RAA07084@apoc.cs.uiuc.edu>


Changes in directory llvm/test:

Makefile.tests updated: 1.67 -> 1.68

---
Log message:

Make the TestRunner tests not print out horrible things when they run


---
Diffs of the changes:

Index: llvm/test/Makefile.tests
diff -u llvm/test/Makefile.tests:1.67 llvm/test/Makefile.tests:1.68
--- llvm/test/Makefile.tests:1.67	Sat Jun 28 17:21:39 2003
+++ llvm/test/Makefile.tests	Sat Jun 28 17:35:46 2003
@@ -54,8 +54,9 @@
 LCXXFLAGS += -O2 -Wall
 LLCFLAGS =
 FAILURE  = $(LEVEL)/test/Failure.sh
-TESTRUNR = PATH=$(LLVMTOOLCURRENT):$(LEVEL)/test/Scripts:$(PATH) \
-               $(LEVEL)/test/TestRunner.sh
+TESTRUNR = @echo Running test: $<; \
+             PATH=$(LLVMTOOLCURRENT):$(LEVEL)/test/Scripts:$(PATH) \
+                  $(LEVEL)/test/TestRunner.sh
 
 # Native Tool Definitions
 NATGCC  = $(CC)




From lattner at cs.uiuc.edu  Sat Jun 28 17:45:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 17:45:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/InstCombine/2002-05-14-SubFailure.ll 2002-12-05-MissedConstProp.ll add.ll and.ll call.ll cast-set.ll cast.ll div.ll getelementptr.ll load.ll not.ll or.ll phi.ll pow.ll rem.ll set.ll setcc-strength-reduce.ll shift.ll sub.ll
Message-ID: <200306282244.RAA08731@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/InstCombine:

2002-05-14-SubFailure.ll updated: 1.2 -> 1.3
2002-12-05-MissedConstProp.ll updated: 1.1 -> 1.2
add.ll updated: 1.8 -> 1.9
and.ll updated: 1.7 -> 1.8
call.ll updated: 1.1 -> 1.2
cast-set.ll updated: 1.5 -> 1.6
cast.ll updated: 1.9 -> 1.10
div.ll updated: 1.4 -> 1.5
getelementptr.ll updated: 1.9 -> 1.10
load.ll updated: 1.1 -> 1.2
not.ll updated: 1.7 -> 1.8
or.ll updated: 1.13 -> 1.14
phi.ll updated: 1.4 -> 1.5
pow.ll updated: 1.2 -> 1.3
rem.ll updated: 1.3 -> 1.4
set.ll updated: 1.8 -> 1.9
setcc-strength-reduce.ll updated: 1.2 -> 1.3
shift.ll updated: 1.7 -> 1.8
sub.ll updated: 1.9 -> 1.10

---
Log message:

Convert scripts from using explicit control flow to use the new grep-not script


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/InstCombine/2002-05-14-SubFailure.ll
diff -u llvm/test/Regression/Transforms/InstCombine/2002-05-14-SubFailure.ll:1.2 llvm/test/Regression/Transforms/InstCombine/2002-05-14-SubFailure.ll:1.3
--- llvm/test/Regression/Transforms/InstCombine/2002-05-14-SubFailure.ll:1.2	Wed Aug 14 12:38:15 2002
+++ llvm/test/Regression/Transforms/InstCombine/2002-05-14-SubFailure.ll	Sat Jun 28 17:44:42 2003
@@ -1,10 +1,7 @@
 ; Instcombine was missing a test that caused it to make illegal transformations
 ; sometimes.  In this case, it transforms the sub into an add:
 ; RUN: echo foo
-; RUN: if as < %s | opt -instcombine | dis | grep add
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -instcombine | dis | grep-not add
 ;
 
 


Index: llvm/test/Regression/Transforms/InstCombine/2002-12-05-MissedConstProp.ll
diff -u llvm/test/Regression/Transforms/InstCombine/2002-12-05-MissedConstProp.ll:1.1 llvm/test/Regression/Transforms/InstCombine/2002-12-05-MissedConstProp.ll:1.2
--- llvm/test/Regression/Transforms/InstCombine/2002-12-05-MissedConstProp.ll:1.1	Thu Dec  5 16:25:47 2002
+++ llvm/test/Regression/Transforms/InstCombine/2002-12-05-MissedConstProp.ll	Sat Jun 28 17:44:42 2003
@@ -1,7 +1,4 @@
-; RUN: if as < %s | opt -instcombine | dis | grep add
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -instcombine | dis | grep-not add
 
 int %test(int %A) {
   %A.neg = sub int 0, %A


Index: llvm/test/Regression/Transforms/InstCombine/add.ll
diff -u llvm/test/Regression/Transforms/InstCombine/add.ll:1.8 llvm/test/Regression/Transforms/InstCombine/add.ll:1.9
--- llvm/test/Regression/Transforms/InstCombine/add.ll:1.8	Mon Mar 10 18:10:59 2003
+++ llvm/test/Regression/Transforms/InstCombine/add.ll	Sat Jun 28 17:44:42 2003
@@ -3,10 +3,7 @@
 ; This also tests that a subtract with a constant is properly converted
 ; to a add w/negative constant
 
-; RUN: if as < %s | opt -instcombine -die | dis | grep add
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -instcombine -die | dis | grep-not add
 
 implementation
 


Index: llvm/test/Regression/Transforms/InstCombine/and.ll
diff -u llvm/test/Regression/Transforms/InstCombine/and.ll:1.7 llvm/test/Regression/Transforms/InstCombine/and.ll:1.8
--- llvm/test/Regression/Transforms/InstCombine/and.ll:1.7	Mon Mar 10 17:52:54 2003
+++ llvm/test/Regression/Transforms/InstCombine/and.ll	Sat Jun 28 17:44:42 2003
@@ -1,10 +1,7 @@
 ; This test makes sure that these instructions are properly eliminated.
 ;
 
-; RUN: if as < %s | opt -instcombine | dis | grep and
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -instcombine | dis | grep-not and
 
 implementation
 


Index: llvm/test/Regression/Transforms/InstCombine/call.ll
diff -u llvm/test/Regression/Transforms/InstCombine/call.ll:1.1 llvm/test/Regression/Transforms/InstCombine/call.ll:1.2
--- llvm/test/Regression/Transforms/InstCombine/call.ll:1.1	Wed Jun 18 17:48:11 2003
+++ llvm/test/Regression/Transforms/InstCombine/call.ll	Sat Jun 28 17:44:42 2003
@@ -1,7 +1,4 @@
-; RUN: if as < %s | opt -instcombine -die | dis | grep call | grep cast
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -instcombine -die | dis | grep call | grep-not cast
 
 implementation
 


Index: llvm/test/Regression/Transforms/InstCombine/cast-set.ll
diff -u llvm/test/Regression/Transforms/InstCombine/cast-set.ll:1.5 llvm/test/Regression/Transforms/InstCombine/cast-set.ll:1.6
--- llvm/test/Regression/Transforms/InstCombine/cast-set.ll:1.5	Wed Mar  5 16:50:55 2003
+++ llvm/test/Regression/Transforms/InstCombine/cast-set.ll	Sat Jun 28 17:44:42 2003
@@ -3,10 +3,7 @@
 ; slow.  Might it be better to make there be an instcombine prepass before
 ; level raise that takes care of the obvious stuff?
 
-; RUN: if as < %s | opt -instcombine | dis | grep cast
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -instcombine | dis | grep-not cast
 
 bool %test1(int %X) {
 	%A = cast int %X to uint


Index: llvm/test/Regression/Transforms/InstCombine/cast.ll
diff -u llvm/test/Regression/Transforms/InstCombine/cast.ll:1.9 llvm/test/Regression/Transforms/InstCombine/cast.ll:1.10
--- llvm/test/Regression/Transforms/InstCombine/cast.ll:1.9	Mon Jun 23 16:52:59 2003
+++ llvm/test/Regression/Transforms/InstCombine/cast.ll	Sat Jun 28 17:44:42 2003
@@ -1,9 +1,6 @@
 ; Tests to make sure elimination of casts is working correctly
 
-; RUN: if as < %s | opt -instcombine -die | dis | grep '%c' | grep cast
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -instcombine -die | dis | grep '%c' | grep-not cast
 
 implementation
 


Index: llvm/test/Regression/Transforms/InstCombine/div.ll
diff -u llvm/test/Regression/Transforms/InstCombine/div.ll:1.4 llvm/test/Regression/Transforms/InstCombine/div.ll:1.5
--- llvm/test/Regression/Transforms/InstCombine/div.ll:1.4	Tue Feb 18 13:28:47 2003
+++ llvm/test/Regression/Transforms/InstCombine/div.ll	Sat Jun 28 17:44:42 2003
@@ -1,10 +1,7 @@
 ; This test makes sure that div instructions are properly eliminated.
 ;
 
-; RUN: if as < %s | opt -instcombine | dis | grep div
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -instcombine | dis | grep-not div
 
 implementation
 


Index: llvm/test/Regression/Transforms/InstCombine/getelementptr.ll
diff -u llvm/test/Regression/Transforms/InstCombine/getelementptr.ll:1.9 llvm/test/Regression/Transforms/InstCombine/getelementptr.ll:1.10
--- llvm/test/Regression/Transforms/InstCombine/getelementptr.ll:1.9	Sat Jun 21 17:59:39 2003
+++ llvm/test/Regression/Transforms/InstCombine/getelementptr.ll	Sat Jun 28 17:44:42 2003
@@ -1,9 +1,6 @@
 ; The %A getelementptr instruction should be eliminated here
 
-; RUN: if as < %s | opt -instcombine -die | dis | grep getelementptr | grep '%A '
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -instcombine -die | dis | grep getelementptr | grep-not '%A '
 
 %Global = constant [10 x sbyte] c"helloworld"
 


Index: llvm/test/Regression/Transforms/InstCombine/load.ll
diff -u llvm/test/Regression/Transforms/InstCombine/load.ll:1.1 llvm/test/Regression/Transforms/InstCombine/load.ll:1.2
--- llvm/test/Regression/Transforms/InstCombine/load.ll:1.1	Thu Jun 26 00:05:51 2003
+++ llvm/test/Regression/Transforms/InstCombine/load.ll	Sat Jun 28 17:44:42 2003
@@ -1,10 +1,7 @@
 ; This test makes sure that these instructions are properly eliminated.
 ;
 
-; RUN: if as < %s | opt -instcombine | dis | grep load
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -instcombine | dis | grep-not load
 
 %X = constant int 42
 %Y = constant [2 x { int, float }] [ { int, float } { int 12, float 1.0 }, 


Index: llvm/test/Regression/Transforms/InstCombine/not.ll
diff -u llvm/test/Regression/Transforms/InstCombine/not.ll:1.7 llvm/test/Regression/Transforms/InstCombine/not.ll:1.8
--- llvm/test/Regression/Transforms/InstCombine/not.ll:1.7	Mon Mar 10 17:13:32 2003
+++ llvm/test/Regression/Transforms/InstCombine/not.ll	Sat Jun 28 17:44:42 2003
@@ -1,10 +1,7 @@
 ; This test makes sure that these instructions are properly eliminated.
 ;
 
-; RUN: if as < %s | opt -instcombine -die | dis | grep xor
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -instcombine -die | dis | grep-not xor
 
 implementation
 


Index: llvm/test/Regression/Transforms/InstCombine/or.ll
diff -u llvm/test/Regression/Transforms/InstCombine/or.ll:1.13 llvm/test/Regression/Transforms/InstCombine/or.ll:1.14
--- llvm/test/Regression/Transforms/InstCombine/or.ll:1.13	Tue Jun  3 23:37:46 2003
+++ llvm/test/Regression/Transforms/InstCombine/or.ll	Sat Jun 28 17:44:42 2003
@@ -1,10 +1,7 @@
 ; This test makes sure that these instructions are properly eliminated.
 ;
 
-; RUN: if as < %s | opt -instcombine | dis | grep -v '%OROK = or' | grep or\ 
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -instcombine | dis | grep -v '%OROK = or' | grep-not 'or '
 
 implementation
 


Index: llvm/test/Regression/Transforms/InstCombine/phi.ll
diff -u llvm/test/Regression/Transforms/InstCombine/phi.ll:1.4 llvm/test/Regression/Transforms/InstCombine/phi.ll:1.5
--- llvm/test/Regression/Transforms/InstCombine/phi.ll:1.4	Thu Aug 22 16:26:15 2002
+++ llvm/test/Regression/Transforms/InstCombine/phi.ll	Sat Jun 28 17:44:42 2003
@@ -1,10 +1,7 @@
 ; This test makes sure that these instructions are properly eliminated.
 ;
 
-; RUN: if as < %s | opt -instcombine -die | dis | grep phi
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -instcombine -die | dis | grep-not phi
 
 implementation
 


Index: llvm/test/Regression/Transforms/InstCombine/pow.ll
diff -u llvm/test/Regression/Transforms/InstCombine/pow.ll:1.2 llvm/test/Regression/Transforms/InstCombine/pow.ll:1.3
--- llvm/test/Regression/Transforms/InstCombine/pow.ll:1.2	Wed Mar  5 17:02:25 2003
+++ llvm/test/Regression/Transforms/InstCombine/pow.ll	Sat Jun 28 17:44:42 2003
@@ -2,10 +2,7 @@
 ;
 ; Equivalent to: http://gcc.gnu.org/ml/gcc-patches/2003-02/msg01786.html
 
-; RUN: if as < %s | opt -instcombine | dis | grep 'call double %pow'
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -instcombine | dis | grep-not 'call double %pow'
 
 declare double %pow(double, double)
 


Index: llvm/test/Regression/Transforms/InstCombine/rem.ll
diff -u llvm/test/Regression/Transforms/InstCombine/rem.ll:1.3 llvm/test/Regression/Transforms/InstCombine/rem.ll:1.4
--- llvm/test/Regression/Transforms/InstCombine/rem.ll:1.3	Tue Feb 18 13:28:47 2003
+++ llvm/test/Regression/Transforms/InstCombine/rem.ll	Sat Jun 28 17:44:42 2003
@@ -1,10 +1,7 @@
 ; This test makes sure that these instructions are properly eliminated.
 ;
 
-; RUN: if as < %s | opt -instcombine | dis | grep rem
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -instcombine | dis | grep-not rem
 
 implementation
 


Index: llvm/test/Regression/Transforms/InstCombine/set.ll
diff -u llvm/test/Regression/Transforms/InstCombine/set.ll:1.8 llvm/test/Regression/Transforms/InstCombine/set.ll:1.9
--- llvm/test/Regression/Transforms/InstCombine/set.ll:1.8	Thu Jun  5 15:11:19 2003
+++ llvm/test/Regression/Transforms/InstCombine/set.ll	Sat Jun 28 17:44:42 2003
@@ -1,10 +1,7 @@
 ; This test makes sure that these instructions are properly eliminated.
 ;
 
-; RUN: if as < %s | opt -instcombine | dis | grep set
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -instcombine | dis | grep-not set
 
 %X = uninitialized global int
 


Index: llvm/test/Regression/Transforms/InstCombine/setcc-strength-reduce.ll
diff -u llvm/test/Regression/Transforms/InstCombine/setcc-strength-reduce.ll:1.2 llvm/test/Regression/Transforms/InstCombine/setcc-strength-reduce.ll:1.3
--- llvm/test/Regression/Transforms/InstCombine/setcc-strength-reduce.ll:1.2	Fri Aug  9 18:33:58 2002
+++ llvm/test/Regression/Transforms/InstCombine/setcc-strength-reduce.ll	Sat Jun 28 17:44:42 2003
@@ -3,10 +3,7 @@
 ; into equivalent setne,eq instructions.
 ;
 
-; RUN: if as < %s | opt -instcombine | dis | grep -v seteq | grep -v setne | grep set
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -instcombine | dis | grep -v seteq | grep -v setne | grep-not set
 
 bool "test1"(uint %A) {
 	%B = setge uint %A, 1   ; setne %A, 0


Index: llvm/test/Regression/Transforms/InstCombine/shift.ll
diff -u llvm/test/Regression/Transforms/InstCombine/shift.ll:1.7 llvm/test/Regression/Transforms/InstCombine/shift.ll:1.8
--- llvm/test/Regression/Transforms/InstCombine/shift.ll:1.7	Mon Mar 10 13:16:20 2003
+++ llvm/test/Regression/Transforms/InstCombine/shift.ll	Sat Jun 28 17:44:42 2003
@@ -1,10 +1,7 @@
 ; This test makes sure that these instructions are properly eliminated.
 ;
 
-; RUN: if as < %s | opt -instcombine | dis | grep sh
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -instcombine | dis | grep-not sh
 
 implementation
 


Index: llvm/test/Regression/Transforms/InstCombine/sub.ll
diff -u llvm/test/Regression/Transforms/InstCombine/sub.ll:1.9 llvm/test/Regression/Transforms/InstCombine/sub.ll:1.10
--- llvm/test/Regression/Transforms/InstCombine/sub.ll:1.9	Mon Mar 10 17:22:50 2003
+++ llvm/test/Regression/Transforms/InstCombine/sub.ll	Sat Jun 28 17:44:42 2003
@@ -1,10 +1,7 @@
 ; This test makes sure that these instructions are properly eliminated.
 ;
 
-; RUN: if as < %s | opt -instcombine -die | dis | grep sub | grep -v 'sub int %Cok, %Bok'
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -instcombine -die | dis | grep sub | grep-not -v 'sub int %Cok, %Bok'
 
 implementation
 




From lattner at cs.uiuc.edu  Sat Jun 28 17:45:14 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 17:45:14 2003
Subject: [llvm-commits] CVS: llvm/test/Scripts/grep-not
Message-ID: <200306282244.RAA08498@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Scripts:

grep-not updated: 1.1 -> 1.2

---
Log message:

Handle quoting right


---
Diffs of the changes:

Index: llvm/test/Scripts/grep-not
diff -u llvm/test/Scripts/grep-not:1.1 llvm/test/Scripts/grep-not:1.2
--- llvm/test/Scripts/grep-not:1.1	Sat Jun 28 17:28:31 2003
+++ llvm/test/Scripts/grep-not	Sat Jun 28 17:43:58 2003
@@ -10,7 +10,7 @@
 #
 # Syntax:   The same as grep
 
-if grep $@
+if grep "$@"
 then exit 1
 else exit 0
 fi




From lattner at cs.uiuc.edu  Sat Jun 28 17:54:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 17:54:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Jello/2003-05-06-LivenessClobber.llx
Message-ID: <200306282253.RAA09037@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Jello:

2003-05-06-LivenessClobber.llx updated: 1.1 -> 1.2

---
Log message:

Get rid of control flow in the test


---
Diffs of the changes:

Index: llvm/test/Regression/Jello/2003-05-06-LivenessClobber.llx
diff -u llvm/test/Regression/Jello/2003-05-06-LivenessClobber.llx:1.1 llvm/test/Regression/Jello/2003-05-06-LivenessClobber.llx:1.2
--- llvm/test/Regression/Jello/2003-05-06-LivenessClobber.llx:1.1	Tue May  6 14:41:25 2003
+++ llvm/test/Regression/Jello/2003-05-06-LivenessClobber.llx	Sat Jun 28 17:53:07 2003
@@ -1,9 +1,6 @@
 ; This testcase shoudl return with an exit code of 1.
 ;
-; RUN: if as < %s | lli -force-interpreter=false
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | not lli -force-interpreter=false
 
 target endian = little
 target pointersize = 32




From lattner at cs.uiuc.edu  Sat Jun 28 17:54:13 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 17:54:13 2003
Subject: [llvm-commits] CVS: llvm/test/Scripts/not
Message-ID: <200306282253.RAA09028@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Scripts:

not added (r1.1)

---
Log message:

New script for testrunner programs


---
Diffs of the changes:

Index: llvm/test/Scripts/not
diff -c /dev/null llvm/test/Scripts/not:1.1
*** /dev/null	Sat Jun 28 17:53:02 2003
--- llvm/test/Scripts/not	Sat Jun 28 17:52:52 2003
***************
*** 0 ****
--- 1,12 ----
+ #!/bin/sh
+ #
+ # Program: not
+ #
+ # Synopsis: Inverse the output of the program specified on the command line
+ #
+ # Syntax:   not command <arguments>
+ 
+ if "$@"
+ then exit 1
+ else exit 0
+ fi




From lattner at cs.uiuc.edu  Sat Jun 28 18:03:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:03:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/LevelRaise/2002-02-11-ArrayShape.ll
Message-ID: <200306282302.SAA11016@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/LevelRaise:

2002-02-11-ArrayShape.ll updated: 1.2 -> 1.3

---
Log message:

Force the interpreter for this testcase


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/LevelRaise/2002-02-11-ArrayShape.ll
diff -u llvm/test/Regression/Transforms/LevelRaise/2002-02-11-ArrayShape.ll:1.2 llvm/test/Regression/Transforms/LevelRaise/2002-02-11-ArrayShape.ll:1.3
--- llvm/test/Regression/Transforms/LevelRaise/2002-02-11-ArrayShape.ll:1.2	Wed May 14 13:37:03 2003
+++ llvm/test/Regression/Transforms/LevelRaise/2002-02-11-ArrayShape.ll	Sat Jun 28 18:02:04 2003
@@ -18,7 +18,7 @@
 ; equivelent form.
 ;
 ; RUN: as < %s | opt -q -raise > Output/%s.raised.bc
-; RUN: lli -array-checks -abort-on-exception < Output/%s.raised.bc
+; RUN: lli -force-interpreter -array-checks -abort-on-exception < Output/%s.raised.bc
 ;
 
 implementation




From lattner at cs.uiuc.edu  Sat Jun 28 18:03:13 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:03:13 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/LevelRaise/2002-03-21-MissedRaise.ll 2002-03-21-MissedRaise2.ll 2002-03-21-MissedRaise3.ll 2002-04-09-MissedRaise.ll 2002-04-16-MissedRaise.ll 2002-05-02-MissedRaise.ll 2002-05-10-LoadPeephole.ll 2002-05-23-MissedRaise.ll 2002-07-16-MissedRaise.ll 2002-07-18-MissedAllocaRaise.ll 2002-10-02-SignExtensionProblem.ll 2002-10-08-VarArgCall.ll 2002-11-13-PointerFunction.ll 2003-05-01-CallCast.ll 2003-06-25-ExprAnalysis.ll
Message-ID: <200306282302.SAA11010@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/LevelRaise:

2002-03-21-MissedRaise.ll updated: 1.1 -> 1.2
2002-03-21-MissedRaise2.ll updated: 1.2 -> 1.3
2002-03-21-MissedRaise3.ll updated: 1.1 -> 1.2
2002-04-09-MissedRaise.ll updated: 1.1 -> 1.2
2002-04-16-MissedRaise.ll updated: 1.1 -> 1.2
2002-05-02-MissedRaise.ll updated: 1.2 -> 1.3
2002-05-10-LoadPeephole.ll updated: 1.1 -> 1.2
2002-05-23-MissedRaise.ll updated: 1.3 -> 1.4
2002-07-16-MissedRaise.ll updated: 1.2 -> 1.3
2002-07-18-MissedAllocaRaise.ll updated: 1.1 -> 1.2
2002-10-02-SignExtensionProblem.ll updated: 1.1 -> 1.2
2002-10-08-VarArgCall.ll updated: 1.1 -> 1.2
2002-11-13-PointerFunction.ll updated: 1.1 -> 1.2
2003-05-01-CallCast.ll updated: 1.1 -> 1.2
2003-06-25-ExprAnalysis.ll updated: 1.1 -> 1.2

---
Log message:

Switch tests over to use the 'not' script, instead of explicit control flow


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/LevelRaise/2002-03-21-MissedRaise.ll
diff -u llvm/test/Regression/Transforms/LevelRaise/2002-03-21-MissedRaise.ll:1.1 llvm/test/Regression/Transforms/LevelRaise/2002-03-21-MissedRaise.ll:1.2
--- llvm/test/Regression/Transforms/LevelRaise/2002-03-21-MissedRaise.ll:1.1	Thu Mar 21 15:20:04 2002
+++ llvm/test/Regression/Transforms/LevelRaise/2002-03-21-MissedRaise.ll	Sat Jun 28 18:02:20 2003
@@ -2,10 +2,7 @@
 ; successful, all cast instructions should be eliminated from this testcase.
 ;
 
-; RUN: if as < %s | opt -raise | dis | grep cast 
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -raise | dis | grep-not cast 
 
 	%Hash = type { { uint, sbyte *, \2 } * *, int (uint) *, int } *
 	%hash = type { { uint, sbyte *, \2 } * *, int (uint) *, int }


Index: llvm/test/Regression/Transforms/LevelRaise/2002-03-21-MissedRaise2.ll
diff -u llvm/test/Regression/Transforms/LevelRaise/2002-03-21-MissedRaise2.ll:1.2 llvm/test/Regression/Transforms/LevelRaise/2002-03-21-MissedRaise2.ll:1.3
--- llvm/test/Regression/Transforms/LevelRaise/2002-03-21-MissedRaise2.ll:1.2	Mon Nov  4 00:09:46 2002
+++ llvm/test/Regression/Transforms/LevelRaise/2002-03-21-MissedRaise2.ll	Sat Jun 28 18:02:20 2003
@@ -3,10 +3,7 @@
 ; prevented reg115 from being able to change.
 ;
 
-; RUN: if as < %s | opt -raise | dis | grep '= cast' | grep \*
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -raise | dis | grep '= cast' | grep-not \*
 	
 	%Hash = type { { uint, sbyte *, \2 } * *, int (uint) *, int } *
 	%HashEntry = type { uint, sbyte *, \2 } *


Index: llvm/test/Regression/Transforms/LevelRaise/2002-03-21-MissedRaise3.ll
diff -u llvm/test/Regression/Transforms/LevelRaise/2002-03-21-MissedRaise3.ll:1.1 llvm/test/Regression/Transforms/LevelRaise/2002-03-21-MissedRaise3.ll:1.2
--- llvm/test/Regression/Transforms/LevelRaise/2002-03-21-MissedRaise3.ll:1.1	Thu Mar 21 17:00:48 2002
+++ llvm/test/Regression/Transforms/LevelRaise/2002-03-21-MissedRaise3.ll	Sat Jun 28 18:02:20 2003
@@ -1,7 +1,4 @@
-; RUN: if as < %s | opt -raise | dis | grep '= cast' | grep \*
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -raise | dis | grep '= cast' | not grep \*
 
 	%Hash = type { { uint, sbyte *, \2 } * *, int (uint) *, int } *
 	%HashEntry = type { uint, sbyte *, \2 } *
@@ -9,9 +6,7 @@
 	%hash_entry = type { uint, sbyte *, \2 * }
 implementation
 
-%Hash "MakeHash"(int %size, int (uint) * %map)
-begin
-bb0:					;[#uses=1]
+%Hash "MakeHash"(int %size, int (uint) * %map) {
 	%reg112 = malloc sbyte * *, uint 3		; <sbyte * * *> [#uses=5]
 	%reg107-uint = cast int %size to uint		; <uint> [#uses=1]
 	%reg115 = malloc sbyte *, uint %reg107-uint		; <sbyte * *> [#uses=1]
@@ -19,5 +14,5 @@
 
 	%cast246 = cast sbyte * * * %reg112 to %Hash		; <%Hash> [#uses=1]
 	ret %Hash %cast246
-end
+}
 


Index: llvm/test/Regression/Transforms/LevelRaise/2002-04-09-MissedRaise.ll
diff -u llvm/test/Regression/Transforms/LevelRaise/2002-04-09-MissedRaise.ll:1.1 llvm/test/Regression/Transforms/LevelRaise/2002-04-09-MissedRaise.ll:1.2
--- llvm/test/Regression/Transforms/LevelRaise/2002-04-09-MissedRaise.ll:1.1	Tue Apr  9 15:43:18 2002
+++ llvm/test/Regression/Transforms/LevelRaise/2002-04-09-MissedRaise.ll	Sat Jun 28 18:02:20 2003
@@ -8,10 +8,7 @@
 ;     x[j*i-12] = j;
 ; }
 
-; RUN: if as < %s | opt -raise | dis | grep ' cast ' | grep '*'
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -raise | dis | grep ' cast ' | not grep '*'
 
 implementation
 


Index: llvm/test/Regression/Transforms/LevelRaise/2002-04-16-MissedRaise.ll
diff -u llvm/test/Regression/Transforms/LevelRaise/2002-04-16-MissedRaise.ll:1.1 llvm/test/Regression/Transforms/LevelRaise/2002-04-16-MissedRaise.ll:1.2
--- llvm/test/Regression/Transforms/LevelRaise/2002-04-16-MissedRaise.ll:1.1	Tue Apr 16 00:44:56 2002
+++ llvm/test/Regression/Transforms/LevelRaise/2002-04-16-MissedRaise.ll	Sat Jun 28 18:02:20 2003
@@ -5,18 +5,11 @@
 ;   return (int*)malloc(i+j);
 ; }
 
-; RUN: if as < %s | opt -raise | dis | grep ' cast ' | grep '*'
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -raise | dis | grep ' cast ' | not grep '*'
 
 implementation
 
-declare sbyte* "malloc"(uint)
-
-int* "test"(uint %N, uint %M)
-begin
-bb0:					;[#uses=0]
+int* %test(uint %N, uint %M) {
 	%reg111 = shl uint %N, ubyte 2		; <uint> [#uses=1]
 	%reg109 = add uint %reg111, 4		; <uint> [#uses=1]
 	%reg114 = shl uint %M, ubyte 2		; <uint> [#uses=1]
@@ -25,4 +18,4 @@
 	%reg117 = malloc sbyte, uint %reg116		; <sbyte*> [#uses=1]
 	%cast221 = cast sbyte* %reg117 to int*		; <int*> [#uses=1]
 	ret int* %cast221
-end
+}


Index: llvm/test/Regression/Transforms/LevelRaise/2002-05-02-MissedRaise.ll
diff -u llvm/test/Regression/Transforms/LevelRaise/2002-05-02-MissedRaise.ll:1.2 llvm/test/Regression/Transforms/LevelRaise/2002-05-02-MissedRaise.ll:1.3
--- llvm/test/Regression/Transforms/LevelRaise/2002-05-02-MissedRaise.ll:1.2	Mon Nov  4 00:07:04 2002
+++ llvm/test/Regression/Transforms/LevelRaise/2002-05-02-MissedRaise.ll	Sat Jun 28 18:02:20 2003
@@ -1,9 +1,6 @@
 ; This testcase is not level raised properly...
 ;
-; RUN: if as < %s | opt -raise | dis | grep ' cast ' | grep '*'
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -raise | dis | grep ' cast ' | not grep '*'
 
 	%List = type { int, %List* }
 


Index: llvm/test/Regression/Transforms/LevelRaise/2002-05-10-LoadPeephole.ll
diff -u llvm/test/Regression/Transforms/LevelRaise/2002-05-10-LoadPeephole.ll:1.1 llvm/test/Regression/Transforms/LevelRaise/2002-05-10-LoadPeephole.ll:1.2
--- llvm/test/Regression/Transforms/LevelRaise/2002-05-10-LoadPeephole.ll:1.1	Fri May 10 10:14:54 2002
+++ llvm/test/Regression/Transforms/LevelRaise/2002-05-10-LoadPeephole.ll	Sat Jun 28 18:02:20 2003
@@ -1,10 +1,7 @@
 ; This testcase should have the cast propogated through the load
 ; just like a store does...
 ;
-; RUN: if as < %s | opt -raise | dis | grep ' cast ' | grep '*'
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -raise | dis | grep ' cast ' | not grep '*'
 
 int "test"(uint * %Ptr) {
 	%P2 = cast uint *%Ptr to int *


Index: llvm/test/Regression/Transforms/LevelRaise/2002-05-23-MissedRaise.ll
diff -u llvm/test/Regression/Transforms/LevelRaise/2002-05-23-MissedRaise.ll:1.3 llvm/test/Regression/Transforms/LevelRaise/2002-05-23-MissedRaise.ll:1.4
--- llvm/test/Regression/Transforms/LevelRaise/2002-05-23-MissedRaise.ll:1.3	Mon Nov  4 00:16:13 2002
+++ llvm/test/Regression/Transforms/LevelRaise/2002-05-23-MissedRaise.ll	Sat Jun 28 18:02:20 2003
@@ -1,7 +1,4 @@
-; RUN: if as < %s | opt -raise | dis | grep '= cast' | grep \*
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -raise | dis | grep '= cast' | not grep \*
 
 %FILE = type { int, ubyte*, ubyte*, ubyte, ubyte, uint, uint, uint }
 


Index: llvm/test/Regression/Transforms/LevelRaise/2002-07-16-MissedRaise.ll
diff -u llvm/test/Regression/Transforms/LevelRaise/2002-07-16-MissedRaise.ll:1.2 llvm/test/Regression/Transforms/LevelRaise/2002-07-16-MissedRaise.ll:1.3
--- llvm/test/Regression/Transforms/LevelRaise/2002-07-16-MissedRaise.ll:1.2	Tue Oct 15 16:39:49 2002
+++ llvm/test/Regression/Transforms/LevelRaise/2002-07-16-MissedRaise.ll	Sat Jun 28 18:02:20 2003
@@ -5,10 +5,7 @@
 ; This could be fixed by making all stores add themselves to a list, and check
 ; their arguments are consistent AFTER all other values are propogated.
 
-; RUN: if as < %s | opt -raise | dis | grep '= cast' 
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -raise | dis | not grep '= cast' 
 
         %Tree = type %struct.tree*
         %struct.tree = type { int, double, double, %Tree, %Tree, %Tree, %Tree }


Index: llvm/test/Regression/Transforms/LevelRaise/2002-07-18-MissedAllocaRaise.ll
diff -u llvm/test/Regression/Transforms/LevelRaise/2002-07-18-MissedAllocaRaise.ll:1.1 llvm/test/Regression/Transforms/LevelRaise/2002-07-18-MissedAllocaRaise.ll:1.2
--- llvm/test/Regression/Transforms/LevelRaise/2002-07-18-MissedAllocaRaise.ll:1.1	Thu Jul 18 22:51:49 2002
+++ llvm/test/Regression/Transforms/LevelRaise/2002-07-18-MissedAllocaRaise.ll	Sat Jun 28 18:02:20 2003
@@ -1,9 +1,6 @@
 ; Looks like we don't raise alloca's like we do mallocs
 ;
-; RUN: if as < %s | opt -raise | dis | grep '= cast' | grep \*
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -raise | dis | grep '= cast' | not grep \*
 
 implementation   ; Functions:
 


Index: llvm/test/Regression/Transforms/LevelRaise/2002-10-02-SignExtensionProblem.ll
diff -u llvm/test/Regression/Transforms/LevelRaise/2002-10-02-SignExtensionProblem.ll:1.1 llvm/test/Regression/Transforms/LevelRaise/2002-10-02-SignExtensionProblem.ll:1.2
--- llvm/test/Regression/Transforms/LevelRaise/2002-10-02-SignExtensionProblem.ll:1.1	Wed Oct  2 13:32:38 2002
+++ llvm/test/Regression/Transforms/LevelRaise/2002-10-02-SignExtensionProblem.ll	Sat Jun 28 18:02:20 2003
@@ -1,7 +1,4 @@
-; RUN: if as < %s | opt -raise | dis | grep 4294967295
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -raise | dis | grep -v uint | not grep 4294967295
 
 %length_code = uninitialized global [256 x ubyte]
 


Index: llvm/test/Regression/Transforms/LevelRaise/2002-10-08-VarArgCall.ll
diff -u llvm/test/Regression/Transforms/LevelRaise/2002-10-08-VarArgCall.ll:1.1 llvm/test/Regression/Transforms/LevelRaise/2002-10-08-VarArgCall.ll:1.2
--- llvm/test/Regression/Transforms/LevelRaise/2002-10-08-VarArgCall.ll:1.1	Tue Oct  8 16:57:49 2002
+++ llvm/test/Regression/Transforms/LevelRaise/2002-10-08-VarArgCall.ll	Sat Jun 28 18:02:20 2003
@@ -1,7 +1,4 @@
-; RUN: if as < %s | opt -raise | dis | grep call | grep \.\.\.
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -raise | dis | grep call | not grep '\.\.\.'
 
 implementation
 


Index: llvm/test/Regression/Transforms/LevelRaise/2002-11-13-PointerFunction.ll
diff -u llvm/test/Regression/Transforms/LevelRaise/2002-11-13-PointerFunction.ll:1.1 llvm/test/Regression/Transforms/LevelRaise/2002-11-13-PointerFunction.ll:1.2
--- llvm/test/Regression/Transforms/LevelRaise/2002-11-13-PointerFunction.ll:1.1	Wed Nov 13 00:54:00 2002
+++ llvm/test/Regression/Transforms/LevelRaise/2002-11-13-PointerFunction.ll	Sat Jun 28 18:02:20 2003
@@ -1,9 +1,6 @@
 ; This testcase should be able to eliminate at least one of the casts.
 ;
-; RUN: if as < %s | opt -raise | dis | grep 'REMOVE'
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -raise | dis | not grep 'REMOVE'
 
 int %foo(sbyte * %PF) {
 	%UPF = cast sbyte* %PF to uint()*


Index: llvm/test/Regression/Transforms/LevelRaise/2003-05-01-CallCast.ll
diff -u llvm/test/Regression/Transforms/LevelRaise/2003-05-01-CallCast.ll:1.1 llvm/test/Regression/Transforms/LevelRaise/2003-05-01-CallCast.ll:1.2
--- llvm/test/Regression/Transforms/LevelRaise/2003-05-01-CallCast.ll:1.1	Thu May  1 16:02:35 2003
+++ llvm/test/Regression/Transforms/LevelRaise/2003-05-01-CallCast.ll	Sat Jun 28 18:02:20 2003
@@ -1,7 +1,4 @@
-; RUN: if as < %s | opt -raise | dis | grep cast
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -raise | dis | not grep cast
 
 void %test(...) { ret void }
 


Index: llvm/test/Regression/Transforms/LevelRaise/2003-06-25-ExprAnalysis.ll
diff -u llvm/test/Regression/Transforms/LevelRaise/2003-06-25-ExprAnalysis.ll:1.1 llvm/test/Regression/Transforms/LevelRaise/2003-06-25-ExprAnalysis.ll:1.2
--- llvm/test/Regression/Transforms/LevelRaise/2003-06-25-ExprAnalysis.ll:1.1	Wed Jun 25 12:08:34 2003
+++ llvm/test/Regression/Transforms/LevelRaise/2003-06-25-ExprAnalysis.ll	Sat Jun 28 18:02:20 2003
@@ -1,9 +1,6 @@
 ; The expr analysis routines were being too aggressive across cast instructions!
 
-; RUN: if as < %s | opt -raise | dis | grep 4294967295
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -raise | dis | not grep 4294967295
 
 target endian = big
 target pointersize = 64




From lattner at cs.uiuc.edu  Sat Jun 28 18:04:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:04:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/LevelRaise/2002-03-21-MissedRaise.ll 2002-03-21-MissedRaise2.ll
Message-ID: <200306282303.SAA11036@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/LevelRaise:

2002-03-21-MissedRaise.ll updated: 1.2 -> 1.3
2002-03-21-MissedRaise2.ll updated: 1.3 -> 1.4

---
Log message:

Use not instead of grep-not


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/LevelRaise/2002-03-21-MissedRaise.ll
diff -u llvm/test/Regression/Transforms/LevelRaise/2002-03-21-MissedRaise.ll:1.2 llvm/test/Regression/Transforms/LevelRaise/2002-03-21-MissedRaise.ll:1.3
--- llvm/test/Regression/Transforms/LevelRaise/2002-03-21-MissedRaise.ll:1.2	Sat Jun 28 18:02:20 2003
+++ llvm/test/Regression/Transforms/LevelRaise/2002-03-21-MissedRaise.ll	Sat Jun 28 18:02:51 2003
@@ -2,7 +2,7 @@
 ; successful, all cast instructions should be eliminated from this testcase.
 ;
 
-; RUN: as < %s | opt -raise | dis | grep-not cast 
+; RUN: as < %s | opt -raise | dis | not grep cast 
 
 	%Hash = type { { uint, sbyte *, \2 } * *, int (uint) *, int } *
 	%hash = type { { uint, sbyte *, \2 } * *, int (uint) *, int }


Index: llvm/test/Regression/Transforms/LevelRaise/2002-03-21-MissedRaise2.ll
diff -u llvm/test/Regression/Transforms/LevelRaise/2002-03-21-MissedRaise2.ll:1.3 llvm/test/Regression/Transforms/LevelRaise/2002-03-21-MissedRaise2.ll:1.4
--- llvm/test/Regression/Transforms/LevelRaise/2002-03-21-MissedRaise2.ll:1.3	Sat Jun 28 18:02:20 2003
+++ llvm/test/Regression/Transforms/LevelRaise/2002-03-21-MissedRaise2.ll	Sat Jun 28 18:02:51 2003
@@ -3,7 +3,7 @@
 ; prevented reg115 from being able to change.
 ;
 
-; RUN: as < %s | opt -raise | dis | grep '= cast' | grep-not \*
+; RUN: as < %s | opt -raise | dis | grep '= cast' | not grep \*
 	
 	%Hash = type { { uint, sbyte *, \2 } * *, int (uint) *, int } *
 	%HashEntry = type { uint, sbyte *, \2 } *




From lattner at cs.uiuc.edu  Sat Jun 28 18:14:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:14:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/CFGSimplify/EqualPHIEdgeBlockMerge.ll PhiBlockMerge.ll PhiBlockMerge2.ll basictest.ll
Message-ID: <200306282313.SAA12323@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/CFGSimplify:

EqualPHIEdgeBlockMerge.ll updated: 1.1 -> 1.2
PhiBlockMerge.ll updated: 1.1 -> 1.2
PhiBlockMerge2.ll updated: 1.1 -> 1.2
basictest.ll updated: 1.2 -> 1.3

---
Log message:

Eliminate explicit control flow in tests


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/CFGSimplify/EqualPHIEdgeBlockMerge.ll
diff -u llvm/test/Regression/Transforms/CFGSimplify/EqualPHIEdgeBlockMerge.ll:1.1 llvm/test/Regression/Transforms/CFGSimplify/EqualPHIEdgeBlockMerge.ll:1.2
--- llvm/test/Regression/Transforms/CFGSimplify/EqualPHIEdgeBlockMerge.ll:1.1	Wed Mar  5 14:35:24 2003
+++ llvm/test/Regression/Transforms/CFGSimplify/EqualPHIEdgeBlockMerge.ll	Sat Jun 28 18:13:12 2003
@@ -1,9 +1,6 @@
 ; Test merging of blocks with phi nodes.
 ;
-; RUN: if as < %s | opt -simplifycfg | dis | grep 'N:'
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -simplifycfg | dis | not grep 'N:'
 ;
 
 int %test(bool %a) {


Index: llvm/test/Regression/Transforms/CFGSimplify/PhiBlockMerge.ll
diff -u llvm/test/Regression/Transforms/CFGSimplify/PhiBlockMerge.ll:1.1 llvm/test/Regression/Transforms/CFGSimplify/PhiBlockMerge.ll:1.2
--- llvm/test/Regression/Transforms/CFGSimplify/PhiBlockMerge.ll:1.1	Wed Mar  5 14:22:50 2003
+++ llvm/test/Regression/Transforms/CFGSimplify/PhiBlockMerge.ll	Sat Jun 28 18:13:12 2003
@@ -1,9 +1,6 @@
 ; Test merging of blocks that only have PHI nodes in them
 ;
-; RUN: if as < %s | opt -simplifycfg | dis | grep 'N:'
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -simplifycfg | dis | not grep 'N:'
 ;
 
 int %test(bool %a, bool %b) {


Index: llvm/test/Regression/Transforms/CFGSimplify/PhiBlockMerge2.ll
diff -u llvm/test/Regression/Transforms/CFGSimplify/PhiBlockMerge2.ll:1.1 llvm/test/Regression/Transforms/CFGSimplify/PhiBlockMerge2.ll:1.2
--- llvm/test/Regression/Transforms/CFGSimplify/PhiBlockMerge2.ll:1.1	Wed Mar  5 15:34:35 2003
+++ llvm/test/Regression/Transforms/CFGSimplify/PhiBlockMerge2.ll	Sat Jun 28 18:13:12 2003
@@ -2,10 +2,7 @@
 ; where the mergedinto block doesn't have any PHI nodes, and is in fact 
 ; dominated by the block-to-be-eliminated
 ;
-; RUN: if as < %s | opt -simplifycfg | dis | grep 'N:'
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -simplifycfg | dis | not grep 'N:'
 ;
 
 int %test(bool %a, bool %b) {


Index: llvm/test/Regression/Transforms/CFGSimplify/basictest.ll
diff -u llvm/test/Regression/Transforms/CFGSimplify/basictest.ll:1.2 llvm/test/Regression/Transforms/CFGSimplify/basictest.ll:1.3
--- llvm/test/Regression/Transforms/CFGSimplify/basictest.ll:1.2	Tue May 21 14:33:17 2002
+++ llvm/test/Regression/Transforms/CFGSimplify/basictest.ll	Sat Jun 28 18:13:12 2003
@@ -1,9 +1,6 @@
 ; Test CFG simplify removal of branch instructions...
 ;
-; RUN: if as < %s | opt -simplifycfg | dis | grep br
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -simplifycfg | dis | not grep br
 
 
 void "test1"() {




From lattner at cs.uiuc.edu  Sat Jun 28 18:14:13 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:14:13 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/SimplifyCFG/2002-06-24-PHINode.ll branch-fold-test.ll
Message-ID: <200306282313.SAA12289@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/SimplifyCFG:

2002-06-24-PHINode.ll updated: 1.2 -> 1.3
branch-fold-test.ll updated: 1.1 -> 1.2

---
Log message:

Eliminate explicit control flow in tests


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/SimplifyCFG/2002-06-24-PHINode.ll
diff -u llvm/test/Regression/Transforms/SimplifyCFG/2002-06-24-PHINode.ll:1.2 llvm/test/Regression/Transforms/SimplifyCFG/2002-06-24-PHINode.ll:1.3
--- llvm/test/Regression/Transforms/SimplifyCFG/2002-06-24-PHINode.ll:1.2	Mon Jul 29 16:12:20 2002
+++ llvm/test/Regression/Transforms/SimplifyCFG/2002-06-24-PHINode.ll	Sat Jun 28 18:13:03 2003
@@ -1,10 +1,7 @@
 ; -simplifycfg is not folding blocks if there is a PHI node involved.  This 
 ; should be fixed eventually
 
-; RUN: if as < %s | opt -simplifycfg | dis | grep br
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -simplifycfg | dis | not grep br
 
 int %main(int %argc) {
         br label %InlinedFunctionReturnNode


Index: llvm/test/Regression/Transforms/SimplifyCFG/branch-fold-test.ll
diff -u llvm/test/Regression/Transforms/SimplifyCFG/branch-fold-test.ll:1.1 llvm/test/Regression/Transforms/SimplifyCFG/branch-fold-test.ll:1.2
--- llvm/test/Regression/Transforms/SimplifyCFG/branch-fold-test.ll:1.1	Tue Sep 24 11:13:20 2002
+++ llvm/test/Regression/Transforms/SimplifyCFG/branch-fold-test.ll	Sat Jun 28 18:13:03 2003
@@ -1,10 +1,7 @@
 ; This test ensures that the simplifycfg pass continues to constant fold
 ; terminator instructions.
 
-; RUN: if as < %s | opt -simplifycfg | grep br
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -simplifycfg | not grep br
 
 int %test(int %A, int %B) {
 J:




From lattner at cs.uiuc.edu  Sat Jun 28 18:15:02 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:15:02 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/Makefile
Message-ID: <200306282314.SAA12357@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms:

Makefile updated: 1.24 -> 1.25

---
Log message:

No use in having CFGSimplify and SimplifyCFG directories!


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/Makefile
diff -u llvm/test/Regression/Transforms/Makefile:1.24 llvm/test/Regression/Transforms/Makefile:1.25
--- llvm/test/Regression/Transforms/Makefile:1.24	Tue Jun 24 13:49:51 2003
+++ llvm/test/Regression/Transforms/Makefile	Sat Jun 28 18:14:35 2003
@@ -1,6 +1,5 @@
 LEVEL = ../../..
 DIRS  = ADCE \
-	CFGSimplify \
 	ConstProp \
 	ConstantMerge \
 	CorrelatedExprs \




From lattner at cs.uiuc.edu  Sat Jun 28 18:18:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:18:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/Makefile
Message-ID: <200306282317.SAA12869@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms:

Makefile updated: 1.25 -> 1.26

---
Log message:

Move DSGraph tests to test/Regression/Analysis/DSGraph


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/Makefile
diff -u llvm/test/Regression/Transforms/Makefile:1.25 llvm/test/Regression/Transforms/Makefile:1.26
--- llvm/test/Regression/Transforms/Makefile:1.25	Sat Jun 28 18:14:35 2003
+++ llvm/test/Regression/Transforms/Makefile	Sat Jun 28 18:17:42 2003
@@ -4,7 +4,6 @@
 	ConstantMerge \
 	CorrelatedExprs \
 	DecomposeMultiDimRefs \
-        DSGraphAA \
 	FunctionResolve \
 	GCSE \
 	GlobalDCE \




From lattner at cs.uiuc.edu  Sat Jun 28 18:18:15 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:18:15 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/DSGraphAA/Makefile
Message-ID: <200306282317.SAA12860@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/DSGraphAA:

Makefile (r1.1) removed

---
Log message:

Move files to test/Regression/Analysis/DSGraph


---
Diffs of the changes:



From lattner at cs.uiuc.edu  Sat Jun 28 18:24:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:24:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/CorrelatedExprs/2002-10-03-PHIPropogate.ll basictest.ll branchtest.ll looptest.ll nullpointer.ll whet.ll
Message-ID: <200306282323.SAA13620@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/CorrelatedExprs:

2002-10-03-PHIPropogate.ll updated: 1.1 -> 1.2
basictest.ll updated: 1.1 -> 1.2
branchtest.ll updated: 1.1 -> 1.2
looptest.ll updated: 1.1 -> 1.2
nullpointer.ll updated: 1.1 -> 1.2
whet.ll updated: 1.1 -> 1.2

---
Log message:

Remove explicit control flow through the use of the 'not' script


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/CorrelatedExprs/2002-10-03-PHIPropogate.ll
diff -u llvm/test/Regression/Transforms/CorrelatedExprs/2002-10-03-PHIPropogate.ll:1.1 llvm/test/Regression/Transforms/CorrelatedExprs/2002-10-03-PHIPropogate.ll:1.2
--- llvm/test/Regression/Transforms/CorrelatedExprs/2002-10-03-PHIPropogate.ll:1.1	Thu Oct  3 15:06:33 2002
+++ llvm/test/Regression/Transforms/CorrelatedExprs/2002-10-03-PHIPropogate.ll	Sat Jun 28 18:23:18 2003
@@ -12,10 +12,7 @@
 ;   }
 ; }
 ;
-; RUN: if as < %s | opt -cee -simplifycfg | dis | grep bb3
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -cee -simplifycfg | dis | not grep bb3
 
 implementation
 declare void %bar()


Index: llvm/test/Regression/Transforms/CorrelatedExprs/basictest.ll
diff -u llvm/test/Regression/Transforms/CorrelatedExprs/basictest.ll:1.1 llvm/test/Regression/Transforms/CorrelatedExprs/basictest.ll:1.2
--- llvm/test/Regression/Transforms/CorrelatedExprs/basictest.ll:1.1	Mon Sep 16 14:08:35 2002
+++ llvm/test/Regression/Transforms/CorrelatedExprs/basictest.ll	Sat Jun 28 18:23:18 2003
@@ -1,7 +1,4 @@
-; RUN: if as < %s | opt -cee -constprop -instcombine -dce | dis | grep 'REMOVE'
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -cee -constprop -instcombine -dce | dis | not grep 'REMOVE'
 
 int %test1(int %A) {
 	%cond = seteq int %A, 40


Index: llvm/test/Regression/Transforms/CorrelatedExprs/branchtest.ll
diff -u llvm/test/Regression/Transforms/CorrelatedExprs/branchtest.ll:1.1 llvm/test/Regression/Transforms/CorrelatedExprs/branchtest.ll:1.2
--- llvm/test/Regression/Transforms/CorrelatedExprs/branchtest.ll:1.1	Mon Sep 16 14:08:36 2002
+++ llvm/test/Regression/Transforms/CorrelatedExprs/branchtest.ll	Sat Jun 28 18:23:18 2003
@@ -1,7 +1,4 @@
-; RUN: if as < %s | opt -cee -simplifycfg | dis | grep 'REMOVE'
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -cee -simplifycfg | dis | not grep 'REMOVE'
 
 implementation
 


Index: llvm/test/Regression/Transforms/CorrelatedExprs/looptest.ll
diff -u llvm/test/Regression/Transforms/CorrelatedExprs/looptest.ll:1.1 llvm/test/Regression/Transforms/CorrelatedExprs/looptest.ll:1.2
--- llvm/test/Regression/Transforms/CorrelatedExprs/looptest.ll:1.1	Sun Oct  6 18:05:38 2002
+++ llvm/test/Regression/Transforms/CorrelatedExprs/looptest.ll	Sat Jun 28 18:23:18 2003
@@ -11,10 +11,7 @@
 ;
 ; Note that this is a "feature" test, not a correctness test.
 ;
-; RUN: if as < %s | opt -cee -simplifycfg | dis | grep cond213
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -cee -simplifycfg | dis | not grep cond213
 ;
 implementation   ; Functions:
 


Index: llvm/test/Regression/Transforms/CorrelatedExprs/nullpointer.ll
diff -u llvm/test/Regression/Transforms/CorrelatedExprs/nullpointer.ll:1.1 llvm/test/Regression/Transforms/CorrelatedExprs/nullpointer.ll:1.2
--- llvm/test/Regression/Transforms/CorrelatedExprs/nullpointer.ll:1.1	Fri Nov  1 11:29:35 2002
+++ llvm/test/Regression/Transforms/CorrelatedExprs/nullpointer.ll	Sat Jun 28 18:23:18 2003
@@ -1,10 +1,7 @@
 ; a load or store of a pointer indicates that the pointer is not null.
 ; Any succeeding uses of the pointer should get this info
 
-; RUN: if as < %s | opt -cee -instcombine -simplifycfg | dis | grep br
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -cee -instcombine -simplifycfg | dis | not grep br
 
 implementation   ; Functions:
 


Index: llvm/test/Regression/Transforms/CorrelatedExprs/whet.ll
diff -u llvm/test/Regression/Transforms/CorrelatedExprs/whet.ll:1.1 llvm/test/Regression/Transforms/CorrelatedExprs/whet.ll:1.2
--- llvm/test/Regression/Transforms/CorrelatedExprs/whet.ll:1.1	Sun Oct  6 18:31:09 2002
+++ llvm/test/Regression/Transforms/CorrelatedExprs/whet.ll	Sat Jun 28 18:23:18 2003
@@ -11,10 +11,7 @@
 ;
 ; This should eliminate all BB's except BB0, BB9, BB10
 ;
-; RUN: if as < %s | opt -cee -instcombine -simplifycfg | dis | grep 'bb[2-8]'
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -cee -instcombine -simplifycfg | dis | not grep 'bb[2-8]'
 
 implementation   ; Functions:
 




From lattner at cs.uiuc.edu  Sat Jun 28 18:24:13 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:24:13 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/DeadArgElim/basictest.ll
Message-ID: <200306282323.SAA13625@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/DeadArgElim:

basictest.ll updated: 1.2 -> 1.3

---
Log message:

Remove explicit control flow through the use of the 'not' script



---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/DeadArgElim/basictest.ll
diff -u llvm/test/Regression/Transforms/DeadArgElim/basictest.ll:1.2 llvm/test/Regression/Transforms/DeadArgElim/basictest.ll:1.3
--- llvm/test/Regression/Transforms/DeadArgElim/basictest.ll:1.2	Tue Jun 17 17:21:55 2003
+++ llvm/test/Regression/Transforms/DeadArgElim/basictest.ll	Sat Jun 28 18:23:19 2003
@@ -1,7 +1,4 @@
-; RUN: if as < %s | opt -deadargelim | dis | grep DEADARG
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -deadargelim | dis | not grep DEADARG
 
 implementation
 




From lattner at cs.uiuc.edu  Sat Jun 28 18:24:25 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:24:25 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/ConstProp/2002-05-03-NotOperator.ll 2002-09-03-SetCC-Bools.ll basictest.ll logicaltest.ll nottest.ll phi.ll remtest.ll
Message-ID: <200306282323.SAA13605@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/ConstProp:

2002-05-03-NotOperator.ll updated: 1.2 -> 1.3
2002-09-03-SetCC-Bools.ll updated: 1.1 -> 1.2
basictest.ll updated: 1.2 -> 1.3
logicaltest.ll updated: 1.1 -> 1.2
nottest.ll updated: 1.3 -> 1.4
phi.ll updated: 1.1 -> 1.2
remtest.ll updated: 1.3 -> 1.4

---
Log message:

Remove explicit control flow through the use of the 'not' script



---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/ConstProp/2002-05-03-NotOperator.ll
diff -u llvm/test/Regression/Transforms/ConstProp/2002-05-03-NotOperator.ll:1.2 llvm/test/Regression/Transforms/ConstProp/2002-05-03-NotOperator.ll:1.3
--- llvm/test/Regression/Transforms/ConstProp/2002-05-03-NotOperator.ll:1.2	Wed Aug 14 14:28:09 2002
+++ llvm/test/Regression/Transforms/ConstProp/2002-05-03-NotOperator.ll	Sat Jun 28 18:23:17 2003
@@ -4,10 +4,7 @@
 
 ; Fix #2: The unary not instruction now no longer exists. Change to xor.
 
-; RUN: if as < %s | opt -constprop | dis | grep 'int 0'
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -constprop | dis | not grep 'int 0'
 
 int "test1"() {
 	%R = xor int 123, -1


Index: llvm/test/Regression/Transforms/ConstProp/2002-09-03-SetCC-Bools.ll
diff -u llvm/test/Regression/Transforms/ConstProp/2002-09-03-SetCC-Bools.ll:1.1 llvm/test/Regression/Transforms/ConstProp/2002-09-03-SetCC-Bools.ll:1.2
--- llvm/test/Regression/Transforms/ConstProp/2002-09-03-SetCC-Bools.ll:1.1	Tue Sep  3 15:09:23 2002
+++ llvm/test/Regression/Transforms/ConstProp/2002-09-03-SetCC-Bools.ll	Sat Jun 28 18:23:17 2003
@@ -1,9 +1,6 @@
 ; SetCC on boolean values was not implemented!
 
-; RUN: if as < %s | opt -constprop -die | dis | grep 'set'
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -constprop -die | dis | not grep 'set'
 
 bool "test1"() {
 	%A = setle bool true, false


Index: llvm/test/Regression/Transforms/ConstProp/basictest.ll
diff -u llvm/test/Regression/Transforms/ConstProp/basictest.ll:1.2 llvm/test/Regression/Transforms/ConstProp/basictest.ll:1.3
--- llvm/test/Regression/Transforms/ConstProp/basictest.ll:1.2	Tue May  7 15:25:43 2002
+++ llvm/test/Regression/Transforms/ConstProp/basictest.ll	Sat Jun 28 18:23:17 2003
@@ -1,13 +1,9 @@
 ; This is a basic sanity check for constant propogation.  The add instruction 
 ; should be eliminated.
 
-; RUN: if as < %s | opt -constprop -die | dis | grep add
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -constprop -die | dis | not grep add
 
-int "test"(bool %B)
-begin
+int %test(bool %B) {
 	br bool %B, label %BB1, label %BB2
 BB1:
 	%Val = add int 0, 0
@@ -17,4 +13,4 @@
 BB3:
 	%Ret = phi int [%Val, %BB1], [1, %BB2]
 	ret int %Ret
-end
+}


Index: llvm/test/Regression/Transforms/ConstProp/logicaltest.ll
diff -u llvm/test/Regression/Transforms/ConstProp/logicaltest.ll:1.1 llvm/test/Regression/Transforms/ConstProp/logicaltest.ll:1.2
--- llvm/test/Regression/Transforms/ConstProp/logicaltest.ll:1.1	Tue Jul 30 11:18:38 2002
+++ llvm/test/Regression/Transforms/ConstProp/logicaltest.ll	Sat Jun 28 18:23:17 2003
@@ -1,9 +1,6 @@
 ; Ensure constant propogation of logical instructions is working correctly.
 
-; RUN: if as < %s | opt -constprop -die | dis | ggrep -E 'and|or|xor'
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -constprop -die | dis | not ggrep -E 'and|or|xor'
 
 
 int  "test1"() { %R = and int 4,1234          ret int  %R }


Index: llvm/test/Regression/Transforms/ConstProp/nottest.ll
diff -u llvm/test/Regression/Transforms/ConstProp/nottest.ll:1.3 llvm/test/Regression/Transforms/ConstProp/nottest.ll:1.4
--- llvm/test/Regression/Transforms/ConstProp/nottest.ll:1.3	Wed Aug 14 14:26:34 2002
+++ llvm/test/Regression/Transforms/ConstProp/nottest.ll	Sat Jun 28 18:23:17 2003
@@ -1,9 +1,6 @@
 ; Ensure constant propogation of 'not' instructions is working correctly.
 
-; RUN: if as < %s | opt -constprop -die | dis | grep xor
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -constprop -die | dis | not grep xor
 
 int "test1"() {
 	%R = xor int 4, -1


Index: llvm/test/Regression/Transforms/ConstProp/phi.ll
diff -u llvm/test/Regression/Transforms/ConstProp/phi.ll:1.1 llvm/test/Regression/Transforms/ConstProp/phi.ll:1.2
--- llvm/test/Regression/Transforms/ConstProp/phi.ll:1.1	Tue May  7 15:32:19 2002
+++ llvm/test/Regression/Transforms/ConstProp/phi.ll	Sat Jun 28 18:23:17 2003
@@ -1,13 +1,9 @@
 ; This is a basic sanity check for constant propogation.  The add instruction 
 ; should be eliminated.
 
-; RUN: if as < %s | opt -constprop -die | dis | grep phi
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -constprop -die | dis | not grep phi
 
-int "test"(bool %B)
-begin
+int %test(bool %B) {
 BB0:
 	br bool %B, label %BB1, label %BB3
 BB1:
@@ -15,4 +11,4 @@
 BB3:
 	%Ret = phi int [1, %BB0], [1, %BB1]
 	ret int %Ret
-end
+}


Index: llvm/test/Regression/Transforms/ConstProp/remtest.ll
diff -u llvm/test/Regression/Transforms/ConstProp/remtest.ll:1.3 llvm/test/Regression/Transforms/ConstProp/remtest.ll:1.4
--- llvm/test/Regression/Transforms/ConstProp/remtest.ll:1.3	Tue May  7 15:25:44 2002
+++ llvm/test/Regression/Transforms/ConstProp/remtest.ll	Sat Jun 28 18:23:17 2003
@@ -1,26 +1,23 @@
-; Ensure constant propogation of remainder instructions is working correctly.
+; Ensure constant propagation of remainder instructions is working correctly.
 
-; RUN: if as < %s | opt -constprop -die | dis | grep rem
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -constprop -die | dis | not grep rem
 
-int "test1"() {
+int %test1() {
 	%R = rem int 4, 3
 	ret int %R
 }
 
-int "test2"() {
+int %test2() {
 	%R = rem int 123, -23
 	ret int %R
 }
 
-float "test3"() {
+float %test3() {
 	%R = rem float 12.45, 123.213
 	ret float %R
 }
 
-double "test4"() {
+double %test4() {
 	%R = rem double 312.20213123, 12.3333412
 	ret double %R
 }




From lattner at cs.uiuc.edu  Sat Jun 28 18:24:37 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:24:37 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/BasicAA/featuretest.ll gcsetest.ll licmtest.ll
Message-ID: <200306282323.SAA13588@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/BasicAA:

featuretest.ll updated: 1.7 -> 1.8
gcsetest.ll updated: 1.3 -> 1.4
licmtest.ll updated: 1.6 -> 1.7

---
Log message:

Remove explicit control flow through the use of the 'not' script


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/BasicAA/featuretest.ll
diff -u llvm/test/Regression/Transforms/BasicAA/featuretest.ll:1.7 llvm/test/Regression/Transforms/BasicAA/featuretest.ll:1.8
--- llvm/test/Regression/Transforms/BasicAA/featuretest.ll:1.7	Tue Jun 17 10:16:35 2003
+++ llvm/test/Regression/Transforms/BasicAA/featuretest.ll	Sat Jun 28 18:23:16 2003
@@ -1,10 +1,7 @@
 ; This testcase tests for various features the basicaa test should be able to 
 ; determine, as noted in the comments.
 
-; RUN: if as < %s | opt -basicaa -load-vn -gcse -instcombine -dce | dis | grep REMOVE
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -basicaa -load-vn -gcse -instcombine -dce | dis | not grep REMOVE
 
 %Global = external global { int }
 


Index: llvm/test/Regression/Transforms/BasicAA/gcsetest.ll
diff -u llvm/test/Regression/Transforms/BasicAA/gcsetest.ll:1.3 llvm/test/Regression/Transforms/BasicAA/gcsetest.ll:1.4
--- llvm/test/Regression/Transforms/BasicAA/gcsetest.ll:1.3	Sat Sep  7 17:50:41 2002
+++ llvm/test/Regression/Transforms/BasicAA/gcsetest.ll	Sat Jun 28 18:23:16 2003
@@ -2,10 +2,7 @@
 ; disambiguating some obvious cases.  All loads should be removable in 
 ; this testcase.
 
-; RUN: if as < %s | opt -basicaa -load-vn -gcse -instcombine -dce | dis | grep load
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -basicaa -load-vn -gcse -instcombine -dce | dis | not grep load
 
 %A = global int 7
 %B = global int 8


Index: llvm/test/Regression/Transforms/BasicAA/licmtest.ll
diff -u llvm/test/Regression/Transforms/BasicAA/licmtest.ll:1.6 llvm/test/Regression/Transforms/BasicAA/licmtest.ll:1.7
--- llvm/test/Regression/Transforms/BasicAA/licmtest.ll:1.6	Mon Mar  3 17:27:15 2003
+++ llvm/test/Regression/Transforms/BasicAA/licmtest.ll	Sat Jun 28 18:23:16 2003
@@ -3,10 +3,7 @@
 ; two pointers, then the load should be hoisted, and the store sunk.  Thus
 ; the loop becomes empty and can be deleted by ADCE. 
 
-; RUN: if as < %s | opt -basicaa -licm --adce | dis | grep Loop
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -basicaa -licm --adce | dis | not grep Loop
 
 %A = global int 7
 %B = global int 8




From lattner at cs.uiuc.edu  Sat Jun 28 18:24:49 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:24:49 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/ADCE/2003-06-24-BasicFunctionality.ll adcetest.ll
Message-ID: <200306282323.SAA13579@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/ADCE:

2003-06-24-BasicFunctionality.ll updated: 1.1 -> 1.2
adcetest.ll updated: 1.2 -> 1.3

---
Log message:

Remove explicit control flow through the use of the 'not' script



---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/ADCE/2003-06-24-BasicFunctionality.ll
diff -u llvm/test/Regression/Transforms/ADCE/2003-06-24-BasicFunctionality.ll:1.1 llvm/test/Regression/Transforms/ADCE/2003-06-24-BasicFunctionality.ll:1.2
--- llvm/test/Regression/Transforms/ADCE/2003-06-24-BasicFunctionality.ll:1.1	Tue Jun 24 16:22:09 2003
+++ llvm/test/Regression/Transforms/ADCE/2003-06-24-BasicFunctionality.ll	Sat Jun 28 18:23:15 2003
@@ -1,7 +1,4 @@
-; RUN: if as < %s | opt -adce -simplifycfg | dis | grep then:
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -adce -simplifycfg | dis | not grep then:
 
 void %dead_test8(int* %data.1, int %idx.1) {
 entry:		; No predecessors!


Index: llvm/test/Regression/Transforms/ADCE/adcetest.ll
diff -u llvm/test/Regression/Transforms/ADCE/adcetest.ll:1.2 llvm/test/Regression/Transforms/ADCE/adcetest.ll:1.3
--- llvm/test/Regression/Transforms/ADCE/adcetest.ll:1.2	Wed May 22 14:22:42 2002
+++ llvm/test/Regression/Transforms/ADCE/adcetest.ll	Sat Jun 28 18:23:15 2003
@@ -1,10 +1,7 @@
 ; This is the test case taken from Appel's book that illustrates a hard case
 ; that SCCP gets right, and when followed by ADCE, is completely eliminated
 ;
-; RUN: if as < %s | opt -sccp -adce -simplifycfg | dis | grep br
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -sccp -adce -simplifycfg | dis | not grep br
 
 int "test function"(int %i0, int %j0) {
 BB1:




From lattner at cs.uiuc.edu  Sat Jun 28 18:25:03 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:25:03 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Linker/2003-05-31-LinkerRename.ll
Message-ID: <200306282323.SAA13572@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Linker:

2003-05-31-LinkerRename.ll updated: 1.1 -> 1.2

---
Log message:

Remove explicit control flow through the use of the 'not' script


---
Diffs of the changes:

Index: llvm/test/Regression/Linker/2003-05-31-LinkerRename.ll
diff -u llvm/test/Regression/Linker/2003-05-31-LinkerRename.ll:1.1 llvm/test/Regression/Linker/2003-05-31-LinkerRename.ll:1.2
--- llvm/test/Regression/Linker/2003-05-31-LinkerRename.ll:1.1	Sat May 31 16:14:45 2003
+++ llvm/test/Regression/Linker/2003-05-31-LinkerRename.ll	Sat Jun 28 18:23:14 2003
@@ -5,10 +5,7 @@
 
 ; RUN: echo "implementation internal int %foo() { ret int 7 }" | as > Output/%s.1.bc
 ; RUN: as < %s > Output/%s.2.bc
-; RUN: if link Output/%s.[12].bc | dis | grep 'internal' | grep '%foo('
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: link Output/%s.[12].bc | dis | grep 'internal' | grep-not '%foo('
 
 implementation
 declare int %foo() 




From lattner at cs.uiuc.edu  Sat Jun 28 18:25:15 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:25:15 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Assembler/2002-04-29-NameBinding.llx ConstantExprFoldCast.llx
Message-ID: <200306282323.SAA13567@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Assembler:

2002-04-29-NameBinding.llx updated: 1.1 -> 1.2
ConstantExprFoldCast.llx updated: 1.1 -> 1.2

---
Log message:

Remove explicit control flow through the use of the 'not' script



---
Diffs of the changes:

Index: llvm/test/Regression/Assembler/2002-04-29-NameBinding.llx
diff -u llvm/test/Regression/Assembler/2002-04-29-NameBinding.llx:1.1 llvm/test/Regression/Assembler/2002-04-29-NameBinding.llx:1.2
--- llvm/test/Regression/Assembler/2002-04-29-NameBinding.llx:1.1	Mon Apr 29 13:12:28 2002
+++ llvm/test/Regression/Assembler/2002-04-29-NameBinding.llx	Sat Jun 28 18:23:13 2003
@@ -4,10 +4,7 @@
 ; Check by running globaldce, which will remove the constant if there are
 ; no references to it!
 ; 
-; RUN: if as < %s | opt -globaldce | dis | grep constant
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -globaldce | dis | grep-not constant
 ;
 	
 %v1 = internal constant int 5
@@ -15,7 +12,6 @@
 implementation
 
 int "createtask"()
-begin
 	%v1 = alloca int           ;; Alloca should have one use!
 	%reg112 = load int* %v1    ;; This load should not use the global!
 	ret int %reg112


Index: llvm/test/Regression/Assembler/ConstantExprFoldCast.llx
diff -u llvm/test/Regression/Assembler/ConstantExprFoldCast.llx:1.1 llvm/test/Regression/Assembler/ConstantExprFoldCast.llx:1.2
--- llvm/test/Regression/Assembler/ConstantExprFoldCast.llx:1.1	Thu Apr 17 14:21:22 2003
+++ llvm/test/Regression/Assembler/ConstantExprFoldCast.llx	Sat Jun 28 18:23:13 2003
@@ -1,9 +1,6 @@
 ; This test checks to make sure that constant exprs fold in some simple situations
 
-; RUN: if as < %s | dis | grep cast
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | dis | grep-not cast
 
 %A = global int* cast (sbyte* null to int*)  ; Cast null -> fold
 %B = global int** cast (int** %A to int**)   ; Cast to same type -> fold




From lattner at cs.uiuc.edu  Sat Jun 28 18:25:28 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:25:28 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Analysis/DSGraph/gcsetest.ll
Message-ID: <200306282323.SAA13560@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Analysis/DSGraph:

gcsetest.ll updated: 1.3 -> 1.4

---
Log message:

Remove explicit control flow through the use of the 'not' script


---
Diffs of the changes:

Index: llvm/test/Regression/Analysis/DSGraph/gcsetest.ll
diff -u llvm/test/Regression/Analysis/DSGraph/gcsetest.ll:1.3 llvm/test/Regression/Analysis/DSGraph/gcsetest.ll:1.4
--- llvm/test/Regression/Analysis/DSGraph/gcsetest.ll:1.3	Mon Feb 24 14:30:14 2003
+++ llvm/test/Regression/Analysis/DSGraph/gcsetest.ll	Sat Jun 28 18:23:11 2003
@@ -1,10 +1,7 @@
 ; Test that GCSE uses ds-aa to do alias analysis, which is capable of 
 ; disambiguating some cases.
 
-; RUN: if as < %s | opt -ds-aa -load-vn -gcse -instcombine -dce | dis | grep ELIM
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -ds-aa -load-vn -gcse -instcombine -dce | dis | not grep ELIM
 
 %intpair = type {int*, int*}
 implementation




From lattner at cs.uiuc.edu  Sat Jun 28 18:25:41 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:25:41 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/ScalarRepl/arraytest.ll basictest.ll
Message-ID: <200306282323.SAA13552@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/ScalarRepl:

arraytest.ll updated: 1.1 -> 1.2
basictest.ll updated: 1.1 -> 1.2

---
Log message:

Remove explicit control flow through the use of the 'not' script


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/ScalarRepl/arraytest.ll
diff -u llvm/test/Regression/Transforms/ScalarRepl/arraytest.ll:1.1 llvm/test/Regression/Transforms/ScalarRepl/arraytest.ll:1.2
--- llvm/test/Regression/Transforms/ScalarRepl/arraytest.ll:1.1	Tue May 27 10:51:17 2003
+++ llvm/test/Regression/Transforms/ScalarRepl/arraytest.ll	Sat Jun 28 18:23:34 2003
@@ -1,7 +1,4 @@
-; RUN: if as < %s | opt -scalarrepl -mem2reg | dis | grep alloca
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -scalarrepl -mem2reg | dis | not grep alloca
 
 int %test() {
   %X = alloca [ 4 x int ]


Index: llvm/test/Regression/Transforms/ScalarRepl/basictest.ll
diff -u llvm/test/Regression/Transforms/ScalarRepl/basictest.ll:1.1 llvm/test/Regression/Transforms/ScalarRepl/basictest.ll:1.2
--- llvm/test/Regression/Transforms/ScalarRepl/basictest.ll:1.1	Tue May 27 10:51:17 2003
+++ llvm/test/Regression/Transforms/ScalarRepl/basictest.ll	Sat Jun 28 18:23:34 2003
@@ -1,7 +1,4 @@
-; RUN: if as < %s | opt -scalarrepl -mem2reg | dis | grep alloca
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -scalarrepl -mem2reg | dis | not grep alloca
 
 int %test() {
   %X = alloca { int, float }




From lattner at cs.uiuc.edu  Sat Jun 28 18:25:53 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:25:53 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/SCCP/2002-05-02-EdgeFailure.ll 2002-05-02-MissSecondInst.ll 2002-08-30-GetElementPtrTest.ll 2003-06-24-OverdefinedPHIValue.ll basictest.ll sccptest.ll
Message-ID: <200306282323.SAA13543@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/SCCP:

2002-05-02-EdgeFailure.ll updated: 1.1 -> 1.2
2002-05-02-MissSecondInst.ll updated: 1.1 -> 1.2
2002-08-30-GetElementPtrTest.ll updated: 1.1 -> 1.2
2003-06-24-OverdefinedPHIValue.ll updated: 1.1 -> 1.2
basictest.ll updated: 1.1 -> 1.2
sccptest.ll updated: 1.3 -> 1.4

---
Log message:

Remove explicit control flow through the use of the 'not' script



---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/SCCP/2002-05-02-EdgeFailure.ll
diff -u llvm/test/Regression/Transforms/SCCP/2002-05-02-EdgeFailure.ll:1.1 llvm/test/Regression/Transforms/SCCP/2002-05-02-EdgeFailure.ll:1.2
--- llvm/test/Regression/Transforms/SCCP/2002-05-02-EdgeFailure.ll:1.1	Thu May  2 15:41:39 2002
+++ llvm/test/Regression/Transforms/SCCP/2002-05-02-EdgeFailure.ll	Sat Jun 28 18:23:33 2003
@@ -5,13 +5,9 @@
 ; real benchmark (mst from Olden benchmark, MakeGraph function).  When SCCP is
 ; fixed, this should be eliminated by a single SCCP application.
 ;
-; RUN: if as < %s | opt -sccp | dis | grep loop
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -sccp | dis | not grep loop
 
-int *"test"()
-begin
+int* %test() {
 bb1:
 	%A = malloc int
 	br label %bb2
@@ -24,5 +20,4 @@
 
 bb3:
 	ret int * %A
-end
-
+}


Index: llvm/test/Regression/Transforms/SCCP/2002-05-02-MissSecondInst.ll
diff -u llvm/test/Regression/Transforms/SCCP/2002-05-02-MissSecondInst.ll:1.1 llvm/test/Regression/Transforms/SCCP/2002-05-02-MissSecondInst.ll:1.2
--- llvm/test/Regression/Transforms/SCCP/2002-05-02-MissSecondInst.ll:1.1	Thu May  2 15:18:36 2002
+++ llvm/test/Regression/Transforms/SCCP/2002-05-02-MissSecondInst.ll	Sat Jun 28 18:23:33 2003
@@ -1,12 +1,7 @@
-; RUN: if as < %s | opt -sccp | dis | grep sub
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -sccp | dis | not grep sub
 
-void "test3"(int, int)
-begin
+void %test3(int, int) {
 	add int 0, 0
         sub int 0, 4
         ret void
-end
-
+}


Index: llvm/test/Regression/Transforms/SCCP/2002-08-30-GetElementPtrTest.ll
diff -u llvm/test/Regression/Transforms/SCCP/2002-08-30-GetElementPtrTest.ll:1.1 llvm/test/Regression/Transforms/SCCP/2002-08-30-GetElementPtrTest.ll:1.2
--- llvm/test/Regression/Transforms/SCCP/2002-08-30-GetElementPtrTest.ll:1.1	Fri Aug 30 18:22:15 2002
+++ llvm/test/Regression/Transforms/SCCP/2002-08-30-GetElementPtrTest.ll	Sat Jun 28 18:23:33 2003
@@ -1,7 +1,4 @@
-; RUN: if as < %s | opt -sccp | dis | grep '%X'
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -sccp | dis | not grep '%X'
 
 %G = uninitialized global [40x int]
 


Index: llvm/test/Regression/Transforms/SCCP/2003-06-24-OverdefinedPHIValue.ll
diff -u llvm/test/Regression/Transforms/SCCP/2003-06-24-OverdefinedPHIValue.ll:1.1 llvm/test/Regression/Transforms/SCCP/2003-06-24-OverdefinedPHIValue.ll:1.2
--- llvm/test/Regression/Transforms/SCCP/2003-06-24-OverdefinedPHIValue.ll:1.1	Tue Jun 24 15:29:36 2003
+++ llvm/test/Regression/Transforms/SCCP/2003-06-24-OverdefinedPHIValue.ll	Sat Jun 28 18:23:33 2003
@@ -1,7 +1,4 @@
-; RUN: if as < %s | opt -sccp -simplifycfg | dis | grep then:
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -sccp -simplifycfg | dis | not grep then:
 
 void %cprop_test11(int* %data.1) {
 entry:		; No predecessors!


Index: llvm/test/Regression/Transforms/SCCP/basictest.ll
diff -u llvm/test/Regression/Transforms/SCCP/basictest.ll:1.1 llvm/test/Regression/Transforms/SCCP/basictest.ll:1.2
--- llvm/test/Regression/Transforms/SCCP/basictest.ll:1.1	Thu May  2 15:41:39 2002
+++ llvm/test/Regression/Transforms/SCCP/basictest.ll	Sat Jun 28 18:23:33 2003
@@ -1,13 +1,9 @@
 ; This is a basic sanity check for constant propogation.  The add instruction 
 ; should be eliminated.
 
-; RUN: if as < %s | opt -sccp | dis | grep add
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -sccp | dis | not grep add
 
-int "test"(bool %B)
-begin
+int %test(bool %B) {
 	br bool %B, label %BB1, label %BB2
 BB1:
 	%Val = add int 0, 0
@@ -17,4 +13,4 @@
 BB3:
 	%Ret = phi int [%Val, %BB1], [1, %BB2]
 	ret int %Ret
-end
+}


Index: llvm/test/Regression/Transforms/SCCP/sccptest.ll
diff -u llvm/test/Regression/Transforms/SCCP/sccptest.ll:1.3 llvm/test/Regression/Transforms/SCCP/sccptest.ll:1.4
--- llvm/test/Regression/Transforms/SCCP/sccptest.ll:1.3	Wed May 22 10:49:14 2002
+++ llvm/test/Regression/Transforms/SCCP/sccptest.ll	Sat Jun 28 18:23:33 2003
@@ -1,13 +1,9 @@
 ; This is the test case taken from appel's book that illustrates a hard case
 ; that SCCP gets right. BB3 should be completely eliminated.
 ;
-; RUN: if as < %s | opt -sccp -constprop -dce -cfgsimplify | dis | grep BB3
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -sccp -constprop -dce -cfgsimplify | dis | not grep BB3
 
-int "test function"(int %i0, int %j0)
-begin
+int %test function(int %i0, int %j0) {
 BB1:
 	br label %BB2
 BB2:
@@ -35,5 +31,4 @@
 	%j4 = phi int [1, %BB5], [%k2, %BB6]
 	%k4 = phi int [%k3, %BB5], [%k5, %BB6]
 	br label %BB2
-end
-
+}




From lattner at cs.uiuc.edu  Sat Jun 28 18:26:06 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:26:06 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/Reassociate/2002-05-15-AgressiveSubMove.ll 2002-05-15-MissedTree.ll 2002-05-15-SubReassociate.ll 2002-05-15-SubReassociate2.ll basictest.ll basictest2.ll looptest.ll otherops.ll subtest.ll
Message-ID: <200306282323.SAA13525@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/Reassociate:

2002-05-15-AgressiveSubMove.ll updated: 1.2 -> 1.3
2002-05-15-MissedTree.ll updated: 1.1 -> 1.2
2002-05-15-SubReassociate.ll updated: 1.2 -> 1.3
2002-05-15-SubReassociate2.ll updated: 1.1 -> 1.2
basictest.ll updated: 1.1 -> 1.2
basictest2.ll updated: 1.1 -> 1.2
looptest.ll updated: 1.2 -> 1.3
otherops.ll updated: 1.1 -> 1.2
subtest.ll updated: 1.2 -> 1.3

---
Log message:

Remove explicit control flow through the use of the 'not' script


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/Reassociate/2002-05-15-AgressiveSubMove.ll
diff -u llvm/test/Regression/Transforms/Reassociate/2002-05-15-AgressiveSubMove.ll:1.2 llvm/test/Regression/Transforms/Reassociate/2002-05-15-AgressiveSubMove.ll:1.3
--- llvm/test/Regression/Transforms/Reassociate/2002-05-15-AgressiveSubMove.ll:1.2	Wed May 15 19:56:59 2002
+++ llvm/test/Regression/Transforms/Reassociate/2002-05-15-AgressiveSubMove.ll	Sat Jun 28 18:23:31 2003
@@ -1,9 +1,6 @@
-; RUN: if as < %s | opt -reassociate -instcombine -constprop -dce | dis | grep add
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -reassociate -instcombine -constprop -dce | dis | not grep add
 
-int "test"(int %A) {
+int %test(int %A) {
 	%X = add int %A, 1
 	%Y = add int %A, 1
 	%r = sub int %X, %Y


Index: llvm/test/Regression/Transforms/Reassociate/2002-05-15-MissedTree.ll
diff -u llvm/test/Regression/Transforms/Reassociate/2002-05-15-MissedTree.ll:1.1 llvm/test/Regression/Transforms/Reassociate/2002-05-15-MissedTree.ll:1.2
--- llvm/test/Regression/Transforms/Reassociate/2002-05-15-MissedTree.ll:1.1	Wed May 15 16:56:51 2002
+++ llvm/test/Regression/Transforms/Reassociate/2002-05-15-MissedTree.ll	Sat Jun 28 18:23:31 2003
@@ -1,9 +1,6 @@
-; RUN: if as < %s | opt -reassociate -instcombine -constprop -die | dis | grep 5
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -reassociate -instcombine -constprop -die | dis | not grep 5
 
-int "test"(int %A, int %B) {
+int %test(int %A, int %B) {
         %W = add int %B, -5
         %Y = add int %A, 5
         %Z = add int %W, %Y


Index: llvm/test/Regression/Transforms/Reassociate/2002-05-15-SubReassociate.ll
diff -u llvm/test/Regression/Transforms/Reassociate/2002-05-15-SubReassociate.ll:1.2 llvm/test/Regression/Transforms/Reassociate/2002-05-15-SubReassociate.ll:1.3
--- llvm/test/Regression/Transforms/Reassociate/2002-05-15-SubReassociate.ll:1.2	Wed May 15 19:56:36 2002
+++ llvm/test/Regression/Transforms/Reassociate/2002-05-15-SubReassociate.ll	Sat Jun 28 18:23:31 2003
@@ -1,11 +1,8 @@
 ; With sub reassociation, constant folding can eliminate all of the constants.
 ;
-; RUN: if as < %s | opt -reassociate -constprop -instcombine -dce | dis | grep add
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -reassociate -constprop -instcombine -dce | dis | not grep add
 
-int "test"(int %A, int %B) {
+int %test(int %A, int %B) {
 	%W = add int 5, %B
 	%X = add int -7, %A
 	%Y = sub int %X, %W


Index: llvm/test/Regression/Transforms/Reassociate/2002-05-15-SubReassociate2.ll
diff -u llvm/test/Regression/Transforms/Reassociate/2002-05-15-SubReassociate2.ll:1.1 llvm/test/Regression/Transforms/Reassociate/2002-05-15-SubReassociate2.ll:1.2
--- llvm/test/Regression/Transforms/Reassociate/2002-05-15-SubReassociate2.ll:1.1	Wed May 15 16:59:31 2002
+++ llvm/test/Regression/Transforms/Reassociate/2002-05-15-SubReassociate2.ll	Sat Jun 28 18:23:31 2003
@@ -1,9 +1,6 @@
 ; With sub reassociation, constant folding can eliminate the two 12 constants.
 ;
-; RUN: if as < %s | opt -reassociate -constprop -dce | dis | grep 12
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -reassociate -constprop -dce | dis | not grep 12
 
 int "test"(int %A, int %B, int %C, int %D) {
 	%M = add int %A, 12


Index: llvm/test/Regression/Transforms/Reassociate/basictest.ll
diff -u llvm/test/Regression/Transforms/Reassociate/basictest.ll:1.1 llvm/test/Regression/Transforms/Reassociate/basictest.ll:1.2
--- llvm/test/Regression/Transforms/Reassociate/basictest.ll:1.1	Wed May  8 16:34:21 2002
+++ llvm/test/Regression/Transforms/Reassociate/basictest.ll	Sat Jun 28 18:23:31 2003
@@ -1,11 +1,8 @@
 ; With reassociation, constant folding can eliminate the 12 and -12 constants.
 ;
-; RUN: if as < %s | opt -reassociate -constprop -instcombine -die | dis | grep add
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -reassociate -constprop -instcombine -die | dis | not grep add
 
-int "test"(int %arg) {
+int %test(int %arg) {
 	%tmp1 = sub int -12, %arg
 	%tmp2 = add int %tmp1, 12
 	ret int %tmp2


Index: llvm/test/Regression/Transforms/Reassociate/basictest2.ll
diff -u llvm/test/Regression/Transforms/Reassociate/basictest2.ll:1.1 llvm/test/Regression/Transforms/Reassociate/basictest2.ll:1.2
--- llvm/test/Regression/Transforms/Reassociate/basictest2.ll:1.1	Wed May  8 16:34:21 2002
+++ llvm/test/Regression/Transforms/Reassociate/basictest2.ll	Sat Jun 28 18:23:31 2003
@@ -1,9 +1,6 @@
 ; With reassociation, constant folding can eliminate the +/- 30 constants.
 ;
-; RUN: if as < %s | opt -reassociate -constprop -instcombine -die | dis | grep 30
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -reassociate -constprop -instcombine -die | dis | not grep 30
 
 int "test"(int %reg109, int %reg1111) {
         %reg115 = add int %reg109, -30           ; <int> [#uses=1]


Index: llvm/test/Regression/Transforms/Reassociate/looptest.ll
diff -u llvm/test/Regression/Transforms/Reassociate/looptest.ll:1.2 llvm/test/Regression/Transforms/Reassociate/looptest.ll:1.3
--- llvm/test/Regression/Transforms/Reassociate/looptest.ll:1.2	Tue Oct 15 16:21:42 2002
+++ llvm/test/Regression/Transforms/Reassociate/looptest.ll	Sat Jun 28 18:23:31 2003
@@ -12,10 +12,7 @@
 ; In this case, we want to reassociate the specified expr so that i+j can be
 ; hoisted out of the inner most loop.
 ;
-; RUN: if as < %s | opt -reassociate | dis | grep 115 | grep 117
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -reassociate | dis | grep 115 | not grep 117
 
 %.LC0 = internal global [4 x sbyte] c"%d\0A\00"		; <[4 x sbyte]*> [#uses=1]
 


Index: llvm/test/Regression/Transforms/Reassociate/otherops.ll
diff -u llvm/test/Regression/Transforms/Reassociate/otherops.ll:1.1 llvm/test/Regression/Transforms/Reassociate/otherops.ll:1.2
--- llvm/test/Regression/Transforms/Reassociate/otherops.ll:1.1	Wed May  8 16:34:22 2002
+++ llvm/test/Regression/Transforms/Reassociate/otherops.ll	Sat Jun 28 18:23:31 2003
@@ -1,9 +1,6 @@
 ; Reassociation should apply to Add, Mul, And, Or, & Xor
 ;
-; RUN: if as < %s | opt -reassociate -constprop -instcombine -die | dis | grep 12
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -reassociate -constprop -instcombine -die | dis | not grep 12
 
 int "test_mul"(int %arg) {
         %tmp1 = mul int 12, %arg


Index: llvm/test/Regression/Transforms/Reassociate/subtest.ll
diff -u llvm/test/Regression/Transforms/Reassociate/subtest.ll:1.2 llvm/test/Regression/Transforms/Reassociate/subtest.ll:1.3
--- llvm/test/Regression/Transforms/Reassociate/subtest.ll:1.2	Wed May 15 19:56:59 2002
+++ llvm/test/Regression/Transforms/Reassociate/subtest.ll	Sat Jun 28 18:23:31 2003
@@ -1,9 +1,6 @@
 ; With sub reassociation, constant folding can eliminate the 12 and -12 constants.
 ;
-; RUN: if as < %s | opt -reassociate -constprop -instcombine -die | dis | grep 12
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -reassociate -constprop -instcombine -die | dis | not grep 12
 
 int "test"(int %A, int %B) {
 	%X = add int -12, %A




From lattner at cs.uiuc.edu  Sat Jun 28 18:26:20 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:26:20 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/PiNodeInserter/substitutetest.ll
Message-ID: <200306282323.SAA13502@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/PiNodeInserter:

substitutetest.ll updated: 1.1 -> 1.2

---
Log message:

Remove explicit control flow through the use of the 'not' script



---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/PiNodeInserter/substitutetest.ll
diff -u llvm/test/Regression/Transforms/PiNodeInserter/substitutetest.ll:1.1 llvm/test/Regression/Transforms/PiNodeInserter/substitutetest.ll:1.2
--- llvm/test/Regression/Transforms/PiNodeInserter/substitutetest.ll:1.1	Fri May 10 00:38:17 2002
+++ llvm/test/Regression/Transforms/PiNodeInserter/substitutetest.ll	Sat Jun 28 18:23:29 2003
@@ -1,7 +1,4 @@
-; RUN: if as < %s | opt -pinodes -instcombine -die | dis | grep add
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -pinodes -instcombine -die | dis | not grep add
 
 int "test"(int %i, int %j) {
 	%c = seteq int %i, 0




From lattner at cs.uiuc.edu  Sat Jun 28 18:26:32 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:26:32 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/Mem2Reg/2003-06-26-IterativePromote.ll DifferingTypes.ll PromoteMemToRegister.ll
Message-ID: <200306282323.SAA13495@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/Mem2Reg:

2003-06-26-IterativePromote.ll updated: 1.1 -> 1.2
DifferingTypes.ll updated: 1.1 -> 1.2
PromoteMemToRegister.ll updated: 1.1 -> 1.2

---
Log message:

Remove explicit control flow through the use of the 'not' script


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/Mem2Reg/2003-06-26-IterativePromote.ll
diff -u llvm/test/Regression/Transforms/Mem2Reg/2003-06-26-IterativePromote.ll:1.1 llvm/test/Regression/Transforms/Mem2Reg/2003-06-26-IterativePromote.ll:1.2
--- llvm/test/Regression/Transforms/Mem2Reg/2003-06-26-IterativePromote.ll:1.1	Wed Jun 25 09:55:30 2003
+++ llvm/test/Regression/Transforms/Mem2Reg/2003-06-26-IterativePromote.ll	Sat Jun 28 18:23:26 2003
@@ -1,8 +1,5 @@
 ; Promoting some values allows promotion of other values.
-; RUN: if as < %s | opt -mem2reg | dis | grep alloca
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -mem2reg | dis | not grep alloca
 
 int %test2() {
         %result = alloca int             ; ty=int*


Index: llvm/test/Regression/Transforms/Mem2Reg/DifferingTypes.ll
diff -u llvm/test/Regression/Transforms/Mem2Reg/DifferingTypes.ll:1.1 llvm/test/Regression/Transforms/Mem2Reg/DifferingTypes.ll:1.2
--- llvm/test/Regression/Transforms/Mem2Reg/DifferingTypes.ll:1.1	Fri Feb 28 13:20:56 2003
+++ llvm/test/Regression/Transforms/Mem2Reg/DifferingTypes.ll	Sat Jun 28 18:23:26 2003
@@ -2,10 +2,7 @@
 ; generated code should perform the appropriate masking operations required 
 ; depending on the endianness of the target...
 
-; RUN: if as < %s | opt -mem2reg | dis | grep 'alloca'
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -mem2reg | dis | not grep 'alloca'
 
 implementation
 


Index: llvm/test/Regression/Transforms/Mem2Reg/PromoteMemToRegister.ll
diff -u llvm/test/Regression/Transforms/Mem2Reg/PromoteMemToRegister.ll:1.1 llvm/test/Regression/Transforms/Mem2Reg/PromoteMemToRegister.ll:1.2
--- llvm/test/Regression/Transforms/Mem2Reg/PromoteMemToRegister.ll:1.1	Tue Apr 30 14:50:47 2002
+++ llvm/test/Regression/Transforms/Mem2Reg/PromoteMemToRegister.ll	Sat Jun 28 18:23:26 2003
@@ -1,8 +1,5 @@
 ; Simple sanity check testcase.  Both alloca's should be eliminated.
-; RUN: if as < %s | opt -mem2reg | dis | grep 'alloca'
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -mem2reg | dis | not grep 'alloca'
 
 implementation
 




From lattner at cs.uiuc.edu  Sat Jun 28 18:26:45 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:26:45 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/LICM/2003-02-27-PreheaderProblem.ll
Message-ID: <200306282323.SAA13484@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/LICM:

2003-02-27-PreheaderProblem.ll updated: 1.1 -> 1.2

---
Log message:

Remove explicit control flow through the use of the 'not' script



---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/LICM/2003-02-27-PreheaderProblem.ll
diff -u llvm/test/Regression/Transforms/LICM/2003-02-27-PreheaderProblem.ll:1.1 llvm/test/Regression/Transforms/LICM/2003-02-27-PreheaderProblem.ll:1.2
--- llvm/test/Regression/Transforms/LICM/2003-02-27-PreheaderProblem.ll:1.1	Thu Feb 27 15:38:56 2003
+++ llvm/test/Regression/Transforms/LICM/2003-02-27-PreheaderProblem.ll	Sat Jun 28 18:23:25 2003
@@ -3,10 +3,7 @@
 ; happens because preheader insertion doesn't insert a preheader for this
 ; case... bad.
 
-; RUN: if as < %s | opt -licm -adce -simplifycfg | dis | grep 'br '
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -licm -adce -simplifycfg | dis | not grep 'br '
 
 int %main(int %argc) {
         br label %bb5




From lattner at cs.uiuc.edu  Sat Jun 28 18:27:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:27:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/FunctionResolve/2002-08-19-ResolveGlobalVars.ll 2002-08-19-ResolveGlobalVarsEasier.ll 2002-11-07-RetMismatch.ll 2002-11-09-ExternFn.ll 2003-04-18-ForwardDeclGlobal.ll 2003-05-31-AllInternalDecls.ll 2003-05-31-FuncPointerResolve.ll 2003-05-31-InternalDecl.ll basictest.ll retmismatch1.ll retmismatch2.ll
Message-ID: <200306282324.SAA13650@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/FunctionResolve:

2002-08-19-ResolveGlobalVars.ll updated: 1.4 -> 1.5
2002-08-19-ResolveGlobalVarsEasier.ll updated: 1.2 -> 1.3
2002-11-07-RetMismatch.ll updated: 1.2 -> 1.3
2002-11-09-ExternFn.ll updated: 1.2 -> 1.3
2003-04-18-ForwardDeclGlobal.ll updated: 1.2 -> 1.3
2003-05-31-AllInternalDecls.ll updated: 1.1 -> 1.2
2003-05-31-FuncPointerResolve.ll updated: 1.2 -> 1.3
2003-05-31-InternalDecl.ll updated: 1.1 -> 1.2
basictest.ll updated: 1.2 -> 1.3
retmismatch1.ll updated: 1.3 -> 1.4
retmismatch2.ll updated: 1.3 -> 1.4

---
Log message:

Remove explicit control flow through the use of the 'not' script


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/FunctionResolve/2002-08-19-ResolveGlobalVars.ll
diff -u llvm/test/Regression/Transforms/FunctionResolve/2002-08-19-ResolveGlobalVars.ll:1.4 llvm/test/Regression/Transforms/FunctionResolve/2002-08-19-ResolveGlobalVars.ll:1.5
--- llvm/test/Regression/Transforms/FunctionResolve/2002-08-19-ResolveGlobalVars.ll:1.4	Fri Apr 18 19:17:11 2003
+++ llvm/test/Regression/Transforms/FunctionResolve/2002-08-19-ResolveGlobalVars.ll	Sat Jun 28 18:23:21 2003
@@ -1,16 +1,7 @@
 ; Test that: extern int X[]  and int X[] = { 1, 2, 3, 4 } are resolved 
 ; correctly.
 ;
-; RUN: if as < %s | opt -funcresolve > /dev/null
-; RUN: then echo "opt ok"
-; RUN: else exit 1   # Make sure opt doesn't abort!
-; RUN: fi
-;
-; RUN: if as < %s | opt -funcresolve | dis | grep external
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
-;
+; RUN: as < %s | opt -funcresolve | dis | not grep external
 
 %X = external global [0 x int]
 %X = global [4 x int] [ int 1, int 2, int 3, int 4 ]


Index: llvm/test/Regression/Transforms/FunctionResolve/2002-08-19-ResolveGlobalVarsEasier.ll
diff -u llvm/test/Regression/Transforms/FunctionResolve/2002-08-19-ResolveGlobalVarsEasier.ll:1.2 llvm/test/Regression/Transforms/FunctionResolve/2002-08-19-ResolveGlobalVarsEasier.ll:1.3
--- llvm/test/Regression/Transforms/FunctionResolve/2002-08-19-ResolveGlobalVarsEasier.ll:1.2	Fri Apr 18 19:17:11 2003
+++ llvm/test/Regression/Transforms/FunctionResolve/2002-08-19-ResolveGlobalVarsEasier.ll	Sat Jun 28 18:23:21 2003
@@ -1,10 +1,7 @@
 ; Test that: extern int X[]  and int X[] = { 1, 2, 3, 4 } are resolved 
 ; correctly.  This doesn't have constantexprs
 ;
-; RUN: if as < %s | opt -funcresolve | dis | grep external
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -funcresolve | dis | not grep external
 ;
 
 %X = external global [0 x int]


Index: llvm/test/Regression/Transforms/FunctionResolve/2002-11-07-RetMismatch.ll
diff -u llvm/test/Regression/Transforms/FunctionResolve/2002-11-07-RetMismatch.ll:1.2 llvm/test/Regression/Transforms/FunctionResolve/2002-11-07-RetMismatch.ll:1.3
--- llvm/test/Regression/Transforms/FunctionResolve/2002-11-07-RetMismatch.ll:1.2	Thu Nov  7 18:31:27 2002
+++ llvm/test/Regression/Transforms/FunctionResolve/2002-11-07-RetMismatch.ll	Sat Jun 28 18:23:21 2003
@@ -1,7 +1,4 @@
-; RUN: if as < %s | opt -funcresolve -funcresolve | dis | grep declare
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -funcresolve -funcresolve | dis | not grep declare
 
 declare void %qsortg(sbyte*, int, int)
 


Index: llvm/test/Regression/Transforms/FunctionResolve/2002-11-09-ExternFn.ll
diff -u llvm/test/Regression/Transforms/FunctionResolve/2002-11-09-ExternFn.ll:1.2 llvm/test/Regression/Transforms/FunctionResolve/2002-11-09-ExternFn.ll:1.3
--- llvm/test/Regression/Transforms/FunctionResolve/2002-11-09-ExternFn.ll:1.2	Thu Jun 19 12:02:06 2003
+++ llvm/test/Regression/Transforms/FunctionResolve/2002-11-09-ExternFn.ll	Sat Jun 28 18:23:21 2003
@@ -1,12 +1,5 @@
-; RUN: if as < %s | opt -funcresolve > /dev/null
-; RUN: then echo "opt ok"
-; RUN: else exit 1   # Make sure opt doesn't abort!
-; RUN: fi
 ;
-; RUN: if as < %s | opt -funcresolve -instcombine | dis | grep '\.\.\.'
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -funcresolve -instcombine | dis | not grep '\.\.\.'
 
 declare int %foo(...)
 declare int %foo(int)


Index: llvm/test/Regression/Transforms/FunctionResolve/2003-04-18-ForwardDeclGlobal.ll
diff -u llvm/test/Regression/Transforms/FunctionResolve/2003-04-18-ForwardDeclGlobal.ll:1.2 llvm/test/Regression/Transforms/FunctionResolve/2003-04-18-ForwardDeclGlobal.ll:1.3
--- llvm/test/Regression/Transforms/FunctionResolve/2003-04-18-ForwardDeclGlobal.ll:1.2	Fri Apr 18 19:15:00 2003
+++ llvm/test/Regression/Transforms/FunctionResolve/2003-04-18-ForwardDeclGlobal.ll	Sat Jun 28 18:23:21 2003
@@ -1,7 +1,5 @@
-; RUN: if as < %s | opt -funcresolve -disable-output 2>&1 | grep WARNING
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -funcresolve -disable-output 2>&1 | not grep WARNING
+
 %__popcount_tab = external constant [0 x ubyte]
 %__popcount_tab = constant [4 x ubyte] c"\00\01\01\02"
 


Index: llvm/test/Regression/Transforms/FunctionResolve/2003-05-31-AllInternalDecls.ll
diff -u llvm/test/Regression/Transforms/FunctionResolve/2003-05-31-AllInternalDecls.ll:1.1 llvm/test/Regression/Transforms/FunctionResolve/2003-05-31-AllInternalDecls.ll:1.2
--- llvm/test/Regression/Transforms/FunctionResolve/2003-05-31-AllInternalDecls.ll:1.1	Sat May 31 16:33:08 2003
+++ llvm/test/Regression/Transforms/FunctionResolve/2003-05-31-AllInternalDecls.ll	Sat Jun 28 18:23:21 2003
@@ -1,9 +1,6 @@
 ; This testcase should not cause a warning!
 
-; RUN: if (as < %s | opt -funcresolve -disable-output) 2>&1 | grep 'WARNING'
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: (as < %s | opt -funcresolve -disable-output) 2>&1 | not grep 'WARNING'
 
 %X = internal global float 1.0
 %X = internal global int 1


Index: llvm/test/Regression/Transforms/FunctionResolve/2003-05-31-FuncPointerResolve.ll
diff -u llvm/test/Regression/Transforms/FunctionResolve/2003-05-31-FuncPointerResolve.ll:1.2 llvm/test/Regression/Transforms/FunctionResolve/2003-05-31-FuncPointerResolve.ll:1.3
--- llvm/test/Regression/Transforms/FunctionResolve/2003-05-31-FuncPointerResolve.ll:1.2	Sat May 31 15:33:09 2003
+++ llvm/test/Regression/Transforms/FunctionResolve/2003-05-31-FuncPointerResolve.ll	Sat Jun 28 18:23:21 2003
@@ -1,7 +1,5 @@
-; RUN: if as < %s | opt -funcresolve | dis | grep declare
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -funcresolve | dis | not grep declare
+
 %Table = constant int(...)* %foo
 
 %Table2 = constant [1 x int(...)* ] [ int(...)* %foo ]


Index: llvm/test/Regression/Transforms/FunctionResolve/2003-05-31-InternalDecl.ll
diff -u llvm/test/Regression/Transforms/FunctionResolve/2003-05-31-InternalDecl.ll:1.1 llvm/test/Regression/Transforms/FunctionResolve/2003-05-31-InternalDecl.ll:1.2
--- llvm/test/Regression/Transforms/FunctionResolve/2003-05-31-InternalDecl.ll:1.1	Sat May 31 16:04:39 2003
+++ llvm/test/Regression/Transforms/FunctionResolve/2003-05-31-InternalDecl.ll	Sat Jun 28 18:23:21 2003
@@ -1,7 +1,4 @@
-; RUN: if as < %s | opt -funcresolve | dis | grep declare
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -funcresolve | dis | not grep declare
 
 declare void %test(...)
 


Index: llvm/test/Regression/Transforms/FunctionResolve/basictest.ll
diff -u llvm/test/Regression/Transforms/FunctionResolve/basictest.ll:1.2 llvm/test/Regression/Transforms/FunctionResolve/basictest.ll:1.3
--- llvm/test/Regression/Transforms/FunctionResolve/basictest.ll:1.2	Thu Jun 19 12:02:06 2003
+++ llvm/test/Regression/Transforms/FunctionResolve/basictest.ll	Sat Jun 28 18:23:21 2003
@@ -1,7 +1,4 @@
-; RUN: if as < %s | opt -funcresolve -instcombine | dis | grep '\.\.\.' | grep call
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -funcresolve -instcombine | dis | grep '\.\.\.' | not grep call
 
 declare int %foo(...)
 


Index: llvm/test/Regression/Transforms/FunctionResolve/retmismatch1.ll
diff -u llvm/test/Regression/Transforms/FunctionResolve/retmismatch1.ll:1.3 llvm/test/Regression/Transforms/FunctionResolve/retmismatch1.ll:1.4
--- llvm/test/Regression/Transforms/FunctionResolve/retmismatch1.ll:1.3	Thu Jun 19 12:02:06 2003
+++ llvm/test/Regression/Transforms/FunctionResolve/retmismatch1.ll	Sat Jun 28 18:23:21 2003
@@ -1,14 +1,6 @@
 ; This shows where the function is called with the prototype indicating a
 ; return type exists, but it really doesn't.
-; RUN: if as < %s | opt -funcresolve > /dev/null
-; RUN: then echo "opt ok"
-; RUN: else exit 1   # Make sure opt doesn't abort!
-; RUN: fi
-;
-; RUN: if as < %s | opt -funcresolve -instcombine | dis | grep '\.\.\.' | grep call
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -funcresolve -instcombine | dis | grep '\.\.\.' | not grep call
 
 declare int %foo(...)
 


Index: llvm/test/Regression/Transforms/FunctionResolve/retmismatch2.ll
diff -u llvm/test/Regression/Transforms/FunctionResolve/retmismatch2.ll:1.3 llvm/test/Regression/Transforms/FunctionResolve/retmismatch2.ll:1.4
--- llvm/test/Regression/Transforms/FunctionResolve/retmismatch2.ll:1.3	Thu Jun 19 12:02:06 2003
+++ llvm/test/Regression/Transforms/FunctionResolve/retmismatch2.ll	Sat Jun 28 18:23:21 2003
@@ -1,15 +1,7 @@
 ; This shows where the function is called with the prototype indicating a
 ; return type doesn't exists, but it really does.
 ;
-; RUN: if as < %s | opt -funcresolve > /dev/null
-; RUN: then echo "opt ok"
-; RUN: else exit 1   # Make sure opt doesn't abort!
-; RUN: fi
-;
-; RUN: if as < %s | opt -funcresolve -instcombine | dis | grep '\.\.\.' | grep call
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -funcresolve -instcombine | dis | grep '\.\.\.' | not grep call
 
 declare void %foo(...)
 




From lattner at cs.uiuc.edu  Sat Jun 28 18:27:15 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:27:15 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/GCSE/2002-05-14-OperandSwap.ll 2003-06-13-LoadStoreEliminate.ll RLE-Eliminate.ll
Message-ID: <200306282324.SAA13663@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/GCSE:

2002-05-14-OperandSwap.ll updated: 1.2 -> 1.3
2003-06-13-LoadStoreEliminate.ll updated: 1.1 -> 1.2
RLE-Eliminate.ll updated: 1.2 -> 1.3

---
Log message:

Remove explicit control flow through the use of the 'not' script



---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/GCSE/2002-05-14-OperandSwap.ll
diff -u llvm/test/Regression/Transforms/GCSE/2002-05-14-OperandSwap.ll:1.2 llvm/test/Regression/Transforms/GCSE/2002-05-14-OperandSwap.ll:1.3
--- llvm/test/Regression/Transforms/GCSE/2002-05-14-OperandSwap.ll:1.2	Thu Aug 22 11:39:10 2002
+++ llvm/test/Regression/Transforms/GCSE/2002-05-14-OperandSwap.ll	Sat Jun 28 18:23:23 2003
@@ -1,10 +1,7 @@
 ; This entire chain of computation should be optimized away, but
 ; wasn't because the two multiplies were not detected as being identical.
 ;
-; RUN: if as < %s  | opt -gcse -instcombine -dce | dis | grep sub
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s  | opt -gcse -instcombine -dce | dis | not grep sub
 
 implementation   ; Functions:
 


Index: llvm/test/Regression/Transforms/GCSE/2003-06-13-LoadStoreEliminate.ll
diff -u llvm/test/Regression/Transforms/GCSE/2003-06-13-LoadStoreEliminate.ll:1.1 llvm/test/Regression/Transforms/GCSE/2003-06-13-LoadStoreEliminate.ll:1.2
--- llvm/test/Regression/Transforms/GCSE/2003-06-13-LoadStoreEliminate.ll:1.1	Mon Jun 16 07:05:38 2003
+++ llvm/test/Regression/Transforms/GCSE/2003-06-13-LoadStoreEliminate.ll	Sat Jun 28 18:23:23 2003
@@ -1,10 +1,7 @@
 ; This testcase shows a bug where an common subexpression exists, but there
 ; is no shared dominator block that the expression can be hoisted out to.
 ;
-; RUN: if as < %s | opt -load-vn -gcse | dis | grep load
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -load-vn -gcse | dis | not grep load
 
 int %test(int* %P) {
 	store int 5, int* %P


Index: llvm/test/Regression/Transforms/GCSE/RLE-Eliminate.ll
diff -u llvm/test/Regression/Transforms/GCSE/RLE-Eliminate.ll:1.2 llvm/test/Regression/Transforms/GCSE/RLE-Eliminate.ll:1.3
--- llvm/test/Regression/Transforms/GCSE/RLE-Eliminate.ll:1.2	Fri Aug 30 15:18:09 2002
+++ llvm/test/Regression/Transforms/GCSE/RLE-Eliminate.ll	Sat Jun 28 18:23:23 2003
@@ -1,10 +1,7 @@
 ; This testcase ensures that redundant loads are eliminated when they should 
 ; be.  All RL variables (redundant loads) should be eliminated.
 ;
-; RUN: if as < %s | opt -load-vn -gcse | dis | grep %RL
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -load-vn -gcse | dis | not grep %RL
 ;
 int "test1"(int* %P) {
 	%A = load int* %P




From lattner at cs.uiuc.edu  Sat Jun 28 18:27:27 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:27:27 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/GlobalDCE/2002-08-17-FunctionDGE.ll 2002-08-17-WorkListTest.ll basicvariabletest.ll
Message-ID: <200306282324.SAA13672@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/GlobalDCE:

2002-08-17-FunctionDGE.ll updated: 1.1 -> 1.2
2002-08-17-WorkListTest.ll updated: 1.2 -> 1.3
basicvariabletest.ll updated: 1.1 -> 1.2

---
Log message:

Remove explicit control flow through the use of the 'not' script


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/GlobalDCE/2002-08-17-FunctionDGE.ll
diff -u llvm/test/Regression/Transforms/GlobalDCE/2002-08-17-FunctionDGE.ll:1.1 llvm/test/Regression/Transforms/GlobalDCE/2002-08-17-FunctionDGE.ll:1.2
--- llvm/test/Regression/Transforms/GlobalDCE/2002-08-17-FunctionDGE.ll:1.1	Sat Aug 17 20:27:10 2002
+++ llvm/test/Regression/Transforms/GlobalDCE/2002-08-17-FunctionDGE.ll	Sat Jun 28 18:23:24 2003
@@ -1,10 +1,7 @@
 ; Make sure that functions are removed successfully if they are referred to by
 ; a global that is dead.  Make sure any globals they refer to die as well.
 
-; RUN: if as < %s | opt -globaldce | dis | grep foo
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -globaldce | dis | not grep foo
 
 %b = internal global int ()* %foo   ;; Unused, kills %foo
 


Index: llvm/test/Regression/Transforms/GlobalDCE/2002-08-17-WorkListTest.ll
diff -u llvm/test/Regression/Transforms/GlobalDCE/2002-08-17-WorkListTest.ll:1.2 llvm/test/Regression/Transforms/GlobalDCE/2002-08-17-WorkListTest.ll:1.3
--- llvm/test/Regression/Transforms/GlobalDCE/2002-08-17-WorkListTest.ll:1.2	Sat Aug 17 18:20:37 2002
+++ llvm/test/Regression/Transforms/GlobalDCE/2002-08-17-WorkListTest.ll	Sat Jun 28 18:23:24 2003
@@ -1,10 +1,7 @@
 ; This testcase tests that a worklist is being used, and that globals can be 
 ; removed if they are the subject of a constexpr and ConstantPointerRef
 
-; RUN: if as < %s | opt -globaldce | dis | grep global
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
+; RUN: as < %s | opt -globaldce | dis | not grep global
 
 %t0 = internal global [4 x sbyte] c"foo\00"
 %t1 = internal global [4 x sbyte] c"bar\00"


Index: llvm/test/Regression/Transforms/GlobalDCE/basicvariabletest.ll
diff -u llvm/test/Regression/Transforms/GlobalDCE/basicvariabletest.ll:1.1 llvm/test/Regression/Transforms/GlobalDCE/basicvariabletest.ll:1.2
--- llvm/test/Regression/Transforms/GlobalDCE/basicvariabletest.ll:1.1	Sat Aug 17 17:36:45 2002
+++ llvm/test/Regression/Transforms/GlobalDCE/basicvariabletest.ll	Sat Jun 28 18:23:24 2003
@@ -1,8 +1,4 @@
-; RUN: if as < %s | opt -globaldce | dis | grep global
-; RUN: then exit 1
-; RUN: else exit 0
-; RUN: fi
-
+; RUN: as < %s | opt -globaldce | dis | not grep global
 
 %X = uninitialized global int
 %Y = internal global int 7




From lattner at cs.uiuc.edu  Sat Jun 28 18:33:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:33:01 2003
Subject: [llvm-commits] CVS: llvm/test/Scripts/grep-not
Message-ID: <200306282332.SAA14400@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Scripts:

grep-not (r1.2) removed

---
Log message:

Remove obsolete script


---
Diffs of the changes:



From lattner at cs.uiuc.edu  Sat Jun 28 18:33:13 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:33:13 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Linker/2003-05-31-LinkerRename.ll
Message-ID: <200306282332.SAA14388@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Linker:

2003-05-31-LinkerRename.ll updated: 1.2 -> 1.3

---
Log message:

Remove usage of grep-not script



---
Diffs of the changes:

Index: llvm/test/Regression/Linker/2003-05-31-LinkerRename.ll
diff -u llvm/test/Regression/Linker/2003-05-31-LinkerRename.ll:1.2 llvm/test/Regression/Linker/2003-05-31-LinkerRename.ll:1.3
--- llvm/test/Regression/Linker/2003-05-31-LinkerRename.ll:1.2	Sat Jun 28 18:23:14 2003
+++ llvm/test/Regression/Linker/2003-05-31-LinkerRename.ll	Sat Jun 28 18:32:04 2003
@@ -5,7 +5,7 @@
 
 ; RUN: echo "implementation internal int %foo() { ret int 7 }" | as > Output/%s.1.bc
 ; RUN: as < %s > Output/%s.2.bc
-; RUN: link Output/%s.[12].bc | dis | grep 'internal' | grep-not '%foo('
+; RUN: link Output/%s.[12].bc | dis | grep 'internal' | not grep '%foo('
 
 implementation
 declare int %foo() 




From lattner at cs.uiuc.edu  Sat Jun 28 18:33:26 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:33:26 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Assembler/2002-04-29-NameBinding.llx ConstantExprFoldCast.llx
Message-ID: <200306282332.SAA14380@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Assembler:

2002-04-29-NameBinding.llx updated: 1.2 -> 1.3
ConstantExprFoldCast.llx updated: 1.2 -> 1.3

---
Log message:

Remove usage of grep-not script


---
Diffs of the changes:

Index: llvm/test/Regression/Assembler/2002-04-29-NameBinding.llx
diff -u llvm/test/Regression/Assembler/2002-04-29-NameBinding.llx:1.2 llvm/test/Regression/Assembler/2002-04-29-NameBinding.llx:1.3
--- llvm/test/Regression/Assembler/2002-04-29-NameBinding.llx:1.2	Sat Jun 28 18:23:13 2003
+++ llvm/test/Regression/Assembler/2002-04-29-NameBinding.llx	Sat Jun 28 18:32:03 2003
@@ -4,7 +4,7 @@
 ; Check by running globaldce, which will remove the constant if there are
 ; no references to it!
 ; 
-; RUN: as < %s | opt -globaldce | dis | grep-not constant
+; RUN: as < %s | opt -globaldce | dis | not grep constant
 ;
 	
 %v1 = internal constant int 5


Index: llvm/test/Regression/Assembler/ConstantExprFoldCast.llx
diff -u llvm/test/Regression/Assembler/ConstantExprFoldCast.llx:1.2 llvm/test/Regression/Assembler/ConstantExprFoldCast.llx:1.3
--- llvm/test/Regression/Assembler/ConstantExprFoldCast.llx:1.2	Sat Jun 28 18:23:13 2003
+++ llvm/test/Regression/Assembler/ConstantExprFoldCast.llx	Sat Jun 28 18:32:03 2003
@@ -1,6 +1,6 @@
 ; This test checks to make sure that constant exprs fold in some simple situations
 
-; RUN: as < %s | dis | grep-not cast
+; RUN: as < %s | dis | not grep cast
 
 %A = global int* cast (sbyte* null to int*)  ; Cast null -> fold
 %B = global int** cast (int** %A to int**)   ; Cast to same type -> fold




From lattner at cs.uiuc.edu  Sat Jun 28 18:33:39 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:33:39 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/InstCombine/2002-05-14-SubFailure.ll 2002-12-05-MissedConstProp.ll add.ll and.ll call.ll cast-set.ll cast.ll div.ll getelementptr.ll load.ll mul.ll not.ll or.ll phi.ll pow.ll rem.ll set.ll setcc-strength-reduce.ll shift.ll sub.ll
Message-ID: <200306282332.SAA14369@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/InstCombine:

2002-05-14-SubFailure.ll updated: 1.3 -> 1.4
2002-12-05-MissedConstProp.ll updated: 1.2 -> 1.3
add.ll updated: 1.9 -> 1.10
and.ll updated: 1.8 -> 1.9
call.ll updated: 1.2 -> 1.3
cast-set.ll updated: 1.6 -> 1.7
cast.ll updated: 1.10 -> 1.11
div.ll updated: 1.5 -> 1.6
getelementptr.ll updated: 1.10 -> 1.11
load.ll updated: 1.2 -> 1.3
mul.ll updated: 1.7 -> 1.8
not.ll updated: 1.8 -> 1.9
or.ll updated: 1.14 -> 1.15
phi.ll updated: 1.5 -> 1.6
pow.ll updated: 1.3 -> 1.4
rem.ll updated: 1.4 -> 1.5
set.ll updated: 1.9 -> 1.10
setcc-strength-reduce.ll updated: 1.3 -> 1.4
shift.ll updated: 1.8 -> 1.9
sub.ll updated: 1.10 -> 1.11

---
Log message:

Remove usage of grep-not script


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/InstCombine/2002-05-14-SubFailure.ll
diff -u llvm/test/Regression/Transforms/InstCombine/2002-05-14-SubFailure.ll:1.3 llvm/test/Regression/Transforms/InstCombine/2002-05-14-SubFailure.ll:1.4
--- llvm/test/Regression/Transforms/InstCombine/2002-05-14-SubFailure.ll:1.3	Sat Jun 28 17:44:42 2003
+++ llvm/test/Regression/Transforms/InstCombine/2002-05-14-SubFailure.ll	Sat Jun 28 18:31:56 2003
@@ -1,7 +1,7 @@
 ; Instcombine was missing a test that caused it to make illegal transformations
 ; sometimes.  In this case, it transforms the sub into an add:
 ; RUN: echo foo
-; RUN: as < %s | opt -instcombine | dis | grep-not add
+; RUN: as < %s | opt -instcombine | dis | not grep add
 ;
 
 


Index: llvm/test/Regression/Transforms/InstCombine/2002-12-05-MissedConstProp.ll
diff -u llvm/test/Regression/Transforms/InstCombine/2002-12-05-MissedConstProp.ll:1.2 llvm/test/Regression/Transforms/InstCombine/2002-12-05-MissedConstProp.ll:1.3
--- llvm/test/Regression/Transforms/InstCombine/2002-12-05-MissedConstProp.ll:1.2	Sat Jun 28 17:44:42 2003
+++ llvm/test/Regression/Transforms/InstCombine/2002-12-05-MissedConstProp.ll	Sat Jun 28 18:31:56 2003
@@ -1,4 +1,4 @@
-; RUN: as < %s | opt -instcombine | dis | grep-not add
+; RUN: as < %s | opt -instcombine | dis | not grep add
 
 int %test(int %A) {
   %A.neg = sub int 0, %A


Index: llvm/test/Regression/Transforms/InstCombine/add.ll
diff -u llvm/test/Regression/Transforms/InstCombine/add.ll:1.9 llvm/test/Regression/Transforms/InstCombine/add.ll:1.10
--- llvm/test/Regression/Transforms/InstCombine/add.ll:1.9	Sat Jun 28 17:44:42 2003
+++ llvm/test/Regression/Transforms/InstCombine/add.ll	Sat Jun 28 18:31:56 2003
@@ -3,7 +3,7 @@
 ; This also tests that a subtract with a constant is properly converted
 ; to a add w/negative constant
 
-; RUN: as < %s | opt -instcombine -die | dis | grep-not add
+; RUN: as < %s | opt -instcombine -die | dis | not grep add
 
 implementation
 


Index: llvm/test/Regression/Transforms/InstCombine/and.ll
diff -u llvm/test/Regression/Transforms/InstCombine/and.ll:1.8 llvm/test/Regression/Transforms/InstCombine/and.ll:1.9
--- llvm/test/Regression/Transforms/InstCombine/and.ll:1.8	Sat Jun 28 17:44:42 2003
+++ llvm/test/Regression/Transforms/InstCombine/and.ll	Sat Jun 28 18:31:56 2003
@@ -1,7 +1,7 @@
 ; This test makes sure that these instructions are properly eliminated.
 ;
 
-; RUN: as < %s | opt -instcombine | dis | grep-not and
+; RUN: as < %s | opt -instcombine | dis | not grep and
 
 implementation
 


Index: llvm/test/Regression/Transforms/InstCombine/call.ll
diff -u llvm/test/Regression/Transforms/InstCombine/call.ll:1.2 llvm/test/Regression/Transforms/InstCombine/call.ll:1.3
--- llvm/test/Regression/Transforms/InstCombine/call.ll:1.2	Sat Jun 28 17:44:42 2003
+++ llvm/test/Regression/Transforms/InstCombine/call.ll	Sat Jun 28 18:31:56 2003
@@ -1,4 +1,4 @@
-; RUN: as < %s | opt -instcombine -die | dis | grep call | grep-not cast
+; RUN: as < %s | opt -instcombine -die | dis | grep call | not grep cast
 
 implementation
 


Index: llvm/test/Regression/Transforms/InstCombine/cast-set.ll
diff -u llvm/test/Regression/Transforms/InstCombine/cast-set.ll:1.6 llvm/test/Regression/Transforms/InstCombine/cast-set.ll:1.7
--- llvm/test/Regression/Transforms/InstCombine/cast-set.ll:1.6	Sat Jun 28 17:44:42 2003
+++ llvm/test/Regression/Transforms/InstCombine/cast-set.ll	Sat Jun 28 18:31:56 2003
@@ -3,7 +3,7 @@
 ; slow.  Might it be better to make there be an instcombine prepass before
 ; level raise that takes care of the obvious stuff?
 
-; RUN: as < %s | opt -instcombine | dis | grep-not cast
+; RUN: as < %s | opt -instcombine | dis | not grep cast
 
 bool %test1(int %X) {
 	%A = cast int %X to uint


Index: llvm/test/Regression/Transforms/InstCombine/cast.ll
diff -u llvm/test/Regression/Transforms/InstCombine/cast.ll:1.10 llvm/test/Regression/Transforms/InstCombine/cast.ll:1.11
--- llvm/test/Regression/Transforms/InstCombine/cast.ll:1.10	Sat Jun 28 17:44:42 2003
+++ llvm/test/Regression/Transforms/InstCombine/cast.ll	Sat Jun 28 18:31:56 2003
@@ -1,6 +1,6 @@
 ; Tests to make sure elimination of casts is working correctly
 
-; RUN: as < %s | opt -instcombine -die | dis | grep '%c' | grep-not cast
+; RUN: as < %s | opt -instcombine -die | dis | grep '%c' | not grep cast
 
 implementation
 


Index: llvm/test/Regression/Transforms/InstCombine/div.ll
diff -u llvm/test/Regression/Transforms/InstCombine/div.ll:1.5 llvm/test/Regression/Transforms/InstCombine/div.ll:1.6
--- llvm/test/Regression/Transforms/InstCombine/div.ll:1.5	Sat Jun 28 17:44:42 2003
+++ llvm/test/Regression/Transforms/InstCombine/div.ll	Sat Jun 28 18:31:56 2003
@@ -1,7 +1,7 @@
 ; This test makes sure that div instructions are properly eliminated.
 ;
 
-; RUN: as < %s | opt -instcombine | dis | grep-not div
+; RUN: as < %s | opt -instcombine | dis | not grep div
 
 implementation
 


Index: llvm/test/Regression/Transforms/InstCombine/getelementptr.ll
diff -u llvm/test/Regression/Transforms/InstCombine/getelementptr.ll:1.10 llvm/test/Regression/Transforms/InstCombine/getelementptr.ll:1.11
--- llvm/test/Regression/Transforms/InstCombine/getelementptr.ll:1.10	Sat Jun 28 17:44:42 2003
+++ llvm/test/Regression/Transforms/InstCombine/getelementptr.ll	Sat Jun 28 18:31:56 2003
@@ -1,6 +1,6 @@
 ; The %A getelementptr instruction should be eliminated here
 
-; RUN: as < %s | opt -instcombine -die | dis | grep getelementptr | grep-not '%A '
+; RUN: as < %s | opt -instcombine -die | dis | grep getelementptr | not grep '%A '
 
 %Global = constant [10 x sbyte] c"helloworld"
 


Index: llvm/test/Regression/Transforms/InstCombine/load.ll
diff -u llvm/test/Regression/Transforms/InstCombine/load.ll:1.2 llvm/test/Regression/Transforms/InstCombine/load.ll:1.3
--- llvm/test/Regression/Transforms/InstCombine/load.ll:1.2	Sat Jun 28 17:44:42 2003
+++ llvm/test/Regression/Transforms/InstCombine/load.ll	Sat Jun 28 18:31:56 2003
@@ -1,7 +1,7 @@
 ; This test makes sure that these instructions are properly eliminated.
 ;
 
-; RUN: as < %s | opt -instcombine | dis | grep-not load
+; RUN: as < %s | opt -instcombine | dis | not grep load
 
 %X = constant int 42
 %Y = constant [2 x { int, float }] [ { int, float } { int 12, float 1.0 }, 


Index: llvm/test/Regression/Transforms/InstCombine/mul.ll
diff -u llvm/test/Regression/Transforms/InstCombine/mul.ll:1.7 llvm/test/Regression/Transforms/InstCombine/mul.ll:1.8
--- llvm/test/Regression/Transforms/InstCombine/mul.ll:1.7	Sat Jun 28 17:31:37 2003
+++ llvm/test/Regression/Transforms/InstCombine/mul.ll	Sat Jun 28 18:31:56 2003
@@ -1,7 +1,7 @@
 ; This test makes sure that mul instructions are properly eliminated.
 ;
 
-; RUN: as < %s | opt -instcombine | dis | grep-not mul
+; RUN: as < %s | opt -instcombine | dis | not grep mul
 
 implementation
 


Index: llvm/test/Regression/Transforms/InstCombine/not.ll
diff -u llvm/test/Regression/Transforms/InstCombine/not.ll:1.8 llvm/test/Regression/Transforms/InstCombine/not.ll:1.9
--- llvm/test/Regression/Transforms/InstCombine/not.ll:1.8	Sat Jun 28 17:44:42 2003
+++ llvm/test/Regression/Transforms/InstCombine/not.ll	Sat Jun 28 18:31:56 2003
@@ -1,7 +1,7 @@
 ; This test makes sure that these instructions are properly eliminated.
 ;
 
-; RUN: as < %s | opt -instcombine -die | dis | grep-not xor
+; RUN: as < %s | opt -instcombine -die | dis | not grep xor
 
 implementation
 


Index: llvm/test/Regression/Transforms/InstCombine/or.ll
diff -u llvm/test/Regression/Transforms/InstCombine/or.ll:1.14 llvm/test/Regression/Transforms/InstCombine/or.ll:1.15
--- llvm/test/Regression/Transforms/InstCombine/or.ll:1.14	Sat Jun 28 17:44:42 2003
+++ llvm/test/Regression/Transforms/InstCombine/or.ll	Sat Jun 28 18:31:56 2003
@@ -1,7 +1,7 @@
 ; This test makes sure that these instructions are properly eliminated.
 ;
 
-; RUN: as < %s | opt -instcombine | dis | grep -v '%OROK = or' | grep-not 'or '
+; RUN: as < %s | opt -instcombine | dis | grep -v '%OROK = or' | not grep 'or '
 
 implementation
 


Index: llvm/test/Regression/Transforms/InstCombine/phi.ll
diff -u llvm/test/Regression/Transforms/InstCombine/phi.ll:1.5 llvm/test/Regression/Transforms/InstCombine/phi.ll:1.6
--- llvm/test/Regression/Transforms/InstCombine/phi.ll:1.5	Sat Jun 28 17:44:42 2003
+++ llvm/test/Regression/Transforms/InstCombine/phi.ll	Sat Jun 28 18:31:56 2003
@@ -1,7 +1,7 @@
 ; This test makes sure that these instructions are properly eliminated.
 ;
 
-; RUN: as < %s | opt -instcombine -die | dis | grep-not phi
+; RUN: as < %s | opt -instcombine -die | dis | not grep phi
 
 implementation
 


Index: llvm/test/Regression/Transforms/InstCombine/pow.ll
diff -u llvm/test/Regression/Transforms/InstCombine/pow.ll:1.3 llvm/test/Regression/Transforms/InstCombine/pow.ll:1.4
--- llvm/test/Regression/Transforms/InstCombine/pow.ll:1.3	Sat Jun 28 17:44:42 2003
+++ llvm/test/Regression/Transforms/InstCombine/pow.ll	Sat Jun 28 18:31:56 2003
@@ -2,7 +2,7 @@
 ;
 ; Equivalent to: http://gcc.gnu.org/ml/gcc-patches/2003-02/msg01786.html
 
-; RUN: as < %s | opt -instcombine | dis | grep-not 'call double %pow'
+; RUN: as < %s | opt -instcombine | dis | not grep 'call double %pow'
 
 declare double %pow(double, double)
 


Index: llvm/test/Regression/Transforms/InstCombine/rem.ll
diff -u llvm/test/Regression/Transforms/InstCombine/rem.ll:1.4 llvm/test/Regression/Transforms/InstCombine/rem.ll:1.5
--- llvm/test/Regression/Transforms/InstCombine/rem.ll:1.4	Sat Jun 28 17:44:42 2003
+++ llvm/test/Regression/Transforms/InstCombine/rem.ll	Sat Jun 28 18:31:56 2003
@@ -1,7 +1,7 @@
 ; This test makes sure that these instructions are properly eliminated.
 ;
 
-; RUN: as < %s | opt -instcombine | dis | grep-not rem
+; RUN: as < %s | opt -instcombine | dis | not grep rem
 
 implementation
 


Index: llvm/test/Regression/Transforms/InstCombine/set.ll
diff -u llvm/test/Regression/Transforms/InstCombine/set.ll:1.9 llvm/test/Regression/Transforms/InstCombine/set.ll:1.10
--- llvm/test/Regression/Transforms/InstCombine/set.ll:1.9	Sat Jun 28 17:44:42 2003
+++ llvm/test/Regression/Transforms/InstCombine/set.ll	Sat Jun 28 18:31:56 2003
@@ -1,7 +1,7 @@
 ; This test makes sure that these instructions are properly eliminated.
 ;
 
-; RUN: as < %s | opt -instcombine | dis | grep-not set
+; RUN: as < %s | opt -instcombine | dis | not grep set
 
 %X = uninitialized global int
 


Index: llvm/test/Regression/Transforms/InstCombine/setcc-strength-reduce.ll
diff -u llvm/test/Regression/Transforms/InstCombine/setcc-strength-reduce.ll:1.3 llvm/test/Regression/Transforms/InstCombine/setcc-strength-reduce.ll:1.4
--- llvm/test/Regression/Transforms/InstCombine/setcc-strength-reduce.ll:1.3	Sat Jun 28 17:44:42 2003
+++ llvm/test/Regression/Transforms/InstCombine/setcc-strength-reduce.ll	Sat Jun 28 18:31:56 2003
@@ -3,7 +3,7 @@
 ; into equivalent setne,eq instructions.
 ;
 
-; RUN: as < %s | opt -instcombine | dis | grep -v seteq | grep -v setne | grep-not set
+; RUN: as < %s | opt -instcombine | dis | grep -v seteq | grep -v setne | not grep set
 
 bool "test1"(uint %A) {
 	%B = setge uint %A, 1   ; setne %A, 0


Index: llvm/test/Regression/Transforms/InstCombine/shift.ll
diff -u llvm/test/Regression/Transforms/InstCombine/shift.ll:1.8 llvm/test/Regression/Transforms/InstCombine/shift.ll:1.9
--- llvm/test/Regression/Transforms/InstCombine/shift.ll:1.8	Sat Jun 28 17:44:42 2003
+++ llvm/test/Regression/Transforms/InstCombine/shift.ll	Sat Jun 28 18:31:56 2003
@@ -1,7 +1,7 @@
 ; This test makes sure that these instructions are properly eliminated.
 ;
 
-; RUN: as < %s | opt -instcombine | dis | grep-not sh
+; RUN: as < %s | opt -instcombine | dis | not grep sh
 
 implementation
 


Index: llvm/test/Regression/Transforms/InstCombine/sub.ll
diff -u llvm/test/Regression/Transforms/InstCombine/sub.ll:1.10 llvm/test/Regression/Transforms/InstCombine/sub.ll:1.11
--- llvm/test/Regression/Transforms/InstCombine/sub.ll:1.10	Sat Jun 28 17:44:42 2003
+++ llvm/test/Regression/Transforms/InstCombine/sub.ll	Sat Jun 28 18:31:56 2003
@@ -1,7 +1,7 @@
 ; This test makes sure that these instructions are properly eliminated.
 ;
 
-; RUN: as < %s | opt -instcombine -die | dis | grep sub | grep-not -v 'sub int %Cok, %Bok'
+; RUN: as < %s | opt -instcombine -die | dis | grep sub | not grep -v 'sub int %Cok, %Bok'
 
 implementation
 




From lattner at cs.uiuc.edu  Sat Jun 28 18:50:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:50:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Analysis/DSGraph/gcsetest.ll
Message-ID: <200306282349.SAA14630@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Analysis/DSGraph:

gcsetest.ll updated: 1.4 -> 1.5

---
Log message:

Do not chain to basic-aa


---
Diffs of the changes:

Index: llvm/test/Regression/Analysis/DSGraph/gcsetest.ll
diff -u llvm/test/Regression/Analysis/DSGraph/gcsetest.ll:1.4 llvm/test/Regression/Analysis/DSGraph/gcsetest.ll:1.5
--- llvm/test/Regression/Analysis/DSGraph/gcsetest.ll:1.4	Sat Jun 28 18:23:11 2003
+++ llvm/test/Regression/Analysis/DSGraph/gcsetest.ll	Sat Jun 28 18:49:06 2003
@@ -1,7 +1,7 @@
 ; Test that GCSE uses ds-aa to do alias analysis, which is capable of 
 ; disambiguating some cases.
 
-; RUN: as < %s | opt -ds-aa -load-vn -gcse -instcombine -dce | dis | not grep ELIM
+; RUN: as < %s | opt -no-aa -ds-aa -load-vn -gcse -instcombine -dce | dis | not grep ELIM
 
 %intpair = type {int*, int*}
 implementation




From lattner at cs.uiuc.edu  Sat Jun 28 18:53:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 18:53:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Analysis/DSGraph/mustalias.ll
Message-ID: <200306282352.SAA14899@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Analysis/DSGraph:

mustalias.ll added (r1.1)

---
Log message:

New testcase


---
Diffs of the changes:

Index: llvm/test/Regression/Analysis/DSGraph/mustalias.ll
diff -c /dev/null llvm/test/Regression/Analysis/DSGraph/mustalias.ll:1.1
*** /dev/null	Sat Jun 28 18:52:45 2003
--- llvm/test/Regression/Analysis/DSGraph/mustalias.ll	Sat Jun 28 18:52:34 2003
***************
*** 0 ****
--- 1,17 ----
+ ; Test that ds-aa is returning must-alias information when it can.
+ 
+ ; RUN: as < %s | opt -no-aa -ds-aa -load-vn -gcse | dis | not grep load
+ 
+ %X = global int 20
+ 
+ implementation
+ 
+ int* %id(int* %P) { ret int* %P }
+ 
+ int %main() {
+ 	store int 0, int* %X
+ 	%XP = call int* %id(int* %X)
+ 	%A = load int* %XP        ; Should eliminate load!
+ 	ret int %A
+ }
+ 




From lattner at cs.uiuc.edu  Sat Jun 28 19:08:02 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 19:08:02 2003
Subject: [llvm-commits] CVS: llvm/lib/Analysis/AliasAnalysisEvaluator.cpp
Message-ID: <200306290007.TAA16073@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis:

AliasAnalysisEvaluator.cpp updated: 1.6 -> 1.7

---
Log message:

Count operands to instructions as well as just results.  This allows for global variables to be checked as well


---
Diffs of the changes:

Index: llvm/lib/Analysis/AliasAnalysisEvaluator.cpp
diff -u llvm/lib/Analysis/AliasAnalysisEvaluator.cpp:1.6 llvm/lib/Analysis/AliasAnalysisEvaluator.cpp:1.7
--- llvm/lib/Analysis/AliasAnalysisEvaluator.cpp:1.6	Wed Feb 26 13:29:16 2003
+++ llvm/lib/Analysis/AliasAnalysisEvaluator.cpp	Sat Jun 28 19:07:11 2003
@@ -17,6 +17,7 @@
 #include "llvm/Support/InstIterator.h"
 #include "llvm/Assembly/Writer.h"
 #include "Support/CommandLine.h"
+#include <set>
 
 namespace {
   cl::opt<bool> PrintNo  ("print-no-aliases", cl::ReallyHidden);
@@ -52,23 +53,27 @@
 bool AAEval::runOnFunction(Function &F) {
   AliasAnalysis &AA = getAnalysis<AliasAnalysis>();
   
-  std::vector<Value *> Pointers;
+  std::set<Value *> Pointers;
 
   for (Function::aiterator I = F.abegin(), E = F.aend(); I != E; ++I)
     if (isa<PointerType>(I->getType()))    // Add all pointer arguments
-      Pointers.push_back(I);
+      Pointers.insert(I);
 
-  for (inst_iterator I = inst_begin(F), E = inst_end(F); I != E; ++I)
+  for (inst_iterator I = inst_begin(F), E = inst_end(F); I != E; ++I) {
     if (isa<PointerType>((*I)->getType())) // Add all pointer instructions
-      Pointers.push_back(*I);
+      Pointers.insert(*I);
+    for (User::op_iterator OI = (*I)->op_begin(); OI != (*I)->op_end(); ++OI)
+      if (isa<PointerType>((*OI)->getType()))
+        Pointers.insert(*OI);
+  }
 
   if (PrintNo || PrintMay || PrintMust)
     std::cerr << "Function: " << F.getName() << "\n";
 
   // iterate over the worklist, and run the full (n^2)/2 disambiguations
-  for (std::vector<Value *>::iterator I1 = Pointers.begin(), E = Pointers.end();
+  for (std::set<Value *>::iterator I1 = Pointers.begin(), E = Pointers.end();
        I1 != E; ++I1)
-    for (std::vector<Value *>::iterator I2 = Pointers.begin(); I2 != I1; ++I2)
+    for (std::set<Value *>::iterator I2 = Pointers.begin(); I2 != I1; ++I2)
       switch (AA.alias(*I1, 0, *I2, 0)) {
       case AliasAnalysis::NoAlias:
         PrintResults("No", PrintNo, *I1, *I2);




From lattner at cs.uiuc.edu  Sat Jun 28 19:24:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 19:24:01 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Analysis/AliasAnalysis.h
Message-ID: <200306290023.TAA16221@apoc.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Analysis:

AliasAnalysis.h updated: 1.4 -> 1.5

---
Log message:

Add support for gathering sets of must aliases


---
Diffs of the changes:

Index: llvm/include/llvm/Analysis/AliasAnalysis.h
diff -u llvm/include/llvm/Analysis/AliasAnalysis.h:1.4 llvm/include/llvm/Analysis/AliasAnalysis.h:1.5
--- llvm/include/llvm/Analysis/AliasAnalysis.h:1.4	Wed Feb 26 13:24:47 2003
+++ llvm/include/llvm/Analysis/AliasAnalysis.h	Sat Jun 28 19:23:11 2003
@@ -76,6 +76,15 @@
     return MayAlias;
   }
 
+  /// getMustAliases - If there are any pointers known that must alias this
+  /// pointer, return them now.  This allows alias-set based alias analyses to
+  /// perform a form a value numbering (which is exposed by load-vn).  If an
+  /// alias analysis supports this, it should ADD any must aliased pointers to
+  /// the specified vector.
+  ///
+  virtual void getMustAliases(Value *P, std::vector<Value*> &RetVals) {}
+
+
   //===--------------------------------------------------------------------===//
   /// Simple mod/ref information...
   ///




From lattner at cs.uiuc.edu  Sat Jun 28 19:54:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 19:54:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Analysis/LoadValueNumbering.cpp
Message-ID: <200306290053.TAA18543@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis:

LoadValueNumbering.cpp updated: 1.3 -> 1.4

---
Log message:

If the alias analysis algorithm we are using can provide MUST alias information,
expose it directly as value numbering information


---
Diffs of the changes:

Index: llvm/lib/Analysis/LoadValueNumbering.cpp
diff -u llvm/lib/Analysis/LoadValueNumbering.cpp:1.3 llvm/lib/Analysis/LoadValueNumbering.cpp:1.4
--- llvm/lib/Analysis/LoadValueNumbering.cpp:1.3	Mon Jun 16 07:06:41 2003
+++ llvm/lib/Analysis/LoadValueNumbering.cpp	Sat Jun 28 19:53:34 2003
@@ -17,6 +17,7 @@
 #include "llvm/Analysis/Dominators.h"
 #include "llvm/Target/TargetData.h"
 #include "llvm/Pass.h"
+#include "llvm/Type.h"
 #include "llvm/iMemory.h"
 #include "llvm/BasicBlock.h"
 #include "llvm/Support/CFG.h"
@@ -83,6 +84,10 @@
 //
 void LoadVN::getEqualNumberNodes(Value *V,
                                  std::vector<Value*> &RetVals) const {
+  // If the alias analysis has any must alias information to share with us, we
+  // can definately use it.
+  if (isa<PointerType>(V->getType()))
+    getAnalysis<AliasAnalysis>().getMustAliases(V, RetVals);
 
   if (LoadInst *LI = dyn_cast<LoadInst>(V)) {
     // If we have a load instruction, find all of the load and store




From lattner at cs.uiuc.edu  Sat Jun 28 19:55:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 19:55:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Analysis/DataStructure/DataStructureAA.cpp
Message-ID: <200306290054.TAA18556@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis/DataStructure:

DataStructureAA.cpp updated: 1.6 -> 1.7

---
Log message:

Expose must alias information for global variables, implementing: DSGraph/mustalias.ll



---
Diffs of the changes:

Index: llvm/lib/Analysis/DataStructure/DataStructureAA.cpp
diff -u llvm/lib/Analysis/DataStructure/DataStructureAA.cpp:1.6 llvm/lib/Analysis/DataStructure/DataStructureAA.cpp:1.7
--- llvm/lib/Analysis/DataStructure/DataStructureAA.cpp:1.6	Sat Jun 28 16:58:28 2003
+++ llvm/lib/Analysis/DataStructure/DataStructureAA.cpp	Sat Jun 28 19:54:08 2003
@@ -40,9 +40,13 @@
     // Implement the AliasAnalysis API
     //  
 
-    // alias - This is the only method here that does anything interesting...
     AliasResult alias(const Value *V1, unsigned V1Size,
                       const Value *V2, unsigned V2Size);
+
+    void getMustAliases(Value *P, std::vector<Value*> &RetVals);
+
+  private:
+    DSGraph *getGraphForValue(const Value *V);
   };
 
   // Register the pass...
@@ -52,66 +56,78 @@
   RegisterAnalysisGroup<AliasAnalysis, DSAA> Y;
 }
 
-// getValueFunction - return the function containing the specified value if
-// available, or null otherwise.
+// getGraphForValue - Return the DSGraph to use for queries about the specified
+// value...
 //
-static const Function *getValueFunction(const Value *V) {
+DSGraph *DSAA::getGraphForValue(const Value *V) {
   if (const Instruction *I = dyn_cast<Instruction>(V))
-    return I->getParent()->getParent();
+    return &TD->getDSGraph(*I->getParent()->getParent());
   else if (const Argument *A = dyn_cast<Argument>(V))
-    return A->getParent();
+    return &TD->getDSGraph(*A->getParent());
   else if (const BasicBlock *BB = dyn_cast<BasicBlock>(V))
-    return BB->getParent();
+    return &TD->getDSGraph(*BB->getParent());
   return 0;
 }
 
+// isSinglePhysicalObject - For now, the only case that we know that there is
+// only one memory object in the node is when there is a single global in the
+// node, and the only composition bit set is Global.
+//
+static bool isSinglePhysicalObject(DSNode *N) {
+  assert(N->isComplete() && "Can only tell if this is a complete object!");
+  return N->isGlobalNode() && N->getGlobals().size() == 1 &&
+         !N->isHeapNode() && !N->isAllocaNode() && !N->isUnknownNode();
+}
+
 // alias - This is the only method here that does anything interesting...
 AliasAnalysis::AliasResult DSAA::alias(const Value *V1, unsigned V1Size,
                                        const Value *V2, unsigned V2Size) {
   if (V1 == V2) return MustAlias;
 
-  const Function *F1 = getValueFunction(V1);
-  const Function *F2 = getValueFunction(V2);
-  assert((!F1 || !F2 || F1 == F2) && "Alias query for 2 different functions?");
+  DSGraph *G1 = getGraphForValue(V1);
+  DSGraph *G2 = getGraphForValue(V2);
+  assert((!G1 || !G2 || G1 == G2) && "Alias query for 2 different functions?");
   
-  if (F2) F1 = F2;
-  if (F1) {
-    // Get the graph for a function...
-    DSGraph &G = TD->getDSGraph(*F1);
-    hash_map<Value*, DSNodeHandle> &GSM = G.getScalarMap();
-    hash_map<Value*, DSNodeHandle>::iterator I = GSM.find((Value*)V1);
-    if (I != GSM.end()) {
-      assert(I->second.getNode() && "Scalar map points to null node?");
-      hash_map<Value*, DSNodeHandle>::iterator J = GSM.find((Value*)V2);
-      if (J != GSM.end()) {
-        assert(J->second.getNode() && "Scalar map points to null node?");
+  // Get the graph to use...
+  DSGraph &G = *(G1 ? G1 : (G2 ? G2 : &TD->getGlobalsGraph()));
+
+  const DSGraph::ScalarMapTy &GSM = G.getScalarMap();
+  DSGraph::ScalarMapTy::const_iterator I = GSM.find((Value*)V1);
+  if (I != GSM.end()) {
+    assert(I->second.getNode() && "Scalar map points to null node?");
+    DSGraph::ScalarMapTy::const_iterator J = GSM.find((Value*)V2);
+    if (J != GSM.end()) {
+      assert(J->second.getNode() && "Scalar map points to null node?");
 
-        DSNode  *N1 = I->second.getNode(),  *N2 = J->second.getNode();
-        unsigned O1 = I->second.getOffset(), O2 = J->second.getOffset();
+      DSNode  *N1 = I->second.getNode(),  *N2 = J->second.getNode();
+      unsigned O1 = I->second.getOffset(), O2 = J->second.getOffset();
         
-        // We can only make a judgement of one of the nodes is complete...
-        if (N1->isComplete() || N2->isComplete()) {
-          if (N1 != N2)
-            return NoAlias;   // Completely different nodes.
-
-          // Both point to the same node and same offset, and there is only one
-          // physical memory object represented in the node, return must alias.
-          //if (O1 == O2 && !N1->isMultiObject())
-          //  return MustAlias; // Exactly the same object & offset
-
-          // See if they point to different offsets...  if so, we may be able to
-          // determine that they do not alias...
-          if (O1 != O2) {
-            if (O2 < O1) {    // Ensure that O1 <= O2
-              std::swap(V1, V2);
-              std::swap(O1, O2);
-              std::swap(V1Size, V2Size);
-            }
-
-            // FIXME: This is not correct because we do not handle array
-            // indexing correctly with this check!
-            //if (O1+V1Size <= O2) return NoAlias;
+      // We can only make a judgement of one of the nodes is complete...
+      if (N1->isComplete() || N2->isComplete()) {
+        if (N1 != N2)
+          return NoAlias;   // Completely different nodes.
+
+        // Both point to the same node and same offset, and there is only one
+        // physical memory object represented in the node, return must alias.
+        //
+        // FIXME: This isn't correct because we do not handle array indexing
+        // correctly.
+
+        if (O1 == O2 && isSinglePhysicalObject(N1))
+          return MustAlias; // Exactly the same object & offset
+
+        // See if they point to different offsets...  if so, we may be able to
+        // determine that they do not alias...
+        if (O1 != O2) {
+          if (O2 < O1) {    // Ensure that O1 <= O2
+            std::swap(V1, V2);
+            std::swap(O1, O2);
+            std::swap(V1Size, V2Size);
           }
+
+          // FIXME: This is not correct because we do not handle array
+          // indexing correctly with this check!
+          //if (O1+V1Size <= O2) return NoAlias;
         }
       }
     }
@@ -120,4 +136,28 @@
   // FIXME: we could improve on this by checking the globals graph for aliased
   // global queries...
   return getAnalysis<AliasAnalysis>().alias(V1, V1Size, V2, V2Size);
+}
+
+
+/// getMustAliases - If there are any pointers known that must alias this
+/// pointer, return them now.  This allows alias-set based alias analyses to
+/// perform a form a value numbering (which is exposed by load-vn).  If an alias
+/// analysis supports this, it should ADD any must aliased pointers to the
+/// specified vector.
+///
+void DSAA::getMustAliases(Value *P, std::vector<Value*> &RetVals) {
+  DSGraph *G = getGraphForValue(P);
+  if (!G) G = &TD->getGlobalsGraph();
+  
+  // The only must alias information we can currently determine occurs when the
+  // node for P is a global node with only one entry.
+  const DSGraph::ScalarMapTy &GSM = G->getScalarMap();
+  DSGraph::ScalarMapTy::const_iterator I = GSM.find(P);
+  if (I != GSM.end()) {
+    DSNode *N = I->second.getNode();
+    if (isSinglePhysicalObject(N))
+      RetVals.push_back(N->getGlobals()[0]);
+  }
+
+  return getAnalysis<AliasAnalysis>().getMustAliases(P, RetVals);
 }




From lattner at cs.uiuc.edu  Sat Jun 28 19:58:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 19:58:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Analysis/DSGraph/mustalias.ll
Message-ID: <200306290057.TAA18643@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Analysis/DSGraph:

mustalias.ll updated: 1.1 -> 1.2

---
Log message:

Make sure that the global is known to be internal


---
Diffs of the changes:

Index: llvm/test/Regression/Analysis/DSGraph/mustalias.ll
diff -u llvm/test/Regression/Analysis/DSGraph/mustalias.ll:1.1 llvm/test/Regression/Analysis/DSGraph/mustalias.ll:1.2
--- llvm/test/Regression/Analysis/DSGraph/mustalias.ll:1.1	Sat Jun 28 18:52:34 2003
+++ llvm/test/Regression/Analysis/DSGraph/mustalias.ll	Sat Jun 28 19:57:34 2003
@@ -2,7 +2,7 @@
 
 ; RUN: as < %s | opt -no-aa -ds-aa -load-vn -gcse | dis | not grep load
 
-%X = global int 20
+%X = internal global int 20
 
 implementation
 




From lattner at cs.uiuc.edu  Sat Jun 28 20:05:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 20:05:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Analysis/DSGraph/incompletenode.ll
Message-ID: <200306290104.UAA18745@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Analysis/DSGraph:

incompletenode.ll added (r1.1)

---
Log message:

New testcase


---
Diffs of the changes:

Index: llvm/test/Regression/Analysis/DSGraph/incompletenode.ll
diff -c /dev/null llvm/test/Regression/Analysis/DSGraph/incompletenode.ll:1.1
*** /dev/null	Sat Jun 28 20:04:03 2003
--- llvm/test/Regression/Analysis/DSGraph/incompletenode.ll	Sat Jun 28 20:03:53 2003
***************
*** 0 ****
--- 1,23 ----
+ ; This test is failing because the globals X and Y are marked incomplete
+ ; in the TD graph for %test
+ 
+ ; RUN: as < %s | opt -no-aa -ds-aa -load-vn -gcse | dis | not grep seteq
+ 
+ %X = internal global int 20
+ %Y = internal global int* null
+ 
+ implementation
+ 
+ bool %test(int** %P) { 
+   %A = load int** %P              ;; We know P == Y!
+   %B = load int** %Y              ;; B = A
+   %c = seteq int* %A, %B          ;; Always true
+   ret bool %c
+ }
+ 
+ int %main() {
+ 	store int* %X, int** %Y
+ 	call bool %test(int** %Y)
+ 	ret int 0
+ }
+ 




From lattner at cs.uiuc.edu  Sat Jun 28 20:10:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Jun 28 20:10:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Analysis/DSGraph/incompletenode.ll
Message-ID: <200306290109.UAA18979@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Analysis/DSGraph:

incompletenode.ll updated: 1.1 -> 1.2

---
Log message:

Fix testcase


---
Diffs of the changes:

Index: llvm/test/Regression/Analysis/DSGraph/incompletenode.ll
diff -u llvm/test/Regression/Analysis/DSGraph/incompletenode.ll:1.1 llvm/test/Regression/Analysis/DSGraph/incompletenode.ll:1.2
--- llvm/test/Regression/Analysis/DSGraph/incompletenode.ll:1.1	Sat Jun 28 20:03:53 2003
+++ llvm/test/Regression/Analysis/DSGraph/incompletenode.ll	Sat Jun 28 20:09:07 2003
@@ -1,14 +1,14 @@
-; This test is failing because the globals X and Y are marked incomplete
+; This test was failing because the globals X and Y are marked incomplete
 ; in the TD graph for %test
 
-; RUN: as < %s | opt -no-aa -ds-aa -load-vn -gcse | dis | not grep seteq
+; RUN: as < %s | opt -no-aa -ds-aa -load-vn -gcse -instcombine | dis | not grep seteq
 
 %X = internal global int 20
 %Y = internal global int* null
 
 implementation
 
-bool %test(int** %P) { 
+internal bool %test(int** %P) { 
   %A = load int** %P              ;; We know P == Y!
   %B = load int** %Y              ;; B = A
   %c = seteq int* %A, %B          ;; Always true




From lattner at cs.uiuc.edu  Sun Jun 29 11:43:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Jun 29 11:43:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Target/X86/InstSelectSimple.cpp
Message-ID: <200306291642.LAA31830@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Target/X86:

InstSelectSimple.cpp updated: 1.110 -> 1.111

---
Log message:

Until there is a setjmp/longjmp transformation pass, codegen setjmp as a noop
and longjmp as an abort!


---
Diffs of the changes:

Index: llvm/lib/Target/X86/InstSelectSimple.cpp
diff -u llvm/lib/Target/X86/InstSelectSimple.cpp:1.110 llvm/lib/Target/X86/InstSelectSimple.cpp:1.111
--- llvm/lib/Target/X86/InstSelectSimple.cpp:1.110	Sat Jun 21 22:31:18 2003
+++ llvm/lib/Target/X86/InstSelectSimple.cpp	Sun Jun 29 11:42:32 2003
@@ -978,6 +978,10 @@
     addDirectMem(BuildMI(BB, X86::MOVrm32, 5), TmpReg2).addReg(TmpReg1);
     return;
 
+  case LLVMIntrinsic::longjmp:
+    BuildMI(X86::CALLpcrel32, 1).addExternalSymbol("abort", true); 
+  case LLVMIntrinsic::setjmp:
+    return;
   default: assert(0 && "Unknown intrinsic for X86!");
   }
 }




From lattner at cs.uiuc.edu  Sun Jun 29 12:20:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Jun 29 12:20:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Analysis/DSGraph/2003-06-29-NodeCollapsing.ll
Message-ID: <200306291719.MAA32433@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Analysis/DSGraph:

2003-06-29-NodeCollapsing.ll added (r1.1)

---
Log message:

New testcase


---
Diffs of the changes:

Index: llvm/test/Regression/Analysis/DSGraph/2003-06-29-NodeCollapsing.ll
diff -c /dev/null llvm/test/Regression/Analysis/DSGraph/2003-06-29-NodeCollapsing.ll:1.1
*** /dev/null	Sun Jun 29 12:19:09 2003
--- llvm/test/Regression/Analysis/DSGraph/2003-06-29-NodeCollapsing.ll	Sun Jun 29 12:18:59 2003
***************
*** 0 ****
--- 1,24 ----
+ ; RUN: as < %s | opt -no-aa -ds-aa -load-vn -gcse | dis | not grep load
+ %T = type { int*, int* }
+ 
+ int %main() {
+ 	%A = alloca %T
+ 	%B = alloca { %T }
+ 	%C = alloca %T*
+ 	%Bp = getelementptr { %T }* %B, long 0, ubyte 0
+ 
+ 	%i0 = alloca int
+ 	%i1 = alloca int
+ 	%Ap0 = getelementptr %T* %A, long 0, ubyte 0
+ 	%Ap1 = getelementptr %T* %A, long 0, ubyte 1
+ 	store int* %i0, int** %Ap0
+ 	store int* %i1, int** %Ap1
+ 
+ 	store int 0, int* %i0
+ 	store int 1, int* %i1
+ 	%RetVal = load int* %i0   ; This load should be deletable
+ 
+ 	store %T* %A, %T** %C
+ 	store %T* %Bp, %T** %C    ; This store was causing merging to happen!
+ 	ret int %RetVal
+ }




From lattner at cs.uiuc.edu  Sun Jun 29 13:18:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Jun 29 13:18:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Analysis/DSGraph/2003-06-29-NodeCollapsing2.ll
Message-ID: <200306291817.NAA02056@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Analysis/DSGraph:

2003-06-29-NodeCollapsing2.ll added (r1.1)

---
Log message:

New, smaller testcase


---
Diffs of the changes:

Index: llvm/test/Regression/Analysis/DSGraph/2003-06-29-NodeCollapsing2.ll
diff -c /dev/null llvm/test/Regression/Analysis/DSGraph/2003-06-29-NodeCollapsing2.ll:1.1
*** /dev/null	Sun Jun 29 13:17:30 2003
--- llvm/test/Regression/Analysis/DSGraph/2003-06-29-NodeCollapsing2.ll	Sun Jun 29 13:17:20 2003
***************
*** 0 ****
--- 1,18 ----
+ ; This is the same testcase as 2003-06-29-NodeCollapsing2.ll, but it uses the 
+ ; graph checker.
+ ;
+ ; RUN: analyze %s -datastructure-gc -dsgc-abort-if-any-collapsed
+ ;
+ %T = type { int}
+ 
+ int %main() {
+ 	%A = alloca %T
+ 	%B = alloca { %T }
+ 	%C = alloca %T*
+ 	%Bp = getelementptr { %T }* %B, long 0, ubyte 0
+ 	%Ap = getelementptr %T* %A, long 0, ubyte 0
+ 
+ 	store %T* %A, %T** %C
+ 	store %T* %Bp, %T** %C    ; This store was causing merging to happen!
+ 	ret int 0
+ }




From lattner at cs.uiuc.edu  Sun Jun 29 13:18:14 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Jun 29 13:18:14 2003
Subject: [llvm-commits] CVS: llvm/lib/Analysis/DataStructure/GraphChecker.cpp
Message-ID: <200306291817.NAA02047@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis/DataStructure:

GraphChecker.cpp added (r1.1)

---
Log message:

New pass which is useful for writing regression tests


---
Diffs of the changes:

Index: llvm/lib/Analysis/DataStructure/GraphChecker.cpp
diff -c /dev/null llvm/lib/Analysis/DataStructure/GraphChecker.cpp:1.1
*** /dev/null	Sun Jun 29 13:17:18 2003
--- llvm/lib/Analysis/DataStructure/GraphChecker.cpp	Sun Jun 29 13:17:07 2003
***************
*** 0 ****
--- 1,150 ----
+ //===- GraphChecker.cpp - Assert that various graph properties hold -------===//
+ //
+ // This pass is used to test DSA with regression tests.  It can be used to check
+ // that certain graph properties hold, such as two nodes being disjoint, whether
+ // or not a node is collapsed, etc.  These are the command line arguments that
+ // it supports:
+ //
+ //   --dsgc-dsapass={local,bu,td}     - Specify what flavor of graph to check
+ //   --dsgc-abort-if-any-collapsed    - Abort if any collapsed nodes are found
+ //   --dsgc-abort-if-collapsed=<list> - Abort if a node pointed to by an SSA
+ //                                      value with name in <list> is collapsed
+ //   --dsgc-abort-if-merged=<list>    - Abort if any of the named SSA values
+ //                                      point to the same node.
+ //
+ //===----------------------------------------------------------------------===//
+ 
+ #include "llvm/Analysis/DataStructure.h"
+ #include "llvm/Analysis/DSGraph.h"
+ #include "Support/CommandLine.h"
+ #include "llvm/Value.h"
+ #include <set>
+ 
+ namespace {
+   enum DSPass { local, bu, td };
+   cl::opt<DSPass>
+   DSPass("dsgc-dspass", cl::Hidden,
+        cl::desc("Specify which DSA pass the -datastructure-gc pass should use"),
+          cl::values(clEnumVal(local, "Local pass"),
+                     clEnumVal(bu,    "Bottom-up pass"),
+                     clEnumVal(td,    "Top-down pass"), 0), cl::init(local));
+ 
+   cl::opt<bool>
+   AbortIfAnyCollapsed("dsgc-abort-if-any-collapsed", cl::Hidden,
+                       cl::desc("Abort if any collapsed nodes are found"));
+   cl::list<std::string>
+   AbortIfCollapsed("dsgc-abort-if-collapsed", cl::Hidden, cl::CommaSeparated,
+                    cl::desc("Abort if any of the named symbols is collapsed"));
+   cl::list<std::string>
+   AbortIfMerged("dsgc-abort-if-merged", cl::Hidden, cl::CommaSeparated,
+              cl::desc("Abort if any of the named symbols are merged together"));
+ 
+   struct DSGC : public FunctionPass {
+     DSGC();
+     bool doFinalization(Module &M);
+     bool runOnFunction(Function &F);
+ 
+     virtual void getAnalysisUsage(AnalysisUsage &AU) const {
+       switch (DSPass) {
+       case local: AU.addRequired<LocalDataStructures>(); break;
+       case bu:    AU.addRequired<BUDataStructures>(); break;
+       case td:    AU.addRequired<TDDataStructures>(); break;
+       }
+       AU.setPreservesAll();
+     }
+ 
+   private:
+     void verify(const DSGraph &G);
+   };
+ 
+   RegisterAnalysis<DSGC> X("datastructure-gc", "DSA Graph Checking Pass");
+ }
+ 
+ DSGC::DSGC() {
+   if (!AbortIfAnyCollapsed && AbortIfCollapsed.empty() &&
+       AbortIfMerged.empty()) {
+     std::cerr << "The -datastructure-gc is useless if you don't specify any"
+                  " -dsgc-* options.  See the -help-hidden output for a list.\n";
+     abort();
+   }
+ }
+ 
+ 
+ /// doFinalization - Verify that the globals graph is in good shape...
+ ///
+ bool DSGC::doFinalization(Module &M) {
+   switch (DSPass) {
+   case local:verify(getAnalysis<LocalDataStructures>().getGlobalsGraph());break;
+   case bu:   verify(getAnalysis<BUDataStructures>().getGlobalsGraph()); break;
+   case td:   verify(getAnalysis<TDDataStructures>().getGlobalsGraph()); break;
+   }
+   return false;
+ }
+ 
+ /// runOnFunction - Get the DSGraph for this function and verify that it is ok.
+ ///
+ bool DSGC::runOnFunction(Function &F) {
+   switch (DSPass) {
+   case local: verify(getAnalysis<LocalDataStructures>().getDSGraph(F)); break;
+   case bu:    verify(getAnalysis<BUDataStructures>().getDSGraph(F)); break;
+   case td:    verify(getAnalysis<TDDataStructures>().getDSGraph(F)); break;
+   }
+ 
+   return false;
+ }
+ 
+ /// verify - This is the function which checks to make sure that all of the
+ /// invariants established on the command line are true.
+ ///
+ void DSGC::verify(const DSGraph &G) {
+   // Loop over all of the nodes, checking to see if any are collapsed...
+   if (AbortIfAnyCollapsed) {
+     const std::vector<DSNode*> &Nodes = G.getNodes();
+     for (unsigned i = 0, e = Nodes.size(); i != e; ++i)
+       if (Nodes[i]->isNodeCompletelyFolded()) {
+         std::cerr << "Node is collapsed: ";
+         Nodes[i]->print(std::cerr, &G);
+         abort();
+       }
+   }
+ 
+   if (!AbortIfCollapsed.empty() || !AbortIfMerged.empty()) {
+     // Convert from a list to a set, because we don't have cl::set's yet.  FIXME
+     std::set<std::string> AbortIfCollapsedS(AbortIfCollapsed.begin(),
+                                             AbortIfCollapsed.end());
+     std::set<std::string> AbortIfMerged(AbortIfMerged.begin(),
+                                         AbortIfMerged.end());
+     
+     
+     // Now we loop over all of the scalars, checking to see if any are collapsed
+     // that are not supposed to be, or if any are merged together.
+     const DSGraph::ScalarMapTy &SM = G.getScalarMap();
+     std::map<DSNode*, std::string> AbortIfMergedNodes;
+     
+     for (DSGraph::ScalarMapTy::const_iterator I = SM.begin(), E = SM.end();
+          I != E; ++I)
+       if (I->first->hasName() && I->second.getNode()) {
+         DSNode *N = I->second.getNode();
+         
+         // Verify it is not collapsed if it is not supposed to be...
+         if (N->isNodeCompletelyFolded() &&
+             AbortIfCollapsedS.count(I->first->getName())) {
+           std::cerr << "Node for value '%" << I->first->getName()
+                     << "' is collapsed: ";
+           N->print(std::cerr, &G);
+           abort();
+         }
+ 
+         // Verify that it is not merged if it is not supposed to be...
+         if (AbortIfMerged.count(I->first->getName())) {
+           if (AbortIfMergedNodes.count(N)) {
+             std::cerr << "Nodes for values '%" << I->first->getName()
+                       << "' and '%" << AbortIfMergedNodes[N] << "' is merged: ";
+             N->print(std::cerr, &G);
+             abort();
+           }
+           AbortIfMergedNodes[N] = I->first->getName();
+         }
+       }
+   }
+ }




From lattner at cs.uiuc.edu  Sun Jun 29 15:28:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Jun 29 15:28:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Analysis/DataStructure/GraphChecker.cpp
Message-ID: <200306292027.PAA04015@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis/DataStructure:

GraphChecker.cpp updated: 1.1 -> 1.2

---
Log message:

Add print method to not get silly warning from analyze


---
Diffs of the changes:

Index: llvm/lib/Analysis/DataStructure/GraphChecker.cpp
diff -u llvm/lib/Analysis/DataStructure/GraphChecker.cpp:1.1 llvm/lib/Analysis/DataStructure/GraphChecker.cpp:1.2
--- llvm/lib/Analysis/DataStructure/GraphChecker.cpp:1.1	Sun Jun 29 13:17:07 2003
+++ llvm/lib/Analysis/DataStructure/GraphChecker.cpp	Sun Jun 29 15:27:16 2003
@@ -52,6 +52,7 @@
       }
       AU.setPreservesAll();
     }
+    void print(std::ostream &O, const Module *M) const {}
 
   private:
     void verify(const DSGraph &G);




From lattner at cs.uiuc.edu  Sun Jun 29 15:28:14 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Jun 29 15:28:14 2003
Subject: [llvm-commits] CVS: llvm/lib/Analysis/DataStructure/DataStructure.cpp
Message-ID: <200306292027.PAA04010@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis/DataStructure:

DataStructure.cpp updated: 1.102 -> 1.103

---
Log message:

Add support for "physical subtyping", which fixes:
DSGraph/2003-06-29-NodeCollapsing2.ll & DSGraph/PhysicalSubtyping.ll


---
Diffs of the changes:

Index: llvm/lib/Analysis/DataStructure/DataStructure.cpp
diff -u llvm/lib/Analysis/DataStructure/DataStructure.cpp:1.102 llvm/lib/Analysis/DataStructure/DataStructure.cpp:1.103
--- llvm/lib/Analysis/DataStructure/DataStructure.cpp:1.102	Sat Jun 28 16:58:28 2003
+++ llvm/lib/Analysis/DataStructure/DataStructure.cpp	Sun Jun 29 15:27:45 2003
@@ -143,6 +143,121 @@
 }
 
 
+namespace {
+  /// TypeElementWalker Class - Used for implementation of physical subtyping...
+  ///
+  class TypeElementWalker {
+    struct StackState {
+      const Type *Ty;
+      unsigned Offset;
+      unsigned Idx;
+      StackState(const Type *T, unsigned Off = 0)
+        : Ty(T), Offset(Off), Idx(0) {}
+    };
+
+    std::vector<StackState> Stack;
+  public:
+    TypeElementWalker(const Type *T) {
+      Stack.push_back(T);
+      StepToLeaf();
+    }
+
+    bool isDone() const { return Stack.empty(); }
+    const Type *getCurrentType()   const { return Stack.back().Ty;     }
+    unsigned    getCurrentOffset() const { return Stack.back().Offset; }
+
+    void StepToNextType() {
+      PopStackAndAdvance();
+      StepToLeaf();
+    }
+
+  private:
+    /// PopStackAndAdvance - Pop the current element off of the stack and
+    /// advance the underlying element to the next contained member.
+    void PopStackAndAdvance() {
+      assert(!Stack.empty() && "Cannot pop an empty stack!");
+      Stack.pop_back();
+      while (!Stack.empty()) {
+        StackState &SS = Stack.back();
+        if (const StructType *ST = dyn_cast<StructType>(SS.Ty)) {
+          ++SS.Idx;
+          if (SS.Idx != ST->getElementTypes().size()) {
+            const StructLayout *SL = TD.getStructLayout(ST);
+            SS.Offset += SL->MemberOffsets[SS.Idx]-SL->MemberOffsets[SS.Idx-1];
+            return;
+          }
+          Stack.pop_back();  // At the end of the structure
+        } else {
+          const ArrayType *AT = cast<ArrayType>(SS.Ty);
+          ++SS.Idx;
+          if (SS.Idx != AT->getNumElements()) {
+            SS.Offset += TD.getTypeSize(AT->getElementType());
+            return;
+          }
+          Stack.pop_back();  // At the end of the array
+        }
+      }
+    }
+
+    /// StepToLeaf - Used by physical subtyping to move to the first leaf node
+    /// on the type stack.
+    void StepToLeaf() {
+      if (Stack.empty()) return;
+      while (!Stack.empty() && !Stack.back().Ty->isFirstClassType()) {
+        StackState &SS = Stack.back();
+        if (const StructType *ST = dyn_cast<StructType>(SS.Ty)) {
+          if (ST->getElementTypes().empty()) {
+            assert(SS.Idx == 0);
+            PopStackAndAdvance();
+          } else {
+            // Step into the structure...
+            assert(SS.Idx < ST->getElementTypes().size());
+            const StructLayout *SL = TD.getStructLayout(ST);
+            Stack.push_back(StackState(ST->getElementTypes()[SS.Idx],
+                                       SS.Offset+SL->MemberOffsets[SS.Idx]));
+          }
+        } else {
+          const ArrayType *AT = cast<ArrayType>(SS.Ty);
+          if (AT->getNumElements() == 0) {
+            assert(SS.Idx == 0);
+            PopStackAndAdvance();
+          } else {
+            // Step into the array...
+            assert(SS.Idx < AT->getNumElements());
+            Stack.push_back(StackState(AT->getElementType(),
+                                       SS.Offset+SS.Idx*
+                                       TD.getTypeSize(AT->getElementType())));
+          }
+        }
+      }
+    }
+  };
+}
+
+/// ElementTypesAreCompatible - Check to see if the specified types are
+/// "physically" compatible.  If so, return true, else return false.  We only
+/// have to check the fields in T1: T2 may be larger than T1.
+///
+static bool ElementTypesAreCompatible(const Type *T1, const Type *T2) {
+  TypeElementWalker T1W(T1), T2W(T2);
+  
+  while (!T1W.isDone() && !T2W.isDone()) {
+    if (T1W.getCurrentOffset() != T2W.getCurrentOffset())
+      return false;
+
+    const Type *T1 = T1W.getCurrentType();
+    const Type *T2 = T2W.getCurrentType();
+    if (T1 != T2 && !T1->isLosslesslyConvertibleTo(T2))
+      return false;
+    
+    T1W.StepToNextType();
+    T2W.StepToNextType();
+  }
+  
+  return T1W.isDone();
+}
+
+
 /// mergeTypeInfo - This method merges the specified type into the current node
 /// at the specified offset.  This may update the current node's type record if
 /// this gives more information to the node, it may do nothing to the node if
@@ -218,8 +333,8 @@
     }
 
     if (Offset) {  // We could handle this case, but we don't for now...
-      DEBUG(std::cerr << "UNIMP: Trying to merge a growth type into "
-                      << "offset != 0: Collapsing!\n");
+      std::cerr << "UNIMP: Trying to merge a growth type into "
+                << "offset != 0: Collapsing!\n";
       if (FoldIfIncompatible) foldNodeCompletely();
       return true;
     }
@@ -287,12 +402,22 @@
   // If we found our type exactly, early exit
   if (SubType == NewTy) return false;
 
+  unsigned SubTypeSize = SubType->isSized() ? TD.getTypeSize(SubType) : 0;
+
+  // Ok, we are getting desperate now.  Check for physical subtyping, where we
+  // just require each element in the node to be compatible.
+  assert(NewTySize <= SubTypeSize &&
+         "Expected smaller type merging into this one!");
+  if (NewTySize && NewTySize < 256 &&
+      SubTypeSize && SubTypeSize < 256 && 
+      ElementTypesAreCompatible(NewTy, SubType))
+    return false;
+
   // Okay, so we found the leader type at the offset requested.  Search the list
   // of types that starts at this offset.  If SubType is currently an array or
   // structure, the type desired may actually be the first element of the
   // composite type...
   //
-  unsigned SubTypeSize = SubType->isSized() ? TD.getTypeSize(SubType) : 0;
   unsigned PadSize = SubTypeSize; // Size, including pad memory which is ignored
   while (SubType != NewTy) {
     const Type *NextSubType = 0;
@@ -351,7 +476,6 @@
     return false;
   }
 
-
   DEBUG(std::cerr << "MergeTypeInfo Folding OrigTy: " << Ty
                   << "\n due to:" << NewTy << " @ " << Offset << "!\n"
                   << "SubType: " << SubType << "\n\n");
@@ -444,7 +568,18 @@
   unsigned NOffset = CurNodeH.getOffset()-NH.getOffset();
   unsigned NSize = NH.getNode()->getSize();
 
-  // Merge the type entries of the two nodes together...
+  // If the two nodes are of different size, and the smaller node has the array
+  // bit set, collapse!
+  if (NSize != CurNodeH.getNode()->getSize()) {
+    if (NSize < CurNodeH.getNode()->getSize()) {
+      if (NH.getNode()->isArray())
+        NH.getNode()->foldNodeCompletely();
+    } else if (CurNodeH.getNode()->isArray()) {
+      NH.getNode()->foldNodeCompletely();
+    }
+  }
+
+  // Merge the type entries of the two nodes together...    
   if (NH.getNode()->Ty != Type::VoidTy)
     CurNodeH.getNode()->mergeTypeInfo(NH.getNode()->Ty, NOffset);
   assert(!CurNodeH.getNode()->isDeadNode());




From lattner at cs.uiuc.edu  Sun Jun 29 17:22:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Jun 29 17:22:00 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/CFrontend/2003-06-29-MultipleFunctionDefinition.c
Message-ID: <200306292221.RAA07818@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/CFrontend:

2003-06-29-MultipleFunctionDefinition.c added (r1.1)

---
Log message:

New testcase


---
Diffs of the changes:

Index: llvm/test/Regression/CFrontend/2003-06-29-MultipleFunctionDefinition.c
diff -c /dev/null llvm/test/Regression/CFrontend/2003-06-29-MultipleFunctionDefinition.c:1.1
*** /dev/null	Sun Jun 29 17:21:03 2003
--- llvm/test/Regression/CFrontend/2003-06-29-MultipleFunctionDefinition.c	Sun Jun 29 17:20:53 2003
***************
*** 0 ****
--- 1,5 ----
+ /* This is apparantly legal C.  */
+ extern __inline__ void test() { }
+ 
+ void test() {
+ }




From lattner at cs.uiuc.edu  Sun Jun 29 17:37:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Jun 29 17:37:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Analysis/DataStructure/DataStructure.cpp
Message-ID: <200306292236.RAA12358@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis/DataStructure:

DataStructure.cpp updated: 1.103 -> 1.104

---
Log message:

Fix minor bug in previous checkin


---
Diffs of the changes:

Index: llvm/lib/Analysis/DataStructure/DataStructure.cpp
diff -u llvm/lib/Analysis/DataStructure/DataStructure.cpp:1.103 llvm/lib/Analysis/DataStructure/DataStructure.cpp:1.104
--- llvm/lib/Analysis/DataStructure/DataStructure.cpp:1.103	Sun Jun 29 15:27:45 2003
+++ llvm/lib/Analysis/DataStructure/DataStructure.cpp	Sun Jun 29 17:36:31 2003
@@ -406,9 +406,7 @@
 
   // Ok, we are getting desperate now.  Check for physical subtyping, where we
   // just require each element in the node to be compatible.
-  assert(NewTySize <= SubTypeSize &&
-         "Expected smaller type merging into this one!");
-  if (NewTySize && NewTySize < 256 &&
+  if (NewTySize <= SubTypeSize && NewTySize && NewTySize < 256 &&
       SubTypeSize && SubTypeSize < 256 && 
       ElementTypesAreCompatible(NewTy, SubType))
     return false;




From lattner at cs.uiuc.edu  Sun Jun 29 17:37:13 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Jun 29 17:37:13 2003
Subject: [llvm-commits] CVS: llvm/lib/Analysis/DataStructure/GraphChecker.cpp
Message-ID: <200306292236.RAA12345@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis/DataStructure:

GraphChecker.cpp updated: 1.2 -> 1.3

---
Log message:

Add support for ensuring that nodes are not incomplete


---
Diffs of the changes:

Index: llvm/lib/Analysis/DataStructure/GraphChecker.cpp
diff -u llvm/lib/Analysis/DataStructure/GraphChecker.cpp:1.2 llvm/lib/Analysis/DataStructure/GraphChecker.cpp:1.3
--- llvm/lib/Analysis/DataStructure/GraphChecker.cpp:1.2	Sun Jun 29 15:27:16 2003
+++ llvm/lib/Analysis/DataStructure/GraphChecker.cpp	Sun Jun 29 17:36:15 2003
@@ -9,6 +9,8 @@
 //   --dsgc-abort-if-any-collapsed    - Abort if any collapsed nodes are found
 //   --dsgc-abort-if-collapsed=<list> - Abort if a node pointed to by an SSA
 //                                      value with name in <list> is collapsed
+//   --dsgc-abort-if-incomplete=<list>- Abort if any of the named SSA values
+//                                      are incomplete.
 //   --dsgc-abort-if-merged=<list>    - Abort if any of the named SSA values
 //                                      point to the same node.
 //
@@ -36,6 +38,9 @@
   AbortIfCollapsed("dsgc-abort-if-collapsed", cl::Hidden, cl::CommaSeparated,
                    cl::desc("Abort if any of the named symbols is collapsed"));
   cl::list<std::string>
+  AbortIfIncomplete("dsgc-abort-if-incomplete", cl::Hidden, cl::CommaSeparated,
+                   cl::desc("Abort if any of the named symbols is incomplete"));
+  cl::list<std::string>
   AbortIfMerged("dsgc-abort-if-merged", cl::Hidden, cl::CommaSeparated,
              cl::desc("Abort if any of the named symbols are merged together"));
 
@@ -63,7 +68,7 @@
 
 DSGC::DSGC() {
   if (!AbortIfAnyCollapsed && AbortIfCollapsed.empty() &&
-      AbortIfMerged.empty()) {
+      AbortIfIncomplete.empty() && AbortIfMerged.empty()) {
     std::cerr << "The -datastructure-gc is useless if you don't specify any"
                  " -dsgc-* options.  See the -help-hidden output for a list.\n";
     abort();
@@ -109,12 +114,15 @@
       }
   }
 
-  if (!AbortIfCollapsed.empty() || !AbortIfMerged.empty()) {
+  if (!AbortIfCollapsed.empty() || !AbortIfIncomplete.empty() ||
+      !AbortIfMerged.empty()) {
     // Convert from a list to a set, because we don't have cl::set's yet.  FIXME
     std::set<std::string> AbortIfCollapsedS(AbortIfCollapsed.begin(),
                                             AbortIfCollapsed.end());
-    std::set<std::string> AbortIfMerged(AbortIfMerged.begin(),
-                                        AbortIfMerged.end());
+    std::set<std::string> AbortIfIncompleteS(AbortIfIncomplete.begin(),
+                                             AbortIfIncomplete.end());
+    std::set<std::string> AbortIfMergedS(AbortIfMerged.begin(),
+                                         AbortIfMerged.end());
     
     
     // Now we loop over all of the scalars, checking to see if any are collapsed
@@ -125,26 +133,31 @@
     for (DSGraph::ScalarMapTy::const_iterator I = SM.begin(), E = SM.end();
          I != E; ++I)
       if (I->first->hasName() && I->second.getNode()) {
+        std::string Name = I->first->getName();
         DSNode *N = I->second.getNode();
         
         // Verify it is not collapsed if it is not supposed to be...
-        if (N->isNodeCompletelyFolded() &&
-            AbortIfCollapsedS.count(I->first->getName())) {
-          std::cerr << "Node for value '%" << I->first->getName()
-                    << "' is collapsed: ";
+        if (N->isNodeCompletelyFolded() && AbortIfCollapsedS.count(Name)) {
+          std::cerr << "Node for value '%" << Name << "' is collapsed: ";
+          N->print(std::cerr, &G);
+          abort();
+        }
+
+        if (N->isIncomplete() && AbortIfIncompleteS.count(Name)) {
+          std::cerr << "Node for value '%" << Name << "' is incomplete: ";
           N->print(std::cerr, &G);
           abort();
         }
 
         // Verify that it is not merged if it is not supposed to be...
-        if (AbortIfMerged.count(I->first->getName())) {
+        if (AbortIfMergedS.count(Name)) {
           if (AbortIfMergedNodes.count(N)) {
-            std::cerr << "Nodes for values '%" << I->first->getName()
-                      << "' and '%" << AbortIfMergedNodes[N] << "' is merged: ";
+            std::cerr << "Nodes for values '%" << Name << "' and '%"
+                      << AbortIfMergedNodes[N] << "' is merged: ";
             N->print(std::cerr, &G);
             abort();
           }
-          AbortIfMergedNodes[N] = I->first->getName();
+          AbortIfMergedNodes[N] = Name;
         }
       }
   }




From lattner at cs.uiuc.edu  Sun Jun 29 17:37:26 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Jun 29 17:37:26 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Analysis/DSGraph/2003-06-29-IncompleteTDPass.ll PhysicalSubtyping.ll
Message-ID: <200306292236.RAA12330@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Analysis/DSGraph:

2003-06-29-IncompleteTDPass.ll added (r1.1)
PhysicalSubtyping.ll added (r1.1)

---
Log message:

This node should not be incomplete!


---
Diffs of the changes:

Index: llvm/test/Regression/Analysis/DSGraph/2003-06-29-IncompleteTDPass.ll
diff -c /dev/null llvm/test/Regression/Analysis/DSGraph/2003-06-29-IncompleteTDPass.ll:1.1
*** /dev/null	Sun Jun 29 17:36:05 2003
--- llvm/test/Regression/Analysis/DSGraph/2003-06-29-IncompleteTDPass.ll	Sun Jun 29 17:35:55 2003
***************
*** 0 ****
--- 1,26 ----
+ ; RUN: analyze %s -datastructure-gc -dsgc-dspass=td -dsgc-abort-if-incomplete=Ptr
+ 
+ 
+ 
+ int %main() {
+ 	call void %A()
+ 	call void %B()
+ 	ret int 0
+ } 
+ 
+ internal void %A() {
+ 	%V = malloc int
+ 	call void %Callee(int* %V)
+ 	ret void
+ }
+ 
+ internal void %B() {
+ 	%V = malloc int
+ 	call void %Callee(int* %V)
+ 	ret void
+ }
+ 
+ internal void %Callee(int* %Ptr) {
+ 	load int* %Ptr
+ 	ret void
+ }


Index: llvm/test/Regression/Analysis/DSGraph/PhysicalSubtyping.ll
diff -c /dev/null llvm/test/Regression/Analysis/DSGraph/PhysicalSubtyping.ll:1.1
*** /dev/null	Sun Jun 29 17:36:05 2003
--- llvm/test/Regression/Analysis/DSGraph/PhysicalSubtyping.ll	Sun Jun 29 17:35:55 2003
***************
*** 0 ****
--- 1,18 ----
+ ; Test to check for support for "physical subtyping"
+ ;
+ ; RUN: analyze %s -datastructure-gc -dsgc-abort-if-any-collapsed
+ ;
+ %S = type { int }
+ %T = type { int, float, double }
+ 
+ int %main() {
+ 	%A = alloca %S
+ 	%Ap = getelementptr %S* %A, long 0, ubyte 0
+ 	%B = alloca %T
+ 	%Bp = getelementptr %T* %B, long 0, ubyte 0
+ 	%C = alloca int*
+ 	
+ 	store int* %Ap, int** %C
+ 	store int* %Bp, int** %C
+ 	ret int 0
+ }




From lattner at cs.uiuc.edu  Sun Jun 29 17:38:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Jun 29 17:38:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Analysis/DataStructure/TopDownClosure.cpp
Message-ID: <200306292237.RAA12369@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis/DataStructure:

TopDownClosure.cpp updated: 1.42 -> 1.43

---
Log message:

Abstract out the predicate which decides whether a function gets complete
arguments or not...


---
Diffs of the changes:

Index: llvm/lib/Analysis/DataStructure/TopDownClosure.cpp
diff -u llvm/lib/Analysis/DataStructure/TopDownClosure.cpp:1.42 llvm/lib/Analysis/DataStructure/TopDownClosure.cpp:1.43
--- llvm/lib/Analysis/DataStructure/TopDownClosure.cpp:1.42	Sat Jun 28 17:14:55 2003
+++ llvm/lib/Analysis/DataStructure/TopDownClosure.cpp	Sun Jun 29 17:37:07 2003
@@ -94,6 +94,18 @@
   return *G;
 }
 
+
+/// FunctionHasCompleteArguments - This function returns true if it is safe not
+/// to mark arguments to the function complete.
+///
+/// FIXME: Need to check if all callers have been found, or rather if a
+/// funcpointer escapes!
+///
+static bool FunctionHasCompleteArguments(Function &F) {
+  return F.hasInternalLinkage();
+}
+
+
 void TDDataStructures::calculateGraph(Function &F) {
   // Make sure this graph has not already been calculated, and that we don't get
   // into an infinite loop with mutually recursive functions.
@@ -106,10 +118,8 @@
 
   // Recompute the Incomplete markers and eliminate unreachable nodes.
   Graph.maskIncompleteMarkers();
-  // FIXME: Need to check if all callers have been found, or rather if a
-  // funcpointer escapes!
-  unsigned Flags = F.hasInternalLinkage() ?
-    DSGraph::IgnoreFormalArgs : DSGraph::MarkFormalArgs;
+  unsigned Flags = FunctionHasCompleteArguments(F) ?
+                            DSGraph::IgnoreFormalArgs : DSGraph::MarkFormalArgs;
   Graph.markIncompleteNodes(Flags | DSGraph::IgnoreGlobals);
   Graph.removeDeadNodes(DSGraph::RemoveUnreachableGlobals);
 
@@ -152,11 +162,11 @@
         ++I;
       } else {
         // For each callee...
-        Function *Callee = I->first;
-        DSGraph &CG = getOrCreateDSGraph(*Callee);  // Get the callee's graph...
+        Function &Callee = *I->first;
+        DSGraph &CG = getOrCreateDSGraph(Callee);  // Get the callee's graph...
       
-        DEBUG(std::cerr << "\t [TD] Inlining into callee '" << Callee->getName()
-              << "'\n");
+        DEBUG(std::cerr << "\t [TD] Inlining into callee '" << Callee.getName()
+                        << "'\n");
       
         // Clone our current graph into the callee...
         hash_map<Value*, DSNodeHandle> OldValMap;
@@ -172,7 +182,7 @@
         // current function calls this callee multiple times with different
         // signatures.
         //
-        for (; I != E && I->first == Callee; ++I) {
+        for (; I != E && I->first == &Callee; ++I) {
           // Map call site into callee graph
           DSCallSite NewCS(*I->second, OldNodeMap);
         
@@ -180,13 +190,13 @@
           NewCS.getRetVal().mergeWith(CG.getRetNode());
         
           // Resolve all of the arguments...
-          Function::aiterator AI = Callee->abegin();
+          Function::aiterator AI = Callee.abegin();
           for (unsigned i = 0, e = NewCS.getNumPtrArgs();
-               i != e && AI != Callee->aend(); ++i, ++AI) {
+               i != e && AI != Callee.aend(); ++i, ++AI) {
             // Advance the argument iterator to the first pointer argument...
-            while (AI != Callee->aend() && !DS::isPointerType(AI->getType()))
+            while (AI != Callee.aend() && !DS::isPointerType(AI->getType()))
               ++AI;
-            if (AI == Callee->aend()) break;
+            if (AI == Callee.aend()) break;
 
             // Add the link from the argument scalar to the provided value
             DSNodeHandle &NH = CG.getNodeForValue(AI);




From lattner at cs.uiuc.edu  Sun Jun 29 18:22:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Jun 29 18:22:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Analysis/DSGraph/Makefile
Message-ID: <200306292321.SAA14544@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Analysis/DSGraph:

Makefile updated: 1.2 -> 1.3

---
Log message:

Add clean target to clean up my mess


---
Diffs of the changes:

Index: llvm/test/Regression/Analysis/DSGraph/Makefile
diff -u llvm/test/Regression/Analysis/DSGraph/Makefile:1.2 llvm/test/Regression/Analysis/DSGraph/Makefile:1.3
--- llvm/test/Regression/Analysis/DSGraph/Makefile:1.2	Sat Jun 28 17:16:59 2003
+++ llvm/test/Regression/Analysis/DSGraph/Makefile	Sun Jun 29 18:20:51 2003
@@ -9,3 +9,6 @@
 $(addprefix Output/, $(TESTS:%.ll=%.ll.out)):
 Output/%.ll.out: %.ll Output/.dir $(LOPT)
 	-$(TESTRUNR) $<
+
+clean::
+	-rm -f *.dot *.ps




From lattner at cs.uiuc.edu  Sun Jun 29 22:14:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Jun 29 22:14:01 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Transforms/PoolAllocate.h
Message-ID: <200306300313.WAA30454@apoc.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Transforms:

PoolAllocate.h updated: 1.3 -> 1.4

---
Log message:

Add argument


---
Diffs of the changes:

Index: llvm/include/llvm/Transforms/PoolAllocate.h
diff -u llvm/include/llvm/Transforms/PoolAllocate.h:1.3 llvm/include/llvm/Transforms/PoolAllocate.h:1.4
--- llvm/include/llvm/Transforms/PoolAllocate.h:1.3	Wed Jun  4 03:01:13 2003
+++ llvm/include/llvm/Transforms/PoolAllocate.h	Sun Jun 29 22:13:36 2003
@@ -139,7 +139,8 @@
   void CreatePools(Function &F, const std::vector<DSNode*> &NodesToPA,
                    std::map<DSNode*, Value*> &PoolDescriptors);
   
-  void TransformFunctionBody(Function &F, DSGraph &G, PA::FuncInfo &FI);
+  void TransformFunctionBody(Function &F, Function &OldF,
+                             DSGraph &G, PA::FuncInfo &FI);
 };
 
 #endif




From lattner at cs.uiuc.edu  Sun Jun 29 22:14:14 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Jun 29 22:14:14 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Analysis/DataStructure.h
Message-ID: <200306300313.WAA30447@apoc.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Analysis:

DataStructure.h updated: 1.64 -> 1.65

---
Log message:

Remove prototype for dead method


---
Diffs of the changes:

Index: llvm/include/llvm/Analysis/DataStructure.h
diff -u llvm/include/llvm/Analysis/DataStructure.h:1.64 llvm/include/llvm/Analysis/DataStructure.h:1.65
--- llvm/include/llvm/Analysis/DataStructure.h:1.64	Sat Jun 21 22:05:45 2003
+++ llvm/include/llvm/Analysis/DataStructure.h	Sun Jun 29 22:13:28 2003
@@ -168,8 +168,6 @@
 private:
   void calculateGraph(Function &F);
   DSGraph &getOrCreateDSGraph(Function &F);
-
-  void ResolveCallSite(DSGraph &Graph, const DSCallSite &CallSite);
 };
 
 #endif




From lattner at cs.uiuc.edu  Sun Jun 29 22:15:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Jun 29 22:15:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/IPO/PoolAllocate.cpp
Message-ID: <200306300314.WAA30491@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/IPO:

PoolAllocate.cpp updated: 1.10 -> 1.11

---
Log message:

Eliminate using declarations, adjust for new DSGraph API


---
Diffs of the changes:

Index: llvm/lib/Transforms/IPO/PoolAllocate.cpp
diff -u llvm/lib/Transforms/IPO/PoolAllocate.cpp:1.10 llvm/lib/Transforms/IPO/PoolAllocate.cpp:1.11
--- llvm/lib/Transforms/IPO/PoolAllocate.cpp:1.10	Fri Jun 20 09:36:52 2003
+++ llvm/lib/Transforms/IPO/PoolAllocate.cpp	Sun Jun 29 22:14:44 2003
@@ -17,9 +17,6 @@
 #include "llvm/Support/InstVisitor.h"
 #include "Support/Statistic.h"
 #include "Support/VectorExtras.h"
-using std::vector;
-using std::map;
-using std::multimap;
 using namespace PA;
 
 namespace {
@@ -32,7 +29,7 @@
   //                        poolfree?)
   const Type *PoolDescType = 
   StructType::get(make_vector<const Type*>(VoidPtrTy, Type::UIntTy, 
-					   Type::UIntTy, 0));
+                                           Type::UIntTy, 0));
   
   const PointerType *PoolDescPtr = PointerType::get(PoolDescType);
   
@@ -49,9 +46,9 @@
 // Prints out the functions mapped to the leader of the equivalence class they
 // belong to.
 void PoolAllocate::printFuncECs() {
-  map<Function*, Function*> &leaderMap = FuncECs.getLeaderMap();
+  std::map<Function*, Function*> &leaderMap = FuncECs.getLeaderMap();
   std::cerr << "Indirect Function Map \n";
-  for (map<Function*, Function*>::iterator LI = leaderMap.begin(),
+  for (std::map<Function*, Function*>::iterator LI = leaderMap.begin(),
 	 LE = leaderMap.end(); LI != LE; ++LI) {
     std::cerr << LI->first->getName() << ": leader is "
 	      << LI->second->getName() << "\n";
@@ -78,7 +75,7 @@
     // For each call site in the function
     // All the functions that can be called at the call site are put in the
     // same equivalence class.
-    for (vector<DSCallSite>::iterator CSI = callSites.begin(), 
+    for (std::vector<DSCallSite>::iterator CSI = callSites.begin(), 
 	   CSE = callSites.end(); CSI != CSE ; ++CSI) {
       if (CSI->isIndirectCall()) {
 	DSNode *DSN = CSI->getCalleeNode();
@@ -234,7 +231,7 @@
         Nodes[i]->markReachableNodes(MarkedNodes);
 
   // Marked the returned node as alive...
-  if (DSNode *RetNode = G.getRetNode().getNode())
+  if (DSNode *RetNode = G.getReturnNodeFor(F).getNode())
     if (RetNode->isHeapNode())
       RetNode->markReachableNodes(MarkedNodes);
 
@@ -423,7 +420,7 @@
   }
   
   // Transform the body of the function now...
-  TransformFunctionBody(NewF, G, FI);
+  TransformFunctionBody(NewF, F, G, FI);
 }
 
 
@@ -540,9 +537,9 @@
   };
 }
 
-void PoolAllocate::TransformFunctionBody(Function &F, DSGraph &G, FuncInfo &FI){
-  DSGraph &TDG = TDDS->getDSGraph(G.getFunction());
-  FuncTransform(*this, G, TDG, FI).visit(F);
+void PoolAllocate::TransformFunctionBody(Function &F, Function &OldF,
+                                         DSGraph &G, FuncInfo &FI) {
+  FuncTransform(*this, G, TDDS->getDSGraph(OldF), FI).visit(F);
 }
 
 // Returns true if V is a function pointer
@@ -672,8 +669,8 @@
   // Remove old malloc instruction
   MI.getParent()->getInstList().erase(&MI);
   
-  hash_map<Value*, DSNodeHandle> &SM = G.getScalarMap();
-  hash_map<Value*, DSNodeHandle>::iterator MII = SM.find(&MI);
+  DSGraph::ScalarMapTy &SM = G.getScalarMap();
+  DSGraph::ScalarMapTy::iterator MII = SM.find(&MI);
   
   // If we are modifying the original function, update the DSGraph... 
   if (MII != SM.end()) {
@@ -754,16 +751,16 @@
   }
 
   std::vector<Value*> Args;  
-  if (!CF) {
-    // Indirect call
-    
+  if (!CF) {   // Indirect call
     DEBUG(std::cerr << "  Handling call: " << CI);
     
     std::map<unsigned, Value*> PoolArgs;
     Function *FuncClass;
     
-    std::pair<multimap<CallInst*, Function*>::iterator, multimap<CallInst*, Function*>::iterator> Targets = PAInfo.CallInstTargets.equal_range(&CI);
-    for (multimap<CallInst*, Function*>::iterator TFI = Targets.first,
+    std::pair<std::multimap<CallInst*, Function*>::iterator,
+              std::multimap<CallInst*, Function*>::iterator> Targets =
+      PAInfo.CallInstTargets.equal_range(&CI);
+    for (std::multimap<CallInst*, Function*>::iterator TFI = Targets.first,
 	   TFE = Targets.second; TFI != TFE; ++TFI) {
       if (TFI == Targets.first) {
 	FuncClass = PAInfo.FuncECs.findClass(TFI->second);
@@ -791,7 +788,8 @@
       assert(OpNum == CI.getNumOperands() && "Varargs calls not handled yet!");
       
       if (CI.getType() != Type::VoidTy)
-	CalcNodeMapping(getDSNodeFor(&CI), CG.getRetNode().getNode(), 
+	CalcNodeMapping(getDSNodeFor(&CI),
+                        CG.getReturnNodeFor(*TFI->second).getNode(), 
 			NodeMapping);
       
       unsigned idx = CFI->PoolArgFirst;
@@ -881,7 +879,7 @@
     
     // Map the return value as well...
     if (CI.getType() != Type::VoidTy)
-      CalcNodeMapping(getDSNodeFor(&CI), CG.getRetNode().getNode(), 
+      CalcNodeMapping(getDSNodeFor(&CI), CG.getReturnNodeFor(*CF).getNode(),
 		      NodeMapping);
 
     // Okay, now that we have established our mapping, we can figure out which




From lattner at cs.uiuc.edu  Sun Jun 29 22:15:16 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Jun 29 22:15:16 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Analysis/DSGraph.h
Message-ID: <200306300314.WAA30469@apoc.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Analysis:

DSGraph.h updated: 1.51 -> 1.52

---
Log message:

Substantial revamp: DSGraphs now may contain the graphs for multiple functions 
in the same graph


---
Diffs of the changes:

Index: llvm/include/llvm/Analysis/DSGraph.h
diff -u llvm/include/llvm/Analysis/DSGraph.h:1.51 llvm/include/llvm/Analysis/DSGraph.h:1.52
--- llvm/include/llvm/Analysis/DSGraph.h:1.51	Sat Jun 28 16:57:13 2003
+++ llvm/include/llvm/Analysis/DSGraph.h	Sun Jun 29 22:14:23 2003
@@ -15,16 +15,25 @@
 struct DSGraph {
   // Public data-type declarations...
   typedef hash_map<Value*, DSNodeHandle> ScalarMapTy;
+  typedef hash_map<Function*, DSNodeHandle> ReturnNodesTy;
 
+  /// NodeMapTy - This data type is used when cloning one graph into another to
+  /// keep track of the correspondence between the nodes in the old and new
+  /// graphs.
+  typedef hash_map<const DSNode*, DSNodeHandle> NodeMapTy;
 private:
-  Function *Func;          // Func - The LLVM function this graph corresponds to
   DSGraph *GlobalsGraph;   // Pointer to the common graph of global objects
   bool PrintAuxCalls;      // Should this graph print the Aux calls vector?
 
-  DSNodeHandle RetNode;    // The node that gets returned...
   std::vector<DSNode*> Nodes;
   ScalarMapTy ScalarMap;
 
+  // ReturnNodes - A return value for every function merged into this graph.
+  // Each DSGraph may have multiple functions merged into it at any time, which
+  // is used for representing SCCs.
+  //
+  ReturnNodesTy ReturnNodes;
+
   // FunctionCalls - This vector maintains a single entry for each call
   // instruction in the current graph.  The first entry in the vector is the
   // scalar that holds the return value for the call, the second is the function
@@ -42,7 +51,7 @@
   void operator=(const DSGraph &); // DO NOT IMPLEMENT
 public:
   // Create a new, empty, DSGraph.
-  DSGraph() : Func(0), GlobalsGraph(0), PrintAuxCalls(false) {}
+  DSGraph() : GlobalsGraph(0), PrintAuxCalls(false) {}
   DSGraph(Function &F, DSGraph *GlobalsGraph); // Compute the local DSGraph
 
   // Copy ctor - If you want to capture the node mapping between the source and
@@ -54,15 +63,9 @@
   // method.
   //
   DSGraph(const DSGraph &DSG);
-  DSGraph(const DSGraph &DSG, hash_map<const DSNode*, DSNodeHandle> &NodeMap);
+  DSGraph(const DSGraph &DSG, NodeMapTy &NodeMap);
   ~DSGraph();
 
-  bool hasFunction() const { return Func != 0; }
-  Function &getFunction() const {
-    assert(hasFunction() && "Cannot call getFunction on graph without a fn!");
-    return *Func;
-  }
-
   DSGraph *getGlobalsGraph() const { return GlobalsGraph; }
   void setGlobalsGraph(DSGraph *G) { GlobalsGraph = G; }
 
@@ -116,14 +119,27 @@
     return I->second;
   }
 
-  const DSNodeHandle &getRetNode() const { return RetNode; }
-        DSNodeHandle &getRetNode()       { return RetNode; }
+  /// getReturnNodes - Return the mapping of functions to their return nodes for
+  /// this graph.
+  const ReturnNodesTy &getReturnNodes() const { return ReturnNodes; }
+        ReturnNodesTy &getReturnNodes()       { return ReturnNodes; }
+
+  /// getReturnNodeFor - Return the return node for the specified function.
+  ///
+  DSNodeHandle &getReturnNodeFor(Function &F) {
+    ReturnNodesTy::iterator I = ReturnNodes.find(&F);
+    assert(I != ReturnNodes.end() && "F not in this DSGraph!");
+    return I->second;
+  }
 
+  /// getGraphSize - Return the number of nodes in this graph.
+  ///
   unsigned getGraphSize() const {
     return Nodes.size();
   }
 
   /// print - Print a dot graph to the specified ostream...
+  ///
   void print(std::ostream &O) const;
 
   /// dump - call print(std::cerr), for use from the debugger...
@@ -170,8 +186,8 @@
   };
   void removeDeadNodes(unsigned Flags);
 
-  // CloneFlags enum - Bits that may be passed into the cloneInto method to
-  // specify how to clone the function graph.
+  /// CloneFlags enum - Bits that may be passed into the cloneInto method to
+  /// specify how to clone the function graph.
   enum CloneFlags {
     StripAllocaBit        = 1 << 0, KeepAllocaBit     = 0 << 0,
     DontCloneCallNodes    = 1 << 1, CloneCallNodes    = 0 << 0,
@@ -179,15 +195,15 @@
     StripModRefBits       = 1 << 3, KeepModRefBits    = 0 << 0,
   };
 
-  // cloneInto - Clone the specified DSGraph into the current graph, returning
-  // the Return node of the graph.  The translated ScalarMap for the old
-  // function is filled into the OldValMap member.  If StripAllocas is set to
-  // 'StripAllocaBit', Alloca markers are removed from the graph as the graph is
-  // being cloned.
-  //
-  DSNodeHandle cloneInto(const DSGraph &G, ScalarMapTy &OldValMap,
-                         hash_map<const DSNode*, DSNodeHandle> &OldNodeMap,
-                         unsigned CloneFlags = 0);
+  /// cloneInto - Clone the specified DSGraph into the current graph.  The
+  /// translated ScalarMap for the old function is filled into the OldValMap
+  /// member, and the translated ReturnNodes map is returned into ReturnNodes.
+  ///
+  /// The CloneFlags member controls various aspects of the cloning process.
+  ///
+  void cloneInto(const DSGraph &G, ScalarMapTy &OldValMap,
+                 ReturnNodesTy &OldReturnNodes, NodeMapTy &OldNodeMap,
+                 unsigned CloneFlags = 0);
 
   /// mergeInGraph - The method is used for merging graphs together.  If the
   /// argument graph is not *this, it makes a clone of the specified graph, then
@@ -195,7 +211,8 @@
   /// the graph.  If the StripAlloca's argument is 'StripAllocaBit' then Alloca
   /// markers are removed from nodes.
   ///
-  void mergeInGraph(DSCallSite &CS, const DSGraph &Graph, unsigned CloneFlags);
+  void mergeInGraph(DSCallSite &CS, Function &F, const DSGraph &Graph,
+                    unsigned CloneFlags);
 
   // Methods for checking to make sure graphs are well formed...
   void AssertNodeInGraph(const DSNode *N) const {




From lattner at cs.uiuc.edu  Sun Jun 29 22:16:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Jun 29 22:16:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Analysis/IPA/IPModRef.cpp
Message-ID: <200306300315.WAA30503@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis/IPA:

IPModRef.cpp updated: 1.14 -> 1.15

---
Log message:

Adjust for new DSGraph API


---
Diffs of the changes:

Index: llvm/lib/Analysis/IPA/IPModRef.cpp
diff -u llvm/lib/Analysis/IPA/IPModRef.cpp:1.14 llvm/lib/Analysis/IPA/IPModRef.cpp:1.15
--- llvm/lib/Analysis/IPA/IPModRef.cpp:1.14	Wed Feb  5 15:59:58 2003
+++ llvm/lib/Analysis/IPA/IPModRef.cpp	Sun Jun 29 22:14:54 2003
@@ -166,7 +166,7 @@
 
       // Perform the merging now of the graph for the callee, which will
       // come with mod/ref bits set...
-      Result->mergeInGraph(CS, IPModRefObj.getBUDSGraph(*F),
+      Result->mergeInGraph(CS, *F, IPModRefObj.getBUDSGraph(*F),
                            DSGraph::StripAllocaBit
                            | DSGraph::DontCloneCallNodes
                            | DSGraph::DontCloneAuxCallNodes);




From lattner at cs.uiuc.edu  Sun Jun 29 22:16:17 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Jun 29 22:16:17 2003
Subject: [llvm-commits] CVS: llvm/lib/Analysis/DataStructure/BottomUpClosure.cpp DataStructure.cpp Local.cpp Printer.cpp Steensgaard.cpp TopDownClosure.cpp
Message-ID: <200306300315.WAA30641@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis/DataStructure:

BottomUpClosure.cpp updated: 1.54 -> 1.55
DataStructure.cpp updated: 1.104 -> 1.105
Local.cpp updated: 1.54 -> 1.55
Printer.cpp updated: 1.53 -> 1.54
Steensgaard.cpp updated: 1.28 -> 1.29
TopDownClosure.cpp updated: 1.43 -> 1.44

---
Log message:

Revamp DSGraphs so that they can support multiple functions in the same
DSGraph at one time


---
Diffs of the changes:

Index: llvm/lib/Analysis/DataStructure/BottomUpClosure.cpp
diff -u llvm/lib/Analysis/DataStructure/BottomUpClosure.cpp:1.54 llvm/lib/Analysis/DataStructure/BottomUpClosure.cpp:1.55
--- llvm/lib/Analysis/DataStructure/BottomUpClosure.cpp:1.54	Sat Jun 28 17:14:55 2003
+++ llvm/lib/Analysis/DataStructure/BottomUpClosure.cpp	Sun Jun 29 22:15:25 2003
@@ -342,7 +342,7 @@
       DEBUG(std::cerr << "    Self Inlining: " << F.getName() << "\n");
 
       // Handle self recursion by resolving the arguments and return value
-      Graph.mergeInGraph(CS, Graph, 0);
+      Graph.mergeInGraph(CS, F, Graph, 0);
 
     } else {
       // Get the data structure graph for the called function.
@@ -360,7 +360,7 @@
 #endif
       
       // Handle self recursion by resolving the arguments and return value
-      Graph.mergeInGraph(CS, GI,
+      Graph.mergeInGraph(CS, *Callee, GI,
                          DSGraph::KeepModRefBits | 
                          DSGraph::StripAllocaBit | DSGraph::DontCloneCallNodes);
 
@@ -423,28 +423,28 @@
     LastCallSiteIdx = I.getCallSiteIdx();
     
     // Resolve the current call...
-    Function *Callee = *I;
+    Function &Callee = **I;
     DSCallSite &CS = I.getCallSite();
 
-    if (Callee->isExternal()) {
+    if (Callee.isExternal()) {
       // Ignore this case, simple varargs functions we cannot stub out!
-    } else if (SCCFunctions.count(Callee)) {
+    } else if (SCCFunctions.count(&Callee)) {
       // Calling a function in the SCC, ignore it for now!
-      DEBUG(std::cerr << "    SCC CallSite for: " << Callee->getName() << "\n");
+      DEBUG(std::cerr << "    SCC CallSite for: " << Callee.getName() << "\n");
       AuxCallsList.push_back(CS);
     } else {
       // Get the data structure graph for the called function.
       //
-      DSGraph &GI = getDSGraph(*Callee);  // Graph to inline
+      DSGraph &GI = getDSGraph(Callee);  // Graph to inline
 
-      DEBUG(std::cerr << "    Inlining graph for " << Callee->getName()
+      DEBUG(std::cerr << "    Inlining graph for " << Callee.getName()
             << "[" << GI.getGraphSize() << "+"
             << GI.getAuxFunctionCalls().size() << "] into: " << F.getName()
             << "[" << Graph.getGraphSize() << "+"
             << Graph.getAuxFunctionCalls().size() << "]\n");
 
       // Handle self recursion by resolving the arguments and return value
-      Graph.mergeInGraph(CS, GI,
+      Graph.mergeInGraph(CS, Callee, GI,
                          DSGraph::KeepModRefBits | DSGraph::StripAllocaBit |
                          DSGraph::DontCloneCallNodes);
     }
@@ -514,7 +514,7 @@
         DEBUG(std::cerr << "    Self Inlining: " << F.getName() << "\n");
         
         // Handle self recursion by resolving the arguments and return value
-        Graph.mergeInGraph(CS, Graph, 0);
+        Graph.mergeInGraph(CS, *Callee, Graph, 0);
       } else if (SCCCallSiteMap.count(Callee)) {
         // We have already seen a call site in the SCC for this function, just
         // merge the two call sites together and we are done.
@@ -530,7 +530,7 @@
               << Graph.getAuxFunctionCalls().size() << "]\n");
         
         // Handle self recursion by resolving the arguments and return value
-        Graph.mergeInGraph(CS, GI,
+        Graph.mergeInGraph(CS, *Callee, GI,
                            DSGraph::KeepModRefBits | DSGraph::StripAllocaBit |
                            DSGraph::DontCloneCallNodes);
 


Index: llvm/lib/Analysis/DataStructure/DataStructure.cpp
diff -u llvm/lib/Analysis/DataStructure/DataStructure.cpp:1.104 llvm/lib/Analysis/DataStructure/DataStructure.cpp:1.105
--- llvm/lib/Analysis/DataStructure/DataStructure.cpp:1.104	Sun Jun 29 17:36:31 2003
+++ llvm/lib/Analysis/DataStructure/DataStructure.cpp	Sun Jun 29 22:15:25 2003
@@ -703,24 +703,23 @@
 // DSGraph Implementation
 //===----------------------------------------------------------------------===//
 
-DSGraph::DSGraph(const DSGraph &G) : Func(G.Func), GlobalsGraph(0) {
+DSGraph::DSGraph(const DSGraph &G) : GlobalsGraph(0) {
   PrintAuxCalls = false;
-  hash_map<const DSNode*, DSNodeHandle> NodeMap;
-  RetNode = cloneInto(G, ScalarMap, NodeMap);
+  NodeMapTy NodeMap;
+  cloneInto(G, ScalarMap, ReturnNodes, NodeMap);
 }
 
-DSGraph::DSGraph(const DSGraph &G,
-                 hash_map<const DSNode*, DSNodeHandle> &NodeMap)
-  : Func(G.Func), GlobalsGraph(0) {
+DSGraph::DSGraph(const DSGraph &G, NodeMapTy &NodeMap)
+  : GlobalsGraph(0) {
   PrintAuxCalls = false;
-  RetNode = cloneInto(G, ScalarMap, NodeMap);
+  cloneInto(G, ScalarMap, ReturnNodes, NodeMap);
 }
 
 DSGraph::~DSGraph() {
   FunctionCalls.clear();
   AuxFunctionCalls.clear();
   ScalarMap.clear();
-  RetNode.setNode(0);
+  ReturnNodes.clear();
 
   // Drop all intra-node references, so that assertions don't fail...
   std::for_each(Nodes.begin(), Nodes.end(),
@@ -746,16 +745,15 @@
 }
 
 
-// cloneInto - Clone the specified DSGraph into the current graph, returning the
-// Return node of the graph.  The translated ScalarMap for the old function is
-// filled into the OldValMap member.  If StripAllocas is set to true, Alloca
-// markers are removed from the graph, as the graph is being cloned into a
-// calling function's graph.
-//
-DSNodeHandle DSGraph::cloneInto(const DSGraph &G, 
-                                hash_map<Value*, DSNodeHandle> &OldValMap,
-                              hash_map<const DSNode*, DSNodeHandle> &OldNodeMap,
-                                unsigned CloneFlags) {
+/// cloneInto - Clone the specified DSGraph into the current graph.  The
+/// translated ScalarMap for the old function is filled into the OldValMap
+/// member, and the translated ReturnNodes map is returned into ReturnNodes.
+///
+/// The CloneFlags member controls various aspects of the cloning process.
+///
+void DSGraph::cloneInto(const DSGraph &G, ScalarMapTy &OldValMap,
+                        ReturnNodesTy &OldReturnNodes, NodeMapTy &OldNodeMap,
+                        unsigned CloneFlags) {
   assert(OldNodeMap.empty() && "Returned OldNodeMap should be empty!");
   assert(&G != this && "Cannot clone graph into itself!");
 
@@ -816,10 +814,15 @@
       AuxFunctionCalls.push_back(DSCallSite(G.AuxFunctionCalls[i], OldNodeMap));
   }
 
-  // Return the returned node pointer...
-  DSNodeHandle &MappedRet = OldNodeMap[G.RetNode.getNode()];
-  return DSNodeHandle(MappedRet.getNode(),
-                      MappedRet.getOffset()+G.RetNode.getOffset());
+  // Map the return node pointers over...
+  for (ReturnNodesTy::const_iterator I = G.getReturnNodes().begin(),
+         E = G.getReturnNodes().end(); I != E; ++I) {
+    const DSNodeHandle &Ret = I->second;
+    DSNodeHandle &MappedRet = OldNodeMap[Ret.getNode()];
+    OldReturnNodes.insert(std::make_pair(I->first,
+                          DSNodeHandle(MappedRet.getNode(),
+                                       MappedRet.getOffset()+Ret.getOffset())));
+  }
 }
 
 /// mergeInGraph - The method is used for merging graphs together.  If the
@@ -827,25 +830,27 @@
 /// merges the nodes specified in the call site with the formal arguments in the
 /// graph.
 ///
-void DSGraph::mergeInGraph(DSCallSite &CS, const DSGraph &Graph,
+void DSGraph::mergeInGraph(DSCallSite &CS, Function &F, const DSGraph &Graph,
                            unsigned CloneFlags) {
-  hash_map<Value*, DSNodeHandle> OldValMap;
+  ScalarMapTy OldValMap;
+  ScalarMapTy *ScalarMap = &OldValMap;
   DSNodeHandle RetVal;
-  hash_map<Value*, DSNodeHandle> *ScalarMap = &OldValMap;
 
   // If this is not a recursive call, clone the graph into this graph...
   if (&Graph != this) {
     // Clone the callee's graph into the current graph, keeping
     // track of where scalars in the old graph _used_ to point,
     // and of the new nodes matching nodes of the old graph.
-    hash_map<const DSNode*, DSNodeHandle> OldNodeMap;
+    NodeMapTy OldNodeMap;
     
     // The clone call may invalidate any of the vectors in the data
     // structure graph.  Strip locals and don't copy the list of callers
-    RetVal = cloneInto(Graph, OldValMap, OldNodeMap, CloneFlags);
+    ReturnNodesTy OldRetNodes;
+    cloneInto(Graph, OldValMap, OldRetNodes, OldNodeMap, CloneFlags);
+    RetVal = OldRetNodes[&F];
     ScalarMap = &OldValMap;
   } else {
-    RetVal = getRetNode();
+    RetVal = getReturnNodeFor(F);
     ScalarMap = &getScalarMap();
   }
 
@@ -853,7 +858,6 @@
   RetVal.mergeWith(CS.getRetVal());
 
   // Resolve all of the function arguments...
-  Function &F = Graph.getFunction();
   Function::aiterator AI = F.abegin();
 
   for (unsigned i = 0, e = CS.getNumPtrArgs(); i != e; ++i, ++AI) {
@@ -916,10 +920,16 @@
 //
 void DSGraph::markIncompleteNodes(unsigned Flags) {
   // Mark any incoming arguments as incomplete...
-  if ((Flags & DSGraph::MarkFormalArgs) && Func && Func->getName() != "main")
-    for (Function::aiterator I = Func->abegin(), E = Func->aend(); I != E; ++I)
-      if (isPointerType(I->getType()) && ScalarMap.find(I) != ScalarMap.end())
-        markIncompleteNode(ScalarMap[I].getNode());
+  if (Flags & DSGraph::MarkFormalArgs)
+    for (ReturnNodesTy::iterator FI = ReturnNodes.begin(), E =ReturnNodes.end();
+         FI != E; ++FI) {
+      Function &F = *FI->first;
+      if (F.getName() != "main")
+        for (Function::aiterator I = F.abegin(), E = F.aend(); I != E; ++I)
+          if (isPointerType(I->getType()) &&
+              ScalarMap.find(I) != ScalarMap.end())
+            markIncompleteNode(ScalarMap[I].getNode());
+    }
 
   // Mark stuff passed into functions calls as being incomplete...
   if (!shouldPrintAuxCalls())
@@ -954,8 +964,7 @@
   return false;
 }
 
-static void removeIdenticalCalls(std::vector<DSCallSite> &Calls,
-                                 const std::string &where) {
+static void removeIdenticalCalls(std::vector<DSCallSite> &Calls) {
   // Remove trivially identical function calls
   unsigned NumFns = Calls.size();
   std::sort(Calls.begin(), Calls.end());  // Sort by callee as primary key!
@@ -1034,8 +1043,7 @@
   NumCallNodesMerged += NumFns-Calls.size();
 
   DEBUG(if (NumFns != Calls.size())
-          std::cerr << "Merged " << (NumFns-Calls.size())
-                    << " call nodes in " << where << "\n";);
+          std::cerr << "Merged " << (NumFns-Calls.size()) << " call nodes.\n";);
 }
 
 
@@ -1045,8 +1053,8 @@
 // we don't have to perform any non-trivial analysis here.
 //
 void DSGraph::removeTriviallyDeadNodes() {
-  removeIdenticalCalls(FunctionCalls, Func ? Func->getName() : "");
-  removeIdenticalCalls(AuxFunctionCalls, Func ? Func->getName() : "");
+  removeIdenticalCalls(FunctionCalls);
+  removeIdenticalCalls(AuxFunctionCalls);
 
   for (unsigned i = 0; i != Nodes.size(); ++i) {
     DSNode *Node = Nodes[i];
@@ -1195,7 +1203,9 @@
     }
 
   // The return value is alive as well...
-  RetNode.getNode()->markReachableNodes(Alive);
+  for (ReturnNodesTy::iterator I = ReturnNodes.begin(), E = ReturnNodes.end();
+       I != E; ++I)
+    I->second.getNode()->markReachableNodes(Alive);
 
   // Mark any nodes reachable by primary calls as alive...
   for (unsigned i = 0, e = FunctionCalls.size(); i != e; ++i)


Index: llvm/lib/Analysis/DataStructure/Local.cpp
diff -u llvm/lib/Analysis/DataStructure/Local.cpp:1.54 llvm/lib/Analysis/DataStructure/Local.cpp:1.55
--- llvm/lib/Analysis/DataStructure/Local.cpp:1.54	Thu Jun 19 16:15:11 2003
+++ llvm/lib/Analysis/DataStructure/Local.cpp	Sun Jun 29 22:15:25 2003
@@ -62,25 +62,26 @@
   /// graph by performing a single pass over the function in question.
   ///
   class GraphBuilder : InstVisitor<GraphBuilder> {
+    Function &F;
     DSGraph &G;
     std::vector<DSNode*> &Nodes;
     DSNodeHandle &RetNode;               // Node that gets returned...
-    hash_map<Value*, DSNodeHandle> &ScalarMap;
+    DSGraph::ScalarMapTy &ScalarMap;
     std::vector<DSCallSite> &FunctionCalls;
 
   public:
-    GraphBuilder(DSGraph &g, std::vector<DSNode*> &nodes, DSNodeHandle &retNode,
-                 hash_map<Value*, DSNodeHandle> &SM,
+    GraphBuilder(Function &f, DSGraph &g, std::vector<DSNode*> &nodes,
+                 DSNodeHandle &retNode, DSGraph::ScalarMapTy &SM,
                  std::vector<DSCallSite> &fc)
-      : G(g), Nodes(nodes), RetNode(retNode), ScalarMap(SM), FunctionCalls(fc) {
+      : F(f), G(g), Nodes(nodes), RetNode(retNode), ScalarMap(SM),
+        FunctionCalls(fc) {
 
       // Create scalar nodes for all pointer arguments...
-      for (Function::aiterator I = G.getFunction().abegin(),
-             E = G.getFunction().aend(); I != E; ++I)
+      for (Function::aiterator I = F.abegin(), E = F.aend(); I != E; ++I)
         if (isPointerType(I->getType()))
           getValueDest(*I);
 
-      visit(G.getFunction());  // Single pass over the function
+      visit(F);  // Single pass over the function
     }
 
   private:
@@ -139,10 +140,10 @@
 //===----------------------------------------------------------------------===//
 // DSGraph constructor - Simply use the GraphBuilder to construct the local
 // graph.
-DSGraph::DSGraph(Function &F, DSGraph *GG) : Func(&F), GlobalsGraph(GG) {
+DSGraph::DSGraph(Function &F, DSGraph *GG) : GlobalsGraph(GG) {
   PrintAuxCalls = false;
   // Use the graph builder to construct the local version of the graph
-  GraphBuilder B(*this, Nodes, RetNode, ScalarMap, FunctionCalls);
+  GraphBuilder B(F, *this, Nodes, ReturnNodes[&F], ScalarMap, FunctionCalls);
 #ifndef NDEBUG
   Timer::addPeakMemoryMeasurement();
 #endif


Index: llvm/lib/Analysis/DataStructure/Printer.cpp
diff -u llvm/lib/Analysis/DataStructure/Printer.cpp:1.53 llvm/lib/Analysis/DataStructure/Printer.cpp:1.54
--- llvm/lib/Analysis/DataStructure/Printer.cpp:1.53	Sat Jun 21 22:03:52 2003
+++ llvm/lib/Analysis/DataStructure/Printer.cpp	Sun Jun 29 22:15:25 2003
@@ -29,7 +29,10 @@
 
 static std::string getCaption(const DSNode *N, const DSGraph *G) {
   std::stringstream OS;
-  Module *M = G && G->hasFunction() ? G->getFunction().getParent() : 0;
+  Module *M = 0;
+  // Get the module from ONE of the functions in the graph it is available.
+  if (G && !G->getReturnNodes().empty())
+    M = G->getReturnNodes().begin()->first->getParent();
 
   if (N->isNodeCompletelyFolded())
     OS << "FOLDED";
@@ -64,10 +67,16 @@
 template<>
 struct DOTGraphTraits<const DSGraph*> : public DefaultDOTGraphTraits {
   static std::string getGraphName(const DSGraph *G) {
-    if (G->hasFunction())
-      return "Function " + G->getFunction().getName();
-    else
-      return "Global graph";
+    switch (G->getReturnNodes().size()) {
+    case 0: return "Global graph";
+    case 1: return "Function " + G->getReturnNodes().begin()->first->getName();
+    default:
+      std::string Return = "Functions: ";
+      for (DSGraph::ReturnNodesTy::const_iterator I=G->getReturnNodes().begin();
+           I != G->getReturnNodes().end(); ++I)
+        Return += I->first->getName() + " ";
+      return Return;
+    }
   }
 
   static const char *getGraphProperties(const DSGraph *G) {
@@ -88,12 +97,13 @@
   ///
   static void addCustomGraphFeatures(const DSGraph *G,
                                      GraphWriter<const DSGraph*> &GW) {
-    Module *CurMod = G->hasFunction() ? G->getFunction().getParent() : 0;
+    Module *CurMod = 0;
+    if (!G->getReturnNodes().empty())
+      CurMod = G->getReturnNodes().begin()->first->getParent();
 
     // Add scalar nodes to the graph...
-    const hash_map<Value*, DSNodeHandle> &VM = G->getScalarMap();
-    for (hash_map<Value*, DSNodeHandle>::const_iterator I = VM.begin();
-         I != VM.end(); ++I)
+    const DSGraph::ScalarMapTy &VM = G->getScalarMap();
+    for (DSGraph::ScalarMapTy::const_iterator I = VM.begin(); I != VM.end();++I)
       if (!isa<GlobalValue>(I->first)) {
         std::stringstream OS;
         WriteAsOperand(OS, I->first, false, true, CurMod);
@@ -108,16 +118,24 @@
 
 
     // Output the returned value pointer...
-    if (G->getRetNode().getNode() != 0) {
-      // Output the return node...
-      GW.emitSimpleNode((void*)1, "plaintext=circle", "returning");
-
-      // Add edge from return node to real destination
-      int RetEdgeDest = G->getRetNode().getOffset() >> DS::PointerShift;;
-      if (RetEdgeDest == 0) RetEdgeDest = -1;
-      GW.emitEdge((void*)1, -1, G->getRetNode().getNode(),
-                  RetEdgeDest, "arrowtail=tee,color=gray63");
-    }
+    const DSGraph::ReturnNodesTy &RetNodes = G->getReturnNodes();
+    for (DSGraph::ReturnNodesTy::const_iterator I = RetNodes.begin(),
+           E = RetNodes.end(); I != E; ++I)
+      if (I->second.getNode()) {
+        std::string Label;
+        if (RetNodes.size() == 1)
+          Label = "returning";
+        else
+          Label = I->first->getName() + " ret node";
+        // Output the return node...
+        GW.emitSimpleNode((void*)1, "plaintext=circle", Label);
+
+        // Add edge from return node to real destination
+        int RetEdgeDest = I->second.getOffset() >> DS::PointerShift;;
+        if (RetEdgeDest == 0) RetEdgeDest = -1;
+        GW.emitEdge((void*)1, -1, I->second.getNode(),
+                    RetEdgeDest, "arrowtail=tee,color=gray63");
+      }
 
     // Output all of the call nodes...
     const std::vector<DSCallSite> &FCs =


Index: llvm/lib/Analysis/DataStructure/Steensgaard.cpp
diff -u llvm/lib/Analysis/DataStructure/Steensgaard.cpp:1.28 llvm/lib/Analysis/DataStructure/Steensgaard.cpp:1.29
--- llvm/lib/Analysis/DataStructure/Steensgaard.cpp:1.28	Wed Feb 26 13:29:36 2003
+++ llvm/lib/Analysis/DataStructure/Steensgaard.cpp	Sun Jun 29 22:15:25 2003
@@ -111,7 +111,7 @@
   // RetValMap - Keep track of the return values for all functions that return
   // valid pointers.
   //
-  hash_map<Function*, DSNodeHandle> RetValMap;
+  DSGraph::ReturnNodesTy RetValMap;
 
   // Loop over the rest of the module, merging graphs for non-external functions
   // into this graph.
@@ -119,22 +119,16 @@
   unsigned Count = 0;
   for (Module::iterator I = M.begin(), E = M.end(); I != E; ++I)
     if (!I->isExternal()) {
-      hash_map<Value*, DSNodeHandle> ValMap;
+      DSGraph::ScalarMapTy ValMap;
       {  // Scope to free NodeMap memory ASAP
-        hash_map<const DSNode*, DSNodeHandle> NodeMap;
+        DSGraph::NodeMapTy NodeMap;
         const DSGraph &FDSG = LDS.getDSGraph(*I);
-        DSNodeHandle RetNode = ResultGraph->cloneInto(FDSG, ValMap, NodeMap);
-
-        // Keep track of the return node of the function's graph if it returns a
-        // value...
-        //
-        if (RetNode.getNode())
-          RetValMap[I] = RetNode;
+        ResultGraph->cloneInto(FDSG, ValMap, RetValMap, NodeMap);
       }
 
       // Incorporate the inlined Function's ScalarMap into the global
       // ScalarMap...
-      hash_map<Value*, DSNodeHandle> &GVM = ResultGraph->getScalarMap();
+      DSGraph::ScalarMapTy &GVM = ResultGraph->getScalarMap();
       for (hash_map<Value*, DSNodeHandle>::iterator I = ValMap.begin(),
              E = ValMap.end(); I != E; ++I)
         GVM[I->first].mergeWith(I->second);


Index: llvm/lib/Analysis/DataStructure/TopDownClosure.cpp
diff -u llvm/lib/Analysis/DataStructure/TopDownClosure.cpp:1.43 llvm/lib/Analysis/DataStructure/TopDownClosure.cpp:1.44
--- llvm/lib/Analysis/DataStructure/TopDownClosure.cpp:1.43	Sun Jun 29 17:37:07 2003
+++ llvm/lib/Analysis/DataStructure/TopDownClosure.cpp	Sun Jun 29 22:15:25 2003
@@ -57,6 +57,7 @@
   GlobalsGraph = 0;
 }
 
+#if 0
 /// ResolveCallSite - This method is used to link the actual arguments together
 /// with the formal arguments for a function call in the top-down closure.  This
 /// method assumes that the call site arguments have been mapped into nodes
@@ -82,6 +83,7 @@
   if (CallSite.getRetVal().getNode() && Graph.getRetNode().getNode())
     Graph.getRetNode().mergeWith(CallSite.getRetVal());
 }
+#endif
 
 DSGraph &TDDataStructures::getOrCreateDSGraph(Function &F) {
   DSGraph *&G = DSInfo[&F];
@@ -169,13 +171,15 @@
                         << "'\n");
       
         // Clone our current graph into the callee...
-        hash_map<Value*, DSNodeHandle> OldValMap;
-        hash_map<const DSNode*, DSNodeHandle> OldNodeMap;
-        CG.cloneInto(Graph, OldValMap, OldNodeMap,
+        DSGraph::ScalarMapTy OldValMap;
+        DSGraph::NodeMapTy OldNodeMap;
+        DSGraph::ReturnNodesTy ReturnNodes;
+        CG.cloneInto(Graph, OldValMap, ReturnNodes, OldNodeMap,
                      DSGraph::StripModRefBits |
                      DSGraph::KeepAllocaBit | DSGraph::DontCloneCallNodes |
                      DSGraph::DontCloneAuxCallNodes);
         OldValMap.clear();  // We don't care about the ValMap
+        ReturnNodes.clear();  // We don't care about return values either
 
         // Loop over all of the invocation sites of the callee, resolving
         // arguments to our graph.  This loop may iterate multiple times if the
@@ -187,7 +191,7 @@
           DSCallSite NewCS(*I->second, OldNodeMap);
         
           // Resolve the return values...
-          NewCS.getRetVal().mergeWith(CG.getRetNode());
+          NewCS.getRetVal().mergeWith(CG.getReturnNodeFor(Callee));
         
           // Resolve all of the arguments...
           Function::aiterator AI = Callee.abegin();




From lattner at cs.uiuc.edu  Sun Jun 29 22:27:03 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Jun 29 22:27:03 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Analysis/DSGraph/SCCSimpleExample.ll
Message-ID: <200306300326.WAA31103@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Analysis/DSGraph:

SCCSimpleExample.ll added (r1.1)

---
Log message:

New testcase


---
Diffs of the changes:

Index: llvm/test/Regression/Analysis/DSGraph/SCCSimpleExample.ll
diff -c /dev/null llvm/test/Regression/Analysis/DSGraph/SCCSimpleExample.ll:1.1
*** /dev/null	Sun Jun 29 22:26:03 2003
--- llvm/test/Regression/Analysis/DSGraph/SCCSimpleExample.ll	Sun Jun 29 22:25:53 2003
***************
*** 0 ****
--- 1,25 ----
+ 
+ ; RUN: analyze %s -datastructure-gc --dsgc-abort-if-merged=Y,BVal
+ 
+ implementation
+ 
+ internal void %F1(int* %X) {
+ 	%Y = alloca int
+ 	store int 4, int* %Y
+ 	%BVal = call int* %F2(int* %Y)
+ 	ret void
+ }
+ 
+ internal int* %F2(int* %A) {
+ 	%B = malloc int
+ 	store int 4, int* %B
+ 	call void %F1(int* %B)
+ 	ret int* %B
+ }
+ 
+ int %main() {
+ 	%Q = malloc int
+ 	store int 4, int* %Q
+ 	call void %F1(int* %Q)
+ 	ret int 0
+ }




From lattner at cs.uiuc.edu  Sun Jun 29 22:37:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Jun 29 22:37:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Analysis/DataStructure/DataStructure.cpp Local.cpp Steensgaard.cpp
Message-ID: <200306300336.WAA31591@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis/DataStructure:

DataStructure.cpp updated: 1.105 -> 1.106
Local.cpp updated: 1.55 -> 1.56
Steensgaard.cpp updated: 1.29 -> 1.30

---
Log message:

Move usages of explicit hash_* datastructures to use typedefs


---
Diffs of the changes:

Index: llvm/lib/Analysis/DataStructure/DataStructure.cpp
diff -u llvm/lib/Analysis/DataStructure/DataStructure.cpp:1.105 llvm/lib/Analysis/DataStructure/DataStructure.cpp:1.106
--- llvm/lib/Analysis/DataStructure/DataStructure.cpp:1.105	Sun Jun 29 22:15:25 2003
+++ llvm/lib/Analysis/DataStructure/DataStructure.cpp	Sun Jun 29 22:36:09 2003
@@ -736,7 +736,7 @@
 /// remapLinks - Change all of the Links in the current node according to the
 /// specified mapping.
 ///
-void DSNode::remapLinks(hash_map<const DSNode*, DSNodeHandle> &OldNodeMap) {
+void DSNode::remapLinks(DSGraph::NodeMapTy &OldNodeMap) {
   for (unsigned i = 0, e = Links.size(); i != e; ++i) {
     DSNodeHandle &H = OldNodeMap[Links[i].getNode()];
     Links[i].setNode(H.getNode());
@@ -782,7 +782,7 @@
     Nodes[i]->remapLinks(OldNodeMap);
 
   // Copy the scalar map... merging all of the global nodes...
-  for (hash_map<Value*, DSNodeHandle>::const_iterator I = G.ScalarMap.begin(),
+  for (ScalarMapTy::const_iterator I = G.ScalarMap.begin(),
          E = G.ScalarMap.end(); I != E; ++I) {
     DSNodeHandle &H = OldValMap[I->first];
     DSNodeHandle &MappedNode = OldNodeMap[I->second.getNode()];
@@ -790,7 +790,7 @@
     H.setNode(MappedNode.getNode());
 
     if (isa<GlobalValue>(I->first)) {  // Is this a global?
-      hash_map<Value*, DSNodeHandle>::iterator GVI = ScalarMap.find(I->first);
+      ScalarMapTy::iterator GVI = ScalarMap.find(I->first);
       if (GVI != ScalarMap.end())     // Is the global value in this fn already?
         GVI->second.mergeWith(H);
       else
@@ -1176,8 +1176,7 @@
   std::vector<std::pair<Value*, DSNode*> > GlobalNodes;
 
   // Mark all nodes reachable by (non-global) scalar nodes as alive...
-  for (hash_map<Value*, DSNodeHandle>::iterator I = ScalarMap.begin(),
-         E = ScalarMap.end(); I != E; )
+  for (ScalarMapTy::iterator I = ScalarMap.begin(), E = ScalarMap.end(); I !=E;)
     if (isa<GlobalValue>(I->first)) {             // Keep track of global nodes
       assert(I->second.getNode() && "Null global node?");
       GlobalNodes.push_back(std::make_pair(I->first, I->second.getNode()));
@@ -1292,7 +1291,7 @@
     //
     for (unsigned i = 0, e = GlobalNodes.size(); i != e; ++i) {
       Value *G = GlobalNodes[i].first;
-      hash_map<Value*, DSNodeHandle>::iterator I = ScalarMap.find(G);
+      ScalarMapTy::iterator I = ScalarMap.find(G);
       assert(I != ScalarMap.end() && "Global not in scalar map anymore?");
       assert(I->second.getNode() && "Global not pointing to anything?");
       assert(!Alive.count(I->second.getNode()) && "Node is alive??");
@@ -1325,7 +1324,7 @@
   for (unsigned i = 0, e = Nodes.size(); i != e; ++i)
     Nodes[i]->assertOK();
   return;  // FIXME: remove
-  for (hash_map<Value*, DSNodeHandle>::const_iterator I = ScalarMap.begin(),
+  for (ScalarMapTy::const_iterator I = ScalarMap.begin(),
          E = ScalarMap.end(); I != E; ++I) {
     assert(I->second.getNode() && "Null node in scalarmap!");
     AssertNodeInGraph(I->second.getNode());


Index: llvm/lib/Analysis/DataStructure/Local.cpp
diff -u llvm/lib/Analysis/DataStructure/Local.cpp:1.55 llvm/lib/Analysis/DataStructure/Local.cpp:1.56
--- llvm/lib/Analysis/DataStructure/Local.cpp:1.55	Sun Jun 29 22:15:25 2003
+++ llvm/lib/Analysis/DataStructure/Local.cpp	Sun Jun 29 22:36:09 2003
@@ -149,10 +149,9 @@
 #endif
 
   // Remove all integral constants from the scalarmap!
-  for (hash_map<Value*, DSNodeHandle>::iterator I = ScalarMap.begin();
-       I != ScalarMap.end();)
+  for (ScalarMapTy::iterator I = ScalarMap.begin(); I != ScalarMap.end();)
     if (isa<ConstantIntegral>(I->first)) {
-      hash_map<Value*, DSNodeHandle>::iterator J = I++;
+      ScalarMapTy::iterator J = I++;
       ScalarMap.erase(J);
     } else
       ++I;
@@ -190,7 +189,7 @@
         NH = getValueDest(*CE->getOperand(0));
       else if (CE->getOpcode() == Instruction::GetElementPtr) {
         visitGetElementPtrInst(*CE);
-        hash_map<Value*, DSNodeHandle>::iterator I = ScalarMap.find(CE);
+        DSGraph::ScalarMapTy::iterator I = ScalarMap.find(CE);
         assert(I != ScalarMap.end() && "GEP didn't get processed right?");
         NH = I->second;
       } else {


Index: llvm/lib/Analysis/DataStructure/Steensgaard.cpp
diff -u llvm/lib/Analysis/DataStructure/Steensgaard.cpp:1.29 llvm/lib/Analysis/DataStructure/Steensgaard.cpp:1.30
--- llvm/lib/Analysis/DataStructure/Steensgaard.cpp:1.29	Sun Jun 29 22:15:25 2003
+++ llvm/lib/Analysis/DataStructure/Steensgaard.cpp	Sun Jun 29 22:36:09 2003
@@ -77,7 +77,7 @@
 void Steens::ResolveFunctionCall(Function *F, const DSCallSite &Call,
                                  DSNodeHandle &RetVal) {
   assert(ResultGraph != 0 && "Result graph not allocated!");
-  hash_map<Value*, DSNodeHandle> &ValMap = ResultGraph->getScalarMap();
+  DSGraph::ScalarMapTy &ValMap = ResultGraph->getScalarMap();
 
   // Handle the return value of the function...
   if (Call.getRetVal().getNode() && RetVal.getNode())
@@ -87,7 +87,7 @@
   unsigned PtrArgIdx = 0;
   for (Function::aiterator AI = F->abegin(), AE = F->aend();
        AI != AE && PtrArgIdx < Call.getNumPtrArgs(); ++AI) {
-    hash_map<Value*, DSNodeHandle>::iterator I = ValMap.find(AI);
+    DSGraph::ScalarMapTy::iterator I = ValMap.find(AI);
     if (I != ValMap.end())    // If its a pointer argument...
       I->second.mergeWith(Call.getPtrArg(PtrArgIdx++));
   }
@@ -129,7 +129,7 @@
       // Incorporate the inlined Function's ScalarMap into the global
       // ScalarMap...
       DSGraph::ScalarMapTy &GVM = ResultGraph->getScalarMap();
-      for (hash_map<Value*, DSNodeHandle>::iterator I = ValMap.begin(),
+      for (DSGraph::ScalarMapTy::iterator I = ValMap.begin(),
              E = ValMap.end(); I != E; ++I)
         GVM[I->first].mergeWith(I->second);
 
@@ -202,12 +202,12 @@
   // FIXME: HANDLE Size argument!
   assert(ResultGraph && "Result graph has not been computed yet!");
 
-  hash_map<Value*, DSNodeHandle> &GSM = ResultGraph->getScalarMap();
+  DSGraph::ScalarMapTy &GSM = ResultGraph->getScalarMap();
 
-  hash_map<Value*, DSNodeHandle>::iterator I = GSM.find(const_cast<Value*>(V1));
+  DSGraph::ScalarMapTy::iterator I = GSM.find(const_cast<Value*>(V1));
   if (I != GSM.end() && I->second.getNode()) {
     DSNodeHandle &V1H = I->second;
-    hash_map<Value*, DSNodeHandle>::iterator J=GSM.find(const_cast<Value*>(V2));
+    DSGraph::ScalarMapTy::iterator J=GSM.find(const_cast<Value*>(V2));
     if (J != GSM.end() && J->second.getNode()) {
       DSNodeHandle &V2H = J->second;
       // If the two pointers point to different data structure graph nodes, they




From lattner at cs.uiuc.edu  Sun Jun 29 23:54:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Jun 29 23:54:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Analysis/DataStructure/TopDownClosure.cpp
Message-ID: <200306300453.XAA02563@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis/DataStructure:

TopDownClosure.cpp updated: 1.45 -> 1.46

---
Log message:

Remove dead method


---
Diffs of the changes:

Index: llvm/lib/Analysis/DataStructure/TopDownClosure.cpp
diff -u llvm/lib/Analysis/DataStructure/TopDownClosure.cpp:1.45 llvm/lib/Analysis/DataStructure/TopDownClosure.cpp:1.46
--- llvm/lib/Analysis/DataStructure/TopDownClosure.cpp:1.45	Sun Jun 29 23:53:08 2003
+++ llvm/lib/Analysis/DataStructure/TopDownClosure.cpp	Sun Jun 29 23:53:39 2003
@@ -59,33 +59,6 @@
   GlobalsGraph = 0;
 }
 
-#if 0
-/// ResolveCallSite - This method is used to link the actual arguments together
-/// with the formal arguments for a function call in the top-down closure.  This
-/// method assumes that the call site arguments have been mapped into nodes
-/// local to the specified graph.
-///
-void TDDataStructures::ResolveCallSite(DSGraph &Graph,
-                                       const DSCallSite &CallSite) {
-  // Resolve all of the function formal arguments...
-  Function &F = Graph.getFunction();
-  Function::aiterator AI = F.abegin();
-
-  for (unsigned i = 0, e = CallSite.getNumPtrArgs(); i != e; ++i, ++AI) {
-    // Advance the argument iterator to the first pointer argument...
-    while (!DS::isPointerType(AI->getType())) ++AI;
-    
-    // TD ...Merge the formal arg scalar with the actual arg node
-    DSNodeHandle &NodeForFormal = Graph.getNodeForValue(AI);
-    assert(NodeForFormal.getNode() && "Pointer argument has no dest node!");
-    NodeForFormal.mergeWith(CallSite.getPtrArg(i));
-  }
-  
-  // Merge returned node in the caller with the "return" node in callee
-  if (CallSite.getRetVal().getNode() && Graph.getRetNode().getNode())
-    Graph.getRetNode().mergeWith(CallSite.getRetVal());
-}
-#endif
 
 DSGraph &TDDataStructures::getOrCreateDSGraph(Function &F) {
   DSGraph *&G = DSInfo[&F];




From lattner at cs.uiuc.edu  Sun Jun 29 23:54:03 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Jun 29 23:54:03 2003
Subject: [llvm-commits] CVS: llvm/lib/Analysis/DataStructure/TopDownClosure.cpp
Message-ID: <200306300453.XAA02556@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis/DataStructure:

TopDownClosure.cpp updated: 1.44 -> 1.45

---
Log message:

Do not multiply delete graphs if functions are sharing graphs


---
Diffs of the changes:

Index: llvm/lib/Analysis/DataStructure/TopDownClosure.cpp
diff -u llvm/lib/Analysis/DataStructure/TopDownClosure.cpp:1.44 llvm/lib/Analysis/DataStructure/TopDownClosure.cpp:1.45
--- llvm/lib/Analysis/DataStructure/TopDownClosure.cpp:1.44	Sun Jun 29 22:15:25 2003
+++ llvm/lib/Analysis/DataStructure/TopDownClosure.cpp	Sun Jun 29 23:53:08 2003
@@ -45,10 +45,12 @@
 // has no way to extend the lifetime of the pass, which screws up ds-aa.
 //
 void TDDataStructures::releaseMyMemory() {
-  return;
-  for (hash_map<const Function*, DSGraph*>::iterator I = DSInfo.begin(),
-         E = DSInfo.end(); I != E; ++I)
-    delete I->second;
+  for (hash_map<Function*, DSGraph*>::iterator I = DSInfo.begin(),
+         E = DSInfo.end(); I != E; ++I) {
+    I->second->getReturnNodes().erase(I->first);
+    if (I->second->getReturnNodes().empty())
+      delete I->second;
+  }
 
   // Empty map so next time memory is released, data structures are not
   // re-deleted.




From lattner at cs.uiuc.edu  Sun Jun 29 23:55:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Jun 29 23:55:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Analysis/DataStructure/Local.cpp
Message-ID: <200306300454.XAA02570@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis/DataStructure:

Local.cpp updated: 1.56 -> 1.57

---
Log message:

Do not delete the same graph multiple times when freeing memory if graphs are sharing


---
Diffs of the changes:

Index: llvm/lib/Analysis/DataStructure/Local.cpp
diff -u llvm/lib/Analysis/DataStructure/Local.cpp:1.56 llvm/lib/Analysis/DataStructure/Local.cpp:1.57
--- llvm/lib/Analysis/DataStructure/Local.cpp:1.56	Sun Jun 29 22:36:09 2003
+++ llvm/lib/Analysis/DataStructure/Local.cpp	Sun Jun 29 23:53:27 2003
@@ -487,9 +487,12 @@
 // our memory... here...
 //
 void LocalDataStructures::releaseMemory() {
-  for (hash_map<const Function*, DSGraph*>::iterator I = DSInfo.begin(),
-         E = DSInfo.end(); I != E; ++I)
-    delete I->second;
+  for (hash_map<Function*, DSGraph*>::iterator I = DSInfo.begin(),
+         E = DSInfo.end(); I != E; ++I) {
+    I->second->getReturnNodes().erase(I->first);
+    if (I->second->getReturnNodes().empty())
+      delete I->second;
+  }
 
   // Empty map so next time memory is released, data structures are not
   // re-deleted.