From gaeke at cs.uiuc.edu  Mon May 17 15:37:02 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Mon May 17 15:37:02 2004
Subject: [llvm-commits] CVS: reopt/lib/TraceToFunction/
Message-ID: <200405172036.PAA21232@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/TraceToFunction:


---
Log message:

Directory /home/vadve/shared/InternalCVS/reopt/lib/TraceToFunction added to the repository


---
Diffs of the changes:  (+0 -0)



From gaeke at cs.uiuc.edu  Mon May 17 15:38:03 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Mon May 17 15:38:03 2004
Subject: [llvm-commits] CVS: reopt/lib/TraceToFunction/Makefile
Message-ID: <200405172037.PAA21245@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/TraceToFunction:

Makefile added (r1.1)

---
Log message:

Splitting traceToFunction transformation out into its own library


---
Diffs of the changes:  (+4 -0)

Index: reopt/lib/TraceToFunction/Makefile
diff -c /dev/null reopt/lib/TraceToFunction/Makefile:1.1
*** /dev/null	Mon May 17 15:37:31 2004
--- reopt/lib/TraceToFunction/Makefile	Mon May 17 15:37:21 2004
***************
*** 0 ****
--- 1,4 ----
+ LEVEL = ../..
+ LIBRARYNAME = tracetofunction
+ 
+ include $(LEVEL)/Makefile.common




From gaeke at cs.uiuc.edu  Mon May 17 15:57:04 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Mon May 17 15:57:04 2004
Subject: [llvm-commits] CVS: reopt/lib/TraceCache/InstrUtils.cpp VirtualMem.cpp
Message-ID: <200405172056.PAA22388@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/TraceCache:

InstrUtils.cpp updated: 1.18 -> 1.19
VirtualMem.cpp updated: 1.17 -> 1.18

---
Log message:

Add debug printouts to doFlush and VirtualMem's write methods


---
Diffs of the changes:  (+7 -0)

Index: reopt/lib/TraceCache/InstrUtils.cpp
diff -u reopt/lib/TraceCache/InstrUtils.cpp:1.18 reopt/lib/TraceCache/InstrUtils.cpp:1.19
--- reopt/lib/TraceCache/InstrUtils.cpp:1.18	Mon Apr 26 15:02:59 2004
+++ reopt/lib/TraceCache/InstrUtils.cpp	Mon May 17 15:56:09 2004
@@ -6,6 +6,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "reopt/InstrUtils.h"
+#include "Support/Debug.h"
 #include <cassert>
 #include <iostream>
 
@@ -201,6 +202,7 @@
 }
 
 void doFlush(uint64_t st_addr, uint64_t end_addr){
+  DEBUG (std::cerr << "doFlush(" << st_addr << ", " << end_addr << ")\n");
 #ifdef __sparc__
   if(st_addr % 8 != 0)
     st_addr &= ~7; // Round down to the next lower doubleword boundary


Index: reopt/lib/TraceCache/VirtualMem.cpp
diff -u reopt/lib/TraceCache/VirtualMem.cpp:1.17 reopt/lib/TraceCache/VirtualMem.cpp:1.18
--- reopt/lib/TraceCache/VirtualMem.cpp:1.17	Mon Apr 26 14:49:59 2004
+++ reopt/lib/TraceCache/VirtualMem.cpp	Mon May 17 15:56:09 2004
@@ -9,6 +9,7 @@
 #include "reopt/InstrUtils.h"
 #include "reopt/TraceCache.h"
 #include "reopt/GetTraceTime.h"
+#include "Support/Debug.h"
 #include <cassert>
 #include <string>
 #include <cstdio>
@@ -54,6 +55,8 @@
 /// writeInstToVM - Emit the word newInstr to memory at address destAddr.
 ///
 void VirtualMem::writeInstToVM (uint64_t destAddr, unsigned int newInstr) {
+  DEBUG (std::cerr << "writeInstToVM (" << std::hex << destAddr << ", "
+                   << newInstr << ");\n" << std::dec);
   lseek (fp, destAddr, SEEK_SET);
   write (fp, &newInstr, sizeof (unsigned int));
 }
@@ -63,6 +66,8 @@
 ///
 void VirtualMem::writeTraceToVM (uint64_t destAddr,
                                  std::vector<unsigned int> &newInstrs) {
+  DEBUG (std::cerr << "writeTraceToVM ("<<std::hex << destAddr << ", ["
+                   << std::dec << newInstrs.size() << " instrs]);\n");
   lseek (fp, destAddr, SEEK_SET);
   write (fp, &newInstrs[0], newInstrs.size () * sizeof (unsigned int));
 }




From gaeke at cs.uiuc.edu  Mon May 17 16:01:01 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Mon May 17 16:01:01 2004
Subject: [llvm-commits] CVS: reopt/lib/TraceToFunction/TraceToFunction.cpp
Message-ID: <200405172100.QAA22467@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/TraceToFunction:

TraceToFunction.cpp updated: 1.35 -> 1.36

---
Log message:

TraceToFunction.h has moved.


---
Diffs of the changes:  (+1 -1)

Index: reopt/lib/TraceToFunction/TraceToFunction.cpp
diff -u reopt/lib/TraceToFunction/TraceToFunction.cpp:1.35 reopt/lib/TraceToFunction/TraceToFunction.cpp:1.36
--- reopt/lib/TraceToFunction/TraceToFunction.cpp:1.35	Sat May 15 19:52:05 2004
+++ reopt/lib/TraceToFunction/TraceToFunction.cpp	Mon May 17 16:00:47 2004
@@ -23,7 +23,7 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "TraceToFunction.h"
+#include "reopt/TraceToFunction.h"
 #include "llvm/Transforms/Utils/Cloning.h"
 #include "llvm/Pass.h"
 #include "llvm/Module.h"




From gaeke at cs.uiuc.edu  Mon May 17 16:01:04 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Mon May 17 16:01:04 2004
Subject: [llvm-commits] CVS: reopt/lib/LightWtProfiling/TraceToFunction.cpp TraceToFunction.h
Message-ID: <200405172100.QAA22419@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/LightWtProfiling:

TraceToFunction.cpp (r1.35) removed
TraceToFunction.h (r1.11) removed

---
Log message:

Moving these files.


---
Diffs of the changes:  (+0 -0)



From gaeke at cs.uiuc.edu  Mon May 17 16:01:05 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Mon May 17 16:01:05 2004
Subject: [llvm-commits] CVS: reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp UnpackTraceFunction.cpp
Message-ID: <200405172100.QAA22412@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/LightWtProfiling:

RuntimeOptimizations.cpp updated: 1.31 -> 1.32
UnpackTraceFunction.cpp updated: 1.60 -> 1.61

---
Log message:

TraceToFunction's header file has been moved.


---
Diffs of the changes:  (+2 -2)

Index: reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp
diff -u reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp:1.31 reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp:1.32
--- reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp:1.31	Wed May  5 14:12:32 2004
+++ reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp	Mon May 17 16:00:19 2004
@@ -14,7 +14,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "ReoptimizerInternal.h"
-#include "TraceToFunction.h"
+#include "reopt/TraceToFunction.h"
 #include "reopt/VirtualMem.h"
 #include "reopt/InstrUtils.h"
 #include "Support/Debug.h"


Index: reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp
diff -u reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp:1.60 reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp:1.61
--- reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp:1.60	Mon May 10 16:06:20 2004
+++ reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp	Mon May 17 16:00:19 2004
@@ -13,7 +13,7 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "TraceToFunction.h"
+#include "reopt/TraceToFunction.h"
 #include "llvm/CodeGen/MachineFunctionPass.h"
 #include "llvm/CodeGen/MachineFunctionInfo.h"
 #include "llvm/CodeGen/MachineInstr.h"




From gaeke at cs.uiuc.edu  Mon May 17 16:16:04 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Mon May 17 16:16:04 2004
Subject: [llvm-commits] CVS: llvm/lib/VMCore/Verifier.cpp
Message-ID: <200405172115.QAA23053@seraph.cs.uiuc.edu>


Changes in directory llvm/lib/VMCore:

Verifier.cpp updated: 1.97 -> 1.98

---
Log message:

Fold two assertions with backwards error messages into one with a
correct error message.


---
Diffs of the changes:  (+3 -6)

Index: llvm/lib/VMCore/Verifier.cpp
diff -u llvm/lib/VMCore/Verifier.cpp:1.97 llvm/lib/VMCore/Verifier.cpp:1.98
--- llvm/lib/VMCore/Verifier.cpp:1.97	Tue May  4 16:46:05 2004
+++ llvm/lib/VMCore/Verifier.cpp	Mon May 17 16:15:18 2004
@@ -303,12 +303,9 @@
       Assert1(PN->getNumIncomingValues() != 0,
               "PHI nodes must have at least one entry.  If the block is dead, "
               "the PHI should be removed!", PN);
-      Assert1(PN->getNumIncomingValues() >= Preds.size(),
-              "PHINode has more entries than the basic block has predecessors!",
-              PN);
-      Assert1(PN->getNumIncomingValues() <= Preds.size(),
-              "PHINode has less entries than the basic block has predecessors!",
-              PN);
+      Assert1(PN->getNumIncomingValues() == Preds.size(),
+              "PHINode should have one entry for each predecessor of its "
+              "parent basic block!", PN);
       
       // Get and sort all incoming values in the PHI node...
       std::vector<std::pair<BasicBlock*, Value*> > Values;




From gaeke at cs.uiuc.edu  Mon May 17 16:18:03 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Mon May 17 16:18:03 2004
Subject: [llvm-commits] CVS: reopt/tools/ttftest/
Message-ID: <200405172118.QAA23080@seraph.cs.uiuc.edu>


Changes in directory reopt/tools/ttftest:


---
Log message:

Directory /home/vadve/shared/InternalCVS/reopt/tools/ttftest added to the repository


---
Diffs of the changes:  (+0 -0)



From gaeke at cs.uiuc.edu  Mon May 17 16:18:06 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Mon May 17 16:18:06 2004
Subject: [llvm-commits] CVS: llvm/include/llvm/Analysis/Trace.h
Message-ID: <200405172117.QAA23068@seraph.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Analysis:

Trace.h updated: 1.2 -> 1.3

---
Log message:

Define erase forwarding method on traces


---
Diffs of the changes:  (+3 -0)

Index: llvm/include/llvm/Analysis/Trace.h
diff -u llvm/include/llvm/Analysis/Trace.h:1.2 llvm/include/llvm/Analysis/Trace.h:1.3
--- llvm/include/llvm/Analysis/Trace.h:1.2	Mon Mar  8 15:07:12 2004
+++ llvm/include/llvm/Analysis/Trace.h	Mon May 17 16:17:21 2004
@@ -101,6 +101,9 @@
   unsigned                 size() const { return BasicBlocks.size(); }
   bool                    empty() const { return BasicBlocks.empty(); }
 
+  iterator erase(iterator q)               { return BasicBlocks.erase (q); }
+  iterator erase(iterator q1, iterator q2) { return BasicBlocks.erase (q1, q2); }
+
   /// print - Write trace to output stream.
   ///
   void print (std::ostream &O) const;




From gaeke at cs.uiuc.edu  Mon May 17 16:26:02 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Mon May 17 16:26:02 2004
Subject: [llvm-commits] CVS: reopt/tools/ttftest/Makefile ttftest.cpp
Message-ID: <200405172125.QAA23283@seraph.cs.uiuc.edu>


Changes in directory reopt/tools/ttftest:

Makefile added (r1.1)
ttftest.cpp added (r1.1)

---
Log message:

Add ttftest (TraceToFunction tester) program


---
Diffs of the changes:  (+114 -0)

Index: reopt/tools/ttftest/Makefile
diff -c /dev/null reopt/tools/ttftest/Makefile:1.1
*** /dev/null	Mon May 17 16:25:35 2004
--- reopt/tools/ttftest/Makefile	Mon May 17 16:25:25 2004
***************
*** 0 ****
--- 1,9 ----
+ LEVEL = ../..
+ TOOLNAME = ttftest
+ USEDLIBS = tracetofunction
+ LLVMLIBS = vmcore bcreader analysis.a transformutils.a support.a
+ 
+ TOOLLINKOPTS = $(PLATFORMLIBDL)
+ 
+ include $(LEVEL)/Makefile.common
+ 


Index: reopt/tools/ttftest/ttftest.cpp
diff -c /dev/null reopt/tools/ttftest/ttftest.cpp:1.1
*** /dev/null	Mon May 17 16:25:35 2004
--- reopt/tools/ttftest/ttftest.cpp	Mon May 17 16:25:25 2004
***************
*** 0 ****
--- 1,105 ----
+ //===- ttftest.cpp - Test TraceToFunction transformation ------------------===//
+ // 
+ //                     The LLVM Compiler Infrastructure
+ //
+ // This file was developed by the LLVM research group and is distributed under
+ // the University of Illinois Open Source License. See LICENSE.TXT for
+ // details.
+ // 
+ //===----------------------------------------------------------------------===//
+ //
+ // A simple program for testing TraceToFunction. This reads in a trace
+ // from a file, converts it to a function, and tries to verify the
+ // function.
+ //
+ //===----------------------------------------------------------------------===//
+ 
+ #include "llvm/Analysis/Trace.h"
+ #include "llvm/Analysis/Verifier.h"
+ #include "llvm/Bytecode/Reader.h"
+ #include "reopt/TraceToFunction.h"
+ #include "llvm/Module.h"
+ #include <iostream>
+ #include <iterator>
+ #include <fstream>
+ #include <cerrno>
+ using namespace llvm;
+ 
+ /// getBasicBlockByNum - Returns the basic block in F whose index is
+ /// bbNum.
+ ///
+ BasicBlock *getBasicBlockByNum (Function *F, int bbNum) {
+   Function::iterator block = F->begin();
+   std::advance (block, bbNum);
+   return block;
+ }
+ 
+ /// ReadTraceFromFile - Given a module, read a trace for some function
+ /// in that module from the named file. Returns the Trace object, or a null
+ /// pointer (and sets ErrorStr) if there was an error.
+ ///
+ Trace *ReadTraceFromFile (Module *M, const char *filename,
+                           std::string *ErrorStr) {
+   std::ifstream in (filename);
+   if (!in.good()) {
+     if (ErrorStr)
+       *ErrorStr = std::string("Can't open '") + filename + "': "
+                 + strerror(errno);
+     return 0;
+   }
+   char functionName[160];
+   in.getline(functionName, 160, '\n');
+   Function *F = M->getNamedFunction (functionName);
+   if (!F) {
+     if (ErrorStr)
+       *ErrorStr = std::string("Function '") + functionName
+                 + "' not found in module";
+     return 0;
+   }
+   std::cerr << "Function name: '" << functionName << "'\n" << "Basic blocks: ";
+   std::vector<BasicBlock *> vBB;
+   do {
+     int basicBlockNum;
+     in >> basicBlockNum;
+     if (!in.eof ()) {
+       std::cerr << basicBlockNum << " ";
+       vBB.push_back (getBasicBlockByNum (F, basicBlockNum));
+     }
+   } while (!in.eof ());
+   std::cerr << "\n";
+   return new Trace (vBB);
+ }
+ 
+ int main (int argc, char **argv) {
+   const char *bytecodeFileName = argv[1];
+   const char *traceFileName = argv[2];
+ 
+   // Read bytecode file
+   std::string err;
+   Module *M = ParseBytecodeFile(bytecodeFileName, &err);
+   if (!M) {
+     std::cerr << "Error reading bytecode: " << err << "\n";
+     exit (1);
+   }
+ 
+   // Read trace file
+   std::vector<BasicBlock *> vBB;
+   Trace *T = ReadTraceFromFile (M, traceFileName, &err);
+   if (!T) {
+     std::cerr << "Error reading trace: " << err << "\n";
+     exit (1);
+   }
+ 
+   // Run TraceToFunction on the Trace
+   TraceFunction *TF = TraceFunction::get (*T);
+   std::cerr << *TF->TraceFn << "\n\n";
+ 
+   // Run Verifier on the resulting TraceFunction
+   if (verifyFunction (*TF->TraceFn, PrintMessageAction)) {
+     std::cerr << "Verifier fails.\n";
+     exit (1);
+   } else {
+     std::cerr << "Verifier passes.\n";
+     exit (0);
+   }
+ }




From gaeke at cs.uiuc.edu  Mon May 17 16:32:02 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Mon May 17 16:32:02 2004
Subject: [llvm-commits] CVS: reopt/lib/TraceToFunction/TraceToFunction.cpp
Message-ID: <200405172131.QAA23494@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/TraceToFunction:

TraceToFunction.cpp updated: 1.36 -> 1.37

---
Log message:

Rename isFirstLevelInstrumentationBlock to getFLIEdgeTarget.
Add corresponding getFLIEdgeSource method.
Wrap long comment lines.
Only get branch successor node once, and store it in a value, in
 fixupFunctionBodyBB().


---
Diffs of the changes:  (+34 -10)

Index: reopt/lib/TraceToFunction/TraceToFunction.cpp
diff -u reopt/lib/TraceToFunction/TraceToFunction.cpp:1.36 reopt/lib/TraceToFunction/TraceToFunction.cpp:1.37
--- reopt/lib/TraceToFunction/TraceToFunction.cpp:1.36	Mon May 17 16:00:47 2004
+++ reopt/lib/TraceToFunction/TraceToFunction.cpp	Mon May 17 16:30:55 2004
@@ -410,24 +410,47 @@
     fixupPhisAndCalls (FI, O2CMap);
 }
 
-/// isFirstLevelInstrumentationBlock - returns a nonzero value iff BB is a
+/// getFLIEdgeTarget - returns a nonzero value iff BB is a
 /// basic block containing only a call to the first-level instrumentation
 /// function.  Basic blocks of this form are inserted by the -instloops pass on
 /// loop back-edges. The return value is the target of the back-edge.
 ///
-static BasicBlock *isFirstLevelInstrumentationBlock (BasicBlock *BB) {
+static BasicBlock *getFLIEdgeTarget (BasicBlock *BB) {
   BasicBlock::iterator i = BB->begin ();
   // starts with llvm_first_trigger call
   if (!isFirstTriggerCall (*i))
-    return false;
-  // Next instr. should be an unconditional branch - return its target, or NULL if
-  // there is a mismatch.
+    return 0;
+  // Next instr. should be an unconditional branch - return its target,
+  // or NULL if there is a mismatch.
   BranchInst *BI = dyn_cast<BranchInst> (++i);
   if (!(BI && BI->isUnconditional()))
     return 0;
   return BI->getSuccessor (0);
 }
 
+/// getFLIEdgeSource - returns a nonzero value iff BB is a
+/// basic block containing only a call to the first-level instrumentation
+/// function.  Basic blocks of this form are inserted by the -instloops pass on
+/// loop back-edges. The return value is the source of the back-edge.
+///
+static BasicBlock *getFLIEdgeSource (BasicBlock *BB) {
+  BasicBlock::iterator i = BB->begin ();
+  // starts with llvm_first_trigger call
+  if (!isFirstTriggerCall (*i))
+    return 0;
+  // Next instr. should be an unconditional branch. Return NULL if
+  // there is a mismatch.
+  BranchInst *BI = dyn_cast<BranchInst> (++i);
+  if (!(BI && BI->isUnconditional()))
+    return 0;
+  // Use CFG predecessor iterator to get the predecessor of the FLI block
+  pred_iterator PI = pred_begin (BB);
+  DEBUG(pred_iterator NextPI = PI; ++NextPI;
+        assert (NextPI == pred_end (BB) &&
+                "FLI block must have only one predecessor"));
+  return *PI;
+}
+
 /// fixupFunctionBodyBB - Given srcB in T and its clone dstB in F, and
 /// the map O2CMap detailing the correspondences between values in T
 /// and values in F, fix up dstB so that its contents are internally
@@ -480,13 +503,14 @@
 	   && "Trace-exiting branch's clone is null, or not a branch?");
     BranchInst *BIinF = cast<BranchInst> (V);
     for (unsigned i = 0; i < BI->getNumSuccessors (); ++i) {
+      BasicBlock *successor = BI->getSuccessor (i);
       if (BasicBlock *backEdgeTarget =
-          isFirstLevelInstrumentationBlock (BI->getSuccessor (i))) {
+          getFLIEdgeTarget (successor)) {
         std::cerr << "Branch instr " << *BI << " looks like an instrumented "
                   << "backedge from " << srcB->getName() << " to "
                   << backEdgeTarget->getName () << "\n";
       }
-      if (!T.contains (BI->getSuccessor (i))) {
+      if (!T.contains (successor)) {
 	// This is a trace-exiting destination of branch BI. Create a new
 	// basic block FB for its destination in the function.
 	std::string name ("exitfixup" + utostr (BranchNumber[BI]) + "_"
@@ -496,7 +520,7 @@
 	BIinF->setSuccessor (i, FB);
     // Remember that FB's "return" instruction would involve a
     // return to the off-trace successor we just found.
-    TF->ReturnBlockForTraceExit[FB] = BI->getSuccessor (i);
+    TF->ReturnBlockForTraceExit[FB] = successor;
 	// Add the getelementptr/store instruction pairs here that
 	// correspond to each live-out variable.
 	unsigned Slot = 0;
@@ -524,13 +548,13 @@
       } else {
 	// This is a non-trace-exiting destination of branch BI. Fix up
 	// its clone's destination to point to its successor's clone.
-	Value *V2 = O2CMap[BI->getSuccessor (i)];
+	Value *V2 = O2CMap[successor];
 	DEBUG(if (!V2) {
 	  std::cerr
 	    << "I'm confused: saw non-trace-exiting dest. of branch "
 	    << *BI << " whose clone is " << *BIinF
 	    << " successor number is " << i << " and successor is "
-	    << *BI->getSuccessor (i) << ", but where's its clone?\n";
+	    << *successor << ", but where's its clone?\n";
 	  assert(V2 && "Clone of basic block on trace is null?");
 	});
 	assert(isa<BasicBlock> (V2)




From gaeke at cs.uiuc.edu  Mon May 17 16:38:02 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Mon May 17 16:38:02 2004
Subject: [llvm-commits] CVS: reopt/lib/TraceToFunction/TraceToFunction.cpp
Message-ID: <200405172137.QAA23768@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/TraceToFunction:

TraceToFunction.cpp updated: 1.37 -> 1.38

---
Log message:

If we find FLI blocks in the trace, remove them. We don't want any FLI in the
optimized trace code.


---
Diffs of the changes:  (+9 -0)

Index: reopt/lib/TraceToFunction/TraceToFunction.cpp
diff -u reopt/lib/TraceToFunction/TraceToFunction.cpp:1.37 reopt/lib/TraceToFunction/TraceToFunction.cpp:1.38
--- reopt/lib/TraceToFunction/TraceToFunction.cpp:1.37	Mon May 17 16:30:55 2004
+++ reopt/lib/TraceToFunction/TraceToFunction.cpp	Mon May 17 16:37:13 2004
@@ -384,6 +384,8 @@
   }
 }
 
+static BasicBlock *getFLIEdgeTarget (BasicBlock *BB);
+
 /// fillInFunctionBody - Clone the BBs of T into F, then do all
 /// necessary fixups to ensure that F's new contents are internally
 /// consistent and that the live-outs So get stored in F's first
@@ -392,6 +394,13 @@
 void TraceFunctionBuilder::fillInFunctionBody (Trace &T, Function *F,
 					  LiveVariableSet &So) {
   ValueMap &O2CMap = TF->O2CMap;
+
+  // Remove any FLI blocks which ended up in the trace.
+  Trace::iterator i;
+  while ((i = std::find_if (T.begin (), T.end (), getFLIEdgeTarget))
+         != T.end ())
+    T.erase (i);
+  
   // First copy the basic blocks from the trace.
   cloneTraceBBsIntoFunction (T, F, O2CMap);
 




From gaeke at cs.uiuc.edu  Mon May 17 16:39:01 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Mon May 17 16:39:01 2004
Subject: [llvm-commits] CVS: reopt/lib/TraceToFunction/TraceToFunction.cpp
Message-ID: <200405172138.QAA23786@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/TraceToFunction:

TraceToFunction.cpp updated: 1.38 -> 1.39

---
Log message:

Removing FLI calls should no longer be necessary at this later stage.


---
Diffs of the changes:  (+0 -6)

Index: reopt/lib/TraceToFunction/TraceToFunction.cpp
diff -u reopt/lib/TraceToFunction/TraceToFunction.cpp:1.38 reopt/lib/TraceToFunction/TraceToFunction.cpp:1.39
--- reopt/lib/TraceToFunction/TraceToFunction.cpp:1.38	Mon May 17 16:37:13 2004
+++ reopt/lib/TraceToFunction/TraceToFunction.cpp	Mon May 17 16:38:18 2004
@@ -364,12 +364,6 @@
 		  "Sorry, fixupPhisAndCalls mishandled a Phi node"));
   }
 
-  // Remove calls to first-level instrumentation if we find them.
-  for (BasicBlock::iterator BI = dstB->begin (), BE = dstB->end ();
-       BI != BE; ++BI)
-    if (isFirstTriggerCall (*BI))
-      goners.push_back (BI);
-
   while (!goners.empty ()) {
     assert (goners.back ()->use_size () == 0
 	    && "Whoops, I was going to delete something which still has uses");




From gaeke at cs.uiuc.edu  Mon May 17 16:50:02 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Mon May 17 16:50:02 2004
Subject: [llvm-commits] CVS: reopt/lib/TraceToFunction/TraceToFunction.cpp
Message-ID: <200405172149.QAA23896@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/TraceToFunction:

TraceToFunction.cpp updated: 1.39 -> 1.40

---
Log message:

Fold branches that go to FLI blocks, and PHI sources that come from FLI blocks.
Untabify.


---
Diffs of the changes:  (+25 -12)

Index: reopt/lib/TraceToFunction/TraceToFunction.cpp
diff -u reopt/lib/TraceToFunction/TraceToFunction.cpp:1.39 reopt/lib/TraceToFunction/TraceToFunction.cpp:1.40
--- reopt/lib/TraceToFunction/TraceToFunction.cpp:1.39	Mon May 17 16:38:18 2004
+++ reopt/lib/TraceToFunction/TraceToFunction.cpp	Mon May 17 16:49:37 2004
@@ -482,17 +482,22 @@
     // to guess which one it is. Assume it's any phi value source in
     // the entry BB which is not in the trace.
     for (BasicBlock::iterator BI = srcB->begin ();
-	 PHINode *PN = dyn_cast<PHINode> (BI); ++BI)
-      for (unsigned i = 0; i < PN->getNumIncomingValues (); ++i)
-	if (!T.contains (PN->getIncomingBlock (i))) {
-	  // FIXME: Assert that O2CMap[PN]'s value i is live in.
-	  Value *V = O2CMap[PN];
-	  assert (V && "Can't find clone of Phi node from trace entry BB");
-	  PHINode *PNinF = dyn_cast<PHINode> (V);
-	  assert (PNinF
-		  && "Clone of Phi node from entry BB is not a Phi node");
-	  PNinF->setIncomingBlock (i, EntryFixup);
-	}
+         PHINode *PN = dyn_cast<PHINode> (BI); ++BI)
+      for (unsigned i = 0; i < PN->getNumIncomingValues (); ++i) {
+        // Fold Phi sources which come from FLI blocks
+        BasicBlock *phiSource = PN->getIncomingBlock (i);
+        if (BasicBlock *realSource = getFLIEdgeSource (phiSource))
+          phiSource = realSource;
+        if (!T.contains (phiSource)) {
+          // FIXME: Assert that O2CMap[PN]'s value i is live in.
+          Value *V = O2CMap[PN];
+          assert (V && "Can't find clone of Phi node from trace entry BB");
+          PHINode *PNinF = dyn_cast<PHINode> (V);
+          assert (PNinF
+                  && "Clone of Phi node from entry BB is not a Phi node");
+          PNinF->setIncomingBlock (i, EntryFixup);
+        }
+     }
   }
 
   // If srcB contains a trace-exiting branch B, fix up B's clone in
@@ -507,11 +512,13 @@
     BranchInst *BIinF = cast<BranchInst> (V);
     for (unsigned i = 0; i < BI->getNumSuccessors (); ++i) {
       BasicBlock *successor = BI->getSuccessor (i);
+      // Fold branches which go to FLI blocks
       if (BasicBlock *backEdgeTarget =
           getFLIEdgeTarget (successor)) {
         std::cerr << "Branch instr " << *BI << " looks like an instrumented "
-                  << "backedge from " << srcB->getName() << " to "
+                  << "back-edge from " << srcB->getName() << " to "
                   << backEdgeTarget->getName () << "\n";
+        successor = backEdgeTarget;
       }
       if (!T.contains (successor)) {
 	// This is a trace-exiting destination of branch BI. Create a new
@@ -573,6 +580,12 @@
     Instruction &I = *BI;
     for (unsigned i = 0; i < I.getNumOperands (); ++i) {
       Value *V = I.getOperand (i);
+      // Fold Phi sources which come from FLI blocks
+      if (isa<PHINode> (&I))
+        if (BasicBlock *phiSource = dyn_cast<BasicBlock> (V))
+          if (BasicBlock *realSource = getFLIEdgeSource (phiSource))
+            V = realSource;
+
       // If the instruction I has an operand which is in the live-in
       // set of the trace, then we must replace that operand with
       // the corresponding argument of F. We can find out which




From brukman at cs.uiuc.edu  Mon May 17 17:25:02 2004
From: brukman at cs.uiuc.edu (Misha Brukman)
Date: Mon May 17 17:25:02 2004
Subject: [llvm-commits] CVS: llvm/include/llvm/BasicBlock.h
Message-ID: <200405172225.RAA19380@zion.cs.uiuc.edu>


Changes in directory llvm/include/llvm:

BasicBlock.h updated: 1.38 -> 1.39

---
Log message:

* Minor aesthetic alignment of iterator declarations
* Removed redundant function comments


---
Diffs of the changes:  (+2 -5)

Index: llvm/include/llvm/BasicBlock.h
diff -u llvm/include/llvm/BasicBlock.h:1.38 llvm/include/llvm/BasicBlock.h:1.39
--- llvm/include/llvm/BasicBlock.h:1.38	Tue Feb 10 19:17:58 2004
+++ llvm/include/llvm/BasicBlock.h	Mon May 17 17:25:12 2004
@@ -61,17 +61,14 @@
 
 public:
   /// Instruction iterators...
-  typedef InstListType::iterator iterator;
-  typedef InstListType::const_iterator const_iterator;
+  typedef InstListType::iterator                              iterator;
+  typedef InstListType::const_iterator                  const_iterator;
   typedef std::reverse_iterator<const_iterator> const_reverse_iterator;
   typedef std::reverse_iterator<iterator>             reverse_iterator;
 
   /// BasicBlock ctor - If the function parameter is specified, the basic block
   /// is automatically inserted at either the end of the function (if
   /// InsertBefore is null), or before the specified basic block.
-  ///
-  /// BasicBlock ctor - If the InsertBefore parameter is specified, the basic
-  /// block is automatically inserted right before the specified block.
   ///
   BasicBlock(const std::string &Name = "", Function *Parent = 0,
              BasicBlock *InsertBefore = 0);




From brukman at cs.uiuc.edu  Mon May 17 17:28:01 2004
From: brukman at cs.uiuc.edu (Misha Brukman)
Date: Mon May 17 17:28:01 2004
Subject: [llvm-commits] CVS: llvm/include/llvm/BasicBlock.h
Message-ID: <200405172228.RAA21428@zion.cs.uiuc.edu>


Changes in directory llvm/include/llvm:

BasicBlock.h updated: 1.39 -> 1.40

---
Log message:

Minor aesthetic alignments; no functional changes.


---
Diffs of the changes:  (+3 -3)

Index: llvm/include/llvm/BasicBlock.h
diff -u llvm/include/llvm/BasicBlock.h:1.39 llvm/include/llvm/BasicBlock.h:1.40
--- llvm/include/llvm/BasicBlock.h:1.39	Mon May 17 17:25:12 2004
+++ llvm/include/llvm/BasicBlock.h	Mon May 17 17:28:21 2004
@@ -108,12 +108,12 @@
   inline reverse_iterator       rend  ()       { return InstList.rend();   }
   inline const_reverse_iterator rend  () const { return InstList.rend();   }
 
-  inline unsigned                 size() const { return InstList.size(); }
+  inline unsigned                 size() const { return InstList.size();  }
   inline bool                    empty() const { return InstList.empty(); }
   inline const Instruction      &front() const { return InstList.front(); }
   inline       Instruction      &front()       { return InstList.front(); }
-  inline const Instruction       &back()  const { return InstList.back(); }
-  inline       Instruction       &back()        { return InstList.back(); }
+  inline const Instruction       &back() const { return InstList.back();  }
+  inline       Instruction       &back()       { return InstList.back();  }
 
   /// getInstList() - Return the underlying instruction list container.  You
   /// need to access it directly if you want to modify it currently.




From brukman at cs.uiuc.edu  Mon May 17 19:21:01 2004
From: brukman at cs.uiuc.edu (Misha Brukman)
Date: Mon May 17 19:21:01 2004
Subject: [llvm-commits] [parallel] CVS: llvm/lib/Transforms/Parallel/BreakJoinBlocks.cpp
Message-ID: <200405180020.TAA17850@zion.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Parallel:

BreakJoinBlocks.cpp updated: 1.1.2.1 -> 1.1.2.2

---
Log message:

* Also consider LLVM thread start calls, not just the intrinsic
* Split before and after the join, so that it's the only remaining part of the
  basic block, apart from the terminator


---
Diffs of the changes:  (+15 -4)

Index: llvm/lib/Transforms/Parallel/BreakJoinBlocks.cpp
diff -u llvm/lib/Transforms/Parallel/BreakJoinBlocks.cpp:1.1.2.1 llvm/lib/Transforms/Parallel/BreakJoinBlocks.cpp:1.1.2.2
--- llvm/lib/Transforms/Parallel/BreakJoinBlocks.cpp:1.1.2.1	Thu May  6 11:34:58 2004
+++ llvm/lib/Transforms/Parallel/BreakJoinBlocks.cpp	Mon May 17 19:20:27 2004
@@ -11,9 +11,10 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "llvm/iOther.h"
-#include "llvm/Function.h"
 #include "llvm/BasicBlock.h"
+#include "llvm/Function.h"
+#include "llvm/iOther.h"
+#include "llvm/iTerminators.h"
 #include "llvm/Pass.h"
 #include "llvm/Transforms/Utils/FunctionUtils.h"
 using namespace llvm;
@@ -44,9 +45,18 @@
     BasicBlock *BB = Worklist.back();
     Worklist.pop_back();
     if (CallInst *CI = findJoin(BB)) {
+      // Break away everything after the join
+      BasicBlock::iterator Next(CI); ++Next;
+      if (!isa<TerminatorInst>(Next) || isa<ReturnInst>(Next)) {
+        BasicBlock *newBB = 
+          BB->splitBasicBlock(Next, BB->getName() + "_postsplit");
+        if (newBB) Worklist.push_back(newBB);
+      }
+
+      // Break away the join from anything before it
       BasicBlock::iterator II(CI);
       if (II != BB->begin()) {
-        BasicBlock *newBB = BB->splitBasicBlock(II, BB->getName() + "_split");
+        BasicBlock *newBB = BB->splitBasicBlock(II, BB->getName() +"_presplit");
         if (newBB) Worklist.push_back(newBB);
       }
     }
@@ -58,7 +68,8 @@
 CallInst* BreakJoin::findJoin(BasicBlock *BB) {
   for (BasicBlock::iterator i = BB->begin(), e = BB->end(); i != e; ++i)
     if (CallInst *CI = dyn_cast<CallInst>(i))
-      if (CI->getCalledFunction()->getName() == "llvm.join")
+      if (CI->getCalledFunction()->getName() == "llvm.join" ||
+          CI->getCalledFunction()->getName() == "__llvm_thread_join") 
         return CI;
 
   return 0;




From brukman at cs.uiuc.edu  Mon May 17 19:25:01 2004
From: brukman at cs.uiuc.edu (Misha Brukman)
Date: Mon May 17 19:25:01 2004
Subject: [llvm-commits] [parallel] CVS: llvm/lib/Transforms/Parallel/LowerParaBr.cpp
Message-ID: <200405180024.TAA17947@zion.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Parallel:

LowerParaBr.cpp updated: 1.1.2.2 -> 1.1.2.3

---
Log message:

Create a unified join block which contains all the join calls for the region


---
Diffs of the changes:  (+75 -8)

Index: llvm/lib/Transforms/Parallel/LowerParaBr.cpp
diff -u llvm/lib/Transforms/Parallel/LowerParaBr.cpp:1.1.2.2 llvm/lib/Transforms/Parallel/LowerParaBr.cpp:1.1.2.3
--- llvm/lib/Transforms/Parallel/LowerParaBr.cpp:1.1.2.2	Fri May  7 16:37:09 2004
+++ llvm/lib/Transforms/Parallel/LowerParaBr.cpp	Mon May 17 19:24:12 2004
@@ -16,10 +16,11 @@
 //
 //===----------------------------------------------------------------------===//
 
+#include "llvm/BasicBlock.h"
 #include "llvm/DerivedTypes.h"
-#include "llvm/Module.h"
 #include "llvm/iOther.h"
 #include "llvm/iTerminators.h"
+#include "llvm/Module.h"
 #include "llvm/Pass.h"
 #include "llvm/Type.h"
 #include "llvm/Analysis/ParallelInfo.h"
@@ -44,6 +45,7 @@
   private:
     void straightenSequence(ParallelSeq *PS);
     Function* getJoinIntrinsic(Module *M);    
+    Function* getFuncThreadJoin(Module &M);
   };
 
   RegisterOpt<LowerParaBr> X("lowerpbr", "Lower pbr to sequential code");
@@ -67,10 +69,21 @@
   return Changed;
 }
 
-static bool contains(std::vector<BasicBlock*> haystack, BasicBlock *needle) {
+static inline bool
+contains(const std::vector<BasicBlock*> &haystack, const BasicBlock *needle) {
   return std::find(haystack.begin(), haystack.end(), needle) != haystack.end();
 }
 
+static inline bool
+containsJoin(const std::vector<CallInst*> &haystack, const CallInst *needle) {  
+  for (std::vector<CallInst*>::const_iterator h = haystack.begin(), 
+         he = haystack.end(); h != he; ++h)
+    if ((*h)->getOperand(1) == needle->getOperand(1)) 
+      return true;
+
+  return false;
+}
+
 /// straightenSequence - Recursively process parallel sequences
 ///
 void LowerParaBr::straightenSequence(ParallelSeq *PS) {
@@ -80,15 +93,16 @@
   {
     PrevPR = PR;
     PR = *i;
+#if 0
     for (ParallelRegion::seqiterator ci = PR->seqbegin(), ce = PR->seqend(); 
          ci != ce; ++ci) 
       straightenSequence(*ci);
-
-    // Stitch previous region to the current one by all branches to the join
-    // block instead branch to the first block of the second region
+#endif
     if (PrevPR) {
+      // Stitch previous region to the current one by all branches to the join
+      // block instead branch to the first block of the second region
       std::vector<BasicBlock*> &PrevJoinBlocks = PrevPR->getJoinBlocks(),
-        &PRBlocks = PR->getBlocks();
+        &PRJoinBlocks = PR->getJoinBlocks(), &PRBlocks = PR->getBlocks();
       BasicBlock *PRFirstBB = PRBlocks[0];
       for (std::vector<BasicBlock*>::iterator j = PrevJoinBlocks.begin(),
              je = PrevJoinBlocks.end(); j != je; ++j) {
@@ -96,9 +110,55 @@
         for (std::vector<User*>::iterator u = JUsers.begin(), ue = JUsers.end();
              u != ue; ++u)
           if (Instruction *I = dyn_cast<Instruction>(*u))
-            if (contains(PR->getBlocks(), I->getParent()))
+            if (contains(PrevPR->getBlocks(), I->getParent()))
               I->replaceUsesOfWith(*j, PRFirstBB);
       }
+
+      // Make a new join block that will house all the (coallesced) joins from
+      // both regions
+      BasicBlock *SumJoins = new BasicBlock("allJoins", PRFirstBB->getParent());
+
+      // Find the unique set of calls to the llvm.join intrinsic
+      std::vector<CallInst*> AllJoins;
+      for (std::vector<BasicBlock*>::iterator i = PrevJoinBlocks.begin(),
+             e = PrevJoinBlocks.end(); i != e; ++i) {
+        BasicBlock *BB = *i;
+        for (BasicBlock::iterator j = BB->begin(), je = BB->end(); j!=je; ++j)
+          if (CallInst *CI = dyn_cast<CallInst>(j)) 
+            if (CI->getCalledFunction()->getName() == "__llvm_thread_join")
+              if (!containsJoin(AllJoins, CI))
+                AllJoins.push_back(CI);
+      }
+      for (std::vector<BasicBlock*>::iterator i = PRJoinBlocks.begin(),
+             e = PRJoinBlocks.end(); i != e; ++i) {
+        BasicBlock *BB = *i;
+        for (BasicBlock::iterator j = BB->begin(), je = BB->end(); j!=je; ++j)
+          if (CallInst *CI = dyn_cast<CallInst>(j)) 
+            if (CI->getCalledFunction()->getName() == "__llvm_thread_join")
+              if (!containsJoin(AllJoins, CI))
+                AllJoins.push_back(CI);
+      }
+
+      // Add all unique joins to the coallesced joins block
+      for (std::vector<CallInst*>::iterator c = AllJoins.begin(), 
+             ce = AllJoins.end(); c != ce; ++c)
+        SumJoins->getInstList().push_back((*c)->clone());
+
+      // Terminate the Joins block with a branch to the post-join code
+      Instruction *TI = PRJoinBlocks[0]->getTerminator();
+      if (TI)
+        SumJoins->getInstList().push_back(TI->clone());
+
+      // Branch to the all-joins block from the main code
+      for (std::vector<BasicBlock*>::iterator j = PRJoinBlocks.begin(),
+             je = PRJoinBlocks.end(); j != je; ++j) {
+        std::vector<User*> JUsers((*j)->use_begin(), (*j)->use_end());
+        for (std::vector<User*>::iterator u = JUsers.begin(), ue = JUsers.end();
+             u != ue; ++u)
+          if (Instruction *I = dyn_cast<Instruction>(*u))
+            if (contains(PR->getBlocks(), I->getParent()))
+              I->replaceUsesOfWith(*j, SumJoins);
+      }
     }
   }
 
@@ -115,7 +175,6 @@
           
   // Remove pbr
   ParaBrInst *pbr = dyn_cast<ParaBrInst>(PS->getHeader()->getTerminator());
-  //assert(pbr && "Terminator of ParaSeq header is not a pbr");
   if (pbr) {
     new BranchInst(pbr->getSuccessor(0), pbr);
     pbr->getParent()->getInstList().erase(pbr);
@@ -125,4 +184,12 @@
 Function* LowerParaBr::getJoinIntrinsic(Module *M) {
   return M->getOrInsertFunction("llvm.join", Type::VoidTy,
                                 PointerType::get(Type::SByteTy), 0);
+}
+
+/// getFuncThreadJoin -
+///
+Function* LowerParaBr::getFuncThreadJoin(Module &M) {
+  // void __llvm_thread_join(int);
+  return M.getOrInsertFunction("__llvm_thread_join", Type::VoidTy, Type::IntTy, 
+                               0);
 }




From brukman at cs.uiuc.edu  Mon May 17 19:25:03 2004
From: brukman at cs.uiuc.edu (Misha Brukman)
Date: Mon May 17 19:25:03 2004
Subject: [llvm-commits] [parallel] CVS: llvm/lib/Transforms/Parallel/ParallelCallsToThreads.cpp
Message-ID: <200405180024.TAA17928@zion.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Parallel:

ParallelCallsToThreads.cpp updated: 1.1.2.3 -> 1.1.2.4

---
Log message:

Fix comment to be up-to-date.


---
Diffs of the changes:  (+1 -1)

Index: llvm/lib/Transforms/Parallel/ParallelCallsToThreads.cpp
diff -u llvm/lib/Transforms/Parallel/ParallelCallsToThreads.cpp:1.1.2.3 llvm/lib/Transforms/Parallel/ParallelCallsToThreads.cpp:1.1.2.4
--- llvm/lib/Transforms/Parallel/ParallelCallsToThreads.cpp:1.1.2.3	Fri May  7 16:37:56 2004
+++ llvm/lib/Transforms/Parallel/ParallelCallsToThreads.cpp	Mon May 17 19:24:30 2004
@@ -77,7 +77,7 @@
   bool Changed = false;
   ParallelInfo &PI = getAnalysis<ParallelInfo>();
 
-  // Convert parallel calls to pthread_create() invocations
+  // Convert parallel calls to __llvm_thread_start() invocations
   for (ParallelInfo::iterator i = PI.begin(), e = PI.end(); i != e; ++i) {
     ParallelSeq *PS = *i;
     for (ParallelSeq::riterator r = PS->rbegin(), re = PS->rend(); 




From gaeke at cs.uiuc.edu  Tue May 18 11:52:01 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Tue May 18 11:52:01 2004
Subject: [llvm-commits] CVS: reopt/tools/ttftest/ttftest.cpp
Message-ID: <200405181651.LAA26137@zion.cs.uiuc.edu>


Changes in directory reopt/tools/ttftest:

ttftest.cpp updated: 1.1 -> 1.2

---
Log message:

Write out traces to files when they are loaded, for testing purposes.


---
Diffs of the changes:  (+34 -0)

Index: reopt/tools/ttftest/ttftest.cpp
diff -u reopt/tools/ttftest/ttftest.cpp:1.1 reopt/tools/ttftest/ttftest.cpp:1.2
--- reopt/tools/ttftest/ttftest.cpp:1.1	Mon May 17 16:25:25 2004
+++ reopt/tools/ttftest/ttftest.cpp	Tue May 18 11:51:30 2004
@@ -19,6 +19,8 @@
 #include "llvm/Bytecode/Reader.h"
 #include "reopt/TraceToFunction.h"
 #include "llvm/Module.h"
+#include "Support/FileUtilities.h"
+#include "Support/StringExtras.h"
 #include <iostream>
 #include <iterator>
 #include <fstream>
@@ -34,6 +36,7 @@
   return block;
 }
 
+
 /// ReadTraceFromFile - Given a module, read a trace for some function
 /// in that module from the named file. Returns the Trace object, or a null
 /// pointer (and sets ErrorStr) if there was an error.
@@ -70,6 +73,36 @@
   return new Trace (vBB);
 }
 
+
+/// getBasicBlockIndex - Returns the index of BB in its parent function.
+///
+int getBasicBlockIndex (BasicBlock *BB) {
+  int bbNum = 0;
+  Function *F = BB->getParent ();
+  Function::iterator block = F->begin();
+  while (&*block != BB) {
+    ++block;
+    ++bbNum;
+  }
+  return bbNum;
+}
+
+void WriteTraceToFile (Trace &T) {
+  std::string filename; 
+  unsigned count = 0;
+  do {
+    ++count;
+    filename = T.getFunction ()->getName () + ".trace" + utostr (count)
+             + ".txt";
+  } while (FileOpenable(filename));
+  std::ofstream out (filename.c_str ());
+  out << T.getFunction ()->getName () << "\n";
+  for (Trace::iterator i = T.begin (), e = T.end (); i != e; ++i) 
+    out << getBasicBlockIndex (*i) << " ";
+  out << "\n";
+  out.close ();
+}
+
 int main (int argc, char **argv) {
   const char *bytecodeFileName = argv[1];
   const char *traceFileName = argv[2];
@@ -89,6 +122,7 @@
     std::cerr << "Error reading trace: " << err << "\n";
     exit (1);
   }
+  WriteTraceToFile (*T);
 
   // Run TraceToFunction on the Trace
   TraceFunction *TF = TraceFunction::get (*T);




From gaeke at cs.uiuc.edu  Tue May 18 11:52:06 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Tue May 18 11:52:06 2004
Subject: [llvm-commits] CVS: reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp
Message-ID: <200405181651.LAA26123@zion.cs.uiuc.edu>


Changes in directory reopt/lib/LightWtProfiling:

RuntimeOptimizations.cpp updated: 1.32 -> 1.33

---
Log message:

Write out traces to files when they are detected, for use with ttftest.


---
Diffs of the changes:  (+32 -0)

Index: reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp
diff -u reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp:1.32 reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp:1.33
--- reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp:1.32	Mon May 17 16:00:19 2004
+++ reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp	Tue May 18 11:51:28 2004
@@ -19,6 +19,7 @@
 #include "reopt/InstrUtils.h"
 #include "Support/Debug.h"
 #include "Support/StringExtras.h"
+#include "Support/FileUtilities.h"
 #include "llvm/Analysis/Verifier.h"
 #include "llvm/Assembly/PrintModulePass.h"
 #include "llvm/CodeGen/MachineFunction.h"
@@ -31,6 +32,7 @@
 #include "llvm/Target/TargetMachine.h"
 #include "llvm/Target/TargetData.h"
 #include "llvm/Target/TargetMachineImpls.h"
+#include <fstream>
 
 namespace llvm {
 
@@ -65,6 +67,35 @@
 
 ///==----------------------------------------------------------------------==///
 
+/// getBasicBlockIndex - Returns the index of BB in its parent function.
+///
+int getBasicBlockIndex (BasicBlock *BB) {
+  int bbNum = 0;
+  Function *F = BB->getParent ();
+  Function::iterator block = F->begin();
+  while (&*block != BB) {
+    ++block;              
+    ++bbNum;
+  }
+  return bbNum;
+}     
+                
+void WriteTraceToFile (Trace &T) {
+  std::string filename;
+  unsigned count = 0;
+  do {
+    ++count;
+    filename = T.getFunction ()->getName () + ".trace" + utostr (count)
+             + ".txt";
+  } while (FileOpenable(filename));
+  std::ofstream out (filename.c_str ());
+  out << T.getFunction ()->getName () << "\n";
+  for (Trace::iterator i = T.begin (), e = T.end (); i != e; ++i)
+    out << getBasicBlockIndex (*i) << " ";
+  out << "\n";
+  out.close ();
+}   
+
 /// This method is called when we have finally constructed a
 /// trace. The first parameter is the vector of basic blocks that form
 /// the trace; the second parameter is the starting
@@ -89,6 +120,7 @@
   // Turn the vector of basic blocks into a Trace, and then turn the Trace into
   // a TraceFunction.
   Trace T (vBB);
+  WriteTraceToFile (T);
   TraceFunction *TF = TraceFunction::get (T);
 
   // Force the SPARCv9 register allocator to save its state into a global




From gaeke at cs.uiuc.edu  Tue May 18 11:52:08 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Tue May 18 11:52:08 2004
Subject: [llvm-commits] CVS: reopt/test/TEST.reopt.Makefile
Message-ID: <200405181651.LAA26130@zion.cs.uiuc.edu>


Changes in directory reopt/test:

TEST.reopt.Makefile updated: 1.11 -> 1.12

---
Log message:

Link in tracetofunction library.


---
Diffs of the changes:  (+1 -1)

Index: reopt/test/TEST.reopt.Makefile
diff -u reopt/test/TEST.reopt.Makefile:1.11 reopt/test/TEST.reopt.Makefile:1.12
--- reopt/test/TEST.reopt.Makefile:1.11	Wed Mar 31 10:36:43 2004
+++ reopt/test/TEST.reopt.Makefile	Tue May 18 11:51:29 2004
@@ -28,7 +28,7 @@
 # Libraries that contain the Reoptimizer itself
 REOPTIMIZER_OBJS = $(REOPTLIBDIR)/firstTrigger.o \
 	$(REOPTLIBDIR)/tracecache.o $(REOPTLIBDIR)/mapinfo.o \
-	$(REOPTLIBDIR)/scratchmemory.o
+	$(REOPTLIBDIR)/scratchmemory.o $(REOPTLIBDIR)/tracetofunction.o
 
 # Object files that contain common LLVM code the Reoptimizer depends on
 REOPTIMIZER_LLVMOBJS = $(DESTLIBCURRENT)/vmcore.o \




From alkis at cs.uiuc.edu  Tue May 18 14:24:01 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Tue May 18 14:24:01 2004
Subject: [llvm-commits] CVS: llvm-java/autoconf/configure.ac
Message-ID: <200405181923.OAA27464@zion.cs.uiuc.edu>


Changes in directory llvm-java/autoconf:

configure.ac updated: 1.5 -> 1.6

---
Log message:

Look for JAVA and JAVAC executables.


---
Diffs of the changes:  (+2 -0)

Index: llvm-java/autoconf/configure.ac
diff -u llvm-java/autoconf/configure.ac:1.5 llvm-java/autoconf/configure.ac:1.6
--- llvm-java/autoconf/configure.ac:1.5	Wed Apr 21 11:32:31 2004
+++ llvm-java/autoconf/configure.ac	Tue May 18 14:23:31 2004
@@ -25,6 +25,8 @@
 dnl * Check for programs.
 dnl **************************************************************************
 AC_CHECK_PROG(JIKES, jikes, jikes)
+AC_CHECK_PROG(JAVAC, javac, javac)
+AC_CHECK_PROG(JAVA, java, java)
 
 dnl Verify that the source directory is valid
 AC_CONFIG_SRCDIR(Makefile.common.in)




From alkis at cs.uiuc.edu  Tue May 18 14:24:05 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Tue May 18 14:24:05 2004
Subject: [llvm-commits] CVS: llvm-java/configure
Message-ID: <200405181923.OAA27465@zion.cs.uiuc.edu>


Changes in directory llvm-java:

configure updated: 1.4 -> 1.5

---
Log message:

Look for JAVA and JAVAC executables.


---
Diffs of the changes:  (+73 -1)

Index: llvm-java/configure
diff -u llvm-java/configure:1.4 llvm-java/configure:1.5
--- llvm-java/configure:1.4	Thu Apr 15 23:37:54 2004
+++ llvm-java/configure	Tue May 18 14:23:31 2004
@@ -274,7 +274,7 @@
 PACKAGE_BUGREPORT='llvmbugs at cs.uiuc.edu'
 
 ac_unique_file="Makefile.common.in"
-ac_subst_vars='SHELL PATH_SEPARATOR PACKAGE_NAME PACKAGE_TARNAME PACKAGE_VERSION PACKAGE_STRING PACKAGE_BUGREPORT exec_prefix prefix program_transform_name bindir sbindir libexecdir datadir sysconfdir sharedstatedir localstatedir libdir includedir oldincludedir infodir mandir build_alias host_alias target_alias DEFS ECHO_C ECHO_N ECHO_T LIBS JIKES LLVM_SRC LLVM_OBJ LIBOBJS LTLIBOBJS'
+ac_subst_vars='SHELL PATH_SEPARATOR PACKAGE_NAME PACKAGE_TARNAME PACKAGE_VERSION PACKAGE_STRING PACKAGE_BUGREPORT exec_prefix prefix program_transform_name bindir sbindir libexecdir datadir sysconfdir sharedstatedir localstatedir libdir includedir oldincludedir infodir mandir build_alias host_alias target_alias DEFS ECHO_C ECHO_N ECHO_T LIBS JIKES JAVAC JAVA LLVM_SRC LLVM_OBJ LIBOBJS LTLIBOBJS'
 ac_subst_files=''
 
 # Initialize some variables set by options.
@@ -1314,6 +1314,76 @@
 echo "${ECHO_T}no" >&6
 fi
 
+# Extract the first word of "javac", so it can be a program name with args.
+set dummy javac; ac_word=$2
+echo "$as_me:$LINENO: checking for $ac_word" >&5
+echo $ECHO_N "checking for $ac_word... $ECHO_C" >&6
+if test "${ac_cv_prog_JAVAC+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  if test -n "$JAVAC"; then
+  ac_cv_prog_JAVAC="$JAVAC" # Let the user override the test.
+else
+as_save_IFS=$IFS; IFS=$PATH_SEPARATOR
+for as_dir in $PATH
+do
+  IFS=$as_save_IFS
+  test -z "$as_dir" && as_dir=.
+  for ac_exec_ext in '' $ac_executable_extensions; do
+  if $as_executable_p "$as_dir/$ac_word$ac_exec_ext"; then
+    ac_cv_prog_JAVAC="javac"
+    echo "$as_me:$LINENO: found $as_dir/$ac_word$ac_exec_ext" >&5
+    break 2
+  fi
+done
+done
+
+fi
+fi
+JAVAC=$ac_cv_prog_JAVAC
+if test -n "$JAVAC"; then
+  echo "$as_me:$LINENO: result: $JAVAC" >&5
+echo "${ECHO_T}$JAVAC" >&6
+else
+  echo "$as_me:$LINENO: result: no" >&5
+echo "${ECHO_T}no" >&6
+fi
+
+# Extract the first word of "java", so it can be a program name with args.
+set dummy java; ac_word=$2
+echo "$as_me:$LINENO: checking for $ac_word" >&5
+echo $ECHO_N "checking for $ac_word... $ECHO_C" >&6
+if test "${ac_cv_prog_JAVA+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  if test -n "$JAVA"; then
+  ac_cv_prog_JAVA="$JAVA" # Let the user override the test.
+else
+as_save_IFS=$IFS; IFS=$PATH_SEPARATOR
+for as_dir in $PATH
+do
+  IFS=$as_save_IFS
+  test -z "$as_dir" && as_dir=.
+  for ac_exec_ext in '' $ac_executable_extensions; do
+  if $as_executable_p "$as_dir/$ac_word$ac_exec_ext"; then
+    ac_cv_prog_JAVA="java"
+    echo "$as_me:$LINENO: found $as_dir/$ac_word$ac_exec_ext" >&5
+    break 2
+  fi
+done
+done
+
+fi
+fi
+JAVA=$ac_cv_prog_JAVA
+if test -n "$JAVA"; then
+  echo "$as_me:$LINENO: result: $JAVA" >&5
+echo "${ECHO_T}$JAVA" >&6
+else
+  echo "$as_me:$LINENO: result: no" >&5
+echo "${ECHO_T}no" >&6
+fi
+
 
 
 
@@ -2001,6 +2071,8 @@
 s, at ECHO_T@,$ECHO_T,;t t
 s, at LIBS@,$LIBS,;t t
 s, at JIKES@,$JIKES,;t t
+s, at JAVAC@,$JAVAC,;t t
+s, at JAVA@,$JAVA,;t t
 s, at LLVM_SRC@,$LLVM_SRC,;t t
 s, at LLVM_OBJ@,$LLVM_OBJ,;t t
 s, at LIBOBJS@,$LIBOBJS,;t t




From alkis at cs.uiuc.edu  Tue May 18 14:27:02 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Tue May 18 14:27:02 2004
Subject: [llvm-commits] CVS: llvm-java/Makefile.rules
Message-ID: <200405181924.OAA27487@zion.cs.uiuc.edu>


Changes in directory llvm-java:

Makefile.rules updated: 1.1 -> 1.2

---
Log message:

Upadate rule to use JAVAC instead of JIKES.


---
Diffs of the changes:  (+2 -3)

Index: llvm-java/Makefile.rules
diff -u llvm-java/Makefile.rules:1.1 llvm-java/Makefile.rules:1.2
--- llvm-java/Makefile.rules:1.1	Wed Apr 21 11:33:48 2004
+++ llvm-java/Makefile.rules	Tue May 18 14:24:48 2004
@@ -1,4 +1,4 @@
-#===-- Makefile.rules - Common make rules for LLVM -------*- makefile -*--====
+##==-- Makefile.rules - Common make rules for Java -------*- makefile -*--====##
 #
 #                     The LLVM Compiler Infrastructure
 #
@@ -10,7 +10,6 @@
 ifdef BUILD_JAVA_SOURCE
 all:: .class-stamp
 .class-stamp: $(shell find . -name '*.java')
-	$(JIKES) --bootclasspath $(BOOTCLASSPATH) $?
+	$(JAVAC) $?
 	touch .class-stamp
 endif
-




From alkis at cs.uiuc.edu  Tue May 18 15:15:10 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Tue May 18 15:15:10 2004
Subject: [llvm-commits] CVS: llvm-java/include/llvm/Java/Bytecode.h
Message-ID: <200405182013.PAA29286@zion.cs.uiuc.edu>


Changes in directory llvm-java/include/llvm/Java:

Bytecode.h added (r1.1)

---
Log message:

Add java bytecode constants header. It currently only contains the java opcodes.


---
Diffs of the changes:  (+229 -0)

Index: llvm-java/include/llvm/Java/Bytecode.h
diff -c /dev/null llvm-java/include/llvm/Java/Bytecode.h:1.1
*** /dev/null	Tue May 18 15:13:47 2004
--- llvm-java/include/llvm/Java/Bytecode.h	Tue May 18 15:13:37 2004
***************
*** 0 ****
--- 1,229 ----
+ //===-- Bytecode.h - Java bytecode constants --------------------*- C++ -*-===//
+ //
+ //                     The LLVM Compiler Infrastructure
+ //
+ // This file was developed by the LLVM research group and is distributed under
+ // the University of Illinois Open Source License. See LICENSE.TXT for details.
+ //
+ //===----------------------------------------------------------------------===//
+ //
+ // This file contains Java bytecode constants.
+ //
+ //===----------------------------------------------------------------------===//
+ 
+ #include <iosfwd>
+ #include <vector>
+ 
+ #include <stdint.h>
+ 
+ namespace llvm { namespace Java {
+ 
+     enum Opcode {
+         NOP = 0x00,
+         ACONST_NULL = 0x01,
+         ICONST_M1 = 0x02,
+         ICONST_0 = 0x03,
+         ICONST_1 = 0x04,
+         ICONST_2 = 0x05,
+         ICONST_3 = 0x06,
+         ICONST_4 = 0x07,
+         ICONST_5 = 0x08,
+         LCONST_0 = 0x09,
+         LCONST_1 = 0x0A,
+         FCONST_0 = 0x0B,
+         FCONST_1 = 0x0C,
+         FCONST_2 = 0x0D,
+         DCONST_0 = 0x0E,
+         DCONST_1 = 0x0F,
+         BIPUSH = 0x10,
+         SIPUSH = 0x11,
+         LDC = 0x12,
+         LDC_W = 0x13,
+         LDC2_W = 0x14,
+         ILOAD = 0x15,
+         LLOAD = 0x16,
+         FLOAD = 0x17,
+         DLOAD = 0x18,
+         ALOAD = 0x19,
+         ILOAD_0 = 0x1A,
+         ILOAD_1 = 0x1B,
+         ILOAD_2 = 0x1C,
+         ILOAD_3 = 0x1D,
+         LLOAD_0 = 0x1E,
+         LLOAD_1 = 0x1F,
+         LLOAD_2 = 0x20,
+         LLOAD_3 = 0x21,
+         FLOAD_0 = 0x22,
+         FLOAD_1 = 0x23,
+         FLOAD_2 = 0x24,
+         FLOAD_3 = 0x25,
+         DLOAD_0 = 0x26,
+         DLOAD_1 = 0x27,
+         DLOAD_2 = 0x28,
+         DLOAD_3 = 0x29,
+         ALOAD_0 = 0x2A,
+         ALOAD_1 = 0x2B,
+         ALOAD_2 = 0x2C,
+         ALOAD_3 = 0x2D,
+         IALOAD = 0x2E,
+         LALOAD = 0x2F,
+         FALOAD = 0x30,
+         DALOAD = 0x31,
+         AALOAD = 0x32,
+         BALOAD = 0x33,
+         CALOAD = 0x34,
+         SALOAD = 0x35,
+         ISTORE = 0x36,
+         LSTORE = 0x37,
+         FSTORE = 0x38,
+         DSTORE = 0x39,
+         ASTORE = 0x3A,
+         ISTORE_0 = 0x3B,
+         ISTORE_1 = 0x3C,
+         ISTORE_2 = 0x3D,
+         ISTORE_3 = 0x3E,
+         LSTORE_0 = 0x3F,
+         LSTORE_1 = 0x40,
+         LSTORE_2 = 0x41,
+         LSTORE_3 = 0x42,
+         FSTORE_0 = 0x43,
+         FSTORE_1 = 0x44,
+         FSTORE_2 = 0x45,
+         FSTORE_3 = 0x46,
+         DSTORE_0 = 0x47,
+         DSTORE_1 = 0x48,
+         DSTORE_2 = 0x49,
+         DSTORE_3 = 0x4A,
+         ASTORE_0 = 0x4B,
+         ASTORE_1 = 0x4C,
+         ASTORE_2 = 0x4D,
+         ASTORE_3 = 0x4E,
+         IASTORE = 0x4F,
+         LASTORE = 0x50,
+         FASTORE = 0x51,
+         DASTORE = 0x52,
+         AASTORE = 0x53,
+         BASTORE = 0x54,
+         CASTORE = 0x55,
+         SASTORE = 0x56,
+         POP = 0x57,
+         POP2 = 0x58,
+         DUP = 0x59,
+         DUP_X1 = 0x5A,
+         DUP_X2 = 0x5B,
+         DUP2 = 0x5C,
+         DUP2_X1 = 0x5D,
+         DUP2_X2 = 0x5E,
+         SWAP = 0x5F,
+         IADD = 0x60,
+         LADD = 0x61,
+         FADD = 0x62,
+         DADD = 0x63,
+         ISUB = 0x64,
+         LSUB = 0x65,
+         FSUB = 0x66,
+         DSUB = 0x67,
+         IMUL = 0x68,
+         LMUL = 0x69,
+         FMUL = 0x6A,
+         DMUL = 0x6B,
+         IDIV = 0x6C,
+         LDIV = 0x6D,
+         FDIV = 0x6E,
+         DDIV = 0x6F,
+         IREM = 0x70,
+         LREM = 0x71,
+         FREM = 0x72,
+         DREM = 0x73,
+         INEG = 0x74,
+         LNEG = 0x75,
+         FNEG = 0x76,
+         DNEG = 0x77,
+         ISHL = 0x78,
+         LSHL = 0x79,
+         ISHR = 0x7A,
+         LSHR = 0x7B,
+         IUSHR = 0x7C,
+         LUSHR = 0x7D,
+         IAND = 0x7E,
+         LAND = 0x7F,
+         IOR = 0x80,
+         LOR = 0x81,
+         IXOR = 0x82,
+         LXOR = 0x83,
+         IINC = 0x84,
+         I2L = 0x85,
+         I2F = 0x86,
+         I2D = 0x87,
+         L2I = 0x88,
+         L2F = 0x89,
+         L2D = 0x8A,
+         F2I = 0x8B,
+         F2L = 0x8C,
+         F2D = 0x8D,
+         D2I = 0x8E,
+         D2L = 0x8F,
+         D2F = 0x90,
+         I2B = 0x91,
+         I2C = 0x92,
+         I2S = 0x93,
+         LCMP = 0x94,
+         FCMPL = 0x95,
+         FCMPG = 0x96,
+         DCMPL = 0x97,
+         DCMPG = 0x98,
+         IFEQ = 0x99,
+         IFNE = 0x9A,
+         IFLT = 0x9B,
+         IFGE = 0x9C,
+         IFGT = 0x9D,
+         IFLE = 0x9E,
+         IF_ICMPEQ = 0x9F,
+         IF_ICMPNE = 0xA0,
+         IF_ICMPLT = 0xA1,
+         IF_ICMPGE = 0xA2,
+         IF_ICMPGT = 0xA3,
+         IF_ICMPLE = 0xA4,
+         IF_ICMPACMPEQ = 0xA5,
+         IF_ICMPACMPNE = 0xA6,
+         GOTO = 0xA7,
+         JSR = 0xA8,
+         RET = 0xA9,
+         TABLESWITCH = 0xAA,
+         LOOKUPSWITCH = 0xAB,
+         IRETURN = 0xAC,
+         LRETURN = 0xAD,
+         FRETURN = 0xAE,
+         DRETURN = 0xAF,
+         ARETURN = 0xB0,
+         RETURN = 0xB1,
+         GETSTATIC = 0xB2,
+         PUTSTATIC = 0xB3,
+         GETFIELD = 0xB4,
+         PUTFIELD = 0xB5,
+         INVOKEVIRTUAL = 0xB6,
+         INVOKESPECIAL = 0xB7,
+         INVOKESTATIC = 0xB8,
+         INVOKEINTERFACE = 0xB9,
+         XXXUNUSEDXXX = 0xBA,
+         NEW = 0xBB,
+         NEWARRAY = 0xBC,
+         ANEWARRAY = 0xBD,
+         ARRAYLENGTH = 0xBE,
+         ATHROW = 0xBF,
+         CHECKCAST = 0xC0,
+         INSTANCEOF = 0xC1,
+         MONITORENTER = 0xC2,
+         MONITOREXIT = 0xC3,
+         WIDE = 0xC4,
+         MULTIANEWARRAY = 0xC5,
+         IFNULL = 0xC6,
+         IFNONNULL = 0xC7,
+         GOTO_W = 0xC8,
+         JSR_W = 0xC9,
+         BREAKPOINT = 0xCA,
+         IMPDEP1 = 0xFE,
+         IMPDEP2 = 0xFF,
+     };
+ 
+ } } // namespace llvm::Java




From gaeke at cs.uiuc.edu  Tue May 18 16:05:02 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Tue May 18 16:05:02 2004
Subject: [llvm-commits] CVS: reopt/lib/TraceToFunction/TraceToFunction.cpp
Message-ID: <200405182104.QAA14294@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/TraceToFunction:

TraceToFunction.cpp updated: 1.40 -> 1.41

---
Log message:

Significant changes to how we deal with FLI: now we make a pass over
the trace's matrix function (before cloning the trace), and find all
the FLI blocks and their associated edges. Then we clone the trace, and
get rid of FLI blocks and thread branches (and their corresponding
PHIs) on back-edges which have been instrumented with FLI.


---
Diffs of the changes:  (+114 -45)

Index: reopt/lib/TraceToFunction/TraceToFunction.cpp
diff -u reopt/lib/TraceToFunction/TraceToFunction.cpp:1.40 reopt/lib/TraceToFunction/TraceToFunction.cpp:1.41
--- reopt/lib/TraceToFunction/TraceToFunction.cpp:1.40	Mon May 17 16:49:37 2004
+++ reopt/lib/TraceToFunction/TraceToFunction.cpp	Tue May 18 16:04:37 2004
@@ -45,18 +45,27 @@
 typedef std::vector<const Type *> TypeVector;
 typedef std::map<BranchInst *, unsigned int> BranchNumberMap;
 
+// maps FLI blocks to the corresponding <edge source, sink> pair:
+typedef std::map<BasicBlock *, std::pair<BasicBlock *, BasicBlock *> > FLIMapTy;
+
 /// TraceFunctionBuilder - a Method Object which encapsulates the algorithm
 /// for building TraceFunctions given Traces.
 ///
 class TraceFunctionBuilder {
   BranchNumberMap BranchNumber;
   TraceFunction *TF;
-  virtual TypeVector createFunctionArgTypeVector (PointerType *ST,
-						  LiveVariableSet S);
-  virtual void fillInFunctionBody (Trace &T, Function *F, LiveVariableSet &So);
-  virtual void fixupFunctionBodyBB (Trace &T, Function *F, BasicBlock *srcB,
-				    BasicBlock *dstB, ValueMap &O2CMap,
-				    LiveVariableSet &So);
+
+  FLIMapTy FLIMap;
+  BasicBlock *getFLIEdgeSource (BasicBlock *BB);
+  BasicBlock *getFLIEdgeTarget (BasicBlock *BB);
+  void threadFLIEdges (Function *F);
+  void buildFLIMap (Trace &T, FLIMapTy &FLIMap);
+
+  TypeVector createFunctionArgTypeVector (PointerType *ST, LiveVariableSet S);
+  void fillInFunctionBody (Trace &T, Function *F, LiveVariableSet &So);
+  void fixupFunctionBodyBB (Trace &T, Function *F, BasicBlock *srcB,
+				            BasicBlock *dstB, ValueMap &O2CMap,
+				            LiveVariableSet &So);
 public:
   TraceFunctionBuilder () { }
   virtual ~TraceFunctionBuilder () { }
@@ -378,8 +387,6 @@
   }
 }
 
-static BasicBlock *getFLIEdgeTarget (BasicBlock *BB);
-
 /// fillInFunctionBody - Clone the BBs of T into F, then do all
 /// necessary fixups to ensure that F's new contents are internally
 /// consistent and that the live-outs So get stored in F's first
@@ -389,15 +396,13 @@
 					  LiveVariableSet &So) {
   ValueMap &O2CMap = TF->O2CMap;
 
-  // Remove any FLI blocks which ended up in the trace.
-  Trace::iterator i;
-  while ((i = std::find_if (T.begin (), T.end (), getFLIEdgeTarget))
-         != T.end ())
-    T.erase (i);
-  
+  buildFLIMap (T, FLIMap);
+
   // First copy the basic blocks from the trace.
   cloneTraceBBsIntoFunction (T, F, O2CMap);
 
+  threadFLIEdges (F);
+
   numberExitBranchesInTrace (T, BranchNumber);
 
   // Fix up the cloned basic blocks of the function so that they are
@@ -413,45 +418,109 @@
     fixupPhisAndCalls (FI, O2CMap);
 }
 
-/// getFLIEdgeTarget - returns a nonzero value iff BB is a
-/// basic block containing only a call to the first-level instrumentation
-/// function.  Basic blocks of this form are inserted by the -instloops pass on
-/// loop back-edges. The return value is the target of the back-edge.
-///
-static BasicBlock *getFLIEdgeTarget (BasicBlock *BB) {
-  BasicBlock::iterator i = BB->begin ();
-  // starts with llvm_first_trigger call
-  if (!isFirstTriggerCall (*i))
-    return 0;
-  // Next instr. should be an unconditional branch - return its target,
-  // or NULL if there is a mismatch.
-  BranchInst *BI = dyn_cast<BranchInst> (++i);
-  if (!(BI && BI->isUnconditional()))
-    return 0;
-  return BI->getSuccessor (0);
-}
-
-/// getFLIEdgeSource - returns a nonzero value iff BB is a
-/// basic block containing only a call to the first-level instrumentation
-/// function.  Basic blocks of this form are inserted by the -instloops pass on
-/// loop back-edges. The return value is the source of the back-edge.
+/// identifyFLIEdge - returns a true value iff BB is a basic block containing
+/// only a call to the first-level instrumentation function.  Basic blocks of
+/// this form are inserted by the -instloops pass on loop back-edges. The side
+/// effect of this function, if it returns true, is to fill in the pair "edge"
+/// with the source and target of the back-edge.
 ///
-static BasicBlock *getFLIEdgeSource (BasicBlock *BB) {
+static bool identifyFLIEdge (BasicBlock *BB, std::pair<BasicBlock *, BasicBlock *> &edge) {
   BasicBlock::iterator i = BB->begin ();
   // starts with llvm_first_trigger call
   if (!isFirstTriggerCall (*i))
-    return 0;
-  // Next instr. should be an unconditional branch. Return NULL if
+    return false;
+  // Next instr. should be an unconditional branch. Return false if
   // there is a mismatch.
   BranchInst *BI = dyn_cast<BranchInst> (++i);
   if (!(BI && BI->isUnconditional()))
-    return 0;
-  // Use CFG predecessor iterator to get the predecessor of the FLI block
+    return false;
+  // Record the FLI edge's sink.
+  edge.second = BI->getSuccessor (0);
+  // Use CFG predecessor iterator to get the predecessor of the FLI block.
+  // FIXME: should check that there is only one predecessor.
   pred_iterator PI = pred_begin (BB);
-  DEBUG(pred_iterator NextPI = PI; ++NextPI;
-        assert (NextPI == pred_end (BB) &&
-                "FLI block must have only one predecessor"));
-  return *PI;
+  edge.first = *PI;
+  return true;
+}
+
+void TraceFunctionBuilder::threadFLIEdges (Function *F) {
+  for (Function::iterator FI = F->begin (), FE = F->end (); FI != FE; ++FI)
+    for (BasicBlock::iterator Inst = FI->begin (), BE = FI->end (); Inst != BE; ++Inst) {
+      if (isa<PHINode> (Inst)) {
+        // In a PHI node, replace references to FLI blocks with the
+        // corresponding edge's source.
+        for (unsigned OI = 0, OE = Inst->getNumOperands(); OI != OE; ++OI) {
+          Value *op = Inst->getOperand (OI);
+          if (isa<BasicBlock> (op)) {
+            BasicBlock *BB = cast<BasicBlock> (op);
+            FLIMapTy::iterator It = FLIMap.find (BB);
+            if (It != FLIMap.end ()) {
+              DEBUG (std::cerr << "threadFLIEdges: threading operand " << OI
+                << " (FLI block " << BB->getName () << " --> <[" 
+                << It->second.first->getName () << "], "
+                << It->second.second->getName () << ">) of instruction: "
+                << *Inst << "\n");
+              Inst->setOperand (OI, It->second.first);
+            }
+          }
+        }
+      } else if (isa<BranchInst> (Inst)) {
+        // If it's a branch, replace references to FLI blocks with the
+        // corresponding edge's target.
+        for (unsigned OI = 0, OE = Inst->getNumOperands(); OI != OE; ++OI) {
+          Value *op = Inst->getOperand (OI);
+          if (isa<BasicBlock> (op)) {
+            BasicBlock *BB = cast<BasicBlock> (op);
+            FLIMapTy::iterator It = FLIMap.find (BB);
+            if (It != FLIMap.end ()) {
+              DEBUG (std::cerr << "threadFLIEdges: threading operand " << OI
+                << " (FLI block " << BB->getName () << " --> <" 
+                << It->second.first->getName () << ", ["
+                << It->second.second->getName () << "]>) of instruction: "
+                << *Inst << "\n");
+              Inst->setOperand (OI, It->second.second);
+            }
+          }
+        }
+      }
+    }
+}
+
+/// buildFLIMap - Removes any FLI blocks which ended up in the trace, and
+/// identifies the source and sink BasicBlocks of all FLI-instrumented
+/// back-edges. Information about FLI edges is stored in the FLIMap.
+///
+void TraceFunctionBuilder::buildFLIMap (Trace &T, FLIMapTy &FLIMap) {
+  Function *F = T.getFunction ();
+  for (Function::iterator i = F->begin (), e = F->end (); i != e; ++i) {
+    std::pair<BasicBlock *, BasicBlock *> edge;
+    if (identifyFLIEdge (i, edge)) {
+      DEBUG (std::cerr << "buildFLIMap: identified FLI block " << i->getName()
+        << " on edge <" << edge.first->getName () << ", "
+        << edge.second->getName () << ">\n");
+      // 1. remove the block from the trace if it is in there
+      BasicBlock *bb = i;
+      Trace::iterator TI = std::find (T.begin (), T.end (), bb);
+      if (TI != T.end ()) { 
+        DEBUG (std::cerr << "buildFLIMap: removing it from the trace\n");
+        T.erase (TI);
+      }
+      // 2. record its source and target in the map
+      FLIMap[i] = edge;
+    }
+  }
+}
+
+BasicBlock *TraceFunctionBuilder::getFLIEdgeSource (BasicBlock *BB) {
+  FLIMapTy::iterator i = FLIMap.find (BB);
+  if (i == FLIMap.end ()) return 0;
+  return i->second.first;
+}
+
+BasicBlock *TraceFunctionBuilder::getFLIEdgeTarget (BasicBlock *BB) {
+  FLIMapTy::iterator i = FLIMap.find (BB);
+  if (i == FLIMap.end ()) return 0;
+  return i->second.second;
 }
 
 /// fixupFunctionBodyBB - Given srcB in T and its clone dstB in F, and




From gaeke at cs.uiuc.edu  Tue May 18 16:24:02 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Tue May 18 16:24:02 2004
Subject: [llvm-commits] CVS: reopt/lib/TraceToFunction/TraceToFunction.cpp
Message-ID: <200405182123.QAA14501@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/TraceToFunction:

TraceToFunction.cpp updated: 1.41 -> 1.42

---
Log message:

Give cloned trace basic blocks more reasonable names.


---
Diffs of the changes:  (+1 -2)

Index: reopt/lib/TraceToFunction/TraceToFunction.cpp
diff -u reopt/lib/TraceToFunction/TraceToFunction.cpp:1.41 reopt/lib/TraceToFunction/TraceToFunction.cpp:1.42
--- reopt/lib/TraceToFunction/TraceToFunction.cpp:1.41	Tue May 18 16:04:37 2004
+++ reopt/lib/TraceToFunction/TraceToFunction.cpp	Tue May 18 16:23:06 2004
@@ -282,13 +282,12 @@
 /// to O2CMap.
 ///
 static void cloneTraceBBsIntoFunction (Trace &T, Function *F, ValueMap &O2CMap){
-  unsigned BBCount = 0;
   // Clone each basic block into the new function.
   for (Trace::iterator TI = T.begin (), TE = T.end (); TI != TE; ++TI) {
     BasicBlock *srcB = *TI;
     // Create a new BasicBlock dstB that corresponds to srcB in T.
     BasicBlock *dstB = CloneBasicBlock (srcB, O2CMap, ".ttf");
-    DEBUG(dstB->setName ("trace" + utostr(BBCount++)));
+    DEBUG(dstB->setName (srcB->getName () + ".ttf"));
     // Add dstB to F.
     F->getBasicBlockList ().push_back (dstB);
     // Remember the correspondence between srcB and dstB. (CloneFunction does




From gaeke at cs.uiuc.edu  Wed May 19 01:52:02 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Wed May 19 01:52:02 2004
Subject: [llvm-commits] CVS: reopt/lib/LightWtProfiling/Makefile
Message-ID: <200405190651.BAA26425@zion.cs.uiuc.edu>


Changes in directory reopt/lib/LightWtProfiling:

Makefile updated: 1.8 -> 1.9

---
Log message:

Makefile changes due to TraceToFunction move.


---
Diffs of the changes:  (+1 -1)

Index: reopt/lib/LightWtProfiling/Makefile
diff -u reopt/lib/LightWtProfiling/Makefile:1.8 reopt/lib/LightWtProfiling/Makefile:1.9
--- reopt/lib/LightWtProfiling/Makefile:1.8	Sun May 16 00:24:37 2004
+++ reopt/lib/LightWtProfiling/Makefile	Wed May 19 01:50:57 2004
@@ -5,7 +5,7 @@
 include $(LEVEL)/Makefile.config
 
 ifneq ($(ARCH),Sparc)
-Source := FirstTrigger.cpp Initialization.cpp RuntimeOptimizations.cpp SLI.cpp TraceOptEmitter.cpp TraceToFunction.cpp UnpackTraceFunction.cpp
+Source := FirstTrigger.cpp Initialization.cpp RuntimeOptimizations.cpp SLI.cpp TraceOptEmitter.cpp UnpackTraceFunction.cpp
 all ::
 	@echo "This code assumes an LP64 machine- please ignore ptr cast warnings!"
 endif




From gaeke at cs.uiuc.edu  Wed May 19 01:52:06 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Wed May 19 01:52:06 2004
Subject: [llvm-commits] CVS: reopt/lib/Makefile
Message-ID: <200405190651.BAA26418@zion.cs.uiuc.edu>


Changes in directory reopt/lib:

Makefile updated: 1.19 -> 1.20

---
Log message:

Makefile changes due to TraceToFunction move.


---
Diffs of the changes:  (+1 -1)

Index: reopt/lib/Makefile
diff -u reopt/lib/Makefile:1.19 reopt/lib/Makefile:1.20
--- reopt/lib/Makefile:1.19	Wed Apr 28 12:27:36 2004
+++ reopt/lib/Makefile	Wed May 19 01:50:56 2004
@@ -1,5 +1,5 @@
 LEVEL = ..
-DIRS := BinInterface Mapping TraceCache Trigger LightWtProfiling Inst
+DIRS := BinInterface Mapping TraceCache Trigger LightWtProfiling Inst TraceToFunction
 
 include $(LEVEL)/Makefile.config
 




From gaeke at cs.uiuc.edu  Wed May 19 03:45:01 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Wed May 19 03:45:01 2004
Subject: [llvm-commits] CVS: reopt/tools/ttftest/ttftest.cpp
Message-ID: <200405190844.DAA19093@seraph.cs.uiuc.edu>


Changes in directory reopt/tools/ttftest:

ttftest.cpp updated: 1.2 -> 1.3

---
Log message:

Change this tool to use the CommandLine library.
Remove the (duplicated) trace-writing code from this file.
Minor edits to comments.


---
Diffs of the changes:  (+18 -29)

Index: reopt/tools/ttftest/ttftest.cpp
diff -u reopt/tools/ttftest/ttftest.cpp:1.2 reopt/tools/ttftest/ttftest.cpp:1.3
--- reopt/tools/ttftest/ttftest.cpp:1.2	Tue May 18 11:51:30 2004
+++ reopt/tools/ttftest/ttftest.cpp	Wed May 19 03:44:44 2004
@@ -19,6 +19,7 @@
 #include "llvm/Bytecode/Reader.h"
 #include "reopt/TraceToFunction.h"
 #include "llvm/Module.h"
+#include "Support/CommandLine.h"
 #include "Support/FileUtilities.h"
 #include "Support/StringExtras.h"
 #include <iostream>
@@ -27,6 +28,13 @@
 #include <cerrno>
 using namespace llvm;
 
+namespace {
+  cl::opt<std::string> TraceFile("trace",
+    cl::desc("Name of file containing trace"), cl::value_desc("traceFileName"));
+  cl::opt<std::string> BytecodeFile("bc",
+    cl::desc("Name of file containing module"), cl::value_desc("bytecodeFileName"));
+};
+
 /// getBasicBlockByNum - Returns the basic block in F whose index is
 /// bbNum.
 ///
@@ -36,14 +44,13 @@
   return block;
 }
 
-
 /// ReadTraceFromFile - Given a module, read a trace for some function
 /// in that module from the named file. Returns the Trace object, or a null
 /// pointer (and sets ErrorStr) if there was an error.
 ///
-Trace *ReadTraceFromFile (Module *M, const char *filename,
+Trace *ReadTraceFromFile (Module *M, const std::string &filename,
                           std::string *ErrorStr) {
-  std::ifstream in (filename);
+  std::ifstream in (filename.c_str ());
   if (!in.good()) {
     if (ErrorStr)
       *ErrorStr = std::string("Can't open '") + filename + "': "
@@ -73,7 +80,6 @@
   return new Trace (vBB);
 }
 
-
 /// getBasicBlockIndex - Returns the index of BB in its parent function.
 ///
 int getBasicBlockIndex (BasicBlock *BB) {
@@ -87,48 +93,31 @@
   return bbNum;
 }
 
-void WriteTraceToFile (Trace &T) {
-  std::string filename; 
-  unsigned count = 0;
-  do {
-    ++count;
-    filename = T.getFunction ()->getName () + ".trace" + utostr (count)
-             + ".txt";
-  } while (FileOpenable(filename));
-  std::ofstream out (filename.c_str ());
-  out << T.getFunction ()->getName () << "\n";
-  for (Trace::iterator i = T.begin (), e = T.end (); i != e; ++i) 
-    out << getBasicBlockIndex (*i) << " ";
-  out << "\n";
-  out.close ();
-}
-
 int main (int argc, char **argv) {
-  const char *bytecodeFileName = argv[1];
-  const char *traceFileName = argv[2];
+  // Get command line arguments.
+  cl::ParseCommandLineOptions(argc, argv, " trace-to-function tester\n");
 
-  // Read bytecode file
+  // Read bytecode file.
   std::string err;
-  Module *M = ParseBytecodeFile(bytecodeFileName, &err);
+  Module *M = ParseBytecodeFile (BytecodeFile, &err);
   if (!M) {
     std::cerr << "Error reading bytecode: " << err << "\n";
     exit (1);
   }
 
-  // Read trace file
+  // Read trace file.
   std::vector<BasicBlock *> vBB;
-  Trace *T = ReadTraceFromFile (M, traceFileName, &err);
+  Trace *T = ReadTraceFromFile (M, TraceFile, &err);
   if (!T) {
     std::cerr << "Error reading trace: " << err << "\n";
     exit (1);
   }
-  WriteTraceToFile (*T);
 
-  // Run TraceToFunction on the Trace
+  // Run TraceToFunction on the Trace.
   TraceFunction *TF = TraceFunction::get (*T);
   std::cerr << *TF->TraceFn << "\n\n";
 
-  // Run Verifier on the resulting TraceFunction
+  // Run the LLVM Verifier on the resulting TraceFunction.
   if (verifyFunction (*TF->TraceFn, PrintMessageAction)) {
     std::cerr << "Verifier fails.\n";
     exit (1);




From gaeke at cs.uiuc.edu  Wed May 19 03:48:03 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Wed May 19 03:48:03 2004
Subject: [llvm-commits] CVS: reopt/tools/Makefile
Message-ID: <200405190847.DAA19112@seraph.cs.uiuc.edu>


Changes in directory reopt/tools:

Makefile updated: 1.3 -> 1.4

---
Log message:

Build ttftest by default


---
Diffs of the changes:  (+1 -1)

Index: reopt/tools/Makefile
diff -u reopt/tools/Makefile:1.3 reopt/tools/Makefile:1.4
--- reopt/tools/Makefile:1.3	Wed Nov 19 16:51:54 2003
+++ reopt/tools/Makefile	Wed May 19 03:47:01 2004
@@ -1,6 +1,6 @@
 
 LEVEL := ..
-DIRS := 
+DIRS := ttftest
 
 include $(LEVEL)/Makefile.config
 




From gaeke at cs.uiuc.edu  Wed May 19 03:49:00 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Wed May 19 03:49:00 2004
Subject: [llvm-commits] CVS: reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp
Message-ID: <200405190848.DAA19130@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/LightWtProfiling:

RuntimeOptimizations.cpp updated: 1.33 -> 1.34

---
Log message:

When writing out a trace, also write out the module into a separate file.
This allows us to test trace-based optimizations based on exactly what the
trace optimizer would get as input.


---
Diffs of the changes:  (+17 -7)

Index: reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp
diff -u reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp:1.33 reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp:1.34
--- reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp:1.33	Tue May 18 11:51:28 2004
+++ reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp	Wed May 19 03:48:36 2004
@@ -22,6 +22,7 @@
 #include "Support/FileUtilities.h"
 #include "llvm/Analysis/Verifier.h"
 #include "llvm/Assembly/PrintModulePass.h"
+#include "llvm/Bytecode/Writer.h"
 #include "llvm/CodeGen/MachineFunction.h"
 #include "llvm/CodeGen/Passes.h"
 #include "llvm/CodeGen/MachineCodeEmitter.h"
@@ -81,20 +82,29 @@
 }     
                 
 void WriteTraceToFile (Trace &T) {
-  std::string filename;
+  // Get unique filename for trace & corresponding module.
+  std::string traceFileName, bytecodeFileName;
+  Function *F = T.getFunction ();
   unsigned count = 0;
   do {
     ++count;
-    filename = T.getFunction ()->getName () + ".trace" + utostr (count)
-             + ".txt";
-  } while (FileOpenable(filename));
-  std::ofstream out (filename.c_str ());
-  out << T.getFunction ()->getName () << "\n";
+    traceFileName = F->getName () + ".trace" + utostr (count) + ".txt";
+    bytecodeFileName = F->getName () + ".trace" + utostr (count) + ".bc";
+  } while (FileOpenable(traceFileName) || FileOpenable(bytecodeFileName));
+
+  // Write out trace.
+  std::ofstream out (traceFileName.c_str ());
+  out << F->getName () << "\n";
   for (Trace::iterator i = T.begin (), e = T.end (); i != e; ++i)
     out << getBasicBlockIndex (*i) << " ";
   out << "\n";
   out.close ();
-}   
+
+  // Write out bytecode.
+  std::ofstream bout (bytecodeFileName.c_str ());
+  WriteBytecodeToFile(F->getParent (), bout);
+  bout.close ();
+}
 
 /// This method is called when we have finally constructed a
 /// trace. The first parameter is the vector of basic blocks that form




From gaeke at cs.uiuc.edu  Wed May 19 03:54:01 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Wed May 19 03:54:01 2004
Subject: [llvm-commits] CVS: reopt/lib/TraceToFunction/TraceToFunction.cpp
Message-ID: <200405190853.DAA19252@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/TraceToFunction:

TraceToFunction.cpp updated: 1.42 -> 1.43

---
Log message:

Use new trace cloner


---
Diffs of the changes:  (+13 -12)

Index: reopt/lib/TraceToFunction/TraceToFunction.cpp
diff -u reopt/lib/TraceToFunction/TraceToFunction.cpp:1.42 reopt/lib/TraceToFunction/TraceToFunction.cpp:1.43
--- reopt/lib/TraceToFunction/TraceToFunction.cpp:1.42	Tue May 18 16:23:06 2004
+++ reopt/lib/TraceToFunction/TraceToFunction.cpp	Wed May 19 03:53:20 2004
@@ -282,18 +282,17 @@
 /// to O2CMap.
 ///
 static void cloneTraceBBsIntoFunction (Trace &T, Function *F, ValueMap &O2CMap){
-  // Clone each basic block into the new function.
-  for (Trace::iterator TI = T.begin (), TE = T.end (); TI != TE; ++TI) {
-    BasicBlock *srcB = *TI;
-    // Create a new BasicBlock dstB that corresponds to srcB in T.
-    BasicBlock *dstB = CloneBasicBlock (srcB, O2CMap, ".ttf");
-    DEBUG(dstB->setName (srcB->getName () + ".ttf"));
-    // Add dstB to F.
-    F->getBasicBlockList ().push_back (dstB);
-    // Remember the correspondence between srcB and dstB. (CloneFunction does
-    // this, but we aren't cloning the whole function.)
-    O2CMap[srcB] = dstB; 
-  }
+  for (Function::aiterator AI = T.getFunction ()->abegin (),
+                           AE = T.getFunction ()->aend (); AI != AE; ++AI)
+    O2CMap[AI] = AI;
+  for (Function::iterator FI = T.getFunction ()->begin (),
+                          FE = T.getFunction ()->end (); FI != FE; ++FI)
+    if (!T.contains (FI)) {
+      O2CMap[FI] = FI;
+      for (BasicBlock::iterator BI = FI->begin (), BE = FI->end (); BI != BE; ++BI)
+        O2CMap[BI] = BI;
+    }
+  CloneTraceInto (F, T, O2CMap, ".ttf");
 }
 
 /// isFirstTriggerCall - Returns true iff I is a call instruction which appears
@@ -403,6 +402,8 @@
   threadFLIEdges (F);
 
   numberExitBranchesInTrace (T, BranchNumber);
+
+  DEBUG (std::cerr << "\n FUNCTION BEFORE FIXUPS: \n" << *F << "\n\n");
 
   // Fix up the cloned basic blocks of the function so that they are
   // internally consistent.




From gaeke at cs.uiuc.edu  Wed May 19 04:09:00 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Wed May 19 04:09:00 2004
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Utils/CloneTrace.cpp
Message-ID: <200405190908.EAA19859@seraph.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Utils:

CloneTrace.cpp updated: 1.6 -> 1.7

---
Log message:

Add CloneTraceInto(), which is based on (and has mostly the same
effects as) CloneFunctionInto().


---
Diffs of the changes:  (+33 -0)

Index: llvm/lib/Transforms/Utils/CloneTrace.cpp
diff -u llvm/lib/Transforms/Utils/CloneTrace.cpp:1.6 llvm/lib/Transforms/Utils/CloneTrace.cpp:1.7
--- llvm/lib/Transforms/Utils/CloneTrace.cpp:1.6	Tue Feb  3 19:19:43 2004
+++ llvm/lib/Transforms/Utils/CloneTrace.cpp	Wed May 19 04:08:14 2004
@@ -15,9 +15,11 @@
 //
 //===----------------------------------------------------------------------===//
 
+#include "llvm/Analysis/Trace.h"
 #include "llvm/Transforms/Utils/Cloning.h"
 #include "llvm/iPHINode.h"
 #include "llvm/Function.h"
+#include "ValueMapper.h"
 using namespace llvm;
 
 //Clones the trace (a vector of basic blocks)
@@ -84,3 +86,34 @@
   //return new vector of basic blocks
   return clonedTrace;
 }
+
+/// CloneTraceInto - Clone T into NewFunc. Original<->clone mapping is
+/// saved in ValueMap.
+///
+void llvm::CloneTraceInto(Function *NewFunc, Trace &T,
+                          std::map<const Value*, Value*> &ValueMap,
+                          const char *NameSuffix) {
+  assert(NameSuffix && "NameSuffix cannot be null!");
+
+  // Loop over all of the basic blocks in the trace, cloning them as
+  // appropriate.
+  //
+  for (Trace::const_iterator BI = T.begin(), BE = T.end(); BI != BE; ++BI) {
+    const BasicBlock *BB = *BI;
+
+    // Create a new basic block and copy instructions into it!
+    BasicBlock *CBB = CloneBasicBlock(BB, ValueMap, NameSuffix, NewFunc);
+    ValueMap[BB] = CBB;                       // Add basic block mapping.
+  }
+
+  // Loop over all of the instructions in the new function, fixing up operand
+  // references as we go.  This uses ValueMap to do all the hard work.
+  //
+  for (Function::iterator BB =
+         cast<BasicBlock>(ValueMap[T.getEntryBasicBlock()]),
+         BE = NewFunc->end(); BB != BE; ++BB)
+    // Loop over all instructions, fixing each one as we find it...
+    for (BasicBlock::iterator II = BB->begin(); II != BB->end(); ++II)
+      RemapInstruction(II, ValueMap);
+}
+




From gaeke at cs.uiuc.edu  Wed May 19 04:09:03 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Wed May 19 04:09:03 2004
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Utils/CloneFunction.cpp ValueMapper.cpp ValueMapper.h
Message-ID: <200405190908.EAA19852@seraph.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Utils:

CloneFunction.cpp updated: 1.21 -> 1.22
ValueMapper.cpp updated: 1.11 -> 1.12
ValueMapper.h updated: 1.4 -> 1.5

---
Log message:

Move RemapInstruction() to ValueMapper, so that it can be shared with
CloneTrace, and because it is primarily an operation on ValueMaps.  It
is now a global (non-static) function which can be pulled in using
ValueMapper.h.


---
Diffs of the changes:  (+26 -22)

Index: llvm/lib/Transforms/Utils/CloneFunction.cpp
diff -u llvm/lib/Transforms/Utils/CloneFunction.cpp:1.21 llvm/lib/Transforms/Utils/CloneFunction.cpp:1.22
--- llvm/lib/Transforms/Utils/CloneFunction.cpp:1.21	Wed Feb  4 15:44:26 2004
+++ llvm/lib/Transforms/Utils/CloneFunction.cpp	Wed May 19 04:08:11 2004
@@ -20,25 +20,6 @@
 #include "ValueMapper.h"
 using namespace llvm;
 
-// RemapInstruction - Convert the instruction operands from referencing the 
-// current values into those specified by ValueMap.
-//
-static inline void RemapInstruction(Instruction *I, 
-                                    std::map<const Value *, Value*> &ValueMap) {
-  for (unsigned op = 0, E = I->getNumOperands(); op != E; ++op) {
-    const Value *Op = I->getOperand(op);
-    Value *V = MapValue(Op, ValueMap);
-#ifndef NDEBUG
-    if (!V) {
-      std::cerr << "Val = \n" << Op << "Addr = " << (void*)Op;
-      std::cerr << "\nInst = " << I;
-    }
-#endif
-    assert(V && "Referenced value not in value map!");
-    I->setOperand(op, V);
-  }
-}
-
 // CloneBasicBlock - See comments in Cloning.h
 BasicBlock *llvm::CloneBasicBlock(const BasicBlock *BB,
                                   std::map<const Value*, Value*> &ValueMap,


Index: llvm/lib/Transforms/Utils/ValueMapper.cpp
diff -u llvm/lib/Transforms/Utils/ValueMapper.cpp:1.11 llvm/lib/Transforms/Utils/ValueMapper.cpp:1.12
--- llvm/lib/Transforms/Utils/ValueMapper.cpp:1.11	Sat Feb 14 23:55:13 2004
+++ llvm/lib/Transforms/Utils/ValueMapper.cpp	Wed May 19 04:08:12 2004
@@ -99,3 +99,22 @@
   assert(0 && "Unknown value type: why didn't it get resolved?!");
   return 0;
 }
+
+/// RemapInstruction - Convert the instruction operands from referencing the
+/// current values into those specified by ValueMap.
+///
+void llvm::RemapInstruction(Instruction *I,
+                            std::map<const Value *, Value*> &ValueMap) {
+  for (unsigned op = 0, E = I->getNumOperands(); op != E; ++op) {
+    const Value *Op = I->getOperand(op);
+    Value *V = MapValue(Op, ValueMap);
+#ifndef NDEBUG
+    if (!V) {
+      std::cerr << "Val = \n" << Op << "Addr = " << (void*)Op;
+      std::cerr << "\nInst = " << I;
+    }
+#endif
+    assert(V && "Referenced value not in value map!");
+    I->setOperand(op, V);
+  }
+}


Index: llvm/lib/Transforms/Utils/ValueMapper.h
diff -u llvm/lib/Transforms/Utils/ValueMapper.h:1.4 llvm/lib/Transforms/Utils/ValueMapper.h:1.5
--- llvm/lib/Transforms/Utils/ValueMapper.h:1.4	Fri Jan  9 00:12:26 2004
+++ llvm/lib/Transforms/Utils/ValueMapper.h	Wed May 19 04:08:12 2004
@@ -12,14 +12,18 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef LIB_TRANSFORMS_UTILS_VALUE_MAPPER_H
-#define LIB_TRANSFORMS_UTILS_VALUE_MAPPER_H
+#ifndef VALUEMAPPER_H
+#define VALUEMAPPER_H
 
 #include <map>
 
 namespace llvm {
   class Value;
-  Value *MapValue(const Value *V, std::map<const Value*, Value*> &VM);
+  class Instruction;
+  typedef std::map<const Value *, Value *> ValueMapTy;
+
+  Value *MapValue(const Value *V, ValueMapTy &VM);
+  void RemapInstruction(Instruction *I, ValueMapTy &VM);
 } // End llvm namespace
 
 #endif




From gaeke at cs.uiuc.edu  Wed May 19 04:38:02 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Wed May 19 04:38:02 2004
Subject: [llvm-commits] CVS: reopt/lib/TraceToFunction/TraceToFunction.cpp
Message-ID: <200405190937.EAA20391@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/TraceToFunction:

TraceToFunction.cpp updated: 1.43 -> 1.44

---
Log message:

Rewrite threadFLIEdges() not to use RTTI so wastefully, and to skip
over most instructions.


---
Diffs of the changes:  (+35 -38)

Index: reopt/lib/TraceToFunction/TraceToFunction.cpp
diff -u reopt/lib/TraceToFunction/TraceToFunction.cpp:1.43 reopt/lib/TraceToFunction/TraceToFunction.cpp:1.44
--- reopt/lib/TraceToFunction/TraceToFunction.cpp:1.43	Wed May 19 03:53:20 2004
+++ reopt/lib/TraceToFunction/TraceToFunction.cpp	Wed May 19 04:37:04 2004
@@ -197,7 +197,7 @@
     for (unsigned i = 0; i < BI->getNumSuccessors (); ++i) {
       BasicBlock *BIT = BI->getSuccessor (i);
       if (!T.contains (BIT))
-	return BI;
+        return BI;
     }
   return NULL;
 }
@@ -444,46 +444,43 @@
 }
 
 void TraceFunctionBuilder::threadFLIEdges (Function *F) {
-  for (Function::iterator FI = F->begin (), FE = F->end (); FI != FE; ++FI)
-    for (BasicBlock::iterator Inst = FI->begin (), BE = FI->end (); Inst != BE; ++Inst) {
-      if (isa<PHINode> (Inst)) {
-        // In a PHI node, replace references to FLI blocks with the
-        // corresponding edge's source.
-        for (unsigned OI = 0, OE = Inst->getNumOperands(); OI != OE; ++OI) {
-          Value *op = Inst->getOperand (OI);
-          if (isa<BasicBlock> (op)) {
-            BasicBlock *BB = cast<BasicBlock> (op);
-            FLIMapTy::iterator It = FLIMap.find (BB);
-            if (It != FLIMap.end ()) {
-              DEBUG (std::cerr << "threadFLIEdges: threading operand " << OI
-                << " (FLI block " << BB->getName () << " --> <[" 
-                << It->second.first->getName () << "], "
-                << It->second.second->getName () << ">) of instruction: "
-                << *Inst << "\n");
-              Inst->setOperand (OI, It->second.first);
-            }
-          }
-        }
-      } else if (isa<BranchInst> (Inst)) {
-        // If it's a branch, replace references to FLI blocks with the
-        // corresponding edge's target.
-        for (unsigned OI = 0, OE = Inst->getNumOperands(); OI != OE; ++OI) {
-          Value *op = Inst->getOperand (OI);
-          if (isa<BasicBlock> (op)) {
-            BasicBlock *BB = cast<BasicBlock> (op);
-            FLIMapTy::iterator It = FLIMap.find (BB);
-            if (It != FLIMap.end ()) {
-              DEBUG (std::cerr << "threadFLIEdges: threading operand " << OI
-                << " (FLI block " << BB->getName () << " --> <" 
-                << It->second.first->getName () << ", ["
-                << It->second.second->getName () << "]>) of instruction: "
-                << *Inst << "\n");
-              Inst->setOperand (OI, It->second.second);
-            }
-          }
+  for (Function::iterator FI = F->begin (), FE = F->end (); FI != FE; ++FI) {
+    // Replace references to FLI blocks in PHI nodes with the
+    // corresponding edge's source.
+    for (BasicBlock::iterator Inst = FI->begin ();
+         PHINode *PN = dyn_cast<PHINode> (Inst); ++Inst)
+      for (unsigned i = 0, e = PN->getNumIncomingValues (); i != e; ++i) {
+        FLIMapTy::iterator It = FLIMap.find (PN->getIncomingBlock (i));
+        if (It != FLIMap.end ()) {
+          DEBUG (std::cerr << "threadFLIEdges: threading incoming block " << i
+            << " (FLI block " << It->first->getName () << " --> <[" 
+            << It->second.first->getName () << "], "
+            << It->second.second->getName () << ">) of PHI node: "
+            << *PN << "\n");
+          PN->setIncomingBlock (i, It->second.first);
         }
       }
+
+    // Look for a branch at the end of the basic block.
+    TerminatorInst *TI = FI->getTerminator ();
+    if (!TI) continue;
+    BranchInst *BI = dyn_cast<BranchInst> (TI);
+    if (!BI) continue;
+
+    // Replace references to FLI blocks in branches with the
+    // corresponding edge's target.
+    for (unsigned i = 0, e = BI->getNumSuccessors (); i != e; ++i) {
+      FLIMapTy::iterator It = FLIMap.find (BI->getSuccessor (i));
+      if (It != FLIMap.end ()) {
+        DEBUG (std::cerr << "threadFLIEdges: threading successor " << i
+          << " (FLI block " << It->first->getName () << " --> <" 
+          << It->second.first->getName () << ", ["
+          << It->second.second->getName () << "]>) of branch: "
+          << *BI << "\n");
+        BI->setSuccessor (i, It->second.second);
+      }
     }
+  }
 }
 
 /// buildFLIMap - Removes any FLI blocks which ended up in the trace, and




From gaeke at cs.uiuc.edu  Wed May 19 05:00:01 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Wed May 19 05:00:01 2004
Subject: [llvm-commits] CVS: reopt/lib/TraceToFunction/TraceToFunction.cpp
Message-ID: <200405190959.EAA20780@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/TraceToFunction:

TraceToFunction.cpp updated: 1.44 -> 1.45

---
Log message:

Make CloneTraceInto do the work of connecting live-in arguments to
their uses. (cloneTraceBBsIntoFunction now takes a TraceFunction arg.)

Give threadFLIEdges a doc comment.


---
Diffs of the changes:  (+21 -13)

Index: reopt/lib/TraceToFunction/TraceToFunction.cpp
diff -u reopt/lib/TraceToFunction/TraceToFunction.cpp:1.44 reopt/lib/TraceToFunction/TraceToFunction.cpp:1.45
--- reopt/lib/TraceToFunction/TraceToFunction.cpp:1.44	Wed May 19 04:37:04 2004
+++ reopt/lib/TraceToFunction/TraceToFunction.cpp	Wed May 19 04:59:45 2004
@@ -276,23 +276,27 @@
   }
 }
 
-/// cloneTraceBBsIntoFunction - Copy the BasicBlocks of the trace T into
-/// the new Function F, which should be initially empty. Correspondences
-/// between Original instructions (in T) and their Clones (in F) are added
-/// to O2CMap.
+/// cloneTraceBBsIntoFunction - Copy the BasicBlocks of the trace into
+/// the new TraceFn, which should be initially empty. Correspondences
+/// between Original instructions (in the trace) and their Clones (in
+/// TraceFn) are added to TF->O2CMap.
 ///
-static void cloneTraceBBsIntoFunction (Trace &T, Function *F, ValueMap &O2CMap){
-  for (Function::aiterator AI = T.getFunction ()->abegin (),
-                           AE = T.getFunction ()->aend (); AI != AE; ++AI)
+static void cloneTraceBBsIntoFunction (TraceFunction *TF) {
+  Function *MF = TF->MatrixFn;
+  LiveVariableSet &Si = TF->LiveInSet;
+  ValueMap &O2CMap = TF->O2CMap;
+  for (Function::aiterator AI = MF->abegin (), AE = MF->aend (); AI != AE; ++AI)
     O2CMap[AI] = AI;
-  for (Function::iterator FI = T.getFunction ()->begin (),
-                          FE = T.getFunction ()->end (); FI != FE; ++FI)
-    if (!T.contains (FI)) {
+  for (Function::iterator FI = MF->begin (), FE = MF->end (); FI != FE; ++FI)
+    if (!TF->T.contains (FI)) {
       O2CMap[FI] = FI;
-      for (BasicBlock::iterator BI = FI->begin (), BE = FI->end (); BI != BE; ++BI)
+      for (BasicBlock::iterator BI = FI->begin (), BE = FI->end (); BI != BE;
+           ++BI)
         O2CMap[BI] = BI;
     }
-  CloneTraceInto (F, T, O2CMap, ".ttf");
+  for (LiveVariableSet::iterator I = Si.begin (), E = Si.end (); I != E; ++I)
+    O2CMap[*I] = getFunctionArg (TF->TraceFn, TF->LiveInToParameterMap[*I]);
+  CloneTraceInto (TF->TraceFn, TF->T, O2CMap, ".ttf");
 }
 
 /// isFirstTriggerCall - Returns true iff I is a call instruction which appears
@@ -397,7 +401,7 @@
   buildFLIMap (T, FLIMap);
 
   // First copy the basic blocks from the trace.
-  cloneTraceBBsIntoFunction (T, F, O2CMap);
+  cloneTraceBBsIntoFunction (TF);
 
   threadFLIEdges (F);
 
@@ -443,6 +447,10 @@
   return true;
 }
 
+/// threadFLIEdges - Replace all references to FLI blocks in the given
+/// function: at PHI nodes, with the corresponding edge's source, and at
+/// branches, with the corresponding edge's target.
+///
 void TraceFunctionBuilder::threadFLIEdges (Function *F) {
   for (Function::iterator FI = F->begin (), FE = F->end (); FI != FE; ++FI) {
     // Replace references to FLI blocks in PHI nodes with the




From gaeke at cs.uiuc.edu  Wed May 19 05:11:01 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Wed May 19 05:11:01 2004
Subject: [llvm-commits] CVS: reopt/lib/TraceToFunction/TraceToFunction.cpp
Message-ID: <200405191010.FAA21030@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/TraceToFunction:

TraceToFunction.cpp updated: 1.45 -> 1.46

---
Log message:

Simplifications allowed by new FLI threader and trace cloner


---
Diffs of the changes:  (+1 -22)

Index: reopt/lib/TraceToFunction/TraceToFunction.cpp
diff -u reopt/lib/TraceToFunction/TraceToFunction.cpp:1.45 reopt/lib/TraceToFunction/TraceToFunction.cpp:1.46
--- reopt/lib/TraceToFunction/TraceToFunction.cpp:1.45	Wed May 19 04:59:45 2004
+++ reopt/lib/TraceToFunction/TraceToFunction.cpp	Wed May 19 05:10:09 2004
@@ -654,33 +654,12 @@
     Instruction &I = *BI;
     for (unsigned i = 0; i < I.getNumOperands (); ++i) {
       Value *V = I.getOperand (i);
-      // Fold Phi sources which come from FLI blocks
-      if (isa<PHINode> (&I))
-        if (BasicBlock *phiSource = dyn_cast<BasicBlock> (V))
-          if (BasicBlock *realSource = getFLIEdgeSource (phiSource))
-            V = realSource;
-
-      // If the instruction I has an operand which is in the live-in
-      // set of the trace, then we must replace that operand with
-      // the corresponding argument of F. We can find out which
-      // operands to replace by looking them up in
-      // TF->LiveInToParameterMap.
-      if (TF->LiveInToParameterMap.find (V) != TF->LiveInToParameterMap.end ()){
-        DEBUG(WriteAsOperand (std::cerr, V, true, true,
-                              TF->MatrixFn->getParent ());
-              std::cerr << " in instruction:\n " << I << " is argument "
-                        << TF->LiveInToParameterMap[V] << " in new function\n");
-        assert (V->getType () ==
-                getFunctionArg (F, TF->LiveInToParameterMap[V])->getType ()
-                && "Live-in Value's type doesn't match corresponding arg type");
-        I.setOperand(i, getFunctionArg (F, TF->LiveInToParameterMap[V]));
-      }
       // If the instruction I has an operand which is in the
       // trace, that operand will have been cloned into the
       // function, and I will still reference the version from
       // outside the function. Replace any reference to an operand
       // which has had a clone made with a reference to its clone.
-      else if (O2CMap.find (V) != O2CMap.end () && O2CMap[V]) {
+      if (O2CMap.find (V) != O2CMap.end () && O2CMap[V]) {
         DEBUG(std::cerr << *V << " in instruction " << I
               << " is value " << O2CMap[V] << " in new function\n");
         assert (V->getType () == O2CMap[V]->getType ()




From gaeke at cs.uiuc.edu  Wed May 19 05:33:01 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Wed May 19 05:33:01 2004
Subject: [llvm-commits] CVS: reopt/lib/TraceToFunction/TraceToFunction.cpp
Message-ID: <200405191032.FAA21420@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/TraceToFunction:

TraceToFunction.cpp updated: 1.46 -> 1.47

---
Log message:

Minor simplification: merge two asserts; remove a dyn_cast


---
Diffs of the changes:  (+3 -5)

Index: reopt/lib/TraceToFunction/TraceToFunction.cpp
diff -u reopt/lib/TraceToFunction/TraceToFunction.cpp:1.46 reopt/lib/TraceToFunction/TraceToFunction.cpp:1.47
--- reopt/lib/TraceToFunction/TraceToFunction.cpp:1.46	Wed May 19 05:10:09 2004
+++ reopt/lib/TraceToFunction/TraceToFunction.cpp	Wed May 19 05:32:12 2004
@@ -565,11 +565,9 @@
         if (!T.contains (phiSource)) {
           // FIXME: Assert that O2CMap[PN]'s value i is live in.
           Value *V = O2CMap[PN];
-          assert (V && "Can't find clone of Phi node from trace entry BB");
-          PHINode *PNinF = dyn_cast<PHINode> (V);
-          assert (PNinF
-                  && "Clone of Phi node from entry BB is not a Phi node");
-          PNinF->setIncomingBlock (i, EntryFixup);
+          assert (V && isa<PHINode> (V)
+                  && "Clone of PHINode from trace entry BB missing or mangled");
+          cast<PHINode> (V)->setIncomingBlock (i, EntryFixup);
         }
      }
   }




From alkis at cs.uiuc.edu  Wed May 19 14:19:03 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Wed May 19 14:19:03 2004
Subject: [llvm-commits] CVS: llvm/include/llvm/Transforms/Utils/Cloning.h
Message-ID: <200405191917.OAA11151@zion.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Transforms/Utils:

Cloning.h updated: 1.14 -> 1.15

---
Log message:

Declare function defined in namespace llvm as gcc-3.4 doesn't accept a definition in a namespace of a non-declared function.


---
Diffs of the changes:  (+8 -0)

Index: llvm/include/llvm/Transforms/Utils/Cloning.h
diff -u llvm/include/llvm/Transforms/Utils/Cloning.h:1.14 llvm/include/llvm/Transforms/Utils/Cloning.h:1.15
--- llvm/include/llvm/Transforms/Utils/Cloning.h:1.14	Tue Feb  3 21:59:08 2004
+++ llvm/include/llvm/Transforms/Utils/Cloning.h	Wed May 19 14:17:48 2004
@@ -31,6 +31,7 @@
 class InvokeInst;
 class ReturnInst;
 class CallSite;
+class Trace;
 
 /// CloneModule - Return an exact copy of the specified module
 ///
@@ -94,6 +95,13 @@
                        std::vector<ReturnInst*> &Returns,
                        const char *NameSuffix = "");
 
+
+/// CloneTraceInto - Clone T into NewFunc. Original<->clone mapping is
+/// saved in ValueMap.
+///
+void CloneTraceInto(Function *NewFunc, Trace &T,
+                    std::map<const Value*, Value*> &ValueMap,
+                    const char *NameSuffix);
 
 /// InlineFunction - This function inlines the called function into the basic
 /// block of the caller.  This returns false if it is not possible to inline




From gaeke at cs.uiuc.edu  Wed May 19 15:17:02 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Wed May 19 15:17:02 2004
Subject: [llvm-commits] CVS: reopt/lib/LightWtProfiling/TraceWriter.cpp RuntimeOptimizations.cpp
Message-ID: <200405192016.PAA24407@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/LightWtProfiling:

TraceWriter.cpp added (r1.1)
RuntimeOptimizations.cpp updated: 1.34 -> 1.35

---
Log message:

Must resist the urge to entangle trace writer with runtime trace optimizer!


---
Diffs of the changes:  (+64 -37)

Index: reopt/lib/LightWtProfiling/TraceWriter.cpp
diff -c /dev/null reopt/lib/LightWtProfiling/TraceWriter.cpp:1.1
*** /dev/null	Wed May 19 15:16:23 2004
--- reopt/lib/LightWtProfiling/TraceWriter.cpp	Wed May 19 15:16:12 2004
***************
*** 0 ****
--- 1,63 ----
+ //===-- LightWtProfiling/TraceWriter.cpp -------------------------*- C++ -*--=//
+ // 
+ //                     The LLVM Compiler Infrastructure
+ //
+ // This file was developed by the LLVM research group and is distributed under
+ // the University of Illinois Open Source License. See LICENSE.TXT for details.
+ // 
+ //===----------------------------------------------------------------------===//
+ //
+ // This file exports a method which can be used to save a trace to
+ // disk, along with the module it refers to.
+ //
+ //===----------------------------------------------------------------------===//
+ 
+ #include "llvm/Analysis/Trace.h"
+ #include "llvm/Bytecode/Writer.h"
+ #include "llvm/Function.h"
+ #include "Support/FileUtilities.h"
+ #include "Support/StringExtras.h"
+ #include <fstream>
+ using namespace llvm;
+ 
+ namespace llvm {
+ 
+ /// getBasicBlockIndex - Returns the index of BB in its parent function.
+ ///
+ static inline int getBasicBlockIndex (BasicBlock *BB) {
+   int bbNum = 0;
+   Function *F = BB->getParent ();
+   Function::iterator block = F->begin();
+   while (&*block != BB) {
+     ++block;              
+     ++bbNum;
+   }
+   return bbNum;
+ }     
+ 
+ void WriteTraceToFile (Trace &T) {
+   // Get unique filename for trace & corresponding module.
+   std::string traceFileName, bytecodeFileName;
+   Function *F = T.getFunction ();
+   unsigned count = 0;
+   do {
+     ++count;
+     traceFileName = F->getName () + ".trace" + utostr (count) + ".txt";
+     bytecodeFileName = F->getName () + ".trace" + utostr (count) + ".bc";
+   } while (FileOpenable(traceFileName) || FileOpenable(bytecodeFileName));
+ 
+   // Write out trace.
+   std::ofstream out (traceFileName.c_str ());
+   out << F->getName () << "\n";
+   for (Trace::iterator i = T.begin (), e = T.end (); i != e; ++i)
+     out << getBasicBlockIndex (*i) << " ";
+   out << "\n";
+   out.close ();
+ 
+   // Write out bytecode.
+   std::ofstream bout (bytecodeFileName.c_str ());
+   WriteBytecodeToFile(F->getParent (), bout);
+   bout.close ();
+ }
+ 
+ } // end namespace llvm


Index: reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp
diff -u reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp:1.34 reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp:1.35
--- reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp:1.34	Wed May 19 03:48:36 2004
+++ reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp	Wed May 19 15:16:11 2004
@@ -68,43 +68,7 @@
 
 ///==----------------------------------------------------------------------==///
 
-/// getBasicBlockIndex - Returns the index of BB in its parent function.
-///
-int getBasicBlockIndex (BasicBlock *BB) {
-  int bbNum = 0;
-  Function *F = BB->getParent ();
-  Function::iterator block = F->begin();
-  while (&*block != BB) {
-    ++block;              
-    ++bbNum;
-  }
-  return bbNum;
-}     
-                
-void WriteTraceToFile (Trace &T) {
-  // Get unique filename for trace & corresponding module.
-  std::string traceFileName, bytecodeFileName;
-  Function *F = T.getFunction ();
-  unsigned count = 0;
-  do {
-    ++count;
-    traceFileName = F->getName () + ".trace" + utostr (count) + ".txt";
-    bytecodeFileName = F->getName () + ".trace" + utostr (count) + ".bc";
-  } while (FileOpenable(traceFileName) || FileOpenable(bytecodeFileName));
-
-  // Write out trace.
-  std::ofstream out (traceFileName.c_str ());
-  out << F->getName () << "\n";
-  for (Trace::iterator i = T.begin (), e = T.end (); i != e; ++i)
-    out << getBasicBlockIndex (*i) << " ";
-  out << "\n";
-  out.close ();
-
-  // Write out bytecode.
-  std::ofstream bout (bytecodeFileName.c_str ());
-  WriteBytecodeToFile(F->getParent (), bout);
-  bout.close ();
-}
+extern void WriteTraceToFile (Trace &T);
 
 /// This method is called when we have finally constructed a
 /// trace. The first parameter is the vector of basic blocks that form




From alkis at cs.uiuc.edu  Wed May 19 15:46:02 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Wed May 19 15:46:02 2004
Subject: [llvm-commits] CVS: llvm-java/tools/classdump/classdump.cpp
Message-ID: <200405192044.PAA07160@zion.cs.uiuc.edu>


Changes in directory llvm-java/tools/classdump:

classdump.cpp updated: 1.3 -> 1.4

---
Log message:

Use an auto_ptr instead of a raw pointer.


---
Diffs of the changes:  (+4 -2)

Index: llvm-java/tools/classdump/classdump.cpp
diff -u llvm-java/tools/classdump/classdump.cpp:1.3 llvm-java/tools/classdump/classdump.cpp:1.4
--- llvm-java/tools/classdump/classdump.cpp:1.3	Fri Apr 16 19:55:08 2004
+++ llvm-java/tools/classdump/classdump.cpp	Wed May 19 15:44:40 2004
@@ -16,15 +16,17 @@
 
 #include <cstddef>
 #include <iostream>
+#include <memory>
 
 using namespace llvm;
 
 int main(int argc, char* argv[])
 {
     try {
-        Java::ClassFile* cf = Java::ClassFile::readClassFile(std::cin);
+        std::auto_ptr<Java::ClassFile> cf(
+            Java::ClassFile::readClassFile(std::cin));
+
         cf->dump(std::cout);
-        delete cf;
     }
     catch (std::exception& e) {
         std::cerr << e.what() << '\n';




From alkis at cs.uiuc.edu  Wed May 19 16:11:01 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Wed May 19 16:11:01 2004
Subject: [llvm-commits] CVS: llvm-java/test/ClassFile/Simple.java
Message-ID: <200405192110.QAA16393@zion.cs.uiuc.edu>


Changes in directory llvm-java/test/ClassFile:

Simple.java updated: 1.1 -> 1.2

---
Log message:

Put Simple class in a package.


---
Diffs of the changes:  (+2 -0)

Index: llvm-java/test/ClassFile/Simple.java
diff -u llvm-java/test/ClassFile/Simple.java:1.1 llvm-java/test/ClassFile/Simple.java:1.2
--- llvm-java/test/ClassFile/Simple.java:1.1	Wed Apr 21 11:36:04 2004
+++ llvm-java/test/ClassFile/Simple.java	Wed May 19 16:10:18 2004
@@ -1,3 +1,5 @@
+package edu.uiuc.cs.llvm;
+
 public class Simple
 {
     public static int main(String[] args) {




From brukman at cs.uiuc.edu  Wed May 19 16:19:02 2004
From: brukman at cs.uiuc.edu (Misha Brukman)
Date: Wed May 19 16:19:02 2004
Subject: [llvm-commits] CVS: llvm/test/Programs/TEST.nightly.Makefile
Message-ID: <200405192117.QAA02148@zion.cs.uiuc.edu>


Changes in directory llvm/test/Programs:

TEST.nightly.Makefile updated: 1.28 -> 1.29

---
Log message:

The syntax for the vim mode is wrong as nothing else can be on the same line.


---
Diffs of the changes:  (+1 -1)

Index: llvm/test/Programs/TEST.nightly.Makefile
diff -u llvm/test/Programs/TEST.nightly.Makefile:1.28 llvm/test/Programs/TEST.nightly.Makefile:1.29
--- llvm/test/Programs/TEST.nightly.Makefile:1.28	Sun Feb 29 22:43:08 2004
+++ llvm/test/Programs/TEST.nightly.Makefile	Wed May 19 16:17:35 2004
@@ -1,4 +1,4 @@
-##===- test/Programs/TEST.nightly.Makefile --*- Makefile -*- vim:ft=make --===##
+##===- test/Programs/TEST.nightly.Makefile ----------------*- Makefile -*--===##
 #
 # This test is used in conjunction with the llvm/utils/NightlyTest* stuff to
 # generate information about program status for the nightly report.




From brukman at cs.uiuc.edu  Wed May 19 16:22:05 2004
From: brukman at cs.uiuc.edu (Misha Brukman)
Date: Wed May 19 16:22:05 2004
Subject: [llvm-commits] CVS: llvm/utils/vim/README
Message-ID: <200405192120.QAA04696@zion.cs.uiuc.edu>


Changes in directory llvm/utils/vim:

README updated: 1.3 -> 1.4

---
Log message:

Add hint about syntax-highlighting non-standardly-named LLVM Makefiles.


---
Diffs of the changes:  (+9 -0)

Index: llvm/utils/vim/README
diff -u llvm/utils/vim/README:1.3 llvm/utils/vim/README:1.4
--- llvm/utils/vim/README:1.3	Mon Aug 11 13:05:19 2003
+++ llvm/utils/vim/README	Wed May 19 16:20:40 2004
@@ -32,3 +32,12 @@
 Note: If you notice missing or incorrect syntax highlighting, please contact
 <llvmbugs [at] cs.uiuc.edu>; if you wish to provide a patch to improve the
 functionality, it will be most appreciated. Thank you.
+
+If you find yourself working with LLVM Makefiles often, but you don't get syntax
+highlighting (because the files have names such as Makefile.rules or
+TEST.nightly.Makefile), add the following to your ~/.vimrc:
+
+  " LLVM Makefile highlighting mode
+  augroup filetype
+    au! BufRead,BufNewFile *Makefile*     set filetype=make
+  augroup END




From gaeke at cs.uiuc.edu  Wed May 19 16:31:01 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Wed May 19 16:31:01 2004
Subject: [llvm-commits] CVS: llvm/lib/Target/SparcV9/SparcV9CodeEmitter.cpp
Message-ID: <200405192130.QAA29459@seraph.cs.uiuc.edu>


Changes in directory llvm/lib/Target/SparcV9:

SparcV9CodeEmitter.cpp updated: 1.60 -> 1.61

---
Log message:

Minor simplification: eliminate a dyn_cast.
Fix a typo in a debug message. 


---
Diffs of the changes:  (+4 -9)

Index: llvm/lib/Target/SparcV9/SparcV9CodeEmitter.cpp
diff -u llvm/lib/Target/SparcV9/SparcV9CodeEmitter.cpp:1.60 llvm/lib/Target/SparcV9/SparcV9CodeEmitter.cpp:1.61
--- llvm/lib/Target/SparcV9/SparcV9CodeEmitter.cpp:1.60	Fri Apr 23 13:10:38 2004
+++ llvm/lib/Target/SparcV9/SparcV9CodeEmitter.cpp	Wed May 19 16:30:01 2004
@@ -596,14 +596,9 @@
       DEBUG(std::cerr << "Saving reference to BB (VReg)\n");
       unsigned* CurrPC = (unsigned*)(intptr_t)MCE.getCurrentPCValue();
       BBRefs.push_back(std::make_pair(BB, std::make_pair(CurrPC, &MI)));
-    } else if (const Constant *C = dyn_cast<Constant>(V)) {
-      if (const ConstantInt *CI = dyn_cast<ConstantInt>(C)) {
-        rv = CI->getRawValue() - MCE.getCurrentPCValue();
-      } else {
-        std::cerr << "Cannot have non-integral const in instruction: "
-                  << *C;
-        abort();
-      }
+    } else if (const ConstantInt *CI = dyn_cast<ConstantInt>(V)) {
+      // Make constant PC-relative by subtracting the PC from it.
+      rv = CI->getRawValue() - MCE.getCurrentPCValue();
     } else if (GlobalValue *GV = dyn_cast<GlobalValue>(V)) {
       // same as MO.isGlobalAddress()
       DEBUG(std::cerr << "GlobalValue: ");
@@ -777,7 +772,7 @@
     if (binCode == (1 << 30)) {
       // this is an invalid call: the addr is out of bounds. that means a code
       // sequence has already been emitted, and this is a no-op
-      DEBUG(std::cerr << "Call supressed: already emitted far call.\n");
+      DEBUG(std::cerr << "Call suppressed: already emitted far call.\n");
     } else {
       emitWord(binCode);
     }




From gaeke at cs.uiuc.edu  Thu May 20 02:44:01 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Thu May 20 02:44:01 2004
Subject: [llvm-commits] CVS: llvm/lib/Target/SparcV9/SparcV9CodeEmitter.cpp
Message-ID: <200405200743.CAA03040@seraph.cs.uiuc.edu>


Changes in directory llvm/lib/Target/SparcV9:

SparcV9CodeEmitter.cpp updated: 1.61 -> 1.62

---
Log message:

Fix a bug in outputting branches to constant PCs. Since the PC is supplied as
a full 64-bit address, it must be adjusted to fit in the branch instruction's
immediate field. (This is only used in the reoptimizer, for now.)


---
Diffs of the changes:  (+3 -2)

Index: llvm/lib/Target/SparcV9/SparcV9CodeEmitter.cpp
diff -u llvm/lib/Target/SparcV9/SparcV9CodeEmitter.cpp:1.61 llvm/lib/Target/SparcV9/SparcV9CodeEmitter.cpp:1.62
--- llvm/lib/Target/SparcV9/SparcV9CodeEmitter.cpp:1.61	Wed May 19 16:30:01 2004
+++ llvm/lib/Target/SparcV9/SparcV9CodeEmitter.cpp	Thu May 20 02:43:40 2004
@@ -597,8 +597,9 @@
       unsigned* CurrPC = (unsigned*)(intptr_t)MCE.getCurrentPCValue();
       BBRefs.push_back(std::make_pair(BB, std::make_pair(CurrPC, &MI)));
     } else if (const ConstantInt *CI = dyn_cast<ConstantInt>(V)) {
-      // Make constant PC-relative by subtracting the PC from it.
-      rv = CI->getRawValue() - MCE.getCurrentPCValue();
+      // The real target of the branch is CI = PC + (rv * 4)
+      // So undo that: give the instruction (CI - PC) / 4
+      rv = (CI->getRawValue() - MCE.getCurrentPCValue()) / 4;
     } else if (GlobalValue *GV = dyn_cast<GlobalValue>(V)) {
       // same as MO.isGlobalAddress()
       DEBUG(std::cerr << "GlobalValue: ");




From gaeke at cs.uiuc.edu  Thu May 20 12:03:02 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Thu May 20 12:03:02 2004
Subject: [llvm-commits] CVS: reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp
Message-ID: <200405201700.MAA26611@zion.cs.uiuc.edu>


Changes in directory reopt/lib/LightWtProfiling:

RuntimeOptimizations.cpp updated: 1.35 -> 1.36

---
Log message:

Add a hook which is easy to set a breakpoint on for debugging purposes.
For extra bonus points, it even prints out gdb breakpoint commands you'll want
to use.


---
Diffs of the changes:  (+7 -0)

Index: reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp
diff -u reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp:1.35 reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp:1.36
--- reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp:1.35	Wed May 19 15:16:11 2004
+++ reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp	Thu May 20 12:00:45 2004
@@ -70,6 +70,12 @@
 
 extern void WriteTraceToFile (Trace &T);
 
+extern "C" void TraceOptimizerDone (uint64_t a, uint64_t traceStartAddr) {
+  std::cerr << "\n*** About to return from optimizeTrace(). GDB breakpoints:\n"
+    << std::hex << "display/i $pc\n" << "b *0x" << a << "\n" << "b *0x"
+    << traceStartAddr << std::dec << "\n";
+}
+
 /// This method is called when we have finally constructed a
 /// trace. The first parameter is the vector of basic blocks that form
 /// the trace; the second parameter is the starting
@@ -132,6 +138,7 @@
                   << " to point to 0x" << traceStartAddr << std::dec << "\n");
   vm->writeBranchInstruction(a, traceStartAddr);
   doFlush (a, a + 4);
+  TraceOptimizerDone(a, traceStartAddr);
 }
 
 } // end namespace llvm




From gaeke at cs.uiuc.edu  Thu May 20 12:13:04 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Thu May 20 12:13:04 2004
Subject: [llvm-commits] CVS: reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp
Message-ID: <200405201711.MAA26886@zion.cs.uiuc.edu>


Changes in directory reopt/lib/LightWtProfiling:

UnpackTraceFunction.cpp updated: 1.61 -> 1.62

---
Log message:

Print out more debugging information about the mapping info we got from LLC
about the return basic block.


---
Diffs of the changes:  (+8 -2)

Index: reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp
diff -u reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp:1.61 reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp:1.62
--- reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp:1.61	Mon May 17 16:00:19 2004
+++ reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp	Thu May 20 12:11:48 2004
@@ -530,8 +530,14 @@
   assert ((TF->ReturnBlockForTraceExit.find (RBB) !=
 		   TF->ReturnBlockForTraceExit.end ())
 		  && "Can't find matrix fn BB address to return to from trace");
-  uint64_t ReturnAddress = 
-	getBasicBlockInfo(TF->ReturnBlockForTraceExit[RBB]).first;
+  std::pair<uint64_t, uint64_t> BlockAddrs =
+    getBasicBlockInfo(TF->ReturnBlockForTraceExit[RBB]);
+  uint64_t ReturnAddress = BlockAddrs.first;
+  DEBUG (std::cerr << "Return block for trace exit path is: "
+         << *TF->ReturnBlockForTraceExit[RBB] << "\n"
+         << "Mapping info says addresses are: return address = 0x"
+         << std::hex << BlockAddrs.first << ", end of block = 0x"
+         << BlockAddrs.second << std::dec << "\n");
   // Insert a branch back to the return BasicBlock of the matrix fn.
   insertBranchMachineInstrs (ReturnAddress, MBB);
 }




From gaeke at cs.uiuc.edu  Thu May 20 13:06:02 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Thu May 20 13:06:02 2004
Subject: [llvm-commits] CVS: reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp
Message-ID: <200405201805.NAA05093@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/LightWtProfiling:

UnpackTraceFunction.cpp updated: 1.62 -> 1.63

---
Log message:

We were restoring fp without saving it in the first place. This was
causing a crash on return from the MatrixFn, when it tried to RESTORE
its caller's stack.


---
Diffs of the changes:  (+5 -0)

Index: reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp
diff -u reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp:1.62 reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp:1.63
--- reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp:1.62	Thu May 20 12:11:48 2004
+++ reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp	Thu May 20 13:05:22 2004
@@ -226,6 +226,10 @@
     regsUsed.insert (SparcV9::fsr);
   if (intCCRegSeen)
     regsUsed.insert (SparcV9::ccr);
+
+  // Always put fp in the set because it is restored unconditionally.
+  static const unsigned fp = SparcV9::i6;
+  regsUsed.insert (fp);
 }
 
 void UnpackTraceFunction::rewriteProlog (MachineFunction &MF,
@@ -485,6 +489,7 @@
   for (std::vector<MachineInstr *>::iterator vi = mvec.begin (),
        ve = mvec.end (); vi != ve; ++vi)
     MBB.push_back (*vi);
+  regsUsed.erase (fp);
 
   // Insert copies from each live-out variable's reg. in the trace
   // to its reg. in the matrix function.




From gaeke at cs.uiuc.edu  Thu May 20 13:10:03 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Thu May 20 13:10:03 2004
Subject: [llvm-commits] CVS: reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp TraceOptEmitter.cpp
Message-ID: <200405201809.NAA05211@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/LightWtProfiling:

RuntimeOptimizations.cpp updated: 1.36 -> 1.37
TraceOptEmitter.cpp updated: 1.8 -> 1.9

---
Log message:

Put DEBUG(...) around more debugging printouts and stuff.


---
Diffs of the changes:  (+8 -8)

Index: reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp
diff -u reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp:1.36 reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp:1.37
--- reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp:1.36	Thu May 20 12:00:45 2004
+++ reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp	Thu May 20 13:09:30 2004
@@ -71,9 +71,9 @@
 extern void WriteTraceToFile (Trace &T);
 
 extern "C" void TraceOptimizerDone (uint64_t a, uint64_t traceStartAddr) {
-  std::cerr << "\n*** About to return from optimizeTrace(). GDB breakpoints:\n"
-    << std::hex << "display/i $pc\n" << "b *0x" << a << "\n" << "b *0x"
-    << traceStartAddr << std::dec << "\n";
+  DEBUG (std::cerr << "\n*** About to return from optimizeTrace(). GDB"
+    << " breakpoints:\n" << std::hex << "display/i $pc\n" << "b *0x" << a
+    << "\n" << "b *0x" << traceStartAddr << std::dec << "\n");
 }
 
 /// This method is called when we have finally constructed a
@@ -100,7 +100,7 @@
   // Turn the vector of basic blocks into a Trace, and then turn the Trace into
   // a TraceFunction.
   Trace T (vBB);
-  WriteTraceToFile (T);
+  DEBUG (WriteTraceToFile (T));
   TraceFunction *TF = TraceFunction::get (T);
 
   // Force the SPARCv9 register allocator to save its state into a global


Index: reopt/lib/LightWtProfiling/TraceOptEmitter.cpp
diff -u reopt/lib/LightWtProfiling/TraceOptEmitter.cpp:1.8 reopt/lib/LightWtProfiling/TraceOptEmitter.cpp:1.9
--- reopt/lib/LightWtProfiling/TraceOptEmitter.cpp:1.8	Wed Apr 28 16:41:09 2004
+++ reopt/lib/LightWtProfiling/TraceOptEmitter.cpp	Thu May 20 13:09:30 2004
@@ -166,16 +166,16 @@
 void TraceOptEmitter::emitWord(unsigned W) {
   // This won't work if the endianness of the host and target don't agree!  (For
   // a JIT this can't happen though.  :)
-  std::cerr << "TraceOptEmitter: emitting word 0x" << std::hex << W
-            << " at cursor: " << CurByte << std::dec << "\n";
+  DEBUG (std::cerr << "TraceOptEmitter: emitting word 0x" << std::hex << W
+    << " at cursor: " << CurByte << std::dec << "\n");
   vm->writeInstToVM (CurByte, W);  // does: *(unsigned*)CurByte = W;
   doFlush (CurByte - sizeof (unsigned), (uint64_t) CurByte + sizeof (unsigned));
   CurByte += sizeof(unsigned);
 }
 
 void TraceOptEmitter::emitWordAt(unsigned W, unsigned *Ptr) {
-  std::cerr << "TraceOptEmitter: emitting word 0x" << std::hex << W
-            << std::dec << " at pointer: " << Ptr << "\n";
+  DEBUG (std::cerr << "TraceOptEmitter: emitting word 0x" << std::hex << W
+    << std::dec << " at pointer: " << Ptr << "\n");
   vm->writeInstToVM ((uint64_t) Ptr, W);  // does: *Ptr = W;
   doFlush ((uint64_t) Ptr - sizeof (unsigned), (uint64_t) Ptr + sizeof (unsigned));
 }




From gaeke at cs.uiuc.edu  Thu May 20 13:24:03 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Thu May 20 13:24:03 2004
Subject: [llvm-commits] CVS: reopt/lib/TraceToFunction/TraceToFunction.cpp
Message-ID: <200405201822.NAA06321@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/TraceToFunction:

TraceToFunction.cpp updated: 1.47 -> 1.48

---
Log message:

Put DEBUG(...) around more debugging printouts and stuff.


---
Diffs of the changes:  (+3 -3)

Index: reopt/lib/TraceToFunction/TraceToFunction.cpp
diff -u reopt/lib/TraceToFunction/TraceToFunction.cpp:1.47 reopt/lib/TraceToFunction/TraceToFunction.cpp:1.48
--- reopt/lib/TraceToFunction/TraceToFunction.cpp:1.47	Wed May 19 05:32:12 2004
+++ reopt/lib/TraceToFunction/TraceToFunction.cpp	Thu May 20 13:22:48 2004
@@ -587,9 +587,9 @@
       // Fold branches which go to FLI blocks
       if (BasicBlock *backEdgeTarget =
           getFLIEdgeTarget (successor)) {
-        std::cerr << "Branch instr " << *BI << " looks like an instrumented "
-                  << "back-edge from " << srcB->getName() << " to "
-                  << backEdgeTarget->getName () << "\n";
+        DEBUG (std::cerr << "Branch instr " << *BI << " looks like an"
+          << " instrumented back-edge from " << srcB->getName()
+          << " to " << backEdgeTarget->getName () << "\n");
         successor = backEdgeTarget;
       }
       if (!T.contains (successor)) {




From gaeke at cs.uiuc.edu  Thu May 20 14:36:12 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Thu May 20 14:36:12 2004
Subject: [llvm-commits] CVS: reopt/lib/TraceToFunction/TraceToFunction.cpp
Message-ID: <200405201934.OAA27641@zion.cs.uiuc.edu>


Changes in directory reopt/lib/TraceToFunction:

TraceToFunction.cpp updated: 1.48 -> 1.49

---
Log message:

Eliminate a dyn_cast. Rename fixupPhisAndCalls to fixupPhis, because it no
longer needs to deal with calls. 


---
Diffs of the changes:  (+5 -6)

Index: reopt/lib/TraceToFunction/TraceToFunction.cpp
diff -u reopt/lib/TraceToFunction/TraceToFunction.cpp:1.48 reopt/lib/TraceToFunction/TraceToFunction.cpp:1.49
--- reopt/lib/TraceToFunction/TraceToFunction.cpp:1.48	Thu May 20 13:22:48 2004
+++ reopt/lib/TraceToFunction/TraceToFunction.cpp	Thu May 20 14:34:26 2004
@@ -316,7 +316,7 @@
   return false;
 }
 
-static void fixupPhisAndCalls (BasicBlock *dstB, ValueMap &O2CMap) {
+static void fixupPhis (BasicBlock *dstB, ValueMap &O2CMap) {
   std::vector<Instruction *> goners;
   Function *F = dstB->getParent ();
   // Fix up Phi nodes:
@@ -366,13 +366,12 @@
 	  }
 	} while (lastOffTraceSrcFound != -1); // Continue until none found.
       }
-    }
     // Make sure that our Phi fixups did the Right Thing.
-    DEBUG(if (PHINode *PN = dyn_cast<PHINode> (&I))
-	  if (PN->use_size () != 0) // 0-use nodes get deleted later
+    DEBUG(if (PN->use_size () != 0) // 0-use nodes get deleted later
 	  for (unsigned i = 0; i < PN->getNumIncomingValues (); ++i)
 	  assert (PN->getIncomingBlock (i)->getParent () == F &&
-		  "Sorry, fixupPhisAndCalls mishandled a Phi node"));
+		  "Sorry, fixupPhis mishandled a Phi node"));
+    }
   }
 
   while (!goners.empty ()) {
@@ -419,7 +418,7 @@
     fixupFunctionBodyBB (T, F, srcB, dstB, O2CMap, So);
   }
   for (Function::iterator FI = F->begin (), FE = F->end (); FI != FE; ++FI)
-    fixupPhisAndCalls (FI, O2CMap);
+    fixupPhis (FI, O2CMap);
 }
 
 /// identifyFLIEdge - returns a true value iff BB is a basic block containing




From gaeke at cs.uiuc.edu  Thu May 20 14:52:03 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Thu May 20 14:52:03 2004
Subject: [llvm-commits] CVS: reopt/lib/TraceToFunction/TraceToFunction.cpp
Message-ID: <200405201950.OAA27892@zion.cs.uiuc.edu>


Changes in directory reopt/lib/TraceToFunction:

TraceToFunction.cpp updated: 1.49 -> 1.50

---
Log message:

In fixupPhis(), only iterate over Phi nodes in each cloned BasicBlock, not the
whole BasicBlock. Also, untabify most of this function.

Merge two asserts in fixupFunctionBodyBB().


---
Diffs of the changes:  (+48 -61)

Index: reopt/lib/TraceToFunction/TraceToFunction.cpp
diff -u reopt/lib/TraceToFunction/TraceToFunction.cpp:1.49 reopt/lib/TraceToFunction/TraceToFunction.cpp:1.50
--- reopt/lib/TraceToFunction/TraceToFunction.cpp:1.49	Thu May 20 14:34:26 2004
+++ reopt/lib/TraceToFunction/TraceToFunction.cpp	Thu May 20 14:50:44 2004
@@ -320,66 +320,61 @@
   std::vector<Instruction *> goners;
   Function *F = dstB->getParent ();
   // Fix up Phi nodes:
-  for (BasicBlock::iterator BI = dstB->begin (), BE = dstB->end ();
-       BI != BE; ++BI) {
-    Instruction &I = *BI;
-    // In all cases, if a Phi node source in T was an on-trace basic
-    // block, then it will already have been fixed up to point to
-    // that block's clone, so we find off-trace sources by looking
-    // for source BBs which are not in F.
-    if (PHINode *PN = dyn_cast<PHINode> (&I)) {
-      unsigned onTraceSources = 0;
-      int lastSrcFound = -1;
-      for (unsigned i = 0; i < PN->getNumIncomingValues (); ++i)
-	if (PN->getIncomingBlock (i)->getParent () == F) {
-	  lastSrcFound = i;
-	  ++onTraceSources;
-	}
-      // Case 0. If the phi node has 0 sources on the trace, abort,
-      // because that should really never happen.
-      assert (onTraceSources != 0
-	      && "Phi node on trace has ALL its sources from off-trace!");
-      // Case 1. If it has 1 source S on the trace, replace its uses
-      // with S.
-      if (onTraceSources == 1) {
-	DEBUG(std::cerr << "Replacing Phi node" << *PN
-	      << " with its lone on-trace input "
-	      << *PN->getIncomingValue (lastSrcFound) << "\n");
-	PN->replaceAllUsesWith (PN->getIncomingValue (lastSrcFound));
-	goners.push_back (BI); // Delete non-used phi node later.
-      } else {
-	// Case N. If the phi node has >1 source on the trace, just
-	// delete sources from the Phi node that are not on the trace.
-	int lastOffTraceSrcFound;
-	do {
-	  lastOffTraceSrcFound = -1;
-	  for (unsigned i = 0; i < PN->getNumIncomingValues (); ++i)
-	    if (PN->getIncomingBlock (i)->getParent () != F) {
-	      lastOffTraceSrcFound = i; // Found an off-trace source.
-	      break;
-	    }
-	  if (lastOffTraceSrcFound != -1) { // Found one?
-	    DEBUG(std::cerr << "Removing off-trace input "
-		  << *PN->getIncomingValue (lastOffTraceSrcFound)
-		  << " from Phi node " << *PN << "\n");
-	    PN->removeIncomingValue (lastOffTraceSrcFound); // Delete it.
-	  }
-	} while (lastOffTraceSrcFound != -1); // Continue until none found.
+  for (BasicBlock::iterator Inst = dstB->begin ();
+       PHINode *PN = dyn_cast<PHINode> (Inst); ++Inst) {
+    // If a Phi node source in T was an on-trace basic block, then it
+    // will already have been fixed up to point to that block's clone, so we
+    // find off-trace sources by looking for source BBs which are not in F.
+    unsigned onTraceSources = 0;
+    int lastSrcFound = -1;
+    for (unsigned i = 0; i < PN->getNumIncomingValues (); ++i)
+      if (PN->getIncomingBlock (i)->getParent () == F) {
+        lastSrcFound = i;
+        ++onTraceSources;
       }
+    // Case 0. If the phi node has 0 sources on the trace, abort,
+    // because that should really never happen.
+    assert (onTraceSources != 0
+            && "Phi node on trace has ALL its sources from off-trace!");
+    // Case 1. If it has 1 source S on the trace, replace its uses
+    // with S.
+    if (onTraceSources == 1) {
+      DEBUG(std::cerr << "Replacing Phi node" << *PN << " with its lone"
+        << " on-trace input " << *PN->getIncomingValue (lastSrcFound) << "\n");
+      PN->replaceAllUsesWith (PN->getIncomingValue (lastSrcFound));
+      goners.push_back (PN); // Delete non-used phi node later.
+    } else {
+      // Case N. If the phi node has >1 source on the trace, just
+      // delete sources from the Phi node that are not on the trace.
+      int lastOffTraceSrcFound;
+      do {
+        lastOffTraceSrcFound = -1;
+        for (unsigned i = 0; i < PN->getNumIncomingValues (); ++i)
+          if (PN->getIncomingBlock (i)->getParent () != F) {
+            lastOffTraceSrcFound = i; // Found an off-trace source.
+            break;
+          }
+        if (lastOffTraceSrcFound != -1) { // Found one?
+          DEBUG(std::cerr << "Removing off-trace input "
+                << *PN->getIncomingValue (lastOffTraceSrcFound)
+                << " from Phi node " << *PN << "\n");
+          PN->removeIncomingValue (lastOffTraceSrcFound); // Delete it.
+        }
+      } while (lastOffTraceSrcFound != -1); // Continue until none found.
+    }
     // Make sure that our Phi fixups did the Right Thing.
     DEBUG(if (PN->use_size () != 0) // 0-use nodes get deleted later
-	  for (unsigned i = 0; i < PN->getNumIncomingValues (); ++i)
-	  assert (PN->getIncomingBlock (i)->getParent () == F &&
-		  "Sorry, fixupPhis mishandled a Phi node"));
-    }
+          for (unsigned i = 0; i < PN->getNumIncomingValues (); ++i)
+          assert (PN->getIncomingBlock (i)->getParent () == F &&
+                  "Sorry, fixupPhis mishandled a Phi node"));
   }
 
   while (!goners.empty ()) {
     assert (goners.back ()->use_size () == 0
-	    && "Whoops, I was going to delete something which still has uses");
+            && "Whoops, I was going to delete something which still has uses");
     dstB->getInstList ().erase (goners.back ());
     for (ValueMap::iterator i = O2CMap.begin(), e = O2CMap.end(); i != e; ++i) {
-	  if (i->second == goners.back ()) {
+      if (i->second == goners.back ()) {
         O2CMap.erase (i);
         break;
       }
@@ -630,16 +625,8 @@
 	// This is a non-trace-exiting destination of branch BI. Fix up
 	// its clone's destination to point to its successor's clone.
 	Value *V2 = O2CMap[successor];
-	DEBUG(if (!V2) {
-	  std::cerr
-	    << "I'm confused: saw non-trace-exiting dest. of branch "
-	    << *BI << " whose clone is " << *BIinF
-	    << " successor number is " << i << " and successor is "
-	    << *successor << ", but where's its clone?\n";
-	  assert(V2 && "Clone of basic block on trace is null?");
-	});
-	assert(isa<BasicBlock> (V2)
-	       && "Clone of basic block on trace is not a basic block?");
+	assert(V2 && isa<BasicBlock> (V2)
+	       && "Clone of basic block on trace not found, or not a basic block");
 	BIinF->setSuccessor (i, cast<BasicBlock> (V2));
       }
     }




From gaeke at cs.uiuc.edu  Thu May 20 15:06:02 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Thu May 20 15:06:02 2004
Subject: [llvm-commits] CVS: reopt/lib/TraceToFunction/TraceToFunction.cpp
Message-ID: <200405202004.PAA28027@zion.cs.uiuc.edu>


Changes in directory reopt/lib/TraceToFunction:

TraceToFunction.cpp updated: 1.50 -> 1.51

---
Log message:

Clone of a constant is just the identity function, so don't fix up uses of
constants


---
Diffs of the changes:  (+1 -1)

Index: reopt/lib/TraceToFunction/TraceToFunction.cpp
diff -u reopt/lib/TraceToFunction/TraceToFunction.cpp:1.50 reopt/lib/TraceToFunction/TraceToFunction.cpp:1.51
--- reopt/lib/TraceToFunction/TraceToFunction.cpp:1.50	Thu May 20 14:50:44 2004
+++ reopt/lib/TraceToFunction/TraceToFunction.cpp	Thu May 20 15:04:27 2004
@@ -643,7 +643,7 @@
       // function, and I will still reference the version from
       // outside the function. Replace any reference to an operand
       // which has had a clone made with a reference to its clone.
-      if (O2CMap.find (V) != O2CMap.end () && O2CMap[V]) {
+      if (O2CMap.find (V) != O2CMap.end () && O2CMap[V] && !isa<Constant> (V)) {
         DEBUG(std::cerr << *V << " in instruction " << I
               << " is value " << O2CMap[V] << " in new function\n");
         assert (V->getType () == O2CMap[V]->getType ()




From lattner at cs.uiuc.edu  Thu May 20 15:27:02 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu May 20 15:27:02 2004
Subject: [llvm-commits] CVS: llvm/lib/Target/CBackend/Writer.cpp
Message-ID: <200405202026.PAA28495@zion.cs.uiuc.edu>


Changes in directory llvm/lib/Target/CBackend:

Writer.cpp updated: 1.174 -> 1.175

---
Log message:

Fix to make the CBE always emit comparisons inline.  Hopefully this will
fix the really bad code we're getting on PPC.


---
Diffs of the changes:  (+4 -0)

Index: llvm/lib/Target/CBackend/Writer.cpp
diff -u llvm/lib/Target/CBackend/Writer.cpp:1.174 llvm/lib/Target/CBackend/Writer.cpp:1.175
--- llvm/lib/Target/CBackend/Writer.cpp:1.174	Sun May  9 15:41:32 2004
+++ llvm/lib/Target/CBackend/Writer.cpp	Thu May 20 15:25:50 2004
@@ -126,6 +126,10 @@
     // printed and an extra copy of the expr is not emitted.
     //
     static bool isInlinableInst(const Instruction &I) {
+      // Always inline setcc instructions, even if they are shared by multiple
+      // expressions.  GCC generates horrible code if we don't.
+      if (isa<SetCondInst>(I)) return true;
+
       // Must be an expression, must be used exactly once.  If it is dead, we
       // emit it inline where it would go.
       if (I.getType() == Type::VoidTy || !I.hasOneUse() ||




From gaeke at cs.uiuc.edu  Thu May 20 15:39:01 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Thu May 20 15:39:01 2004
Subject: [llvm-commits] CVS: reopt/lib/TraceToFunction/TraceToFunction.cpp
Message-ID: <200405202038.PAA20868@zion.cs.uiuc.edu>


Changes in directory reopt/lib/TraceToFunction:

TraceToFunction.cpp updated: 1.51 -> 1.52

---
Log message:

Make cloneMapDump() into a method that is easier to call from the debugger.
Don't call it by default, because it takes a long time on big modules.


---
Diffs of the changes:  (+12 -11)

Index: reopt/lib/TraceToFunction/TraceToFunction.cpp
diff -u reopt/lib/TraceToFunction/TraceToFunction.cpp:1.51 reopt/lib/TraceToFunction/TraceToFunction.cpp:1.52
--- reopt/lib/TraceToFunction/TraceToFunction.cpp:1.51	Thu May 20 15:04:27 2004
+++ reopt/lib/TraceToFunction/TraceToFunction.cpp	Thu May 20 15:38:04 2004
@@ -66,6 +66,8 @@
   void fixupFunctionBodyBB (Trace &T, Function *F, BasicBlock *srcB,
 				            BasicBlock *dstB, ValueMap &O2CMap,
 				            LiveVariableSet &So);
+
+  void cloneMapDump ();
 public:
   TraceFunctionBuilder () { }
   virtual ~TraceFunctionBuilder () { }
@@ -668,20 +670,21 @@
 }
 
 /// Dump out the O2CMap for the given TraceFunction to stderr.
-/// Only called when debugging.
+/// You can call this from the debugger if you want to see the gory details,
+/// because it takes a really long time on big modules.
 ///
-static void cloneMapDump (TraceFunction *TF, std::ostream &OS) {
-  OS << "\n; TraceFunctionBuilder Original-->Clone map follows:\n";
+void TraceFunctionBuilder::cloneMapDump () {
+  std::cerr << "\n; TraceFunctionBuilder Original-->Clone map follows:\n";
   for (ValueMap::const_iterator i = TF->O2CMap.begin(), e = TF->O2CMap.end();
        i != e; ++i) {
     const std::pair<const Value *, Value *> &elem = *i;
-    OS << "(original-value \"";
-    WriteAsOperand (OS, i->first, true, true, TF->MatrixFn->getParent ());
-    OS << "\" maps-to \"";
-    WriteAsOperand (OS, i->second, true, true, TF->TraceFn->getParent ());
-    OS << "\")\n";
+    std::cerr << "(original-value \"";
+    WriteAsOperand (std::cerr, i->first, true, true, TF->MatrixFn->getParent ());
+    std::cerr << "\" maps-to \"";
+    WriteAsOperand (std::cerr, i->second, true, true, TF->TraceFn->getParent ());
+    std::cerr << "\")\n";
   }
-  OS << "\n";
+  std::cerr << "\n";
 }
 
 TraceFunction *TraceFunctionBuilder::buildTraceFunction (Trace &T) {
@@ -709,8 +712,6 @@
 			      T.getModule ());
   DEBUG(giveNamesToFunctionArgs (TF->LiveInSet, TF->TraceFn));
   fillInFunctionBody (T, TF->TraceFn, TF->LiveOutSet);
-  
-  DEBUG(cloneMapDump(TF, std::cerr));
   return TF;
 }
 




From alkis at cs.uiuc.edu  Thu May 20 16:20:01 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Thu May 20 16:20:01 2004
Subject: [llvm-commits] CVS: llvm-java/lib/ClassFile/ClassFile.cpp
Message-ID: <200405202119.QAA32651@zion.cs.uiuc.edu>


Changes in directory llvm-java/lib/ClassFile:

ClassFile.cpp updated: 1.3 -> 1.4

---
Log message:

Pull commonly used typedefs out of the ClassFile class.


---
Diffs of the changes:  (+25 -41)

Index: llvm-java/lib/ClassFile/ClassFile.cpp
diff -u llvm-java/lib/ClassFile/ClassFile.cpp:1.3 llvm-java/lib/ClassFile/ClassFile.cpp:1.4
--- llvm-java/lib/ClassFile/ClassFile.cpp:1.3	Sat Apr 17 02:52:15 2004
+++ llvm-java/lib/ClassFile/ClassFile.cpp	Thu May 20 16:19:06 2004
@@ -60,7 +60,7 @@
         return tmp.out;
     }
 
-    void readConstantPool(ClassFile::ConstantPool& cp, std::istream& is)
+    void readConstantPool(ConstantPool& cp, std::istream& is)
     {
         assert(cp.empty() && "Should not call with a non-empty constant pool");
         uint16_t count = readU2(is);
@@ -71,9 +71,7 @@
             cp.push_back(Constant::readConstant(cp, is));
     }
 
-    void readInterfaces(ClassFile::Interfaces& i,
-                        const ClassFile::ConstantPool& cp,
-                        std::istream& is)
+    void readInterfaces(Classes& i, const ConstantPool& cp, std::istream& is)
     {
         assert(i.empty() &&
                "Should not call with a non-empty interfaces vector");
@@ -86,9 +84,7 @@
         }
     }
 
-    void readFields(ClassFile::Fields& f,
-                    const ClassFile::ConstantPool& cp,
-                    std::istream& is)
+    void readFields(Fields& f, const ConstantPool& cp, std::istream& is)
     {
         assert(f.empty() && "Should not call with a non-empty fields vector");
         uint16_t count = readU2(is);
@@ -97,9 +93,7 @@
             f.push_back(Field::readField(cp, is));
     }
 
-    void readMethods(ClassFile::Methods& m,
-                     const ClassFile::ConstantPool& cp,
-                     std::istream& is)
+    void readMethods(Methods& m, const ConstantPool& cp, std::istream& is)
     {
         assert(m.empty() && "Should not call with a non-empty methods vector");
         uint16_t count = readU2(is);
@@ -108,8 +102,8 @@
             m.push_back(Method::readMethod(cp, is));
     }
 
-    void readAttributes(ClassFile::Attributes& a,
-                        const ClassFile::ConstantPool& cp,
+    void readAttributes(Attributes& a,
+                        const ConstantPool& cp,
                         std::istream& is)
     {
         assert(a.empty() &&
@@ -212,7 +206,7 @@
 
 //===----------------------------------------------------------------------===//
 // Constant implementation
-Constant* Constant::readConstant(const ClassFile::ConstantPool& cp,
+Constant* Constant::readConstant(const ConstantPool& cp,
                                  std::istream& is)
 {
     Constant::Tag tag = static_cast<Constant::Tag>(readU1(is));
@@ -251,8 +245,7 @@
 
 }
 
-ConstantMemberRef::ConstantMemberRef(const ClassFile::ConstantPool&cp,
-                                     std::istream& is)
+ConstantMemberRef::ConstantMemberRef(const ConstantPool&cp, std::istream& is)
     : Constant(cp),
       classIdx_(readU2(is)),
       nameAndTypeIdx_(readU2(is))
@@ -265,8 +258,7 @@
     return os << *getNameAndType() << '(' << *getClass() << ')';
 }
 
-ConstantClass::ConstantClass(const ClassFile::ConstantPool& cp,
-                             std::istream& is)
+ConstantClass::ConstantClass(const ConstantPool& cp, std::istream& is)
     : Constant(cp),
       nameIdx_(readU2(is))
 {
@@ -278,8 +270,7 @@
     return os << *getName();
 }
 
-ConstantString::ConstantString(const ClassFile::ConstantPool& cp,
-                               std::istream& is)
+ConstantString::ConstantString(const ConstantPool& cp, std::istream& is)
     : Constant(cp),
       stringIdx_(readU2(is))
 {
@@ -291,8 +282,7 @@
     return os << "string " << *getValue();
 }
 
-ConstantInteger::ConstantInteger(const ClassFile::ConstantPool& cp,
-                                 std::istream& is)
+ConstantInteger::ConstantInteger(const ConstantPool& cp, std::istream& is)
     : Constant(cp),
       value_(static_cast<int32_t>(readU4(is)))
 {
@@ -304,8 +294,7 @@
     return os << value_;
 }
 
-ConstantFloat::ConstantFloat(const ClassFile::ConstantPool& cp,
-                             std::istream& is)
+ConstantFloat::ConstantFloat(const ConstantPool& cp, std::istream& is)
     : Constant(cp),
       value_(int2float(readU4(is)))
 {
@@ -317,8 +306,7 @@
     return os << value_;
 }
 
-ConstantLong::ConstantLong(const ClassFile::ConstantPool& cp,
-                           std::istream& is)
+ConstantLong::ConstantLong(const ConstantPool& cp, std::istream& is)
     : Constant(cp),
       value_(static_cast<int64_t>(readU8(is)))
 {
@@ -330,8 +318,7 @@
     return os << value_;
 }
 
-ConstantDouble::ConstantDouble(const ClassFile::ConstantPool& cp,
-                               std::istream& is)
+ConstantDouble::ConstantDouble(const ConstantPool& cp, std::istream& is)
     : Constant(cp),
       value_(long2double(readU8(is)))
 {
@@ -343,7 +330,7 @@
     return os << value_;
 }
 
-ConstantNameAndType::ConstantNameAndType(const ClassFile::ConstantPool& cp,
+ConstantNameAndType::ConstantNameAndType(const ConstantPool& cp,
                                          std::istream& is)
     : Constant(cp),
       nameIdx_(readU2(is)),
@@ -357,8 +344,7 @@
     return os << *getDescriptor() << ' ' << *getName();
 }
 
-ConstantUtf8::ConstantUtf8(const ClassFile::ConstantPool& cp,
-                           std::istream& is)
+ConstantUtf8::ConstantUtf8(const ConstantPool& cp, std::istream& is)
     : Constant(cp)
 {
     uint16_t length = readU2(is);
@@ -377,7 +363,7 @@
 
 //===----------------------------------------------------------------------===//
 // Field implementation
-Field::Field(const ClassFile::ConstantPool& cp, std::istream& is)
+Field::Field(const ConstantPool& cp, std::istream& is)
 {
     accessFlags_ = readU2(is);
     if (!name_)
@@ -414,7 +400,7 @@
 
 //===----------------------------------------------------------------------===//
 // Method implementation
-Method::Method(const ClassFile::ConstantPool& cp, std::istream& is)
+Method::Method(const ConstantPool& cp, std::istream& is)
 {
     accessFlags_ = readU2(is);
     name_ = dynamic_cast<ConstantUtf8*>(cp[readU2(is)]);
@@ -453,8 +439,7 @@
 
 //===----------------------------------------------------------------------===//
 // Attribute implementation
-Attribute* Attribute::readAttribute(const ClassFile::ConstantPool& cp,
-                                    std::istream& is)
+Attribute* Attribute::readAttribute(const ConstantPool& cp, std::istream& is)
 {
     ConstantUtf8* name = dynamic_cast<ConstantUtf8*>(cp[readU2(is)]);
     if (!name)
@@ -473,7 +458,7 @@
 }
 
 Attribute::Attribute(ConstantUtf8* name,
-                     const ClassFile::ConstantPool& cp,
+                     const ConstantPool& cp,
                      std::istream& is)
     : name_(name)
 {
@@ -492,10 +477,9 @@
 
 //===----------------------------------------------------------------------===//
 // AttributeConstantValue implementation
-AttributeConstantValue::AttributeConstantValue(
-    ConstantUtf8* name,
-    const ClassFile::ConstantPool& cp,
-    std::istream& is)
+AttributeConstantValue::AttributeConstantValue(ConstantUtf8* name,
+                                               const ConstantPool& cp,
+                                               std::istream& is)
     : Attribute(name, cp, is)
 {
     uint32_t length = readU4(is);
@@ -513,7 +497,7 @@
 //===----------------------------------------------------------------------===//
 // Attribute code
 AttributeCode::AttributeCode(ConstantUtf8* name,
-                             const ClassFile::ConstantPool& cp,
+                             const ConstantPool& cp,
                              std::istream& is)
     : Attribute(name, cp, is)
 {
@@ -553,7 +537,7 @@
     return os;
 }
 
-AttributeCode::Exception::Exception(const ClassFile::ConstantPool& cp,
+AttributeCode::Exception::Exception(const ConstantPool& cp,
                                     std::istream& is)
 {
     startPc_ = readU2(is);




From alkis at cs.uiuc.edu  Thu May 20 16:20:07 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Thu May 20 16:20:07 2004
Subject: [llvm-commits] CVS: llvm-java/include/llvm/Java/ClassFile.h
Message-ID: <200405202119.QAA32658@zion.cs.uiuc.edu>


Changes in directory llvm-java/include/llvm/Java:

ClassFile.h updated: 1.4 -> 1.5

---
Log message:

Pull commonly used typedefs out of the ClassFile class.


---
Diffs of the changes:  (+38 -48)

Index: llvm-java/include/llvm/Java/ClassFile.h
diff -u llvm-java/include/llvm/Java/ClassFile.h:1.4 llvm-java/include/llvm/Java/ClassFile.h:1.5
--- llvm-java/include/llvm/Java/ClassFile.h:1.4	Sat Apr 17 02:52:16 2004
+++ llvm-java/include/llvm/Java/ClassFile.h	Thu May 20 16:19:06 2004
@@ -46,14 +46,13 @@
         ACC_STRICT       = 0x0800,
     };
 
-    class ClassFile {
-    public:
-        typedef std::vector<Constant*> ConstantPool;
-        typedef std::vector<ConstantClass*> Interfaces;
-        typedef std::vector<Field*> Fields;
-        typedef std::vector<Method*> Methods;
-        typedef std::vector<Attribute*> Attributes;
+    typedef std::vector<Constant*> ConstantPool;
+    typedef std::vector<ConstantClass*> Classes;
+    typedef std::vector<Field*> Fields;
+    typedef std::vector<Method*> Methods;
+    typedef std::vector<Attribute*> Attributes;
 
+    class ClassFile {
     public:
         static ClassFile* readClassFile(std::istream& is);
 
@@ -73,7 +72,7 @@
         const ConstantClass* getThisClass() const { return thisClass_; }
         const ConstantClass* getSuperClass() const { return superClass_; }
 
-        const Interfaces& getInterfaces() const { return interfaces_; }
+        const Classes& getInterfaces() const { return interfaces_; }
 
         const Fields& getFields() const { return fields_; }
 
@@ -90,7 +89,7 @@
         uint16_t accessFlags_;
         ConstantClass* thisClass_;
         ConstantClass* superClass_;
-        Interfaces interfaces_;
+        Classes interfaces_;
         Fields fields_;
         Methods methods_;
         Attributes attributes_;
@@ -104,9 +103,9 @@
 
     class Constant {
     protected:
-        const ClassFile::ConstantPool& cPool_;
+        const ConstantPool& cPool_;
 
-        Constant(const ClassFile::ConstantPool& cp)
+        Constant(const ConstantPool& cp)
             : cPool_(cp) { }
 
     public:
@@ -124,7 +123,7 @@
             UTF8 = 1
         };
 
-        static Constant* readConstant(const ClassFile::ConstantPool& cp,
+        static Constant* readConstant(const ConstantPool& cp,
                                       std::istream& is);
 
         virtual ~Constant();
@@ -139,7 +138,7 @@
     class ConstantClass : public Constant {
         uint16_t nameIdx_;
     public:
-        ConstantClass(const ClassFile::ConstantPool& cp, std::istream& is);
+        ConstantClass(const ConstantPool& cp, std::istream& is);
         const ConstantUtf8* getName() const {
             return (const ConstantUtf8*) cPool_[nameIdx_];
         }
@@ -150,7 +149,7 @@
         uint16_t classIdx_;
         uint16_t nameAndTypeIdx_;
     protected:
-        ConstantMemberRef(const ClassFile::ConstantPool& cp, std::istream& is);
+        ConstantMemberRef(const ConstantPool& cp, std::istream& is);
 
     public:
         const ConstantClass* getClass() const {
@@ -163,25 +162,24 @@
     };
 
     struct ConstantFieldRef : public ConstantMemberRef {
-        ConstantFieldRef(const ClassFile::ConstantPool& cp, std::istream& is)
+        ConstantFieldRef(const ConstantPool& cp, std::istream& is)
             : ConstantMemberRef(cp, is) { }
     };
 
     struct ConstantMethodRef : public ConstantMemberRef {
-        ConstantMethodRef(const ClassFile::ConstantPool& cp, std::istream& is)
+        ConstantMethodRef(const ConstantPool& cp, std::istream& is)
             : ConstantMemberRef(cp, is) { }
     };
 
     struct ConstantInterfaceMethodRef : public ConstantMemberRef {
-        ConstantInterfaceMethodRef(const ClassFile::ConstantPool& cp,
-                                   std::istream& is)
+        ConstantInterfaceMethodRef(const ConstantPool& cp, std::istream& is)
             : ConstantMemberRef(cp, is) { }
     };
 
     class ConstantString : public Constant {
         uint16_t stringIdx_;
     public:
-        ConstantString(const ClassFile::ConstantPool& cp, std::istream& is);
+        ConstantString(const ConstantPool& cp, std::istream& is);
         const ConstantUtf8* getValue() const {
             return  (const ConstantUtf8*) cPool_[stringIdx_];
         }
@@ -191,7 +189,7 @@
     class ConstantInteger : public Constant {
         int32_t value_;
     public:
-        ConstantInteger(const ClassFile::ConstantPool& cp, std::istream& is);
+        ConstantInteger(const ConstantPool& cp, std::istream& is);
         int32_t getValue() const { return value_; }
         std::ostream& dump(std::ostream& os) const;
     };
@@ -199,7 +197,7 @@
     class ConstantFloat : public Constant {
         float value_;
     public:
-        ConstantFloat(const ClassFile::ConstantPool& cp, std::istream& is);
+        ConstantFloat(const ConstantPool& cp, std::istream& is);
         float getValue() const { return value_; }
         std::ostream& dump(std::ostream& os) const;
     };
@@ -207,7 +205,7 @@
     class ConstantLong : public Constant {
         int64_t value_;
     public:
-        ConstantLong(const ClassFile::ConstantPool& cp, std::istream& is);
+        ConstantLong(const ConstantPool& cp, std::istream& is);
         int64_t getValue() const { return value_; }
         std::ostream& dump(std::ostream& os) const;
     };
@@ -215,7 +213,7 @@
     class ConstantDouble : public Constant {
         double value_;
     public:
-        ConstantDouble(const ClassFile::ConstantPool& cp, std::istream& is);
+        ConstantDouble(const ConstantPool& cp, std::istream& is);
         double getValue() const { return value_; }
         std::ostream& dump(std::ostream& os) const;
     };
@@ -224,7 +222,7 @@
         uint16_t nameIdx_;
         uint16_t descriptorIdx_;
     public:
-        ConstantNameAndType(const ClassFile::ConstantPool& cp, std::istream& is);
+        ConstantNameAndType(const ConstantPool& cp, std::istream& is);
         const ConstantUtf8* getName() const {
             return (const ConstantUtf8*) cPool_[nameIdx_];
         }
@@ -237,7 +235,7 @@
     class ConstantUtf8 : public Constant {
         std::string utf8_;
     public:
-        ConstantUtf8(const ClassFile::ConstantPool& cp, std::istream& is);
+        ConstantUtf8(const ConstantPool& cp, std::istream& is);
         operator const char* const() const { return utf8_.c_str(); }
         operator const std::string&() const { return utf8_; }
 
@@ -249,13 +247,12 @@
         uint16_t accessFlags_;
         ConstantUtf8* name_;
         ConstantUtf8* descriptor_;
-        ClassFile::Attributes attributes_;
+        Attributes attributes_;
 
-        Field(const ClassFile::ConstantPool& cp, std::istream& is);
+        Field(const ConstantPool& cp, std::istream& is);
 
     public:
-        static Field* readField(const ClassFile::ConstantPool& cp,
-                                std::istream& is) {
+        static Field* readField(const ConstantPool& cp, std::istream& is) {
             return new Field(cp, is);
         }
 
@@ -271,9 +268,7 @@
 
         ConstantUtf8* getName() const { return name_; }
         ConstantUtf8* getDescriptor() const { return descriptor_; }
-        const ClassFile::Attributes& getAttributes() const {
-            return attributes_;
-        }
+        const Attributes& getAttributes() const { return attributes_; }
 
         std::ostream& dump(std::ostream& os) const;
     };
@@ -286,13 +281,12 @@
         uint16_t accessFlags_;
         ConstantUtf8* name_;
         ConstantUtf8* descriptor_;
-        ClassFile::Attributes attributes_;
+        Attributes attributes_;
 
-        Method(const ClassFile::ConstantPool& cp, std::istream& is);
+        Method(const ConstantPool& cp, std::istream& is);
 
     public:
-        static Method* readMethod(const ClassFile::ConstantPool& cp,
-                                  std::istream& is) {
+        static Method* readMethod(const ConstantPool& cp, std::istream& is) {
             return new Method(cp, is);
         }
 
@@ -310,9 +304,7 @@
 
         ConstantUtf8* getName() const { return name_; }
         ConstantUtf8* getDescriptor() const { return descriptor_; }
-        const ClassFile::Attributes& getAttributes() const {
-            return attributes_;
-        }
+        const Attributes& getAttributes() const { return attributes_; }
 
         std::ostream& dump(std::ostream& os) const;
     };
@@ -326,11 +318,11 @@
 
     protected:
         Attribute(ConstantUtf8* name,
-                  const ClassFile::ConstantPool& cp,
+                  const ConstantPool& cp,
                   std::istream& is);
 
     public:
-        static Attribute* readAttribute(const ClassFile::ConstantPool& cp,
+        static Attribute* readAttribute(const ConstantPool& cp,
                                         std::istream& is);
 
         virtual ~Attribute();
@@ -348,7 +340,7 @@
         Constant* value_;
     public:
         AttributeConstantValue(ConstantUtf8* name,
-                               const ClassFile::ConstantPool& cp,
+                               const ConstantPool& cp,
                                std::istream& is);
 
         Constant* getValue() const { return value_; }
@@ -365,7 +357,7 @@
             ConstantClass* catchType_;
 
         public:
-            Exception(const ClassFile::ConstantPool& cp, std::istream& is);
+            Exception(const ConstantPool& cp, std::istream& is);
 
             uint16_t getStartPc() const { return startPc_; }
             uint16_t getEndPc() const { return endPc_; }
@@ -383,11 +375,11 @@
         uint32_t codeSize_;
         char* code_;
         Exceptions exceptions_;
-        ClassFile::Attributes attributes_;
+        Attributes attributes_;
 
     public:
         AttributeCode(ConstantUtf8* name,
-                      const ClassFile::ConstantPool& cp,
+                      const ConstantPool& cp,
                       std::istream& is);
         ~AttributeCode();
         uint16_t getMaxStack() const { return maxStack_; }
@@ -395,9 +387,7 @@
         const char* getCode() const { return code_; }
         uint32_t getCodeSize() const { return codeSize_; }
         const Exceptions& getExceptions() const { return exceptions_; }
-        const ClassFile::Attributes& getAttributes() const {
-            return attributes_;
-        }
+        const Attributes& getAttributes() const { return attributes_; }
 
         std::ostream& dump(std::ostream& os) const;
 




From alkis at cs.uiuc.edu  Thu May 20 16:33:01 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Thu May 20 16:33:01 2004
Subject: [llvm-commits] CVS: llvm/Makefile.rules
Message-ID: <200405202131.QAA00391@zion.cs.uiuc.edu>


Changes in directory llvm:

Makefile.rules updated: 1.181 -> 1.182

---
Log message:

Change DEPRECATED macro to ATTR_DEPRECATED as this conflicts with some
java constants.


---
Diffs of the changes:  (+1 -1)

Index: llvm/Makefile.rules
diff -u llvm/Makefile.rules:1.181 llvm/Makefile.rules:1.182
--- llvm/Makefile.rules:1.181	Fri Apr 23 19:10:56 2004
+++ llvm/Makefile.rules	Thu May 20 16:31:43 2004
@@ -294,7 +294,7 @@
 CPPFLAGS += -D__STDC_LIMIT_MACROS
 
 ### FIXME: this is GCC specific
-CPPFLAGS += -DDEPRECATED='__attribute__ ((deprecated))'
+CPPFLAGS += -DATTR_DEPRECATED='__attribute__ ((deprecated))'
 
 CompileCommonOpts := -Wall -W -Wwrite-strings -Wno-unused
 CompileOptimizeOpts := -O3 -DNDEBUG -finline-functions




From alkis at cs.uiuc.edu  Thu May 20 17:22:03 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Thu May 20 17:22:03 2004
Subject: [llvm-commits] CVS: llvm-java/lib/ClassFile/ClassFile.cpp
Message-ID: <200405202220.RAA00704@zion.cs.uiuc.edu>


Changes in directory llvm-java/lib/ClassFile:

ClassFile.cpp updated: 1.4 -> 1.5

---
Log message:

Remove implicit conversions for ConstantUtf8.
Add attribute strings for known attributes.


---
Diffs of the changes:  (+12 -2)

Index: llvm-java/lib/ClassFile/ClassFile.cpp
diff -u llvm-java/lib/ClassFile/ClassFile.cpp:1.4 llvm-java/lib/ClassFile/ClassFile.cpp:1.5
--- llvm-java/lib/ClassFile/ClassFile.cpp:1.4	Thu May 20 16:19:06 2004
+++ llvm-java/lib/ClassFile/ClassFile.cpp	Thu May 20 17:20:15 2004
@@ -439,6 +439,16 @@
 
 //===----------------------------------------------------------------------===//
 // Attribute implementation
+const std::string Attribute::CONSTANT_VALUE = "ConstantValue";
+const std::string Attribute::CODE = "Code";
+const std::string Attribute::EXCEPTIONS = "Exceptions";
+const std::string Attribute::INNER_CLASSES = "InnerClasses";
+const std::string Attribute::SYNTHETIC = "Synthetic";
+const std::string Attribute::SOURCE_FILE = "SourceFile";
+const std::string Attribute::LINE_NUMBER_TABLE = "LineNumberTable";
+const std::string Attribute::LOCAL_VARIABLE_TABLE = "LocalVariableTable";
+const std::string Attribute::DEPRECATED = "Deprecated";
+
 Attribute* Attribute::readAttribute(const ConstantPool& cp, std::istream& is)
 {
     ConstantUtf8* name = dynamic_cast<ConstantUtf8*>(cp[readU2(is)]);
@@ -446,9 +456,9 @@
         throw ClassFileSemanticError(
             "Representation of attribute name is not of type ConstantUtf8");
 
-    if (strcmp(*name, "ConstantValue") == 0)
+    if (CONSTANT_VALUE == name->str())
         return new AttributeConstantValue(name, cp, is);
-    else if (strcmp(*name, "Code") == 0)
+    else if (CODE == name->str())
         return new AttributeCode(name, cp, is);
     else {
         uint32_t length = readU4(is);




From alkis at cs.uiuc.edu  Thu May 20 17:22:07 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Thu May 20 17:22:07 2004
Subject: [llvm-commits] CVS: llvm-java/include/llvm/Java/ClassFile.h
Message-ID: <200405202220.RAA00711@zion.cs.uiuc.edu>


Changes in directory llvm-java/include/llvm/Java:

ClassFile.h updated: 1.5 -> 1.6

---
Log message:

Remove implicit conversions for ConstantUtf8.
Add attribute strings for known attributes.


---
Diffs of the changes:  (+11 -2)

Index: llvm-java/include/llvm/Java/ClassFile.h
diff -u llvm-java/include/llvm/Java/ClassFile.h:1.5 llvm-java/include/llvm/Java/ClassFile.h:1.6
--- llvm-java/include/llvm/Java/ClassFile.h:1.5	Thu May 20 16:19:06 2004
+++ llvm-java/include/llvm/Java/ClassFile.h	Thu May 20 17:20:15 2004
@@ -236,8 +236,7 @@
         std::string utf8_;
     public:
         ConstantUtf8(const ConstantPool& cp, std::istream& is);
-        operator const char* const() const { return utf8_.c_str(); }
-        operator const std::string&() const { return utf8_; }
+        const std::string& str() const { return utf8_; }
 
         std::ostream& dump(std::ostream& os) const;
     };
@@ -330,6 +329,16 @@
         ConstantUtf8* getName() const { return name_; }
 
         virtual std::ostream& dump(std::ostream& os) const;
+
+        static const std::string CONSTANT_VALUE;
+        static const std::string CODE;
+        static const std::string EXCEPTIONS;
+        static const std::string INNER_CLASSES;
+        static const std::string SYNTHETIC;
+        static const std::string SOURCE_FILE;
+        static const std::string LINE_NUMBER_TABLE;
+        static const std::string LOCAL_VARIABLE_TABLE;
+        static const std::string DEPRECATED;
     };
 
     inline std::ostream& operator<<(std::ostream& os, const Attribute& a) {




From alkis at cs.uiuc.edu  Thu May 20 17:44:03 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Thu May 20 17:44:03 2004
Subject: [llvm-commits] CVS: llvm-java/lib/ClassFile/ClassFile.cpp
Message-ID: <200405202242.RAA00857@zion.cs.uiuc.edu>


Changes in directory llvm-java/lib/ClassFile:

ClassFile.cpp updated: 1.5 -> 1.6

---
Log message:

Implement Exceptions attribute.


---
Diffs of the changes:  (+25 -5)

Index: llvm-java/lib/ClassFile/ClassFile.cpp
diff -u llvm-java/lib/ClassFile/ClassFile.cpp:1.5 llvm-java/lib/ClassFile/ClassFile.cpp:1.6
--- llvm-java/lib/ClassFile/ClassFile.cpp:1.5	Thu May 20 17:20:15 2004
+++ llvm-java/lib/ClassFile/ClassFile.cpp	Thu May 20 17:42:15 2004
@@ -71,15 +71,16 @@
             cp.push_back(Constant::readConstant(cp, is));
     }
 
-    void readInterfaces(Classes& i, const ConstantPool& cp, std::istream& is)
+    void readClasses(Classes& i, const ConstantPool& cp, std::istream& is)
     {
         assert(i.empty() &&
-               "Should not call with a non-empty interfaces vector");
+               "Should not call with a non-empty classes vector");
         uint16_t count = readU2(is);
         i.reserve(count);
         while (count--) {
             ConstantClass* c = dynamic_cast<ConstantClass*>(cp[readU2(is)]);
-            if (!c) throw "FIXME: give better error message";
+            if (!c)
+                throw ClassFileSemanticError("ConstantClass expected");
             i.push_back(c);
         }
     }
@@ -156,7 +157,7 @@
     if (!superClass_)
         throw ClassFileSemanticError(
             "Representation of super class is not of type ConstantClass");
-    readInterfaces(interfaces_, cPool_, is);
+    readClasses(interfaces_, cPool_, is);
     readFields(fields_, cPool_, is);
     readMethods(methods_, cPool_, is);
     readAttributes(attributes_, cPool_, is);
@@ -505,7 +506,7 @@
 }
 
 //===----------------------------------------------------------------------===//
-// Attribute code
+// AttributeCode implementation
 AttributeCode::AttributeCode(ConstantUtf8* name,
                              const ConstantPool& cp,
                              std::istream& is)
@@ -565,4 +566,23 @@
               << "Start PC: " << startPc_ << '\n'
               << "End PC: " << endPc_ << '\n'
               << "Handler PC: " << handlerPc_;
+}
+
+//===----------------------------------------------------------------------===//
+// AttributeExceptions implementation
+AttributeExceptions::AttributeExceptions(ConstantUtf8* name,
+                                         const ConstantPool& cp,
+                                         std::istream& is)
+    : Attribute(name, cp, is)
+{
+    uint32_t length = readU4(is);
+    readClasses(exceptions_, cp, is);
+}
+
+std::ostream& AttributeExceptions::dump(std::ostream& is) const
+{
+    for (Classes::const_iterator
+             i = exceptions_.begin(), e = exceptions_.end(); i != e; ++i)
+        is << *i << ' ';
+    return is;
 }




From alkis at cs.uiuc.edu  Thu May 20 17:44:07 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Thu May 20 17:44:07 2004
Subject: [llvm-commits] CVS: llvm-java/include/llvm/Java/ClassFile.h
Message-ID: <200405202242.RAA00865@zion.cs.uiuc.edu>


Changes in directory llvm-java/include/llvm/Java:

ClassFile.h updated: 1.6 -> 1.7

---
Log message:

Implement Exceptions attribute.


---
Diffs of the changes:  (+15 -1)

Index: llvm-java/include/llvm/Java/ClassFile.h
diff -u llvm-java/include/llvm/Java/ClassFile.h:1.6 llvm-java/include/llvm/Java/ClassFile.h:1.7
--- llvm-java/include/llvm/Java/ClassFile.h:1.6	Thu May 20 17:20:15 2004
+++ llvm-java/include/llvm/Java/ClassFile.h	Thu May 20 17:42:15 2004
@@ -399,13 +399,27 @@
         const Attributes& getAttributes() const { return attributes_; }
 
         std::ostream& dump(std::ostream& os) const;
-
     };
 
     inline std::ostream& operator<<(std::ostream& os,
                                     const AttributeCode::Exception& e) {
         return e.dump(os);
     }
+
+    class AttributeExceptions : public Attribute {
+    private:
+        ConstantUtf8* name_;
+        Classes exceptions_;
+
+    public:
+        AttributeExceptions(ConstantUtf8* name,
+                            const ConstantPool& cp,
+                            std::istream& is);
+
+        const Classes& getExceptions() const { return exceptions_; }
+
+        std::ostream& dump(std::ostream& os) const;
+    };
 
     class ClassFileParseError : public std::exception {
         std::string msg_;




From alkis at cs.uiuc.edu  Thu May 20 18:52:02 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Thu May 20 18:52:02 2004
Subject: [llvm-commits] CVS: llvm-java/include/llvm/Java/ClassFile.h
Message-ID: <200405202350.SAA17983@zion.cs.uiuc.edu>


Changes in directory llvm-java/include/llvm/Java:

ClassFile.h updated: 1.7 -> 1.8

---
Log message:

Rename subclasses of Attribute from AttributeSomething to
SomethingAttribute.


---
Diffs of the changes:  (+8 -8)

Index: llvm-java/include/llvm/Java/ClassFile.h
diff -u llvm-java/include/llvm/Java/ClassFile.h:1.7 llvm-java/include/llvm/Java/ClassFile.h:1.8
--- llvm-java/include/llvm/Java/ClassFile.h:1.7	Thu May 20 17:42:15 2004
+++ llvm-java/include/llvm/Java/ClassFile.h	Thu May 20 18:50:30 2004
@@ -345,10 +345,10 @@
         return a.dump(os);
     }
 
-    class AttributeConstantValue : public Attribute {
+    class ConstantValueAttribute : public Attribute {
         Constant* value_;
     public:
-        AttributeConstantValue(ConstantUtf8* name,
+        ConstantValueAttribute(ConstantUtf8* name,
                                const ConstantPool& cp,
                                std::istream& is);
 
@@ -357,7 +357,7 @@
         std::ostream& dump(std::ostream& os) const;
     };
 
-    class AttributeCode : public Attribute {
+    class CodeAttribute : public Attribute {
     public:
         class Exception {
             uint16_t startPc_;
@@ -387,10 +387,10 @@
         Attributes attributes_;
 
     public:
-        AttributeCode(ConstantUtf8* name,
+        CodeAttribute(ConstantUtf8* name,
                       const ConstantPool& cp,
                       std::istream& is);
-        ~AttributeCode();
+        ~CodeAttribute();
         uint16_t getMaxStack() const { return maxStack_; }
         uint16_t getMaxLocals() const { return maxLocals_; }
         const char* getCode() const { return code_; }
@@ -402,17 +402,17 @@
     };
 
     inline std::ostream& operator<<(std::ostream& os,
-                                    const AttributeCode::Exception& e) {
+                                    const CodeAttribute::Exception& e) {
         return e.dump(os);
     }
 
-    class AttributeExceptions : public Attribute {
+    class ExceptionsAttribute : public Attribute {
     private:
         ConstantUtf8* name_;
         Classes exceptions_;
 
     public:
-        AttributeExceptions(ConstantUtf8* name,
+        ExceptionsAttribute(ConstantUtf8* name,
                             const ConstantPool& cp,
                             std::istream& is);
 




From alkis at cs.uiuc.edu  Thu May 20 18:52:06 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Thu May 20 18:52:06 2004
Subject: [llvm-commits] CVS: llvm-java/lib/ClassFile/ClassFile.cpp
Message-ID: <200405202350.SAA17978@zion.cs.uiuc.edu>


Changes in directory llvm-java/lib/ClassFile:

ClassFile.cpp updated: 1.6 -> 1.7

---
Log message:

Rename subclasses of Attribute from AttributeSomething to
SomethingAttribute.


---
Diffs of the changes:  (+15 -14)

Index: llvm-java/lib/ClassFile/ClassFile.cpp
diff -u llvm-java/lib/ClassFile/ClassFile.cpp:1.6 llvm-java/lib/ClassFile/ClassFile.cpp:1.7
--- llvm-java/lib/ClassFile/ClassFile.cpp:1.6	Thu May 20 17:42:15 2004
+++ llvm-java/lib/ClassFile/ClassFile.cpp	Thu May 20 18:50:30 2004
@@ -458,9 +458,9 @@
             "Representation of attribute name is not of type ConstantUtf8");
 
     if (CONSTANT_VALUE == name->str())
-        return new AttributeConstantValue(name, cp, is);
+        return new ConstantValueAttribute(name, cp, is);
     else if (CODE == name->str())
-        return new AttributeCode(name, cp, is);
+        return new CodeAttribute(name, cp, is);
     else {
         uint32_t length = readU4(is);
         is.ignore(length);
@@ -488,7 +488,7 @@
 
 //===----------------------------------------------------------------------===//
 // AttributeConstantValue implementation
-AttributeConstantValue::AttributeConstantValue(ConstantUtf8* name,
+ConstantValueAttribute::ConstantValueAttribute(ConstantUtf8* name,
                                                const ConstantPool& cp,
                                                std::istream& is)
     : Attribute(name, cp, is)
@@ -496,18 +496,18 @@
     uint32_t length = readU4(is);
     if (length != 2)
         throw ClassFileSemanticError(
-            "Length of AttributeConstantValue is not 2");
+            "Length of ConstantValueAttribute is not 2");
     value_ = cp[readU2(is)];
 }
 
-std::ostream& AttributeConstantValue::dump(std::ostream& os) const
+std::ostream& ConstantValueAttribute::dump(std::ostream& os) const
 {
     return Attribute::dump(os) << ": " << *value_;
 }
 
 //===----------------------------------------------------------------------===//
 // AttributeCode implementation
-AttributeCode::AttributeCode(ConstantUtf8* name,
+CodeAttribute::CodeAttribute(ConstantUtf8* name,
                              const ConstantPool& cp,
                              std::istream& is)
     : Attribute(name, cp, is)
@@ -528,14 +528,14 @@
     readAttributes(attributes_, cp, is);
 }
 
-AttributeCode::~AttributeCode()
+CodeAttribute::~CodeAttribute()
 {
     delete[] code_;
     for_each(exceptions_.begin(), exceptions_.end(), deleter<Exception>);
     for_each(attributes_.begin(), attributes_.end(), deleter<Attribute>);
 }
 
-std::ostream& AttributeCode::dump(std::ostream& os) const
+std::ostream& CodeAttribute::dump(std::ostream& os) const
 {
     Attribute::dump(os)
         << '\n'
@@ -548,7 +548,7 @@
     return os;
 }
 
-AttributeCode::Exception::Exception(const ConstantPool& cp,
+CodeAttribute::Exception::Exception(const ConstantPool& cp,
                                     std::istream& is)
 {
     startPc_ = readU2(is);
@@ -560,7 +560,7 @@
             "Representation of catch type is not of type ConstantClass");
 }
 
-std::ostream& AttributeCode::Exception::dump(std::ostream& os) const
+std::ostream& CodeAttribute::Exception::dump(std::ostream& os) const
 {
     return os << *getCatchType() << '\n'
               << "Start PC: " << startPc_ << '\n'
@@ -570,7 +570,7 @@
 
 //===----------------------------------------------------------------------===//
 // AttributeExceptions implementation
-AttributeExceptions::AttributeExceptions(ConstantUtf8* name,
+ExceptionsAttribute::ExceptionsAttribute(ConstantUtf8* name,
                                          const ConstantPool& cp,
                                          std::istream& is)
     : Attribute(name, cp, is)
@@ -579,10 +579,11 @@
     readClasses(exceptions_, cp, is);
 }
 
-std::ostream& AttributeExceptions::dump(std::ostream& is) const
+std::ostream& ExceptionsAttribute::dump(std::ostream& os) const
 {
+    os << Attribute::dump(os) << ": ";
     for (Classes::const_iterator
              i = exceptions_.begin(), e = exceptions_.end(); i != e; ++i)
-        is << *i << ' ';
-    return is;
+        os << *i << ' ';
+    return os;
 }




From alkis at cs.uiuc.edu  Thu May 20 19:07:03 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Thu May 20 19:07:03 2004
Subject: [llvm-commits] CVS: llvm-java/include/llvm/Java/ClassFile.h
Message-ID: <200405210006.TAA25554@zion.cs.uiuc.edu>


Changes in directory llvm-java/include/llvm/Java:

ClassFile.h updated: 1.8 -> 1.9

---
Log message:

Add utility functions to easily access attributes.


---
Diffs of the changes:  (+5 -0)

Index: llvm-java/include/llvm/Java/ClassFile.h
diff -u llvm-java/include/llvm/Java/ClassFile.h:1.8 llvm-java/include/llvm/Java/ClassFile.h:1.9
--- llvm-java/include/llvm/Java/ClassFile.h:1.8	Thu May 20 18:50:30 2004
+++ llvm-java/include/llvm/Java/ClassFile.h	Thu May 20 19:05:59 2004
@@ -22,6 +22,7 @@
 
     // Forward declarations
     class Attribute;
+    class CodeAttribute;
     class Constant;
     class ConstantClass;
     class ConstantNameAndType;
@@ -51,6 +52,10 @@
     typedef std::vector<Field*> Fields;
     typedef std::vector<Method*> Methods;
     typedef std::vector<Attribute*> Attributes;
+
+    const Attribute* getAttribute(const Attributes& attrs,
+                                  const std::string& name);
+    const CodeAttribute* getCodeAttribute(const Attributes& attrs);
 
     class ClassFile {
     public:




From alkis at cs.uiuc.edu  Thu May 20 19:07:08 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Thu May 20 19:07:08 2004
Subject: [llvm-commits] CVS: llvm-java/lib/ClassFile/ClassFile.cpp
Message-ID: <200405210006.TAA25549@zion.cs.uiuc.edu>


Changes in directory llvm-java/lib/ClassFile:

ClassFile.cpp updated: 1.7 -> 1.8

---
Log message:

Add utility functions to easily access attributes.


---
Diffs of the changes:  (+22 -1)

Index: llvm-java/lib/ClassFile/ClassFile.cpp
diff -u llvm-java/lib/ClassFile/ClassFile.cpp:1.7 llvm-java/lib/ClassFile/ClassFile.cpp:1.8
--- llvm-java/lib/ClassFile/ClassFile.cpp:1.7	Thu May 20 18:50:30 2004
+++ llvm-java/lib/ClassFile/ClassFile.cpp	Thu May 20 19:05:59 2004
@@ -23,7 +23,7 @@
 using namespace llvm::Java;
 
 //===----------------------------------------------------------------------===//
-// Utility functions
+// Internal utility functions
 namespace {
 
     uint8_t readU1(std::istream& is) {
@@ -189,6 +189,27 @@
     dumpCollection(attributes_, "Attribute", os);
 
     return os;
+}
+
+//===----------------------------------------------------------------------===//
+// Utility functions
+const Attribute* llvm::Java::getAttribute(const Attributes& attrs,
+                                          const std::string& name)
+{
+    for (Attributes::const_iterator
+             i = attrs.begin(), e = attrs.end(); i != e; ++i) {
+        const Attribute* attr = *i;
+        if (attr->getName()->str() == name)
+            return attr;
+    }
+
+    return NULL;
+}
+
+const CodeAttribute* llvm::Java::getCodeAttribute(const Attributes& attrs)
+{
+    return static_cast<const CodeAttribute*>(
+        getAttribute(attrs, Attribute::CODE));
 }
 
 //===----------------------------------------------------------------------===//




From brukman at cs.uiuc.edu  Thu May 20 19:11:01 2004
From: brukman at cs.uiuc.edu (Misha Brukman)
Date: Thu May 20 19:11:01 2004
Subject: [llvm-commits] CVS: llvm/Makefile.rules
Message-ID: <200405210009.TAA25630@zion.cs.uiuc.edu>


Changes in directory llvm:

Makefile.rules updated: 1.182 -> 1.183

---
Log message:

Pass given flags on to recursive sub-makes.


---
Diffs of the changes:  (+1 -1)

Index: llvm/Makefile.rules
diff -u llvm/Makefile.rules:1.182 llvm/Makefile.rules:1.183
--- llvm/Makefile.rules:1.182	Thu May 20 16:31:43 2004
+++ llvm/Makefile.rules	Thu May 20 19:09:21 2004
@@ -402,7 +402,7 @@
 			$(MKDIR) $$dir; \
 			cp $(SourceDir)/$$dir/Makefile $$dir/Makefile; \
 		fi; \
-		($(MAKE) -C $$dir $@) || exit 1; \
+		($(MAKE) -C $$dir $@ $(MFLAGS)) || exit 1; \
 	done
 endif
 




From gaeke at cs.uiuc.edu  Fri May 21 02:46:01 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Fri May 21 02:46:01 2004
Subject: [llvm-commits] CVS: reopt/lib/TraceToFunction/TraceToFunction.cpp
Message-ID: <200405210745.CAA14218@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/TraceToFunction:

TraceToFunction.cpp updated: 1.52 -> 1.53

---
Log message:

Simplify isFirstTriggerCall(); eliminate debug message.


---
Diffs of the changes:  (+2 -6)

Index: reopt/lib/TraceToFunction/TraceToFunction.cpp
diff -u reopt/lib/TraceToFunction/TraceToFunction.cpp:1.52 reopt/lib/TraceToFunction/TraceToFunction.cpp:1.53
--- reopt/lib/TraceToFunction/TraceToFunction.cpp:1.52	Thu May 20 15:38:04 2004
+++ reopt/lib/TraceToFunction/TraceToFunction.cpp	Fri May 21 02:45:05 2004
@@ -308,12 +308,8 @@
   if (CallInst *CI = dyn_cast<CallInst> (&I)) {
     Function *CF = CI->getCalledFunction ();
     // Look for an external function call with no args to "llvm_first_trigger".
-    if (CF && CF->isExternal () && CF->hasName()
-        && CF->getName () == "llvm_first_trigger" && CF->asize () == 0) {
-      DEBUG(std::cerr << " (Found a call instruction " << *CI
-            << " ... Smells like llvm_first_trigger.)\n");
-      return true;
-    }
+    return (CF && CF->isExternal () && CF->hasName()
+            && CF->getName () == "llvm_first_trigger" && CF->asize () == 0);
   }
   return false;
 }




From gaeke at cs.uiuc.edu  Fri May 21 02:59:01 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Fri May 21 02:59:01 2004
Subject: [llvm-commits] CVS: reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp
Message-ID: <200405210758.CAA14281@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/LightWtProfiling:

UnpackTraceFunction.cpp updated: 1.63 -> 1.64

---
Log message:

Don't worry about saving or restoring %g0.
Don't print out a line of spam for every piece of register allocator
state we scan through.


---
Diffs of the changes:  (+2 -3)

Index: reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp
diff -u reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp:1.63 reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp:1.64
--- reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp:1.63	Thu May 20 13:05:22 2004
+++ reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp	Fri May 21 02:58:43 2004
@@ -212,8 +212,8 @@
           } else if (TRI.getRegType(regNo) == SparcV9RegInfo::FloatCCRegType
                   || TRI.getRegType(regNo) == SparcV9RegInfo::SpecialRegType) {
             floatCCRegSeen = true;
-          } else {
-            regsUsed.insert (MO.getReg ());
+          } else if (regNo != SparcV9::g0) { // Ignore defs of the zero register
+            regsUsed.insert (regNo);
           }
         }
       }
@@ -456,7 +456,6 @@
                   << " tuples to search.\n");
   for (unsigned i = 0, s = FState.size (); i < s; ++i) {
     AllocInfo &T = FState[i];
-    DEBUG(std::cerr << "Considering: " << T << "\n");
     if (T.Instruction == InstructionKey && T.Operand == OperandKey)
       return T;
   }




From gaeke at cs.uiuc.edu  Fri May 21 05:06:01 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Fri May 21 05:06:01 2004
Subject: [llvm-commits] CVS: reopt/lib/TraceToFunction/TraceToFunction.cpp
Message-ID: <200405211005.FAA14956@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/TraceToFunction:

TraceToFunction.cpp updated: 1.53 -> 1.54

---
Log message:

Give fixupFunctionBodyBB much better debugging printouts (and make them
less incredibly verbose).  Untabify.


---
Diffs of the changes:  (+59 -50)

Index: reopt/lib/TraceToFunction/TraceToFunction.cpp
diff -u reopt/lib/TraceToFunction/TraceToFunction.cpp:1.53 reopt/lib/TraceToFunction/TraceToFunction.cpp:1.54
--- reopt/lib/TraceToFunction/TraceToFunction.cpp:1.53	Fri May 21 02:45:05 2004
+++ reopt/lib/TraceToFunction/TraceToFunction.cpp	Fri May 21 05:05:48 2004
@@ -455,8 +455,7 @@
           DEBUG (std::cerr << "threadFLIEdges: threading incoming block " << i
             << " (FLI block " << It->first->getName () << " --> <[" 
             << It->second.first->getName () << "], "
-            << It->second.second->getName () << ">) of PHI node: "
-            << *PN << "\n");
+            << It->second.second->getName () << ">) of PHI node:\n" << *PN);
           PN->setIncomingBlock (i, It->second.first);
         }
       }
@@ -475,8 +474,7 @@
         DEBUG (std::cerr << "threadFLIEdges: threading successor " << i
           << " (FLI block " << It->first->getName () << " --> <" 
           << It->second.first->getName () << ", ["
-          << It->second.second->getName () << "]>) of branch: "
-          << *BI << "\n");
+          << It->second.second->getName () << "]>) of branch:\n" << *BI);
         BI->setSuccessor (i, It->second.second);
       }
     }
@@ -572,61 +570,67 @@
     // Fix up this exit branch.
     Value *V = O2CMap[BI];
     assert(V && isa<BranchInst> (V)
-	   && "Trace-exiting branch's clone is null, or not a branch?");
+           && "Trace-exiting branch's clone is null, or not a branch?");
     BranchInst *BIinF = cast<BranchInst> (V);
     for (unsigned i = 0; i < BI->getNumSuccessors (); ++i) {
       BasicBlock *successor = BI->getSuccessor (i);
+      BasicBlock *newSuccessor = 0;
       // Fold branches which go to FLI blocks
       if (BasicBlock *backEdgeTarget =
           getFLIEdgeTarget (successor)) {
-        DEBUG (std::cerr << "Branch instr " << *BI << " looks like an"
-          << " instrumented back-edge from " << srcB->getName()
+        DEBUG (std::cerr << "fixupFunctionBodyBB: Branch instruction:\n" << *BI
+          << " looks like an instrumented back-edge from " << srcB->getName()
           << " to " << backEdgeTarget->getName () << "\n");
         successor = backEdgeTarget;
       }
       if (!T.contains (successor)) {
-	// This is a trace-exiting destination of branch BI. Create a new
-	// basic block FB for its destination in the function.
-	std::string name ("exitfixup" + utostr (BranchNumber[BI]) + "_"
-			  + utostr (i));
-	BasicBlock *FB = new BasicBlock (name, F);
-	// Change BI's clone's destination to FB.
-	BIinF->setSuccessor (i, FB);
-    // Remember that FB's "return" instruction would involve a
-    // return to the off-trace successor we just found.
-    TF->ReturnBlockForTraceExit[FB] = successor;
-	// Add the getelementptr/store instruction pairs here that
-	// correspond to each live-out variable.
-	unsigned Slot = 0;
-	for (LiveVariableSet::iterator SI = So.begin (), SE = So.end ();
-	     SI != SE; ++SI) {
-      if (dominates (T, cast<Instruction> ((*SI))->getParent (),
-                     BI->getParent ())) {
-	    std::vector<Value *> Index;
-	    Index.push_back (Constant::getNullValue (Type::LongTy));  //long 0
-	    Index.push_back (ConstantUInt::get (Type::UIntTy, Slot));//uint Slot
-	    GetElementPtrInst *GEP =
-	      new GetElementPtrInst (getFunctionArg (F, 0), Index,
-				     "liveOutP" + utostr (BranchNumber[BI])
-				     + "_" + utostr (i) + "_" + utostr (Slot));
-	    FB->getInstList ().push_back (GEP);
-	    FB->getInstList ().push_back (new StoreInst (O2CMap[*SI], GEP));
-      }
-	  ++Slot;
-	}
-	// Make FB contain a return instruction that returns the
-	// number of the taken exit-branch. Add it to the end of FB:
-	FB->getInstList ().push_back
-	  (new ReturnInst (ConstantUInt::get (Type::UIntTy,
-					      BranchNumber[BI])));
+        // This is a trace-exiting destination of branch BI. Create a new
+        // basic block FB for its destination in the function.
+        std::string name ("exitfixup" + utostr (BranchNumber[BI]) + "_"
+                          + utostr (i));
+        BasicBlock *FB = new BasicBlock (name, F);
+        // Remember to change BI's clone's destination to FB.
+        newSuccessor = FB;
+        // Remember that FB's "return" instruction would involve a
+        // return to the off-trace successor we just found.
+        TF->ReturnBlockForTraceExit[FB] = successor;
+        // Add the getelementptr/store instruction pairs here that
+        // correspond to each live-out variable.
+        unsigned Slot = 0;
+        for (LiveVariableSet::iterator SI = So.begin (), SE = So.end ();
+             SI != SE; ++SI) {
+          if (dominates (T, cast<Instruction> ((*SI))->getParent (),
+                         BI->getParent ())) {
+            std::vector<Value *> Index;
+            Index.push_back (Constant::getNullValue (Type::LongTy));  //long 0
+            Index.push_back (ConstantUInt::get (Type::UIntTy, Slot));//uint Slot
+            GetElementPtrInst *GEP =
+              new GetElementPtrInst (getFunctionArg (F, 0), Index,
+                                     "liveOutP" + utostr (BranchNumber[BI])
+                                     + "_" + utostr (i) + "_" + utostr (Slot));
+            FB->getInstList ().push_back (GEP);
+            FB->getInstList ().push_back (new StoreInst (O2CMap[*SI], GEP));
+          }
+          ++Slot;
+        }
+        // Make FB contain a return instruction that returns the
+        // number of the taken exit-branch. Add it to the end of FB:
+        FB->getInstList ().push_back
+          (new ReturnInst (ConstantUInt::get (Type::UIntTy,
+                                              BranchNumber[BI])));
       } else {
-	// This is a non-trace-exiting destination of branch BI. Fix up
-	// its clone's destination to point to its successor's clone.
-	Value *V2 = O2CMap[successor];
-	assert(V2 && isa<BasicBlock> (V2)
-	       && "Clone of basic block on trace not found, or not a basic block");
-	BIinF->setSuccessor (i, cast<BasicBlock> (V2));
-      }
+        // This is a non-trace-exiting destination of branch BI. Fix up
+        // its clone's destination to point to its successor's clone.
+        Value *V2 = O2CMap[successor];
+        assert(V2 && isa<BasicBlock> (V2)
+               && "Clone of BasicBlock on trace missing, or not a BasicBlock");
+        newSuccessor = cast<BasicBlock> (V2);
+      }
+      DEBUG(std::cerr << "fixupFunctionBodyBB: Changing successor " << i
+        << " of trace-exiting branch instruction:\n" << *BIinF << " from "
+        << BIinF->getSuccessor(i)->getName () <<  " to "
+        << newSuccessor->getName () << "\n");
+        BIinF->setSuccessor (i, newSuccessor);
     }
   }
 
@@ -642,8 +646,13 @@
       // outside the function. Replace any reference to an operand
       // which has had a clone made with a reference to its clone.
       if (O2CMap.find (V) != O2CMap.end () && O2CMap[V] && !isa<Constant> (V)) {
-        DEBUG(std::cerr << *V << " in instruction " << I
-              << " is value " << O2CMap[V] << " in new function\n");
+        DEBUG(std::cerr << "fixupFunctionBodyBB: Replacing ";
+              WriteAsOperand (std::cerr, V, true, true,
+                              TF->TraceFn->getParent ());
+              std::cerr << " in instruction:\n" << I << " with value ";
+              WriteAsOperand (std::cerr, O2CMap[V], true, true,
+                              TF->TraceFn->getParent ());
+              std::cerr << "\n");
         assert (V->getType () == O2CMap[V]->getType ()
                 && "Value's type doesn't match clone's type");
         I.setOperand(i, O2CMap[V]);




From gaeke at cs.uiuc.edu  Fri May 21 05:29:02 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Fri May 21 05:29:02 2004
Subject: [llvm-commits] CVS: reopt/lib/TraceToFunction/TraceToFunction.cpp
Message-ID: <200405211028.FAA15437@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/TraceToFunction:

TraceToFunction.cpp updated: 1.54 -> 1.55

---
Log message:

Minor cleanups in fillInFunctionBody():  Make 'function before fixups'
debug message stand out more.  Eliminate a dyn_cast.

In fixupFunctionBodyBB(), reduce the number of times we look each
operand up in O2CMap from 5 to 1, and don't look constants or global
values up at all.


---
Diffs of the changes:  (+12 -10)

Index: reopt/lib/TraceToFunction/TraceToFunction.cpp
diff -u reopt/lib/TraceToFunction/TraceToFunction.cpp:1.54 reopt/lib/TraceToFunction/TraceToFunction.cpp:1.55
--- reopt/lib/TraceToFunction/TraceToFunction.cpp:1.54	Fri May 21 05:05:48 2004
+++ reopt/lib/TraceToFunction/TraceToFunction.cpp	Fri May 21 05:28:42 2004
@@ -388,8 +388,6 @@
 ///
 void TraceFunctionBuilder::fillInFunctionBody (Trace &T, Function *F,
 					  LiveVariableSet &So) {
-  ValueMap &O2CMap = TF->O2CMap;
-
   buildFLIMap (T, FLIMap);
 
   // First copy the basic blocks from the trace.
@@ -399,15 +397,16 @@
 
   numberExitBranchesInTrace (T, BranchNumber);
 
-  DEBUG (std::cerr << "\n FUNCTION BEFORE FIXUPS: \n" << *F << "\n\n");
+  DEBUG (std::cerr << "\n ---- FUNCTION BEFORE FIXUPS: --- \n" << *F << "\n\n");
 
   // Fix up the cloned basic blocks of the function so that they are
   // internally consistent.
+  ValueMap &O2CMap = TF->O2CMap;
   for (Trace::iterator TI = T.begin (), TE = T.end (); TI != TE; ++TI) {
     BasicBlock *srcB = *TI;
-    assert (O2CMap[srcB] && "Can't find clone of basic block I just cloned");
-    BasicBlock *dstB = dyn_cast<BasicBlock> (O2CMap[srcB]);
-    assert (dstB && "Clone of basic block I just cloned is not a basic block");
+    assert (O2CMap[srcB] && isa<BasicBlock> (O2CMap[srcB])
+            && "Clone of trace BasicBlock is missing or not a BasicBlock");
+    BasicBlock *dstB = cast<BasicBlock> (O2CMap[srcB]);
     fixupFunctionBodyBB (T, F, srcB, dstB, O2CMap, So);
   }
   for (Function::iterator FI = F->begin (), FE = F->end (); FI != FE; ++FI)
@@ -645,17 +644,20 @@
       // function, and I will still reference the version from
       // outside the function. Replace any reference to an operand
       // which has had a clone made with a reference to its clone.
-      if (O2CMap.find (V) != O2CMap.end () && O2CMap[V] && !isa<Constant> (V)) {
+      if (isa<Constant> (V) || isa<GlobalValue> (V)) continue;
+      ValueMap::iterator It = O2CMap.find (V);
+      if (It != O2CMap.end ()) {
+        Value *newV = It->second;
         DEBUG(std::cerr << "fixupFunctionBodyBB: Replacing ";
               WriteAsOperand (std::cerr, V, true, true,
                               TF->TraceFn->getParent ());
               std::cerr << " in instruction:\n" << I << " with value ";
-              WriteAsOperand (std::cerr, O2CMap[V], true, true,
+              WriteAsOperand (std::cerr, newV, true, true,
                               TF->TraceFn->getParent ());
               std::cerr << "\n");
-        assert (V->getType () == O2CMap[V]->getType ()
+        assert (V->getType () == newV->getType ()
                 && "Value's type doesn't match clone's type");
-        I.setOperand(i, O2CMap[V]);
+        I.setOperand(i, newV);
       }
     }
     




From lattner at cs.uiuc.edu  Fri May 21 11:48:01 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri May 21 11:48:01 2004
Subject: [llvm-commits] CVS: llvm/test/Regression/Verifier/2004-05-21-SwitchConstantMismatch.ll
Message-ID: <200405211647.LAA10878@zion.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Verifier:

2004-05-21-SwitchConstantMismatch.ll added (r1.1)

---
Log message:

Testcase for PR340: http://llvm.cs.uiuc.edu/PR340 : Verifier misses malformed switch instruction

This testcase was reduced by John, thanks!



---
Diffs of the changes:  (+11 -0)

Index: llvm/test/Regression/Verifier/2004-05-21-SwitchConstantMismatch.ll
diff -c /dev/null llvm/test/Regression/Verifier/2004-05-21-SwitchConstantMismatch.ll:1.1
*** /dev/null	Fri May 21 11:47:13 2004
--- llvm/test/Regression/Verifier/2004-05-21-SwitchConstantMismatch.ll	Fri May 21 11:47:02 2004
***************
*** 0 ****
--- 1,11 ----
+ 
+ 
+ int %main() {  
+ start1:
+   switch uint 0, label %brt0 [int 3, label %brt1  ]
+ brt0:
+   ret int 0
+ brt1:
+   ret int 0
+ }
+ 




From lattner at cs.uiuc.edu  Fri May 21 11:49:01 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri May 21 11:49:01 2004
Subject: [llvm-commits] CVS: llvm/lib/VMCore/Verifier.cpp
Message-ID: <200405211647.LAA10890@zion.cs.uiuc.edu>


Changes in directory llvm/lib/VMCore:

Verifier.cpp updated: 1.98 -> 1.99

---
Log message:

Fix for PR340: http://llvm.cs.uiuc.edu/PR340 : Verifier misses malformed switch instruction


---
Diffs of the changes:  (+12 -0)

Index: llvm/lib/VMCore/Verifier.cpp
diff -u llvm/lib/VMCore/Verifier.cpp:1.98 llvm/lib/VMCore/Verifier.cpp:1.99
--- llvm/lib/VMCore/Verifier.cpp:1.98	Mon May 17 16:15:18 2004
+++ llvm/lib/VMCore/Verifier.cpp	Fri May 21 11:47:21 2004
@@ -173,6 +173,7 @@
     void visitInstruction(Instruction &I);
     void visitTerminatorInst(TerminatorInst &I);
     void visitReturnInst(ReturnInst &RI);
+    void visitSwitchInst(SwitchInst &SI);
     void visitSelectInst(SelectInst &SI);
     void visitUserOp1(Instruction &I);
     void visitUserOp2(Instruction &I) { visitUserOp1(I); }
@@ -360,6 +361,17 @@
   // Check to make sure that the return value has necessary properties for
   // terminators...
   visitTerminatorInst(RI);
+}
+
+void Verifier::visitSwitchInst(SwitchInst &SI) {
+  // Check to make sure that all of the constants in the switch instruction
+  // have the same type as the switched-on value.
+  const Type *SwitchTy = SI.getCondition()->getType();
+  for (unsigned i = 1, e = SI.getNumCases(); i != e; ++i)
+    Assert1(SI.getCaseValue(i)->getType() == SwitchTy,
+            "Switch constants must all be same type as switch value!", &SI);
+
+  visitTerminatorInst(SI);
 }
 
 void Verifier::visitSelectInst(SelectInst &SI) {




From lattner at cs.uiuc.edu  Fri May 21 11:50:01 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri May 21 11:50:01 2004
Subject: [llvm-commits] CVS: llvm/docs/ReleaseNotes.html
Message-ID: <200405211648.LAA11492@zion.cs.uiuc.edu>


Changes in directory llvm/docs:

ReleaseNotes.html updated: 1.179 -> 1.180

---
Log message:

Bug fixed


---
Diffs of the changes:  (+2 -1)

Index: llvm/docs/ReleaseNotes.html
diff -u llvm/docs/ReleaseNotes.html:1.179 llvm/docs/ReleaseNotes.html:1.180
--- llvm/docs/ReleaseNotes.html:1.179	Wed May 12 16:46:05 2004
+++ llvm/docs/ReleaseNotes.html	Fri May 21 11:48:20 2004
@@ -213,6 +213,7 @@
 of the same name, both having constant pools</a></li>
 <li><a href="http://llvm.cs.uiuc.edu/PR337">[livevar] Live variables missed
 physical register use of aliased definition</a></li>
+<li><a href="http://llvm.cs.uiuc.edu/PR340">Verifier misses malformed switch instruction</a></li>
 </ol>
 
 <p>Bugs in the C/C++ front-end:</p>
@@ -645,7 +646,7 @@
   src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
 
   <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/05/12 21:46:05 $
+  Last modified: $Date: 2004/05/21 16:48:20 $
 </address>
 
 </body>




From lattner at cs.uiuc.edu  Fri May 21 11:50:07 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri May 21 11:50:07 2004
Subject: [llvm-commits] CVS: llvm-www/releases/1.2/docs/ReleaseNotes.html
Message-ID: <200405211649.LAA12094@zion.cs.uiuc.edu>


Changes in directory llvm-www/releases/1.2/docs:

ReleaseNotes.html updated: 1.16 -> 1.17

---
Log message:

Bug found


---
Diffs of the changes:  (+4 -1)

Index: llvm-www/releases/1.2/docs/ReleaseNotes.html
diff -u llvm-www/releases/1.2/docs/ReleaseNotes.html:1.16 llvm-www/releases/1.2/docs/ReleaseNotes.html:1.17
--- llvm-www/releases/1.2/docs/ReleaseNotes.html:1.16	Mon May 10 09:29:17 2004
+++ llvm-www/releases/1.2/docs/ReleaseNotes.html	Fri May 21 11:48:51 2004
@@ -356,6 +356,9 @@
 <li><a href="http://llvm.cs.uiuc.edu/PR332">[adce] Crash handling unreachable code that unwinds</a></li>
 
 <li><a href="http://llvm.cs.uiuc.edu/PR330">[vmcore] Linker causes erroneous asssertion</a></li>
+
+<li><a href="http://llvm.cs.uiuc.edu/PR340">Verifier misses malformed switch instruction</a></li>
+
 </ul>
 </div>
 
@@ -683,7 +686,7 @@
   src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
 
   <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/05/10 14:29:17 $
+  Last modified: $Date: 2004/05/21 16:48:51 $
 </address>
 
 </body>




From lattner at cs.uiuc.edu  Fri May 21 13:39:02 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri May 21 13:39:02 2004
Subject: [llvm-commits] CVS: llvm/include/llvm/Support/CFG.h
Message-ID: <200405211837.NAA13655@zion.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Support:

CFG.h updated: 1.20 -> 1.21

---
Log message:

Starting an identifier with an _ is not legal C/C++ code.



---
Diffs of the changes:  (+7 -7)

Index: llvm/include/llvm/Support/CFG.h
diff -u llvm/include/llvm/Support/CFG.h:1.20 llvm/include/llvm/Support/CFG.h:1.21
--- llvm/include/llvm/Support/CFG.h:1.20	Sat May  1 16:05:34 2004
+++ llvm/include/llvm/Support/CFG.h	Fri May 21 13:37:13 2004
@@ -90,20 +90,20 @@
 // BasicBlock succ_iterator definition
 //===--------------------------------------------------------------------===//
 
-template <class _Term, class _BB>           // Successor Iterator
-class SuccIterator : public bidirectional_iterator<_BB, ptrdiff_t> {
-  const _Term Term;
+template <class Term_, class BB_>           // Successor Iterator
+class SuccIterator : public bidirectional_iterator<BB_, ptrdiff_t> {
+  const Term_ Term;
   unsigned idx;
-  typedef bidirectional_iterator<_BB, ptrdiff_t> super;
+  typedef bidirectional_iterator<BB_, ptrdiff_t> super;
 public:
-  typedef SuccIterator<_Term, _BB> _Self;
+  typedef SuccIterator<Term_, BB_> _Self;
   typedef typename super::pointer pointer;
   // TODO: This can be random access iterator, need operator+ and stuff tho
     
-  inline SuccIterator(_Term T) : Term(T), idx(0) {         // begin iterator
+  inline SuccIterator(Term_ T) : Term(T), idx(0) {         // begin iterator
     assert(T && "getTerminator returned null!");
   }
-  inline SuccIterator(_Term T, bool)                       // end iterator
+  inline SuccIterator(Term_ T, bool)                       // end iterator
     : Term(T), idx(Term->getNumSuccessors()) {
     assert(T && "getTerminator returned null!");
   }




From lattner at cs.uiuc.edu  Fri May 21 13:40:03 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri May 21 13:40:03 2004
Subject: [llvm-commits] CVS: llvm/include/llvm/Analysis/Dominators.h
Message-ID: <200405211838.NAA13730@zion.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Analysis:

Dominators.h updated: 1.44 -> 1.45

---
Log message:

It's not clear to me whether the old version was correct C++ code, but in
any case it's not portable.



---
Diffs of the changes:  (+3 -3)

Index: llvm/include/llvm/Analysis/Dominators.h
diff -u llvm/include/llvm/Analysis/Dominators.h:1.44 llvm/include/llvm/Analysis/Dominators.h:1.45
--- llvm/include/llvm/Analysis/Dominators.h:1.44	Thu Mar 11 17:08:20 2004
+++ llvm/include/llvm/Analysis/Dominators.h	Fri May 21 13:38:16 2004
@@ -365,9 +365,9 @@
   /// changeImmediateDominator - This method is used to update the dominator
   /// tree information when a node's immediate dominator changes.
   ///
-  void changeImmediateDominator(Node *Node, Node *NewIDom) {
-    assert(Node && NewIDom && "Cannot change null node pointers!");
-    Node->setIDom(NewIDom);
+  void changeImmediateDominator(Node *N, Node *NewIDom) {
+    assert(N && NewIDom && "Cannot change null node pointers!");
+    N->setIDom(NewIDom);
   }
 
   /// print - Convert to human readable form




From gaeke at cs.uiuc.edu  Fri May 21 15:55:01 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Fri May 21 15:55:01 2004
Subject: [llvm-commits] CVS: reopt/lib/TraceToFunction/TraceToFunction.cpp
Message-ID: <200405212054.PAA23761@kain.cs.uiuc.edu>


Changes in directory reopt/lib/TraceToFunction:

TraceToFunction.cpp updated: 1.55 -> 1.56

---
Log message:

Fix two important bugs in live-in set calculation: 
1) phis which reference values defined in the trace are not live-in 
2) phis which define values in the entry basic block are live-in
   (still need to add fixup code in entry BB for these though!)


---
Diffs of the changes:  (+7 -0)

Index: reopt/lib/TraceToFunction/TraceToFunction.cpp
diff -u reopt/lib/TraceToFunction/TraceToFunction.cpp:1.55 reopt/lib/TraceToFunction/TraceToFunction.cpp:1.56
--- reopt/lib/TraceToFunction/TraceToFunction.cpp:1.55	Fri May 21 05:28:42 2004
+++ reopt/lib/TraceToFunction/TraceToFunction.cpp	Fri May 21 15:54:44 2004
@@ -113,6 +113,9 @@
     Instruction *UInst = dyn_cast<Instruction> (*ui);
     if (!UInst)                      // Non-Instruction Users scare me, mommy
       return false;
+    if (isa<PHINode> (UInst))
+      continue; // If a Phi node uses a value that's defined in the
+                // trace, the def. must have originally dominated the Phi.
     BasicBlock *UBlock = UInst->getParent ();
     if (Block == UBlock) {
       // Start at Block->begin() traversing next pointers; if UInst
@@ -163,6 +166,10 @@
       }
     }
   }
+  // Add phis from entry BB
+  for (BasicBlock::iterator Inst = T.getEntryBasicBlock ()->begin ();
+       PHINode *PN = dyn_cast<PHINode> (Inst); ++Inst)
+    S.insert (PN);
   return S;
 }
 




From alkis at cs.uiuc.edu  Fri May 21 16:37:01 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Fri May 21 16:37:01 2004
Subject: [llvm-commits] CVS: llvm-java/include/llvm/Java/Bytecode.h
Message-ID: <200405212136.QAA29631@zion.cs.uiuc.edu>


Changes in directory llvm-java/include/llvm/Java:

Bytecode.h updated: 1.1 -> 1.2

---
Log message:

Use nested Opcode namespace for opcodes.


---
Diffs of the changes:  (+3 -3)

Index: llvm-java/include/llvm/Java/Bytecode.h
diff -u llvm-java/include/llvm/Java/Bytecode.h:1.1 llvm-java/include/llvm/Java/Bytecode.h:1.2
--- llvm-java/include/llvm/Java/Bytecode.h:1.1	Tue May 18 15:13:37 2004
+++ llvm-java/include/llvm/Java/Bytecode.h	Fri May 21 16:35:56 2004
@@ -16,9 +16,9 @@
 
 #include <stdint.h>
 
-namespace llvm { namespace Java {
+namespace llvm { namespace Java { namespace Opcode {
 
-    enum Opcode {
+    enum {
         NOP = 0x00,
         ACONST_NULL = 0x01,
         ICONST_M1 = 0x02,
@@ -226,4 +226,4 @@
         IMPDEP2 = 0xFF,
     };
 
-} } // namespace llvm::Java
+} } } // namespace llvm::Java::BC




From brukman at cs.uiuc.edu  Fri May 21 18:23:00 2004
From: brukman at cs.uiuc.edu (Misha Brukman)
Date: Fri May 21 18:23:00 2004
Subject: [llvm-commits] CVS: llvm/Makefile.rules
Message-ID: <200405212321.SAA21412@zion.cs.uiuc.edu>


Changes in directory llvm:

Makefile.rules updated: 1.183 -> 1.184

---
Log message:

Pass the Makefile flags to recursive makes in {PARALLEL,OPTIONAL}_DIRS targets.


---
Diffs of the changes:  (+2 -2)

Index: llvm/Makefile.rules
diff -u llvm/Makefile.rules:1.183 llvm/Makefile.rules:1.184
--- llvm/Makefile.rules:1.183	Thu May 20 19:09:21 2004
+++ llvm/Makefile.rules	Fri May 21 18:21:11 2004
@@ -422,7 +422,7 @@
 		$(MKDIR) $(@D); \
 		cp $(SourceDir)/$(@D)/Makefile $(@D)/Makefile; \
 	fi; \
-	$(MAKE) -C $(@D) $(subst $(@D)/.make,,$@)
+	$(MAKE) -C $(@D) $(subst $(@D)/.make,,$@) $(MFLAGS)
 endif
 
 # Handle directories that may or may not exist
@@ -436,7 +436,7 @@
 				$(MKDIR) $$dir; \
 				cp $(SourceDir)/$$dir/Makefile $$dir/Makefile; \
 			fi; \
-			($(MAKE) -C$$dir $@) || exit 1; \
+			($(MAKE) -C$$dir $@ $(MFLAGS)) || exit 1; \
 		fi \
 	done
 endif




From llvm at cs.uiuc.edu  Fri May 21 20:06:01 2004
From: llvm at cs.uiuc.edu (LLVM)
Date: Fri May 21 20:06:01 2004
Subject: [llvm-commits] CVS: llvm/docs/Stacker.html
Message-ID: <200405220104.UAA30209@zion.cs.uiuc.edu>


Changes in directory llvm/docs:

Stacker.html updated: 1.12 -> 1.13

---
Log message:

Removed the 64-bit fix from the "to do" list since the bug is fixed.


---
Diffs of the changes:  (+1 -6)

Index: llvm/docs/Stacker.html
diff -u llvm/docs/Stacker.html:1.12 llvm/docs/Stacker.html:1.13
--- llvm/docs/Stacker.html:1.12	Wed May 12 19:37:23 2004
+++ llvm/docs/Stacker.html	Fri May 21 20:04:19 2004
@@ -1374,11 +1374,6 @@
     <li>Write an LLVM pass to compute the correct stack depth needed by the
     program. Currently the stack is set to a fixed number which means programs
     with large numbers of definitions might fail.</li>
-    <li>Enhance to run on 64-bit platforms like SPARC. Right now the size of a
-    pointer on 64-bit machines will cause incorrect results because of the
-    32-bit size of a stack element currently supported. This feature was not
-    implemented because LLVM needs a union type to be able to support the
-    different sizes correctly (portably and efficiently).</li>
     <li>Write an LLVM pass to optimize the use of the global stack. The code
     emitted currently is somewhat wasteful. It gets cleaned up a lot by existing
     passes but more could be done.</li>
@@ -1410,7 +1405,7 @@
 
   <a href="mailto:rspencer at x10sys.com">Reid Spencer</a><br>
   <a href="http://llvm.cs.uiuc.edu">LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/05/13 00:37:23 $
+  Last modified: $Date: 2004/05/22 01:04:19 $
 </address>
 
 </body>




From llvm at cs.uiuc.edu  Fri May 21 21:30:01 2004
From: llvm at cs.uiuc.edu (LLVM)
Date: Fri May 21 21:30:01 2004
Subject: [llvm-commits] CVS: llvm/docs/BytecodeFormat.html index.html
Message-ID: <200405220228.VAA30722@zion.cs.uiuc.edu>


Changes in directory llvm/docs:

BytecodeFormat.html added (r1.1)
index.html updated: 1.14 -> 1.15

---
Log message:

Added an initial template/overview for documentation of the LLVM Bytecode
file format. This format is being extracted from the source code and the
small recesses of Chris' brain.


---
Diffs of the changes:  (+250 -1)

Index: llvm/docs/BytecodeFormat.html
diff -c /dev/null llvm/docs/BytecodeFormat.html:1.1
*** /dev/null	Fri May 21 21:28:46 2004
--- llvm/docs/BytecodeFormat.html	Fri May 21 21:28:36 2004
***************
*** 0 ****
--- 1,248 ----
+ <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
+                       "http://www.w3.org/TR/html4/strict.dtd">
+ <html>
+ <head>
+   <title>LLVM Bytecode File Format</title>
+   <link rel="stylesheet" href="llvm.css" type="text/css">
+   <style>
+     table, tr, td { border: 2px solid gray }
+     th { border: 2px sold gray; font-weight: bold; }
+     table { border-collapse: collapse; margin-top: 1em margin-bottom: 1em }
+   </style>
+ </head>
+ <body>
+   <div class="doc_title"> LLVM Bytecode File Format </div>
+ <ol>
+   <li><a href="#abstract">Abstract</a></li>
+   <li><a href="#general">General Concepts</a></li>
+     <ol>
+       <li><a href="#blocks">Blocks</a></li>
+       <li><a href="#lists">Lists</a></li>
+       <li><a href="#fields">Fields</a></li>
+       <li><a href="#align">Alignment</a></li>
+     </ol>
+   <li><a href="#details">Detailed Layout</a>
+     <ol>
+       <li><a href="#notation">Notation</a></li>
+       <li><a href="#blocktypes">Blocks Types</a></li>
+       <li><a href="#header">Header Block</a></li>
+       <li><a href="#typeool">Global Type Pool</a></li>
+       <li><a href="#modinfo">Module Info Block</a></li>
+       <li><a href="#constants">Global Constant Pool</a></li>
+       <li><a href="#functions">Function Blocks</a><li>
+       <li><a href="#symtab">Module Symbol Table</a><li>
+     </ol>
+   </li>
+ </ol>
+ <div class="doc_text">
+ <p><b>Written by <a href="mailto:rspencer at x10sys.com">Reid Spencer</a>
+ and <a href="mailto:sabre at nondot.org">Chris Lattner</a></b></p>
+ <p> </p>
+ </div>
+ <!-- *********************************************************************** -->
+ <div class="doc_section"> <a name="abstract">Abstract </a></div>
+ <!-- *********************************************************************** -->
+ <div class="doc_text">
+ <p>This document is an (after the fact) specification of the LLVM bytecode
+ file format. It documents the binary encoding rules of the bytecode file format
+ so that equivalent systems can encode bytecode files correctly.  The LLVM 
+ bytecode representation is used to store the intermediate representation on 
+ disk in compacted form.
+ </p>
+ </div>
+ <!-- *********************************************************************** -->
+ <div class="doc_section"> <a name="general">General Concepts</a> </div>
+ <!-- *********************************************************************** -->
+ <div class="doc_text">
+ <p>This section describes the general concepts of the bytecode file format 
+ without getting into bit and byte level specifics.</p>
+ </div>
+ <!-- _______________________________________________________________________ -->
+ <div class="doc_subsection"><a name="blocks">Blocks</a> </div>
+ <div class="doc_text">
+ <p>LLVM bytecode files consist simply of a sequence of blocks of bytes. 
+ Each block begins with an identification value that determines the type of 
+ the next block.  The possible types of blocks are described below in the section 
+ <a href="#blockstypes">Block Types</a>. The block identifier is used because
+ it is possible for entire blocks to be omitted from the file if they are
+ empty. The block identifier helps the reader determine which kind of block is
+ next in the file.</p>
+ <p>
+ Except for the <a href="#header">Header Block</a> all blocks are variable
+ length. The consume just enough bytes to express their contents.</p>
+ </div>
+ <!-- _______________________________________________________________________ -->
+ <div class="doc_subsection"><a name="lists">Lists</a> </div>
+ <div class="doc_text">
+ <p>Most blocks are constructed of lists of information. Lists can be constructed
+ of other lists, etc. This decomposition of information follows the containment
+ hierarchy of the LLVM Intermediate Representation. For example, a function is
+ composed of a list of basic blocks. Each basic block is composed of a set of
+ instructions. This list of list nesting and hierarchy is maintained in the
+ bytecode file.</p>
+ <p>A list is encoded into the file simply by encoding the number of entries as
+ an integer followed by each of the entries. The reader knows when the list is
+ done because it will have filled the list with the required numbe of entries.
+ </p>
+ </div>
+ <!-- _______________________________________________________________________ -->
+ <div class="doc_subsection"><a name="fields">Fields</a> </div>
+ <div class="doc_text">
+ <p>Fields are units of information that LLVM knows how to write atomically.
+ Most fields have a uniform length or some kind of length indication built into
+ their encoding. For example, a constant string (array of SByte or UByte) is
+ written simply as the length followed by the characters. Although this is 
+ similar to a list, constant strings are treated atomically and are thus
+ fields.</p>
+ <p>Fields use a condensed bit format specific to the type of information
+ they must contain. As few bits as possible are written for each field. The
+ sections that follow will provide the details on how these fields are 
+ written and how the bits are to be interpreted.</p>
+ </div>
+ <!-- _______________________________________________________________________ -->
+ <div class="doc_subsection"><a name="align">Alignment</a> </div>
+ <div class="doc_text">
+ <p>To support cross-platform differences, the bytecode file is aligned on 
+ certain boundaries. This means that a small amount of padding (at most 3 bytes) 
+ will be added to ensure that the next entry is aligned to a 32-bit boundary.
+ </p>
+ </div>
+ <!-- *********************************************************************** -->
+ <div class="doc_section"> <a name="details">Detailed Layout</a> </div>
+ <!-- *********************************************************************** -->
+ <div class="doc_text">
+   <p>This section provides the detailed layout of the LLVM bytecode file format.
+   bit and byte level specifics.</p>
+ </div>
+ <!-- _______________________________________________________________________ -->
+ <div class="doc_subsection"><a name="notation">Notation</a></div>
+ <div class="doc_text">
+   <p>The descriptions of the bytecode format that follow describe the bit
+   fields in detail. These descriptions are provided in tabular form. Each table
+   has four columns that specify:</p>
+   <ol>
+     <li><b>Byte(s)</b>. The offset in bytes of the field from the start of
+     its container (block, list, other field).<li>
+     <li><b>Bit(s)</b>. The offset in bits of the field from the start of
+     the byte field. Bits are always little endian. That is, bit addresses with
+     smaller values have smaller address (i.e. 2^0 is at bit 0, 2^1 at 1, etc.)
+     </li>
+     <li><b>Align?</b> Indicates if this field is aligned to 32 bits or not.
+     This indicates where the <em>next</em> field starts, always on a 32 bit
+     boundary.</li>
+     <li><b>Description</b>. Descripts the contents of the field.</li>
+   </ol>
+ </div>
+ <!-- _______________________________________________________________________ -->
+ <div class="doc_subsection"><a name="blocktypes">Block Types</a></div>
+ <div class="doc_text">
+   <p>The bytecode format encodes the intermediate representation into groups
+   of bytes known as blocks. The blocks are written sequentially to the file in
+   the following order:</p>
+ <ol>
+   <li><a href="#header">Header</a>. This block contains the file signature 
+   (magic number), machine description and file format version (not LLVM 
+   version).</li>
+   <li><a href="#gtypepool">Global Type Pool</a>. This block contains all the
+   global (module) level types.</li>
+   <li><a href="#modinfo">Module Info</a>. This block contains the types of the
+   global variables and functions in the module as well as the constant
+   initializers for the global variables</li>
+   <li><a href="#constants">Constants</a>. This block contains all the global
+   constants except function arguments, global values and constant strings.</li>
+   <li><a href="#functions">Functions</a>. One function block is written for
+   each function in the module. </li>
+   <li><a href="$symtab">Symbol Table</a>. The module level symbol table that
+   provides names for the various other entries in the file is the final block 
+   written.</li>
+ </ol>
+ </div>
+ <!-- _______________________________________________________________________ -->
+ <div class="doc_subsection"><a name="header">Header Block</a> </div>
+ <div class="doc_text">
+ <p>The Header Block occurs in every LLVM bytecode file and is always first. It
+ simply provides a few bytes of data to identify the file, its format, and the
+ bytecode version. This block is fixed length and always eight bytes, as follows:
+ <table class="doc_table" width="90%">
+   <tr>
+     <th><b>Byte(s)</b></th>
+     <th><b>Bit(s)</b></th>
+     <th><b>Align?</b></th>
+     <th align="left"><b>Field Description</b></th>
+   </tr>
+   <tr>
+     <td>00</td><td>00-07</td><td>No</td>
+     <td align="left">Constant "l"</td>
+   </tr>
+   <tr>
+     <td>01</td><td>00-07</td><td>No</td>
+     <td align="left">Constant "l"</td>
+   </tr>
+   <tr>
+     <td>02</td><td>00-07</td><td>No</td>
+     <td align="left">Constant "v"</td>
+   </tr>
+   <tr>
+     <td>03</td><td>00-07</td><td>No</td>
+     <td align="left">Constant "m"</td>
+   </tr>
+   <tr>
+     <td>04-07</td><td>00</td><td>No</td>
+     <td align="left">Target is big endian</td>
+   </tr>
+   <tr>
+     <td>04-07</td><td>01</td><td>No</td>
+     <td align="left">Target has long pointers</td>
+   </tr>
+   <tr>
+     <td>04-07</td><td>02</td><td>No</td>
+     <td align="left">Target has no endianess</td>
+   </tr>
+   <tr>
+     <td>04-07</td><td>03</td><td>No</td>
+     <td align="left">Target has no pointer size</td>
+   </tr>
+   <tr>
+     <td>04-07</td><td>04-31</td><td>Yes</td>
+     <td align="left">The LLVM bytecode format version number</td>
+   </tr>
+ </table>
+ </div>
+ <!-- _______________________________________________________________________ -->
+ <div class="doc_subsection"><a name="gtypepool">Global Type Pool</a> </div>
+ <div class="doc_text">
+ </div>
+ <!-- _______________________________________________________________________ -->
+ <div class="doc_subsection"><a name="modinfo">Module Info</a> </div>
+ <div class="doc_text">
+ </div>
+ <!-- _______________________________________________________________________ -->
+ <div class="doc_subsection"><a name="constants">Constants</a> </div>
+ <div class="doc_text">
+ </div>
+ <!-- _______________________________________________________________________ -->
+ <div class="doc_subsection"><a name="functions">Functions</a> </div>
+ <div class="doc_text">
+ </div>
+ <!-- _______________________________________________________________________ -->
+ <div class="doc_subsection"><a name="symtab">Module Symbol Table</a> </div>
+ <div class="doc_text">
+ </div>
+ 
+ <!-- *********************************************************************** -->
+ <hr>
+ <address>
+   <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
+   src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
+   <a href="http://validator.w3.org/check/referer"><img
+   src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
+ 
+   <a href="mailto:rspencer at x10sys.com">Reid Spencer</a> and 
+   <a href="mailto:sabre at nondot.org">Chris Lattner</a><br>
+   <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
+   Last modified: $Date: 2004/05/22 02:28:36 $
+ </address>
+ </body>
+ </html>
+ <!-- vim: sw=2
+ -->


Index: llvm/docs/index.html
diff -u llvm/docs/index.html:1.14 llvm/docs/index.html:1.15
--- llvm/docs/index.html:1.14	Fri May 14 13:54:04 2004
+++ llvm/docs/index.html	Fri May 21 21:28:36 2004
@@ -114,6 +114,7 @@
 <li><a href="CommandLine.html">Command Line Library</a></li>
 <li><a href="ExtendingLLVM.html">Extending LLVM</a></li>
 <li><a href="CodingStandards.html">Coding Standards</a></li>
+<li><a href="BytecodeFormat.html">LLVM Bytecode File Format</a></li>
 </ul>
 </div>
 
@@ -178,7 +179,7 @@
 
   <a href="mailto:criswell [at] cs.uiuc.edu">John Criswell</a><br>
   <a href="http://llvm.cs.uiuc.edu">LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/05/14 18:54:04 $
+  Last modified: $Date: 2004/05/22 02:28:36 $
 </address>
 
 </body>




From llvm at cs.uiuc.edu  Sat May 22 00:58:01 2004
From: llvm at cs.uiuc.edu (LLVM)
Date: Sat May 22 00:58:01 2004
Subject: [llvm-commits] CVS: llvm/docs/BytecodeFormat.html
Message-ID: <200405220556.AAA31202@zion.cs.uiuc.edu>


Changes in directory llvm/docs:

BytecodeFormat.html updated: 1.1 -> 1.2

---
Log message:

Some additional data added to the bytecode specification.


---
Diffs of the changes:  (+36 -21)

Index: llvm/docs/BytecodeFormat.html
diff -u llvm/docs/BytecodeFormat.html:1.1 llvm/docs/BytecodeFormat.html:1.2
--- llvm/docs/BytecodeFormat.html:1.1	Fri May 21 21:28:36 2004
+++ llvm/docs/BytecodeFormat.html	Sat May 22 00:56:41 2004
@@ -67,9 +67,21 @@
 it is possible for entire blocks to be omitted from the file if they are
 empty. The block identifier helps the reader determine which kind of block is
 next in the file.</p>
-<p>
-Except for the <a href="#header">Header Block</a> all blocks are variable
-length. The consume just enough bytes to express their contents.</p>
+<p>The following block identifiers are currently in use 
+(from llvm/Bytecode/Format.h):</p>
+<ol>
+  <li><b>Module (0x01)</b>.</li>
+  <li><b>Function (0x11)</b>.</li>
+  <li><b>ConstantPool (0x12)</b>.</li>
+  <li><b>SymbolTable (0x13)</b>.</li>
+  <li><b>ModuleGlobalInfo (0x14)</b>.</li>
+  <li><b>GlobalTypePlane (0x15)</b>.</li>
+  <li><b>BasicBlock (0x31)</b>.</li>
+  <li><b>InstructionList (0x32)</b>.</li>
+  <li><b>CompactionTable (0x33)</b>.</li>
+</ol>
+<p> Except for the <a href="#header">Header Block</a> all blocks are variable
+length. They consume just enough bytes to express their contents.</p>
 </div>
 <!-- _______________________________________________________________________ -->
 <div class="doc_subsection"><a name="lists">Lists</a> </div>
@@ -130,6 +142,7 @@
     <li><b>Align?</b> Indicates if this field is aligned to 32 bits or not.
     This indicates where the <em>next</em> field starts, always on a 32 bit
     boundary.</li>
+    <li><b>Type</b>. The basic type of information contained in the field.</li>
     <li><b>Description</b>. Descripts the contents of the field.</li>
   </ol>
 </div>
@@ -168,42 +181,43 @@
     <th><b>Byte(s)</b></th>
     <th><b>Bit(s)</b></th>
     <th><b>Align?</b></th>
+    <th><b>Type</b></th>
     <th align="left"><b>Field Description</b></th>
   </tr>
   <tr>
-    <td>00</td><td>00-07</td><td>No</td>
-    <td align="left">Constant "l"</td>
+    <td>00</td><td>00-07</td><td>No</td><td>Char</td>
+    <td align="left">Constant "l" (0x6C)</td>
   </tr>
   <tr>
-    <td>01</td><td>00-07</td><td>No</td>
-    <td align="left">Constant "l"</td>
+    <td>01</td><td>00-07</td><td>No</td><td>Char</td>
+    <td align="left">Constant "l" (0x6C)</td>
   </tr>
   <tr>
-    <td>02</td><td>00-07</td><td>No</td>
-    <td align="left">Constant "v"</td>
+    <td>02</td><td>00-07</td><td>No</td><td>Char</td>
+    <td align="left">Constant "v" (0x76)</td>
   </tr>
   <tr>
-    <td>03</td><td>00-07</td><td>No</td>
-    <td align="left">Constant "m"</td>
+    <td>03</td><td>00-07</td><td>No</td><td>Char</td>
+    <td align="left">Constant "m" (0x6D)</td>
   </tr>
   <tr>
-    <td>04-07</td><td>00</td><td>No</td>
-    <td align="left">Target is big endian</td>
+    <td>04-07</td><td>00</td><td>No</td><td>Bool</td>
+    <td align="left">Target is big endian?</td>
   </tr>
   <tr>
-    <td>04-07</td><td>01</td><td>No</td>
-    <td align="left">Target has long pointers</td>
+    <td>04-07</td><td>01</td><td>No</td><td>Bool</td>
+    <td align="left">Target has long pointers?</td>
   </tr>
   <tr>
-    <td>04-07</td><td>02</td><td>No</td>
-    <td align="left">Target has no endianess</td>
+    <td>04-07</td><td>02</td><td>No</td><td>Bool</td>
+    <td align="left">Target has no endianess?</td>
   </tr>
   <tr>
-    <td>04-07</td><td>03</td><td>No</td>
-    <td align="left">Target has no pointer size</td>
+    <td>04-07</td><td>03</td><td>No</td><td>Bool</td>
+    <td align="left">Target has no pointer size?</td>
   </tr>
   <tr>
-    <td>04-07</td><td>04-31</td><td>Yes</td>
+    <td>04-07</td><td>04-31</td><td>Yes</td><td>Unsigned</td>
     <td align="left">The LLVM bytecode format version number</td>
   </tr>
 </table>
@@ -227,6 +241,7 @@
 <!-- _______________________________________________________________________ -->
 <div class="doc_subsection"><a name="symtab">Module Symbol Table</a> </div>
 <div class="doc_text">
+<p>The module symbol table is a list of
 </div>
 
 <!-- *********************************************************************** -->
@@ -240,7 +255,7 @@
   <a href="mailto:rspencer at x10sys.com">Reid Spencer</a> and 
   <a href="mailto:sabre at nondot.org">Chris Lattner</a><br>
   <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/05/22 02:28:36 $
+  Last modified: $Date: 2004/05/22 05:56:41 $
 </address>
 </body>
 </html>




From alkis at cs.uiuc.edu  Sat May 22 00:59:03 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sat May 22 00:59:03 2004
Subject: [llvm-commits] CVS: llvm-java/include/llvm/Java/ClassFile.h
Message-ID: <200405220557.AAA31230@zion.cs.uiuc.edu>


Changes in directory llvm-java/include/llvm/Java:

ClassFile.h updated: 1.9 -> 1.10

---
Log message:

Make code be an array of uint8_t's instead of char's.


---
Diffs of the changes:  (+2 -2)

Index: llvm-java/include/llvm/Java/ClassFile.h
diff -u llvm-java/include/llvm/Java/ClassFile.h:1.9 llvm-java/include/llvm/Java/ClassFile.h:1.10
--- llvm-java/include/llvm/Java/ClassFile.h:1.9	Thu May 20 19:05:59 2004
+++ llvm-java/include/llvm/Java/ClassFile.h	Sat May 22 00:57:09 2004
@@ -387,7 +387,7 @@
         uint16_t maxStack_;
         uint16_t maxLocals_;
         uint32_t codeSize_;
-        char* code_;
+        uint8_t* code_;
         Exceptions exceptions_;
         Attributes attributes_;
 
@@ -398,7 +398,7 @@
         ~CodeAttribute();
         uint16_t getMaxStack() const { return maxStack_; }
         uint16_t getMaxLocals() const { return maxLocals_; }
-        const char* getCode() const { return code_; }
+        const uint8_t* getCode() const { return code_; }
         uint32_t getCodeSize() const { return codeSize_; }
         const Exceptions& getExceptions() const { return exceptions_; }
         const Attributes& getAttributes() const { return attributes_; }




From alkis at cs.uiuc.edu  Sat May 22 00:59:09 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sat May 22 00:59:09 2004
Subject: [llvm-commits] CVS: llvm-java/lib/ClassFile/ClassFile.cpp
Message-ID: <200405220557.AAA31223@zion.cs.uiuc.edu>


Changes in directory llvm-java/lib/ClassFile:

ClassFile.cpp updated: 1.8 -> 1.9

---
Log message:

Make code be an array of uint8_t's instead of char's.


---
Diffs of the changes:  (+2 -2)

Index: llvm-java/lib/ClassFile/ClassFile.cpp
diff -u llvm-java/lib/ClassFile/ClassFile.cpp:1.8 llvm-java/lib/ClassFile/ClassFile.cpp:1.9
--- llvm-java/lib/ClassFile/ClassFile.cpp:1.8	Thu May 20 19:05:59 2004
+++ llvm-java/lib/ClassFile/ClassFile.cpp	Sat May 22 00:57:09 2004
@@ -537,8 +537,8 @@
     maxStack_ = readU2(is);
     maxLocals_ = readU2(is);
     codeSize_ = readU4(is);
-    code_ = new char[codeSize_];
-    std::streamsize s = is.rdbuf()->sgetn(code_, codeSize_);
+    code_ = new uint8_t[codeSize_];
+    std::streamsize s = is.rdbuf()->sgetn(reinterpret_cast<char*>(code_), codeSize_);
     if (s != (std::streamsize) codeSize_)
         throw ClassFileParseError(
             "Could not read code from input stream");




From alkis at cs.uiuc.edu  Sat May 22 00:59:15 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sat May 22 00:59:15 2004
Subject: [llvm-commits] CVS: llvm-java/lib/Compiler/
Message-ID: <200405220557.AAA31277@zion.cs.uiuc.edu>


Changes in directory llvm-java/lib/Compiler:


---
Log message:

Directory /home/vadve/shared/PublicCVS/llvm-java/lib/Compiler added to the repository


---
Diffs of the changes:  (+0 -0)



From alkis at cs.uiuc.edu  Sat May 22 01:00:03 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sat May 22 01:00:03 2004
Subject: [llvm-commits] CVS: llvm-java/lib/Compiler/Compiler.cpp
Message-ID: <200405220558.AAA31330@zion.cs.uiuc.edu>


Changes in directory llvm-java/lib/Compiler:

Compiler.cpp added (r1.1)

---
Log message:

Add an initial implementation of a Java bytecode to LLVM compiler.


---
Diffs of the changes:  (+428 -0)

Index: llvm-java/lib/Compiler/Compiler.cpp
diff -c /dev/null llvm-java/lib/Compiler/Compiler.cpp:1.1
*** /dev/null	Sat May 22 00:58:36 2004
--- llvm-java/lib/Compiler/Compiler.cpp	Sat May 22 00:58:26 2004
***************
*** 0 ****
--- 1,428 ----
+ //===-- Compiler.cpp - Java bytecode compiler -------------------*- C++ -*-===//
+ //
+ //                     The LLVM Compiler Infrastructure
+ //
+ // This file was developed by the LLVM research group and is distributed under
+ // the University of Illinois Open Source License. See LICENSE.TXT for details.
+ //
+ //===----------------------------------------------------------------------===//
+ //
+ // This file contains Java bytecode to LLVM bytecode compiler.
+ //
+ //===----------------------------------------------------------------------===//
+ 
+ #include <llvm/Java/Bytecode.h>
+ #include <llvm/Java/ClassFile.h>
+ #include <llvm/Java/Compiler.h>
+ #include <llvm/Constants.h>
+ #include <llvm/DerivedTypes.h>
+ #include <llvm/Instructions.h>
+ #include <llvm/Value.h>
+ #include <llvm/Type.h>
+ #include <stack>
+ #include <vector>
+ 
+ using namespace llvm;
+ 
+ namespace {
+ 
+     inline bool isTwoSlotValue(const Value* v) {
+         return v->getType() == Type::LongTy | v->getType() == Type::DoubleTy;
+     }
+ 
+     inline bool isOneSlotValue(const Value* v) {
+         return !isTwoSlotValue(v);
+     }
+ 
+     void compileMethod(Module& module, const Java::Method& method) {
+         using namespace llvm::Java::Opcode;
+ 
+         Function* function =
+             module.getOrInsertFunction(method.getName()->str(), Type::VoidTy);
+ 
+         const Java::CodeAttribute* codeAttr =
+             Java::getCodeAttribute(method.getAttributes());
+ 
+         std::stack<Value*> opStack;
+         std::vector<Value*> locals(codeAttr->getMaxLocals(), 0);
+ 
+         const uint8_t* code = codeAttr->getCode();
+         for (unsigned i = 0; i < codeAttr->getCodeSize(); ++i) {
+             bool wide = code[i] == WIDE;
+             i += wide;
+             switch (code[i]) {
+             case ACONST_NULL:
+                 // FIXME: should push a null pointer of type Object*
+                 opStack.push(
+                     ConstantPointerNull::get(PointerType::get(Type::VoidTy)));
+                 break;
+             case ICONST_M1:
+             case ICONST_0:
+             case ICONST_1:
+             case ICONST_2:
+             case ICONST_3:
+             case ICONST_4:
+             case ICONST_5:
+                 opStack.push(ConstantInt::get(Type::IntTy, code[i]-ICONST_0));
+                 break;
+             case LCONST_0:
+             case LCONST_1:
+                 opStack.push(ConstantInt::get(Type::LongTy, code[i]-LCONST_0));
+                 break;
+             case FCONST_0:
+             case FCONST_1:
+             case FCONST_2:
+                 opStack.push(ConstantFP::get(Type::FloatTy, code[i]-FCONST_0));
+                 break;
+             case DCONST_0:
+             case DCONST_1:
+                 opStack.push(ConstantFP::get(Type::DoubleTy, code[i]-DCONST_0));
+                 break;
+             case BIPUSH: {
+                 int imm = code[++i];
+                 opStack.push(ConstantInt::get(Type::IntTy, imm));
+                 break;
+             }
+             case SIPUSH: {
+                 int imm = (code[++i] << 8) | code[++i];
+                 opStack.push(ConstantInt::get(Type::IntTy, imm));
+                 break;
+             }
+             case LDC: {
+                 unsigned index = code[++i];
+                 // FIXME: load constant from constant pool
+             }
+             case LDC_W: {
+                 unsigned index = (code[++i] << 8) | code[++i];
+                 // FIXME: load constant from constant pool
+             }
+             case LDC2_W: {
+                 unsigned index = (code[++i] << 8) | code[++i];
+                 // FIXME: load constant from constant pool
+             }
+             case ILOAD:
+             case LLOAD:
+             case FLOAD:
+             case DLOAD:
+             case ALOAD: {
+                 // FIXME: use opcodes to perform type checking
+                 unsigned index = code[++i];
+                 opStack.push(locals[index]);
+                 break;
+             }
+             case ILOAD_0:
+             case ILOAD_1:
+             case ILOAD_2:
+             case ILOAD_3:
+                 opStack.push(locals[code[i]-ILOAD_0]);
+                 break;
+             case LLOAD_0:
+             case LLOAD_1:
+             case LLOAD_2:
+             case LLOAD_3:
+                 opStack.push(locals[code[i]-LLOAD_0]);
+                 break;
+             case FLOAD_0:
+             case FLOAD_1:
+             case FLOAD_2:
+             case FLOAD_3:
+                 opStack.push(locals[code[i]-FLOAD_0]);
+                 break;
+             case DLOAD_0:
+             case DLOAD_1:
+             case DLOAD_2:
+             case DLOAD_3:
+                 opStack.push(locals[code[i]-DLOAD_0]);
+                 break;
+             case ALOAD_0:
+             case ALOAD_1:
+             case ALOAD_2:
+             case ALOAD_3:
+                 opStack.push(locals[code[i]-ALOAD_0]);
+                 break;
+             case IALOAD:
+             case LALOAD:
+             case FALOAD:
+             case DALOAD:
+             case AALOAD:
+             case BALOAD:
+             case CALOAD:
+             case SALOAD:
+             case ISTORE:
+             case LSTORE:
+             case FSTORE:
+             case DSTORE:
+             case ASTORE:
+             case ISTORE_0:
+             case ISTORE_1:
+             case ISTORE_2:
+             case ISTORE_3:
+             case LSTORE_0:
+             case LSTORE_1:
+             case LSTORE_2:
+             case LSTORE_3:
+             case FSTORE_0:
+             case FSTORE_1:
+             case FSTORE_2:
+             case FSTORE_3:
+             case DSTORE_0:
+             case DSTORE_1:
+             case DSTORE_2:
+             case DSTORE_3:
+             case ASTORE_0:
+             case ASTORE_1:
+             case ASTORE_2:
+             case ASTORE_3:
+             case IASTORE:
+             case LASTORE:
+             case FASTORE:
+             case DASTORE:
+             case AASTORE:
+             case BASTORE:
+             case CASTORE:
+             case SASTORE:
+                 assert(0 && "not implemented");
+             case POP:
+                 opStack.pop();
+                 break;
+             case POP2:
+                 opStack.pop();
+                 opStack.pop();
+                 break;
+             case DUP:
+                 opStack.push(opStack.top());
+                 break;
+             case DUP_X1: {
+                 Value* v1 = opStack.top(); opStack.pop();
+                 Value* v2 = opStack.top(); opStack.pop();
+                 opStack.push(v1);
+                 opStack.push(v2);
+                 opStack.push(v1);
+                 break;
+             }
+             case DUP_X2: {
+                 Value* v1 = opStack.top(); opStack.pop();
+                 Value* v2 = opStack.top(); opStack.pop();
+                 if (isOneSlotValue(v2)) {
+                     Value* v3 = opStack.top(); opStack.pop();
+                     opStack.push(v1);
+                     opStack.push(v3);
+                     opStack.push(v2);
+                     opStack.push(v1);
+                 }
+                 else {
+                     opStack.push(v1);
+                     opStack.push(v2);
+                     opStack.push(v1);
+                 }
+                 break;
+             }
+             case DUP2: {
+                 Value* v1 = opStack.top(); opStack.pop();
+                 if (isOneSlotValue(v1)) {
+                     Value* v2 = opStack.top(); opStack.pop();
+                     opStack.push(v2);
+                     opStack.push(v1);
+                     opStack.push(v2);
+                     opStack.push(v1);
+                 }
+                 else {
+                     opStack.push(v1);
+                     opStack.push(v1);
+                 }
+                 break;
+             }
+             case DUP2_X1: {
+                 Value* v1 = opStack.top(); opStack.pop();
+                 Value* v2 = opStack.top(); opStack.pop();
+                 if (isOneSlotValue(v1)) {
+                     Value* v3 = opStack.top(); opStack.pop();
+                     opStack.push(v2);
+                     opStack.push(v1);
+                     opStack.push(v3);
+                     opStack.push(v2);
+                     opStack.push(v1);
+                 }
+                 else {
+                     opStack.push(v1);
+                     opStack.push(v2);
+                     opStack.push(v1);
+                 }
+                 break;
+             }
+             case DUP2_X2: {
+                 Value* v1 = opStack.top(); opStack.pop();
+                 Value* v2 = opStack.top(); opStack.pop();
+                 if (isOneSlotValue(v1)) {
+                     Value* v3 = opStack.top(); opStack.pop();
+                     if (isOneSlotValue(v3)) {
+                         Value* v4 = opStack.top(); opStack.pop();
+                         opStack.push(v2);
+                         opStack.push(v1);
+                         opStack.push(v4);
+                         opStack.push(v3);
+                         opStack.push(v2);
+                         opStack.push(v1);
+                     }
+                     else {
+                         opStack.push(v2);
+                         opStack.push(v1);
+                         opStack.push(v3);
+                         opStack.push(v2);
+                         opStack.push(v1);
+                     }
+                 }
+                 else {
+                     if (isOneSlotValue(v2)) {
+                         Value* v3 = opStack.top(); opStack.pop();
+                         opStack.push(v1);
+                         opStack.push(v3);
+                         opStack.push(v2);
+                         opStack.push(v1);
+                     }
+                     else {
+                         opStack.push(v1);
+                         opStack.push(v2);
+                         opStack.push(v1);
+                     }
+                 }
+                 break;
+             }
+             case SWAP: {
+                 Value* v1 = opStack.top(); opStack.pop();
+                 Value* v2 = opStack.top(); opStack.pop();
+                 opStack.push(v1);
+                 opStack.push(v2);
+                 break;
+             }
+             case IADD:
+             case LADD:
+             case FADD:
+             case DADD:
+             case ISUB:
+             case LSUB:
+             case FSUB:
+             case DSUB:
+             case IMUL:
+             case LMUL:
+             case FMUL:
+             case DMUL:
+             case IDIV:
+             case LDIV:
+             case FDIV:
+             case DDIV:
+             case IREM:
+             case LREM:
+             case FREM:
+             case DREM:
+             case INEG:
+             case LNEG:
+             case FNEG:
+             case DNEG:
+             case ISHL:
+             case LSHL:
+             case ISHR:
+             case LSHR:
+             case IUSHR:
+             case LUSHR:
+             case IAND:
+             case LAND:
+             case IOR:
+             case LOR:
+             case IXOR:
+             case LXOR:
+             case IINC:
+             case I2L:
+             case I2F:
+             case I2D:
+             case L2I:
+             case L2F:
+             case L2D:
+             case F2I:
+             case F2L:
+             case F2D:
+             case D2I:
+             case D2L:
+             case D2F:
+             case I2B:
+             case I2C:
+             case I2S:
+             case LCMP:
+             case FCMPL:
+             case FCMPG:
+             case DCMPL:
+             case DCMPG:
+             case IFEQ:
+             case IFNE:
+             case IFLT:
+             case IFGE:
+             case IFGT:
+             case IFLE:
+             case IF_ICMPEQ:
+             case IF_ICMPNE:
+             case IF_ICMPLT:
+             case IF_ICMPGE:
+             case IF_ICMPGT:
+             case IF_ICMPLE:
+             case IF_ICMPACMPEQ:
+             case IF_ICMPACMPNE:
+             case GOTO:
+             case JSR:
+             case RET:
+             case TABLESWITCH:
+             case LOOKUPSWITCH:
+             case IRETURN:
+             case LRETURN:
+             case FRETURN:
+             case DRETURN:
+             case ARETURN:
+             case RETURN:
+             case GETSTATIC:
+             case PUTSTATIC:
+             case GETFIELD:
+             case PUTFIELD:
+             case INVOKEVIRTUAL:
+             case INVOKESPECIAL:
+             case INVOKESTATIC:
+             case INVOKEINTERFACE:
+             case XXXUNUSEDXXX:
+             case NEW:
+             case NEWARRAY:
+             case ANEWARRAY:
+             case ARRAYLENGTH:
+             case ATHROW:
+             case CHECKCAST:
+             case INSTANCEOF:
+             case MONITORENTER:
+             case MONITOREXIT:
+             case WIDE:
+             case MULTIANEWARRAY:
+             case IFNULL:
+             case IFNONNULL:
+             case GOTO_W:
+             case JSR_W:
+             case BREAKPOINT:
+             case IMPDEP1:
+             case IMPDEP2:
+             case NOP:
+                 break;
+             }
+         }
+     }
+ 
+ } // namespace
+ 
+ Module* llvm::Java::compile(const ClassFile& cf)
+ {
+     Module* module = new Module(cf.getThisClass()->getName()->str());
+ 
+     std::vector<Value*> opStack;
+ 
+     const Java::Methods& methods = cf.getMethods();
+     for (Java::Methods::const_iterator
+              i = methods.begin(), e = methods.end(); i != e; ++i) {
+         compileMethod(*module, **i);
+     }
+ 
+     return module;
+ }




From alkis at cs.uiuc.edu  Sat May 22 01:00:12 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sat May 22 01:00:12 2004
Subject: [llvm-commits] CVS: llvm-java/include/llvm/Java/Compiler.h
Message-ID: <200405220558.AAA31337@zion.cs.uiuc.edu>


Changes in directory llvm-java/include/llvm/Java:

Compiler.h added (r1.1)

---
Log message:

Add an initial implementation of a Java bytecode to LLVM compiler.


---
Diffs of the changes:  (+25 -0)

Index: llvm-java/include/llvm/Java/Compiler.h
diff -c /dev/null llvm-java/include/llvm/Java/Compiler.h:1.1
*** /dev/null	Sat May 22 00:58:36 2004
--- llvm-java/include/llvm/Java/Compiler.h	Sat May 22 00:58:26 2004
***************
*** 0 ****
--- 1,25 ----
+ //===-- Compiler.h - Java bytecode compiler ---------------------*- C++ -*-===//
+ //
+ //                     The LLVM Compiler Infrastructure
+ //
+ // This file was developed by the LLVM research group and is distributed under
+ // the University of Illinois Open Source License. See LICENSE.TXT for details.
+ //
+ //===----------------------------------------------------------------------===//
+ //
+ // This file contains Java bytecode to LLVM bytecode compiler.
+ //
+ //===----------------------------------------------------------------------===//
+ 
+ #ifndef LLVM_JAVA_COMPILER_H
+ #define LLVM_JAVA_COMPILER_H
+ 
+ #include <llvm/Module.h>
+ 
+ namespace llvm { namespace Java {
+ 
+     Module* compile(const ClassFile& cf);
+ 
+ } } // namespace llvm::Java
+ 
+ #endif//LLVM_JAVA_COMPILER_H




From alkis at cs.uiuc.edu  Sat May 22 01:11:02 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sat May 22 01:11:02 2004
Subject: [llvm-commits] CVS: llvm-java/tools/classdump/classdump.cpp
Message-ID: <200405220609.BAA31428@zion.cs.uiuc.edu>


Changes in directory llvm-java/tools/classdump:

classdump.cpp updated: 1.4 -> 1.5

---
Log message:

Fix comment.


---
Diffs of the changes:  (+1 -1)

Index: llvm-java/tools/classdump/classdump.cpp
diff -u llvm-java/tools/classdump/classdump.cpp:1.4 llvm-java/tools/classdump/classdump.cpp:1.5
--- llvm-java/tools/classdump/classdump.cpp:1.4	Wed May 19 15:44:40 2004
+++ llvm-java/tools/classdump/classdump.cpp	Sat May 22 01:09:27 2004
@@ -7,7 +7,7 @@
 //
 //===----------------------------------------------------------------------===//
 //
-// This file a sample class reader driver. It is used to drive class
+// This is a sample class reader driver. It is used to drive class
 // reader tests.
 //
 //===----------------------------------------------------------------------===//




From alkis at cs.uiuc.edu  Sat May 22 01:11:11 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sat May 22 01:11:11 2004
Subject: [llvm-commits] CVS: llvm-java/lib/Makefile
Message-ID: <200405220609.BAA31447@zion.cs.uiuc.edu>


Changes in directory llvm-java/lib:

Makefile updated: 1.2 -> 1.3

---
Log message:

Build the compiler library by default.


---
Diffs of the changes:  (+1 -1)

Index: llvm-java/lib/Makefile
diff -u llvm-java/lib/Makefile:1.2 llvm-java/lib/Makefile:1.3
--- llvm-java/lib/Makefile:1.2	Thu Apr 15 23:03:45 2004
+++ llvm-java/lib/Makefile	Sat May 22 01:09:42 2004
@@ -8,6 +8,6 @@
 ##===----------------------------------------------------------------------===##
 LEVEL := ..
 
-PARALLEL_DIRS := ClassFile
+PARALLEL_DIRS := ClassFile Compiler
 
 include $(LEVEL)/Makefile.common




From alkis at cs.uiuc.edu  Sat May 22 01:23:03 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sat May 22 01:23:03 2004
Subject: [llvm-commits] CVS: llvm-java/Makefile.common.in
Message-ID: <200405220621.BAA31571@zion.cs.uiuc.edu>


Changes in directory llvm-java:

Makefile.common.in updated: 1.2 -> 1.3

---
Log message:

Add JAVAC and JAVA variables.


---
Diffs of the changes:  (+2 -0)

Index: llvm-java/Makefile.common.in
diff -u llvm-java/Makefile.common.in:1.2 llvm-java/Makefile.common.in:1.3
--- llvm-java/Makefile.common.in:1.2	Wed Apr 21 11:33:48 2004
+++ llvm-java/Makefile.common.in	Sat May 22 01:21:36 2004
@@ -28,6 +28,8 @@
 # Additional utilities
 #
 JIKES=@JIKES@
+JAVAC=@JAVAC@
+JAVA=@JAVA@
 
 #
 # Include LLVM's build rules.




From alkis at cs.uiuc.edu  Sat May 22 01:50:02 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sat May 22 01:50:02 2004
Subject: [llvm-commits] CVS: llvm-java/tools/classdump/classdump.cpp Makefile
Message-ID: <200405220648.BAA31723@zion.cs.uiuc.edu>


Changes in directory llvm-java/tools/classdump:

classdump.cpp updated: 1.5 -> 1.6
Makefile updated: 1.2 -> 1.3

---
Log message:

Print a stack trace on a crash.


---
Diffs of the changes:  (+4 -0)

Index: llvm-java/tools/classdump/classdump.cpp
diff -u llvm-java/tools/classdump/classdump.cpp:1.5 llvm-java/tools/classdump/classdump.cpp:1.6
--- llvm-java/tools/classdump/classdump.cpp:1.5	Sat May 22 01:09:27 2004
+++ llvm-java/tools/classdump/classdump.cpp	Sat May 22 01:48:35 2004
@@ -13,6 +13,7 @@
 //===----------------------------------------------------------------------===//
 
 #include <llvm/Java/ClassFile.h>
+#include <Support/Signals.h>
 
 #include <cstddef>
 #include <iostream>
@@ -22,6 +23,8 @@
 
 int main(int argc, char* argv[])
 {
+    PrintStackTraceOnErrorSignal();
+
     try {
         std::auto_ptr<Java::ClassFile> cf(
             Java::ClassFile::readClassFile(std::cin));


Index: llvm-java/tools/classdump/Makefile
diff -u llvm-java/tools/classdump/Makefile:1.2 llvm-java/tools/classdump/Makefile:1.3
--- llvm-java/tools/classdump/Makefile:1.2	Thu Apr 15 23:03:45 2004
+++ llvm-java/tools/classdump/Makefile	Sat May 22 01:48:35 2004
@@ -11,5 +11,6 @@
 TOOLNAME := classdump
 
 USEDLIBS := classfile
+LLVMLIBS := support vmcore
 
 include $(LEVEL)/Makefile.common




From gaeke at cs.uiuc.edu  Sat May 22 01:57:04 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Sat May 22 01:57:04 2004
Subject: [llvm-commits] CVS: reopt/lib/LightWtProfiling/TraceWriter.cpp
Message-ID: <200405220656.BAA00935@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/LightWtProfiling:

TraceWriter.cpp updated: 1.1 -> 1.2

---
Log message:

Add a FIXME comment.


---
Diffs of the changes:  (+2 -1)

Index: reopt/lib/LightWtProfiling/TraceWriter.cpp
diff -u reopt/lib/LightWtProfiling/TraceWriter.cpp:1.1 reopt/lib/LightWtProfiling/TraceWriter.cpp:1.2
--- reopt/lib/LightWtProfiling/TraceWriter.cpp:1.1	Wed May 19 15:16:12 2004
+++ reopt/lib/LightWtProfiling/TraceWriter.cpp	Sat May 22 01:56:38 2004
@@ -54,7 +54,8 @@
   out << "\n";
   out.close ();
 
-  // Write out bytecode.
+  // Write out bytecode. FIXME: Should extract function F (as w/
+  // 'extract' tool) and write out only F, not its entire module.
   std::ofstream bout (bytecodeFileName.c_str ());
   WriteBytecodeToFile(F->getParent (), bout);
   bout.close ();




From gaeke at cs.uiuc.edu  Sat May 22 01:57:14 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Sat May 22 01:57:14 2004
Subject: [llvm-commits] CVS: reopt/lib/TraceToFunction/TraceToFunction.cpp
Message-ID: <200405220656.BAA00942@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/TraceToFunction:

TraceToFunction.cpp updated: 1.56 -> 1.57

---
Log message:

Three important changes regarding trace entry-BB Phi nodes:
1) In DEBUG mode, change their names to reflect their special status, and add
better debug messages.
2) Assert that they are live-in (fixing a longstanding FIXME).
3) Rewrite their off-trace incoming values to be the argument of the TraceFn
which carries their live-in value.


---
Diffs of the changes:  (+28 -12)

Index: reopt/lib/TraceToFunction/TraceToFunction.cpp
diff -u reopt/lib/TraceToFunction/TraceToFunction.cpp:1.56 reopt/lib/TraceToFunction/TraceToFunction.cpp:1.57
--- reopt/lib/TraceToFunction/TraceToFunction.cpp:1.56	Fri May 21 15:54:44 2004
+++ reopt/lib/TraceToFunction/TraceToFunction.cpp	Sat May 22 01:56:39 2004
@@ -546,26 +546,42 @@
   if (srcB == T.getEntryBasicBlock ()) {
     BasicBlock *EntryFixup = new BasicBlock ("entryfixup", F, dstB);
     EntryFixup->getInstList ().push_back (new BranchInst (dstB));
-    // Replace the references to the trace's predecessor with a
-    // reference to EntryFixup now. This is kind of a dodge because we
-    // don't have a pointer to the trace's predecessor, so we have
-    // to guess which one it is. Assume it's any phi value source in
-    // the entry BB which is not in the trace.
+    // Rewrite phi nodes in the entry basic block, which carry values
+    // that are live-in to the trace: Off-trace phi sources must be
+    // rewritten to take their value from the entryfixup block, and get
+    // their value from the argument which carries the live-in value.
     for (BasicBlock::iterator BI = srcB->begin ();
-         PHINode *PN = dyn_cast<PHINode> (BI); ++BI)
-      for (unsigned i = 0; i < PN->getNumIncomingValues (); ++i) {
+         PHINode *oldPN = dyn_cast<PHINode> (BI); ++BI) {
+      DEBUG (std::cerr << "fixupFunctionBodyBB: Changing name of trace"
+        << " entry block Phi node:\n" << O2CMap[oldPN] << " to "
+        << O2CMap[oldPN]->getName() << ".phifixup\n";
+        O2CMap[oldPN]->setName (O2CMap[oldPN]->getName () + ".phifixup"));
+      assert (TF->LiveInSet.find (oldPN) != TF->LiveInSet.end ()
+              && "Phi nodes in trace entry BB must be live-in");
+      unsigned argNum = TF->LiveInToParameterMap [oldPN];
+      Argument *phiLiveIn = getFunctionArg (F, argNum);
+      for (unsigned i = 0; i < oldPN->getNumIncomingValues (); ++i) {
         // Fold Phi sources which come from FLI blocks
-        BasicBlock *phiSource = PN->getIncomingBlock (i);
+        BasicBlock *phiSource = oldPN->getIncomingBlock (i);
         if (BasicBlock *realSource = getFLIEdgeSource (phiSource))
           phiSource = realSource;
         if (!T.contains (phiSource)) {
-          // FIXME: Assert that O2CMap[PN]'s value i is live in.
-          Value *V = O2CMap[PN];
+          Value *V = O2CMap[oldPN];
           assert (V && isa<PHINode> (V)
                   && "Clone of PHINode from trace entry BB missing or mangled");
-          cast<PHINode> (V)->setIncomingBlock (i, EntryFixup);
+          PHINode *newPN = cast<PHINode> (V);
+          DEBUG (std::cerr << "fixupFunctionBodyBB: Changing incoming block "
+            << i << " of trace entry block Phi node:\n" << *newPN << " from "
+            << newPN->getIncomingBlock (i)->getName () << " to entry-fixup "
+            << "block, and incoming value to argument " << argNum << " (";
+            WriteAsOperand (std::cerr, phiLiveIn, true, true,
+                            TF->TraceFn->getParent ());
+            std::cerr << ")\n");
+          newPN->setIncomingBlock (i, EntryFixup);
+          newPN->setIncomingValue (i, phiLiveIn);
         }
-     }
+      }
+    }
   }
 
   // If srcB contains a trace-exiting branch B, fix up B's clone in




From gaeke at cs.uiuc.edu  Sat May 22 01:57:21 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Sat May 22 01:57:21 2004
Subject: [llvm-commits] CVS: reopt/test/TEST.reopt.Makefile
Message-ID: <200405220656.BAA00949@seraph.cs.uiuc.edu>


Changes in directory reopt/test:

TEST.reopt.Makefile updated: 1.12 -> 1.13

---
Log message:

Don't dump the wholereoptimizer.o object file in llvm/lib/Debug; 
put it in llvm/projects/reopt/lib/Debug instead.


---
Diffs of the changes:  (+2 -2)

Index: reopt/test/TEST.reopt.Makefile
diff -u reopt/test/TEST.reopt.Makefile:1.12 reopt/test/TEST.reopt.Makefile:1.13
--- reopt/test/TEST.reopt.Makefile:1.12	Tue May 18 11:51:29 2004
+++ reopt/test/TEST.reopt.Makefile	Sat May 22 01:56:40 2004
@@ -23,7 +23,7 @@
 
 TESTNAME = $*
 
-REOPTLIBDIR = $(PROJECT_DIR)/lib/Debug
+REOPTLIBDIR = $(PROJECT_DIR)/lib/$(CONFIGURATION)
 
 # Libraries that contain the Reoptimizer itself
 REOPTIMIZER_OBJS = $(REOPTLIBDIR)/firstTrigger.o \
@@ -46,7 +46,7 @@
 # Solaris libraries that the Reoptimizer depends on
 REOPTIMIZER_SOLARISLIBS = -lcpc -lm -lrt -lmalloc -ldl
 
-WHOLE_REOPTIMIZER = $(DESTLIBCURRENT)/wholereoptimizer.o
+WHOLE_REOPTIMIZER = $(REOPTLIBDIR)/wholereoptimizer.o
 $(WHOLE_REOPTIMIZER): $(REOPTIMIZER_OBJS) $(REOPTIMIZER_LLVMOBJS)
 	/usr/ccs/bin/ld -r -o $@ $+
 




From gaeke at cs.uiuc.edu  Sat May 22 01:57:26 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Sat May 22 01:57:26 2004
Subject: [llvm-commits] CVS: reopt/tools/ttftest/ttftest.cpp
Message-ID: <200405220656.BAA00956@seraph.cs.uiuc.edu>


Changes in directory reopt/tools/ttftest:

ttftest.cpp updated: 1.3 -> 1.4

---
Log message:

Add -q (quiet) flag, to suppress output unless it is absolutely necessary,
for the benefit of the test script.


---
Diffs of the changes:  (+12 -7)

Index: reopt/tools/ttftest/ttftest.cpp
diff -u reopt/tools/ttftest/ttftest.cpp:1.3 reopt/tools/ttftest/ttftest.cpp:1.4
--- reopt/tools/ttftest/ttftest.cpp:1.3	Wed May 19 03:44:44 2004
+++ reopt/tools/ttftest/ttftest.cpp	Sat May 22 01:56:41 2004
@@ -33,6 +33,7 @@
     cl::desc("Name of file containing trace"), cl::value_desc("traceFileName"));
   cl::opt<std::string> BytecodeFile("bc",
     cl::desc("Name of file containing module"), cl::value_desc("bytecodeFileName"));
+  cl::opt<bool> Quiet("q", cl::desc("Be quiet"), cl::init(false));
 };
 
 /// getBasicBlockByNum - Returns the basic block in F whose index is
@@ -66,17 +67,19 @@
                 + "' not found in module";
     return 0;
   }
-  std::cerr << "Function name: '" << functionName << "'\n" << "Basic blocks: ";
+  if (!Quiet)
+    std::cerr << "Function name: '" << functionName << "'\n"
+              << "Basic blocks: ";
   std::vector<BasicBlock *> vBB;
   do {
     int basicBlockNum;
     in >> basicBlockNum;
     if (!in.eof ()) {
-      std::cerr << basicBlockNum << " ";
+      if (!Quiet) std::cerr << basicBlockNum << " ";
       vBB.push_back (getBasicBlockByNum (F, basicBlockNum));
     }
   } while (!in.eof ());
-  std::cerr << "\n";
+  if (!Quiet) std::cerr << "\n";
   return new Trace (vBB);
 }
 
@@ -115,14 +118,16 @@
 
   // Run TraceToFunction on the Trace.
   TraceFunction *TF = TraceFunction::get (*T);
-  std::cerr << *TF->TraceFn << "\n\n";
+  if (!Quiet) { std::cerr << *TF->TraceFn << "\n\n"; }
 
   // Run the LLVM Verifier on the resulting TraceFunction.
-  if (verifyFunction (*TF->TraceFn, PrintMessageAction)) {
-    std::cerr << "Verifier fails.\n";
+  VerifierFailureAction action =
+    (Quiet ? ReturnStatusAction : PrintMessageAction);
+  if (verifyFunction (*TF->TraceFn, action)) {
+    if (!Quiet) { std::cerr << "Verifier fails.\n"; }
     exit (1);
   } else {
-    std::cerr << "Verifier passes.\n";
+    if (!Quiet) { std::cerr << "Verifier passes.\n"; }
     exit (0);
   }
 }




From alkis at cs.uiuc.edu  Sat May 22 02:02:06 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sat May 22 02:02:06 2004
Subject: [llvm-commits] CVS: llvm-java/lib/Compiler/Makefile
Message-ID: <200405220700.CAA31832@zion.cs.uiuc.edu>


Changes in directory llvm-java/lib/Compiler:

Makefile added (r1.1)

---
Log message:

Add Compiler library makefile.


---
Diffs of the changes:  (+15 -0)

Index: llvm-java/lib/Compiler/Makefile
diff -c /dev/null llvm-java/lib/Compiler/Makefile:1.1
*** /dev/null	Sat May 22 02:00:26 2004
--- llvm-java/lib/Compiler/Makefile	Sat May 22 02:00:16 2004
***************
*** 0 ****
--- 1,15 ----
+ ##===- lib/Compiler/Makefile -------------------------------*- Makefile -*-===##
+ #
+ #                     The LLVM Compiler Infrastructure
+ #
+ # This file was developed by the LLVM research group and is distributed under
+ # the University of Illinois Open Source License. See LICENSE.TXT for details.
+ #
+ ##===----------------------------------------------------------------------===##
+ LEVEL := ../..
+ 
+ LIBRARYNAME := compiler
+ 
+ BUILD_ARCHIVE = 1
+ 
+ include $(LEVEL)/Makefile.common




From alkis at cs.uiuc.edu  Sat May 22 02:09:02 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sat May 22 02:09:02 2004
Subject: [llvm-commits] CVS: llvm-java/tools/class2llvm/
Message-ID: <200405220707.CAA02092@zion.cs.uiuc.edu>


Changes in directory llvm-java/tools/class2llvm:


---
Log message:

Directory /home/vadve/shared/PublicCVS/llvm-java/tools/class2llvm added to the repository


---
Diffs of the changes:  (+0 -0)



From alkis at cs.uiuc.edu  Sat May 22 02:10:02 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sat May 22 02:10:02 2004
Subject: [llvm-commits] CVS: llvm-java/tools/class2llvm/class2llvm.cpp Makefile
Message-ID: <200405220708.CAA03405@zion.cs.uiuc.edu>


Changes in directory llvm-java/tools/class2llvm:

class2llvm.cpp added (r1.1)
Makefile added (r1.1)

---
Log message:

Add class2llvm utility.


---
Diffs of the changes:  (+63 -0)

Index: llvm-java/tools/class2llvm/class2llvm.cpp
diff -c /dev/null llvm-java/tools/class2llvm/class2llvm.cpp:1.1
*** /dev/null	Sat May 22 02:08:41 2004
--- llvm-java/tools/class2llvm/class2llvm.cpp	Sat May 22 02:08:31 2004
***************
*** 0 ****
--- 1,47 ----
+ //===-- class2llvm.cpp - class2llvm utility ---------------------*- C++ -*-===//
+ //
+ //                     The LLVM Compiler Infrastructure
+ //
+ // This file was developed by the LLVM research group and is distributed under
+ // the University of Illinois Open Source License. See LICENSE.TXT for details.
+ //
+ //===----------------------------------------------------------------------===//
+ //
+ // This is a sample class reader driver. It is used to drive class
+ // reader tests.
+ //
+ //===----------------------------------------------------------------------===//
+ 
+ #include <llvm/Java/ClassFile.h>
+ #include <llvm/Java/Compiler.h>
+ #include <llvm/PassManager.h>
+ #include <llvm/Assembly/PrintModulePass.h>
+ #include <Support/Signals.h>
+ 
+ #include <cstddef>
+ #include <iostream>
+ #include <memory>
+ 
+ using namespace llvm;
+ 
+ int main(int argc, char* argv[])
+ {
+     PrintStackTraceOnErrorSignal();
+ 
+     try {
+         std::auto_ptr<Java::ClassFile> cf(
+             Java::ClassFile::readClassFile(std::cin));
+ 
+         Module* module = Java::compile(*cf);
+ 
+         PassManager passes;
+         passes.add(new PrintModulePass(&std::cout));
+         passes.run(*module);
+     }
+     catch (std::exception& e) {
+         std::cerr << e.what() << '\n';
+         return EXIT_FAILURE;
+     }
+ 
+     return EXIT_SUCCESS;
+ }


Index: llvm-java/tools/class2llvm/Makefile
diff -c /dev/null llvm-java/tools/class2llvm/Makefile:1.1
*** /dev/null	Sat May 22 02:08:41 2004
--- llvm-java/tools/class2llvm/Makefile	Sat May 22 02:08:31 2004
***************
*** 0 ****
--- 1,16 ----
+ #===- tools/class2llvm/Makefile ----------------------------*- Makefile -*-===##
+ #
+ #                     The LLVM Compiler Infrastructure
+ #
+ # This file was developed by the LLVM research group and is distributed under
+ # the University of Illinois Open Source License. See LICENSE.TXT for details.
+ #
+ ##===----------------------------------------------------------------------===##
+ LEVEL := ../..
+ 
+ TOOLNAME := class2llvm
+ 
+ USEDLIBS := classfile compiler
+ LLVMLIBS := support vmcore
+ 
+ include $(LEVEL)/Makefile.common




From alkis at cs.uiuc.edu  Sat May 22 02:10:09 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sat May 22 02:10:09 2004
Subject: [llvm-commits] CVS: llvm-java/tools/Makefile
Message-ID: <200405220708.CAA03411@zion.cs.uiuc.edu>


Changes in directory llvm-java/tools:

Makefile updated: 1.2 -> 1.3

---
Log message:

Add class2llvm utility.


---
Diffs of the changes:  (+1 -1)

Index: llvm-java/tools/Makefile
diff -u llvm-java/tools/Makefile:1.2 llvm-java/tools/Makefile:1.3
--- llvm-java/tools/Makefile:1.2	Thu Apr 15 23:03:45 2004
+++ llvm-java/tools/Makefile	Sat May 22 02:08:31 2004
@@ -8,6 +8,6 @@
 ##===----------------------------------------------------------------------===##
 LEVEL := ..
 
-PARALLEL_DIRS := classdump
+PARALLEL_DIRS := classdump class2llvm
 
 include $(LEVEL)/Makefile.common




From alkis at cs.uiuc.edu  Sat May 22 02:23:03 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sat May 22 02:23:03 2004
Subject: [llvm-commits] CVS: llvm-java/test/Makefile
Message-ID: <200405220721.CAA09408@zion.cs.uiuc.edu>


Changes in directory llvm-java/test:

Makefile added (r1.1)

---
Log message:

Add test Makefile.


---
Diffs of the changes:  (+13 -0)

Index: llvm-java/test/Makefile
diff -c /dev/null llvm-java/test/Makefile:1.1
*** /dev/null	Sat May 22 02:21:26 2004
--- llvm-java/test/Makefile	Sat May 22 02:21:16 2004
***************
*** 0 ****
--- 1,13 ----
+ ##===- tests/Makefile --------------------------------------*- Makefile -*-===##
+ #
+ #                     The LLVM Compiler Infrastructure
+ #
+ # This file was developed by the LLVM research group and is distributed under
+ # the University of Illinois Open Source License. See LICENSE.TXT for details.
+ #
+ ##===----------------------------------------------------------------------===##
+ LEVEL := ..
+ 
+ BUILD_JAVA_SOURCE=1
+ 
+ include $(LEVEL)/Makefile.common




From gaeke at cs.uiuc.edu  Sat May 22 15:43:01 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Sat May 22 15:43:01 2004
Subject: [llvm-commits] CVS: reopt/lib/Optimizations/Makefile
Message-ID: <200405222042.PAA01760@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/Optimizations:

Makefile added (r1.1)

---
Log message:

Add makefile for runtimelicm.


---
Diffs of the changes:  (+4 -0)

Index: reopt/lib/Optimizations/Makefile
diff -c /dev/null reopt/lib/Optimizations/Makefile:1.1
*** /dev/null	Sat May 22 15:42:05 2004
--- reopt/lib/Optimizations/Makefile	Sat May 22 15:41:55 2004
***************
*** 0 ****
--- 1,4 ----
+ LEVEL = ../..
+ LIBRARYNAME = runtimelicm
+ 
+ include $(LEVEL)/Makefile.common




From gaeke at cs.uiuc.edu  Sat May 22 15:43:07 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Sat May 22 15:43:07 2004
Subject: [llvm-commits] CVS: reopt/lib/Optimizations/RuntimeLICM.cpp
Message-ID: <200405222042.PAA01806@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/Optimizations:

RuntimeLICM.cpp updated: 1.2 -> 1.3

---
Log message:

Make this file compile again.  Put code into llvm namespace.
Get rid of using std:: directives.  Use DEBUG() for debugging msgs.


---
Diffs of the changes:  (+10 -9)

Index: reopt/lib/Optimizations/RuntimeLICM.cpp
diff -u reopt/lib/Optimizations/RuntimeLICM.cpp:1.2 reopt/lib/Optimizations/RuntimeLICM.cpp:1.3
--- reopt/lib/Optimizations/RuntimeLICM.cpp:1.2	Thu Aug 21 15:55:17 2003
+++ reopt/lib/Optimizations/RuntimeLICM.cpp	Sat May 22 15:41:41 2004
@@ -11,15 +11,14 @@
 #include "llvm/iMemory.h"
 #include "llvm/iPHINode.h"
 #include "reopt/BinInterface/LLVMTrace.h"
-
+#include "Support/Debug.h"
 #include <iostream>
-#include <vector>
 #include <set>
+using namespace llvm;
 
-using std::vector;
-using std::set;
+namespace llvm {
 
-void RuntimeLICM(vector<BasicBlock *> &trace, LLVMTrace &ltrace) {
+void RuntimeLICM(std::vector<BasicBlock *> &trace, LLVMTrace &ltrace) {
   
   //Put the trace in a set for faster searching
   std::set<BasicBlock*> traceSet(trace.begin(), trace.end());
@@ -28,7 +27,7 @@
   unsigned numLIC = 0;
   
   //loop over all the basic blocks in my trace
-  for(vector<BasicBlock *>::iterator T = trace.begin(), End = trace.end(); 
+  for(std::vector<BasicBlock *>::iterator T = trace.begin(), End = trace.end(); 
       T != End; ++T) {
     
     //loop over all the instructions in the basic block
@@ -57,14 +56,16 @@
 	}
 
 	if(isLoopInvariant) {
-	  std::cerr << "Loop Invariant Instruction: " << i << "\n";
+	  DEBUG (std::cerr << "Loop Invariant Instruction: " << i << "\n";
+	         ++numLIC);
 	  //move to header block
 	  ltrace.moveInstrToSec(&(*i), 0);
-	  ++numLIC;
 	}
       }
     }
   }
   
-  std::cerr << "Number Loop Invariant: " << numLIC << "\n";
+  DEBUG (std::cerr << "Number Loop Invariant: " << numLIC << "\n");
 }
+
+} // end namespace llvm




From gaeke at cs.uiuc.edu  Sat May 22 15:43:11 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Sat May 22 15:43:11 2004
Subject: [llvm-commits] CVS: reopt/lib/Makefile
Message-ID: <200405222042.PAA01818@seraph.cs.uiuc.edu>


Changes in directory reopt/lib:

Makefile updated: 1.20 -> 1.21

---
Log message:

Add Optimizations directory to the build.


---
Diffs of the changes:  (+1 -1)

Index: reopt/lib/Makefile
diff -u reopt/lib/Makefile:1.20 reopt/lib/Makefile:1.21
--- reopt/lib/Makefile:1.20	Wed May 19 01:50:56 2004
+++ reopt/lib/Makefile	Sat May 22 15:42:40 2004
@@ -1,5 +1,5 @@
 LEVEL = ..
-DIRS := BinInterface Mapping TraceCache Trigger LightWtProfiling Inst TraceToFunction
+DIRS := BinInterface Mapping TraceCache Trigger LightWtProfiling Inst Optimizations TraceToFunction
 
 include $(LEVEL)/Makefile.config
 




From alkis at cs.uiuc.edu  Sat May 22 17:52:01 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sat May 22 17:52:01 2004
Subject: [llvm-commits] CVS: llvm-java/lib/Compiler/Compiler.cpp
Message-ID: <200405222249.RAA07727@zion.cs.uiuc.edu>


Changes in directory llvm-java/lib/Compiler:

Compiler.cpp updated: 1.1 -> 1.2

---
Log message:

Add utility functions to read immediates and indices from bytecode.


---
Diffs of the changes:  (+27 -8)

Index: llvm-java/lib/Compiler/Compiler.cpp
diff -u llvm-java/lib/Compiler/Compiler.cpp:1.1 llvm-java/lib/Compiler/Compiler.cpp:1.2
--- llvm-java/lib/Compiler/Compiler.cpp:1.1	Sat May 22 00:58:26 2004
+++ llvm-java/lib/Compiler/Compiler.cpp	Sat May 22 17:49:14 2004
@@ -34,6 +34,22 @@
         return !isTwoSlotValue(v);
     }
 
+    inline int readByteImmediate(const uint8_t* code, unsigned& i) {
+        return code[++i];
+    }
+
+    inline int readShortImmediate(const uint8_t* code, unsigned& i) {
+        return (code[++i] << 8) | code[++i];
+    }
+
+    inline unsigned readByteIndex(const uint8_t* code, unsigned& i) {
+        return code[++i];
+    }
+
+    inline unsigned readShortIndex(const uint8_t* code, unsigned& i) {
+        return (code[++i] << 8) | code[++i];
+    }
+
     void compileMethod(Module& module, const Java::Method& method) {
         using namespace llvm::Java::Opcode;
 
@@ -43,11 +59,14 @@
         const Java::CodeAttribute* codeAttr =
             Java::getCodeAttribute(method.getAttributes());
 
-        std::stack<Value*> opStack;
-        std::vector<Value*> locals(codeAttr->getMaxLocals(), 0);
+        typedef std::stack<Value*> OperandStack;
+        OperandStack opStack;
+        typedef std::vector<Value*> Locals;
+        Locals locals(codeAttr->getMaxLocals(), 0);
 
         const uint8_t* code = codeAttr->getCode();
         for (unsigned i = 0; i < codeAttr->getCodeSize(); ++i) {
+            unsigned bcStart = i;
             bool wide = code[i] == WIDE;
             i += wide;
             switch (code[i]) {
@@ -79,25 +98,25 @@
                 opStack.push(ConstantFP::get(Type::DoubleTy, code[i]-DCONST_0));
                 break;
             case BIPUSH: {
-                int imm = code[++i];
+                int imm = readByteImmediate(code, i);
                 opStack.push(ConstantInt::get(Type::IntTy, imm));
                 break;
             }
             case SIPUSH: {
-                int imm = (code[++i] << 8) | code[++i];
+                int imm = readShortImmediate(code, i);
                 opStack.push(ConstantInt::get(Type::IntTy, imm));
                 break;
             }
             case LDC: {
-                unsigned index = code[++i];
+                unsigned index = readByteIndex(code, i);
                 // FIXME: load constant from constant pool
             }
             case LDC_W: {
-                unsigned index = (code[++i] << 8) | code[++i];
+                unsigned index = readShortIndex(code, i);
                 // FIXME: load constant from constant pool
             }
             case LDC2_W: {
-                unsigned index = (code[++i] << 8) | code[++i];
+                unsigned index = readShortIndex(code, i);
                 // FIXME: load constant from constant pool
             }
             case ILOAD:
@@ -106,7 +125,7 @@
             case DLOAD:
             case ALOAD: {
                 // FIXME: use opcodes to perform type checking
-                unsigned index = code[++i];
+                unsigned index = readByteIndex(code, i);
                 opStack.push(locals[index]);
                 break;
             }




From alkis at cs.uiuc.edu  Sat May 22 18:27:01 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sat May 22 18:27:01 2004
Subject: [llvm-commits] CVS: llvm-java/include/llvm/Java/Compiler.h
Message-ID: <200405222325.SAA07939@zion.cs.uiuc.edu>


Changes in directory llvm-java/include/llvm/Java:

Compiler.h updated: 1.1 -> 1.2

---
Log message:

Put Java::compile free function in a Java::Compiler object.


---
Diffs of the changes:  (+17 -1)

Index: llvm-java/include/llvm/Java/Compiler.h
diff -u llvm-java/include/llvm/Java/Compiler.h:1.1 llvm-java/include/llvm/Java/Compiler.h:1.2
--- llvm-java/include/llvm/Java/Compiler.h:1.1	Sat May 22 00:58:26 2004
+++ llvm-java/include/llvm/Java/Compiler.h	Sat May 22 18:25:22 2004
@@ -15,10 +15,26 @@
 #define LLVM_JAVA_COMPILER_H
 
 #include <llvm/Module.h>
+#include <stack>
+#include <vector>
 
 namespace llvm { namespace Java {
 
-    Module* compile(const ClassFile& cf);
+    class Compiler {
+    public:
+        Module* compile(const ClassFile& cf);
+
+    private:
+        void compileMethod(Module& module, const Method& method);
+        void initForCode(const CodeAttribute& codeAttr);
+
+    private:
+        typedef std::stack<Value*, std::vector<Value*> > OperandStack;
+        typedef std::vector<Value*> Locals;
+
+        OperandStack opStack_;
+        Locals locals_;
+    };
 
 } } // namespace llvm::Java
 




From alkis at cs.uiuc.edu  Sat May 22 18:27:06 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sat May 22 18:27:06 2004
Subject: [llvm-commits] CVS: llvm-java/tools/class2llvm/class2llvm.cpp
Message-ID: <200405222325.SAA07927@zion.cs.uiuc.edu>


Changes in directory llvm-java/tools/class2llvm:

class2llvm.cpp updated: 1.1 -> 1.2

---
Log message:

Put Java::compile free function in a Java::Compiler object.


---
Diffs of the changes:  (+3 -1)

Index: llvm-java/tools/class2llvm/class2llvm.cpp
diff -u llvm-java/tools/class2llvm/class2llvm.cpp:1.1 llvm-java/tools/class2llvm/class2llvm.cpp:1.2
--- llvm-java/tools/class2llvm/class2llvm.cpp:1.1	Sat May 22 02:08:31 2004
+++ llvm-java/tools/class2llvm/class2llvm.cpp	Sat May 22 18:25:22 2004
@@ -32,7 +32,9 @@
         std::auto_ptr<Java::ClassFile> cf(
             Java::ClassFile::readClassFile(std::cin));
 
-        Module* module = Java::compile(*cf);
+        Java::Compiler compiler;
+
+        Module* module = compiler.compile(*cf);
 
         PassManager passes;
         passes.add(new PrintModulePass(&std::cout));




From alkis at cs.uiuc.edu  Sat May 22 18:27:11 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sat May 22 18:27:11 2004
Subject: [llvm-commits] CVS: llvm-java/lib/Compiler/Compiler.cpp
Message-ID: <200405222325.SAA07934@zion.cs.uiuc.edu>


Changes in directory llvm-java/lib/Compiler:

Compiler.cpp updated: 1.2 -> 1.3

---
Log message:

Put Java::compile free function in a Java::Compiler object.


---
Diffs of the changes:  (+374 -372)

Index: llvm-java/lib/Compiler/Compiler.cpp
diff -u llvm-java/lib/Compiler/Compiler.cpp:1.2 llvm-java/lib/Compiler/Compiler.cpp:1.3
--- llvm-java/lib/Compiler/Compiler.cpp:1.2	Sat May 22 17:49:14 2004
+++ llvm-java/lib/Compiler/Compiler.cpp	Sat May 22 18:25:22 2004
@@ -19,10 +19,9 @@
 #include <llvm/Instructions.h>
 #include <llvm/Value.h>
 #include <llvm/Type.h>
-#include <stack>
-#include <vector>
 
 using namespace llvm;
+using namespace llvm::Java;
 
 namespace {
 
@@ -50,398 +49,401 @@
         return (code[++i] << 8) | code[++i];
     }
 
-    void compileMethod(Module& module, const Java::Method& method) {
-        using namespace llvm::Java::Opcode;
-
-        Function* function =
-            module.getOrInsertFunction(method.getName()->str(), Type::VoidTy);
+} // namespace
 
-        const Java::CodeAttribute* codeAttr =
-            Java::getCodeAttribute(method.getAttributes());
+void Compiler::initForCode(const CodeAttribute& codeAttr)
+{
+    while (!opStack_.empty())
+        opStack_.pop();
 
-        typedef std::stack<Value*> OperandStack;
-        OperandStack opStack;
-        typedef std::vector<Value*> Locals;
-        Locals locals(codeAttr->getMaxLocals(), 0);
-
-        const uint8_t* code = codeAttr->getCode();
-        for (unsigned i = 0; i < codeAttr->getCodeSize(); ++i) {
-            unsigned bcStart = i;
-            bool wide = code[i] == WIDE;
-            i += wide;
-            switch (code[i]) {
-            case ACONST_NULL:
-                // FIXME: should push a null pointer of type Object*
-                opStack.push(
-                    ConstantPointerNull::get(PointerType::get(Type::VoidTy)));
-                break;
-            case ICONST_M1:
-            case ICONST_0:
-            case ICONST_1:
-            case ICONST_2:
-            case ICONST_3:
-            case ICONST_4:
-            case ICONST_5:
-                opStack.push(ConstantInt::get(Type::IntTy, code[i]-ICONST_0));
-                break;
-            case LCONST_0:
-            case LCONST_1:
-                opStack.push(ConstantInt::get(Type::LongTy, code[i]-LCONST_0));
-                break;
-            case FCONST_0:
-            case FCONST_1:
-            case FCONST_2:
-                opStack.push(ConstantFP::get(Type::FloatTy, code[i]-FCONST_0));
-                break;
-            case DCONST_0:
-            case DCONST_1:
-                opStack.push(ConstantFP::get(Type::DoubleTy, code[i]-DCONST_0));
-                break;
-            case BIPUSH: {
-                int imm = readByteImmediate(code, i);
-                opStack.push(ConstantInt::get(Type::IntTy, imm));
-                break;
-            }
-            case SIPUSH: {
-                int imm = readShortImmediate(code, i);
-                opStack.push(ConstantInt::get(Type::IntTy, imm));
-                break;
-            }
-            case LDC: {
-                unsigned index = readByteIndex(code, i);
-                // FIXME: load constant from constant pool
-            }
-            case LDC_W: {
-                unsigned index = readShortIndex(code, i);
-                // FIXME: load constant from constant pool
-            }
-            case LDC2_W: {
-                unsigned index = readShortIndex(code, i);
-                // FIXME: load constant from constant pool
-            }
-            case ILOAD:
-            case LLOAD:
-            case FLOAD:
-            case DLOAD:
-            case ALOAD: {
-                // FIXME: use opcodes to perform type checking
-                unsigned index = readByteIndex(code, i);
-                opStack.push(locals[index]);
-                break;
-            }
-            case ILOAD_0:
-            case ILOAD_1:
-            case ILOAD_2:
-            case ILOAD_3:
-                opStack.push(locals[code[i]-ILOAD_0]);
-                break;
-            case LLOAD_0:
-            case LLOAD_1:
-            case LLOAD_2:
-            case LLOAD_3:
-                opStack.push(locals[code[i]-LLOAD_0]);
-                break;
-            case FLOAD_0:
-            case FLOAD_1:
-            case FLOAD_2:
-            case FLOAD_3:
-                opStack.push(locals[code[i]-FLOAD_0]);
-                break;
-            case DLOAD_0:
-            case DLOAD_1:
-            case DLOAD_2:
-            case DLOAD_3:
-                opStack.push(locals[code[i]-DLOAD_0]);
-                break;
-            case ALOAD_0:
-            case ALOAD_1:
-            case ALOAD_2:
-            case ALOAD_3:
-                opStack.push(locals[code[i]-ALOAD_0]);
-                break;
-            case IALOAD:
-            case LALOAD:
-            case FALOAD:
-            case DALOAD:
-            case AALOAD:
-            case BALOAD:
-            case CALOAD:
-            case SALOAD:
-            case ISTORE:
-            case LSTORE:
-            case FSTORE:
-            case DSTORE:
-            case ASTORE:
-            case ISTORE_0:
-            case ISTORE_1:
-            case ISTORE_2:
-            case ISTORE_3:
-            case LSTORE_0:
-            case LSTORE_1:
-            case LSTORE_2:
-            case LSTORE_3:
-            case FSTORE_0:
-            case FSTORE_1:
-            case FSTORE_2:
-            case FSTORE_3:
-            case DSTORE_0:
-            case DSTORE_1:
-            case DSTORE_2:
-            case DSTORE_3:
-            case ASTORE_0:
-            case ASTORE_1:
-            case ASTORE_2:
-            case ASTORE_3:
-            case IASTORE:
-            case LASTORE:
-            case FASTORE:
-            case DASTORE:
-            case AASTORE:
-            case BASTORE:
-            case CASTORE:
-            case SASTORE:
-                assert(0 && "not implemented");
-            case POP:
-                opStack.pop();
-                break;
-            case POP2:
-                opStack.pop();
-                opStack.pop();
-                break;
-            case DUP:
-                opStack.push(opStack.top());
-                break;
-            case DUP_X1: {
-                Value* v1 = opStack.top(); opStack.pop();
-                Value* v2 = opStack.top(); opStack.pop();
-                opStack.push(v1);
-                opStack.push(v2);
-                opStack.push(v1);
-                break;
+    locals_.clear();
+    locals_.assign(codeAttr.getMaxLocals(), NULL);
+}
+
+void Compiler::compileMethod(Module& module, const Java::Method& method) {
+    using namespace llvm::Java::Opcode;
+
+    const Java::CodeAttribute* codeAttr =
+        Java::getCodeAttribute(method.getAttributes());
+
+    initForCode(*codeAttr);
+
+    Function* function =
+        module.getOrInsertFunction(method.getName()->str(), Type::VoidTy);
+
+    const uint8_t* code = codeAttr->getCode();
+    for (unsigned i = 0; i < codeAttr->getCodeSize(); ++i) {
+        unsigned bcStart = i;
+        bool wide = code[i] == WIDE;
+        i += wide;
+        switch (code[i]) {
+        case ACONST_NULL:
+            // FIXME: should push a null pointer of type Object*
+            opStack_.push(
+                ConstantPointerNull::get(PointerType::get(Type::VoidTy)));
+            break;
+        case ICONST_M1:
+        case ICONST_0:
+        case ICONST_1:
+        case ICONST_2:
+        case ICONST_3:
+        case ICONST_4:
+        case ICONST_5:
+            opStack_.push(ConstantInt::get(Type::IntTy, code[i]-ICONST_0));
+            break;
+        case LCONST_0:
+        case LCONST_1:
+            opStack_.push(ConstantInt::get(Type::LongTy, code[i]-LCONST_0));
+            break;
+        case FCONST_0:
+        case FCONST_1:
+        case FCONST_2:
+            opStack_.push(ConstantFP::get(Type::FloatTy, code[i]-FCONST_0));
+            break;
+        case DCONST_0:
+        case DCONST_1:
+            opStack_.push(ConstantFP::get(Type::DoubleTy, code[i]-DCONST_0));
+            break;
+        case BIPUSH: {
+            int imm = readByteImmediate(code, i);
+            opStack_.push(ConstantInt::get(Type::IntTy, imm));
+            break;
+        }
+        case SIPUSH: {
+            int imm = readShortImmediate(code, i);
+            opStack_.push(ConstantInt::get(Type::IntTy, imm));
+            break;
+        }
+        case LDC: {
+            unsigned index = readByteIndex(code, i);
+            // FIXME: load constant from constant pool
+        }
+        case LDC_W: {
+            unsigned index = readShortIndex(code, i);
+            // FIXME: load constant from constant pool
+        }
+        case LDC2_W: {
+            unsigned index = readShortIndex(code, i);
+            // FIXME: load constant from constant pool
+        }
+        case ILOAD:
+        case LLOAD:
+        case FLOAD:
+        case DLOAD:
+        case ALOAD: {
+            // FIXME: use opcodes to perform type checking
+            unsigned index = readByteIndex(code, i);
+            opStack_.push(locals_[index]);
+            break;
+        }
+        case ILOAD_0:
+        case ILOAD_1:
+        case ILOAD_2:
+        case ILOAD_3:
+            opStack_.push(locals_[code[i]-ILOAD_0]);
+            break;
+        case LLOAD_0:
+        case LLOAD_1:
+        case LLOAD_2:
+        case LLOAD_3:
+            opStack_.push(locals_[code[i]-LLOAD_0]);
+            break;
+        case FLOAD_0:
+        case FLOAD_1:
+        case FLOAD_2:
+        case FLOAD_3:
+            opStack_.push(locals_[code[i]-FLOAD_0]);
+            break;
+        case DLOAD_0:
+        case DLOAD_1:
+        case DLOAD_2:
+        case DLOAD_3:
+            opStack_.push(locals_[code[i]-DLOAD_0]);
+            break;
+        case ALOAD_0:
+        case ALOAD_1:
+        case ALOAD_2:
+        case ALOAD_3:
+            opStack_.push(locals_[code[i]-ALOAD_0]);
+            break;
+        case IALOAD:
+        case LALOAD:
+        case FALOAD:
+        case DALOAD:
+        case AALOAD:
+        case BALOAD:
+        case CALOAD:
+        case SALOAD:
+        case ISTORE:
+        case LSTORE:
+        case FSTORE:
+        case DSTORE:
+        case ASTORE:
+        case ISTORE_0:
+        case ISTORE_1:
+        case ISTORE_2:
+        case ISTORE_3:
+        case LSTORE_0:
+        case LSTORE_1:
+        case LSTORE_2:
+        case LSTORE_3:
+        case FSTORE_0:
+        case FSTORE_1:
+        case FSTORE_2:
+        case FSTORE_3:
+        case DSTORE_0:
+        case DSTORE_1:
+        case DSTORE_2:
+        case DSTORE_3:
+        case ASTORE_0:
+        case ASTORE_1:
+        case ASTORE_2:
+        case ASTORE_3:
+        case IASTORE:
+        case LASTORE:
+        case FASTORE:
+        case DASTORE:
+        case AASTORE:
+        case BASTORE:
+        case CASTORE:
+        case SASTORE:
+            assert(0 && "not implemented");
+        case POP:
+            opStack_.pop();
+            break;
+        case POP2:
+            opStack_.pop();
+            opStack_.pop();
+            break;
+        case DUP:
+            opStack_.push(opStack_.top());
+            break;
+        case DUP_X1: {
+            Value* v1 = opStack_.top(); opStack_.pop();
+            Value* v2 = opStack_.top(); opStack_.pop();
+            opStack_.push(v1);
+            opStack_.push(v2);
+            opStack_.push(v1);
+            break;
+        }
+        case DUP_X2: {
+            Value* v1 = opStack_.top(); opStack_.pop();
+            Value* v2 = opStack_.top(); opStack_.pop();
+            if (isOneSlotValue(v2)) {
+                Value* v3 = opStack_.top(); opStack_.pop();
+                opStack_.push(v1);
+                opStack_.push(v3);
+                opStack_.push(v2);
+                opStack_.push(v1);
+            }
+            else {
+                opStack_.push(v1);
+                opStack_.push(v2);
+                opStack_.push(v1);
             }
-            case DUP_X2: {
-                Value* v1 = opStack.top(); opStack.pop();
-                Value* v2 = opStack.top(); opStack.pop();
-                if (isOneSlotValue(v2)) {
-                    Value* v3 = opStack.top(); opStack.pop();
-                    opStack.push(v1);
-                    opStack.push(v3);
-                    opStack.push(v2);
-                    opStack.push(v1);
-                }
-                else {
-                    opStack.push(v1);
-                    opStack.push(v2);
-                    opStack.push(v1);
-                }
-                break;
+            break;
+        }
+        case DUP2: {
+            Value* v1 = opStack_.top(); opStack_.pop();
+            if (isOneSlotValue(v1)) {
+                Value* v2 = opStack_.top(); opStack_.pop();
+                opStack_.push(v2);
+                opStack_.push(v1);
+                opStack_.push(v2);
+                opStack_.push(v1);
+            }
+            else {
+                opStack_.push(v1);
+                opStack_.push(v1);
             }
-            case DUP2: {
-                Value* v1 = opStack.top(); opStack.pop();
-                if (isOneSlotValue(v1)) {
-                    Value* v2 = opStack.top(); opStack.pop();
-                    opStack.push(v2);
-                    opStack.push(v1);
-                    opStack.push(v2);
-                    opStack.push(v1);
-                }
-                else {
-                    opStack.push(v1);
-                    opStack.push(v1);
-                }
-                break;
+            break;
+        }
+        case DUP2_X1: {
+            Value* v1 = opStack_.top(); opStack_.pop();
+            Value* v2 = opStack_.top(); opStack_.pop();
+            if (isOneSlotValue(v1)) {
+                Value* v3 = opStack_.top(); opStack_.pop();
+                opStack_.push(v2);
+                opStack_.push(v1);
+                opStack_.push(v3);
+                opStack_.push(v2);
+                opStack_.push(v1);
+            }
+            else {
+                opStack_.push(v1);
+                opStack_.push(v2);
+                opStack_.push(v1);
             }
-            case DUP2_X1: {
-                Value* v1 = opStack.top(); opStack.pop();
-                Value* v2 = opStack.top(); opStack.pop();
-                if (isOneSlotValue(v1)) {
-                    Value* v3 = opStack.top(); opStack.pop();
-                    opStack.push(v2);
-                    opStack.push(v1);
-                    opStack.push(v3);
-                    opStack.push(v2);
-                    opStack.push(v1);
+            break;
+        }
+        case DUP2_X2: {
+            Value* v1 = opStack_.top(); opStack_.pop();
+            Value* v2 = opStack_.top(); opStack_.pop();
+            if (isOneSlotValue(v1)) {
+                Value* v3 = opStack_.top(); opStack_.pop();
+                if (isOneSlotValue(v3)) {
+                    Value* v4 = opStack_.top(); opStack_.pop();
+                    opStack_.push(v2);
+                    opStack_.push(v1);
+                    opStack_.push(v4);
+                    opStack_.push(v3);
+                    opStack_.push(v2);
+                    opStack_.push(v1);
                 }
                 else {
-                    opStack.push(v1);
-                    opStack.push(v2);
-                    opStack.push(v1);
+                    opStack_.push(v2);
+                    opStack_.push(v1);
+                    opStack_.push(v3);
+                    opStack_.push(v2);
+                    opStack_.push(v1);
                 }
-                break;
             }
-            case DUP2_X2: {
-                Value* v1 = opStack.top(); opStack.pop();
-                Value* v2 = opStack.top(); opStack.pop();
-                if (isOneSlotValue(v1)) {
-                    Value* v3 = opStack.top(); opStack.pop();
-                    if (isOneSlotValue(v3)) {
-                        Value* v4 = opStack.top(); opStack.pop();
-                        opStack.push(v2);
-                        opStack.push(v1);
-                        opStack.push(v4);
-                        opStack.push(v3);
-                        opStack.push(v2);
-                        opStack.push(v1);
-                    }
-                    else {
-                        opStack.push(v2);
-                        opStack.push(v1);
-                        opStack.push(v3);
-                        opStack.push(v2);
-                        opStack.push(v1);
-                    }
+            else {
+                if (isOneSlotValue(v2)) {
+                    Value* v3 = opStack_.top(); opStack_.pop();
+                    opStack_.push(v1);
+                    opStack_.push(v3);
+                    opStack_.push(v2);
+                    opStack_.push(v1);
                 }
                 else {
-                    if (isOneSlotValue(v2)) {
-                        Value* v3 = opStack.top(); opStack.pop();
-                        opStack.push(v1);
-                        opStack.push(v3);
-                        opStack.push(v2);
-                        opStack.push(v1);
-                    }
-                    else {
-                        opStack.push(v1);
-                        opStack.push(v2);
-                        opStack.push(v1);
-                    }
+                    opStack_.push(v1);
+                    opStack_.push(v2);
+                    opStack_.push(v1);
                 }
-                break;
-            }
-            case SWAP: {
-                Value* v1 = opStack.top(); opStack.pop();
-                Value* v2 = opStack.top(); opStack.pop();
-                opStack.push(v1);
-                opStack.push(v2);
-                break;
-            }
-            case IADD:
-            case LADD:
-            case FADD:
-            case DADD:
-            case ISUB:
-            case LSUB:
-            case FSUB:
-            case DSUB:
-            case IMUL:
-            case LMUL:
-            case FMUL:
-            case DMUL:
-            case IDIV:
-            case LDIV:
-            case FDIV:
-            case DDIV:
-            case IREM:
-            case LREM:
-            case FREM:
-            case DREM:
-            case INEG:
-            case LNEG:
-            case FNEG:
-            case DNEG:
-            case ISHL:
-            case LSHL:
-            case ISHR:
-            case LSHR:
-            case IUSHR:
-            case LUSHR:
-            case IAND:
-            case LAND:
-            case IOR:
-            case LOR:
-            case IXOR:
-            case LXOR:
-            case IINC:
-            case I2L:
-            case I2F:
-            case I2D:
-            case L2I:
-            case L2F:
-            case L2D:
-            case F2I:
-            case F2L:
-            case F2D:
-            case D2I:
-            case D2L:
-            case D2F:
-            case I2B:
-            case I2C:
-            case I2S:
-            case LCMP:
-            case FCMPL:
-            case FCMPG:
-            case DCMPL:
-            case DCMPG:
-            case IFEQ:
-            case IFNE:
-            case IFLT:
-            case IFGE:
-            case IFGT:
-            case IFLE:
-            case IF_ICMPEQ:
-            case IF_ICMPNE:
-            case IF_ICMPLT:
-            case IF_ICMPGE:
-            case IF_ICMPGT:
-            case IF_ICMPLE:
-            case IF_ICMPACMPEQ:
-            case IF_ICMPACMPNE:
-            case GOTO:
-            case JSR:
-            case RET:
-            case TABLESWITCH:
-            case LOOKUPSWITCH:
-            case IRETURN:
-            case LRETURN:
-            case FRETURN:
-            case DRETURN:
-            case ARETURN:
-            case RETURN:
-            case GETSTATIC:
-            case PUTSTATIC:
-            case GETFIELD:
-            case PUTFIELD:
-            case INVOKEVIRTUAL:
-            case INVOKESPECIAL:
-            case INVOKESTATIC:
-            case INVOKEINTERFACE:
-            case XXXUNUSEDXXX:
-            case NEW:
-            case NEWARRAY:
-            case ANEWARRAY:
-            case ARRAYLENGTH:
-            case ATHROW:
-            case CHECKCAST:
-            case INSTANCEOF:
-            case MONITORENTER:
-            case MONITOREXIT:
-            case WIDE:
-            case MULTIANEWARRAY:
-            case IFNULL:
-            case IFNONNULL:
-            case GOTO_W:
-            case JSR_W:
-            case BREAKPOINT:
-            case IMPDEP1:
-            case IMPDEP2:
-            case NOP:
-                break;
             }
+            break;
+        }
+        case SWAP: {
+            Value* v1 = opStack_.top(); opStack_.pop();
+            Value* v2 = opStack_.top(); opStack_.pop();
+            opStack_.push(v1);
+            opStack_.push(v2);
+            break;
+        }
+        case IADD:
+        case LADD:
+        case FADD:
+        case DADD:
+        case ISUB:
+        case LSUB:
+        case FSUB:
+        case DSUB:
+        case IMUL:
+        case LMUL:
+        case FMUL:
+        case DMUL:
+        case IDIV:
+        case LDIV:
+        case FDIV:
+        case DDIV:
+        case IREM:
+        case LREM:
+        case FREM:
+        case DREM:
+        case INEG:
+        case LNEG:
+        case FNEG:
+        case DNEG:
+        case ISHL:
+        case LSHL:
+        case ISHR:
+        case LSHR:
+        case IUSHR:
+        case LUSHR:
+        case IAND:
+        case LAND:
+        case IOR:
+        case LOR:
+        case IXOR:
+        case LXOR:
+        case IINC:
+        case I2L:
+        case I2F:
+        case I2D:
+        case L2I:
+        case L2F:
+        case L2D:
+        case F2I:
+        case F2L:
+        case F2D:
+        case D2I:
+        case D2L:
+        case D2F:
+        case I2B:
+        case I2C:
+        case I2S:
+        case LCMP:
+        case FCMPL:
+        case FCMPG:
+        case DCMPL:
+        case DCMPG:
+        case IFEQ:
+        case IFNE:
+        case IFLT:
+        case IFGE:
+        case IFGT:
+        case IFLE:
+        case IF_ICMPEQ:
+        case IF_ICMPNE:
+        case IF_ICMPLT:
+        case IF_ICMPGE:
+        case IF_ICMPGT:
+        case IF_ICMPLE:
+        case IF_ICMPACMPEQ:
+        case IF_ICMPACMPNE:
+        case GOTO:
+        case JSR:
+        case RET:
+        case TABLESWITCH:
+        case LOOKUPSWITCH:
+        case IRETURN:
+        case LRETURN:
+        case FRETURN:
+        case DRETURN:
+        case ARETURN:
+        case RETURN:
+        case GETSTATIC:
+        case PUTSTATIC:
+        case GETFIELD:
+        case PUTFIELD:
+        case INVOKEVIRTUAL:
+        case INVOKESPECIAL:
+        case INVOKESTATIC:
+        case INVOKEINTERFACE:
+        case XXXUNUSEDXXX:
+        case NEW:
+        case NEWARRAY:
+        case ANEWARRAY:
+        case ARRAYLENGTH:
+        case ATHROW:
+        case CHECKCAST:
+        case INSTANCEOF:
+        case MONITORENTER:
+        case MONITOREXIT:
+        case WIDE:
+        case MULTIANEWARRAY:
+        case IFNULL:
+        case IFNONNULL:
+        case GOTO_W:
+        case JSR_W:
+        case BREAKPOINT:
+        case IMPDEP1:
+        case IMPDEP2:
+        case NOP:
+            break;
         }
     }
+}
 
-} // namespace
-
-Module* llvm::Java::compile(const ClassFile& cf)
+Module* Compiler::compile(const ClassFile& cf)
 {
     Module* module = new Module(cf.getThisClass()->getName()->str());
 
-    std::vector<Value*> opStack;
-
     const Java::Methods& methods = cf.getMethods();
     for (Java::Methods::const_iterator
-             i = methods.begin(), e = methods.end(); i != e; ++i) {
+             i = methods.begin(), e = methods.end(); i != e; ++i)
         compileMethod(*module, **i);
-    }
 
     return module;
 }




From gaeke at cs.uiuc.edu  Sat May 22 19:11:01 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Sat May 22 19:11:01 2004
Subject: [llvm-commits] CVS: reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp
Message-ID: <200405230010.TAA03707@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/LightWtProfiling:

UnpackTraceFunction.cpp updated: 1.64 -> 1.65

---
Log message:

Simplify the interface to retrieving register allocation state.
Remove a dead reference to TF->LiveOutSet.


---
Diffs of the changes:  (+13 -9)

Index: reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp
diff -u reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp:1.64 reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp:1.65
--- reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp:1.64	Fri May 21 02:58:43 2004
+++ reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp	Sat May 22 19:09:59 2004
@@ -232,6 +232,12 @@
   regsUsed.insert (fp);
 }
 
+static std::pair<AllocInfo, AllocInfo>
+getValueAllocState (TraceFunction *TF, Value *V) {
+  return std::make_pair (getValueAllocStateFromModule (TF->MatrixFn, V),
+    getValueAllocStateFromGlobal (TF->TraceFn, TF->getCorrespondingValue (V)));
+}
+
 void UnpackTraceFunction::rewriteProlog (MachineFunction &MF,
                                          MachineBasicBlock &E) {
   const TargetRegInfo &TRI = TM->getRegInfo ();
@@ -297,7 +303,7 @@
   // 4. Insert copies from each live-in variable's reg. in the matrix fn.
   // to its reg. in the trace.
   Function *TraceF = TF->TraceFn, *MatrixF = TF->MatrixFn;
-  LiveVariableSet &Si = TF->LiveInSet, &So = TF->LiveOutSet;
+  LiveVariableSet &Si = TF->LiveInSet;
   std::vector<CopyInfo> EntryCopies;
   for (LiveVariableSet::iterator SI = Si.begin (), SE = Si.end (); SI != SE;
        ++SI) {
@@ -306,11 +312,10 @@
           << "rewriteProlog: Getting alloc state for live-in value ";
           WriteAsOperand (std::cerr, V, true, true, MatrixF->getParent ());
           std::cerr << "\n");
-    AllocInfo Source = getValueAllocStateFromModule (MatrixF, V),
-              Target = getValueAllocStateFromGlobal (TraceF,
-                         TF->getCorrespondingValue (V));
-    DEBUG(std::cerr << "rewriteProlog: Source = " << Source << "\n"
-                    << "rewriteProlog: Target = " << Target << "\n");
+    std::pair<AllocInfo, AllocInfo> ai = getValueAllocState (TF, V);
+    AllocInfo &Source = ai.first, &Target = ai.second;
+    DEBUG(std::cerr << "rewriteProlog: Source in MatrixFn = " << Source << "\n"
+                    << "rewriteProlog: Target in TraceFn = " << Target << "\n");
     if (Source != Target)
       EntryCopies.push_back (CopyInfo (Source, Target, &E, V->getType ()));
   }
@@ -495,9 +500,8 @@
   for (LiveVariableSet::iterator SI = So.begin (), SE = So.end ();
 	   SI != SE; ++SI) {
 	Value *V = *SI;
-	AllocInfo Source =
-	  getValueAllocStateFromGlobal (TraceF, TF->getCorrespondingValue (V));
-	AllocInfo Target = getValueAllocStateFromModule (MatrixF, V);
+    std::pair<AllocInfo, AllocInfo> ai = getValueAllocState (TF, V);
+    AllocInfo &Target = ai.first, &Source = ai.second;
 	if (Source != Target)
 	  insertCopyMachineInstrs (Source, Target, MBB, V->getType ());
   }




From alkis at cs.uiuc.edu  Sat May 22 19:20:02 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sat May 22 19:20:02 2004
Subject: [llvm-commits] CVS: llvm-java/lib/Compiler/Compiler.cpp
Message-ID: <200405230018.TAA08130@zion.cs.uiuc.edu>


Changes in directory llvm-java/lib/Compiler:

Compiler.cpp updated: 1.3 -> 1.4

---
Log message:

Rename utility functions for reading immediates from bytecode. Add
functions to read ints as well.


---
Diffs of the changes:  (+24 -13)

Index: llvm-java/lib/Compiler/Compiler.cpp
diff -u llvm-java/lib/Compiler/Compiler.cpp:1.3 llvm-java/lib/Compiler/Compiler.cpp:1.4
--- llvm-java/lib/Compiler/Compiler.cpp:1.3	Sat May 22 18:25:22 2004
+++ llvm-java/lib/Compiler/Compiler.cpp	Sat May 22 19:18:36 2004
@@ -33,20 +33,31 @@
         return !isTwoSlotValue(v);
     }
 
-    inline int readByteImmediate(const uint8_t* code, unsigned& i) {
+    inline int readByteSigned(const uint8_t* code, unsigned& i) {
         return code[++i];
     }
 
-    inline int readShortImmediate(const uint8_t* code, unsigned& i) {
-        return (code[++i] << 8) | code[++i];
+    inline unsigned readByteUnsigned(const uint8_t* code, unsigned& i) {
+        return code[++i];
     }
 
-    inline unsigned readByteIndex(const uint8_t* code, unsigned& i) {
-        return code[++i];
+    inline int readShortSigned(const uint8_t* code, unsigned& i) {
+        return (readByteSigned(code, i) << 8) | readByteUnsigned(code, i);
+    }
+
+    inline unsigned readShortUnsigned(const uint8_t* code, unsigned& i) {
+        return (readByteUnsigned(code, i) << 8) | readByteUnsigned(code, i);
+    }
+
+    inline int readIntSigned(const uint8_t* code, unsigned& i) {
+        return ((readByteUnsigned(code, i) << 24) |
+                (readByteUnsigned(code, i) << 16) |
+                (readByteUnsigned(code, i) << 8) |
+                readByteUnsigned(code, i));
     }
 
-    inline unsigned readShortIndex(const uint8_t* code, unsigned& i) {
-        return (code[++i] << 8) | code[++i];
+    inline unsigned readIntUnsigned(const uint8_t* code, unsigned& i) {
+        return readIntSigned(code, i);
     }
 
 } // namespace
@@ -105,25 +116,25 @@
             opStack_.push(ConstantFP::get(Type::DoubleTy, code[i]-DCONST_0));
             break;
         case BIPUSH: {
-            int imm = readByteImmediate(code, i);
+            int imm = readByteSigned(code, i);
             opStack_.push(ConstantInt::get(Type::IntTy, imm));
             break;
         }
         case SIPUSH: {
-            int imm = readShortImmediate(code, i);
+            int imm = readShortSigned(code, i);
             opStack_.push(ConstantInt::get(Type::IntTy, imm));
             break;
         }
         case LDC: {
-            unsigned index = readByteIndex(code, i);
+            unsigned index = readByteUnsigned(code, i);
             // FIXME: load constant from constant pool
         }
         case LDC_W: {
-            unsigned index = readShortIndex(code, i);
+            unsigned index = readShortUnsigned(code, i);
             // FIXME: load constant from constant pool
         }
         case LDC2_W: {
-            unsigned index = readShortIndex(code, i);
+            unsigned index = readShortUnsigned(code, i);
             // FIXME: load constant from constant pool
         }
         case ILOAD:
@@ -132,7 +143,7 @@
         case DLOAD:
         case ALOAD: {
             // FIXME: use opcodes to perform type checking
-            unsigned index = readByteIndex(code, i);
+            unsigned index = readByteUnsigned(code, i);
             opStack_.push(locals_[index]);
             break;
         }




From alkis at cs.uiuc.edu  Sat May 22 20:18:02 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sat May 22 20:18:02 2004
Subject: [llvm-commits] CVS: llvm-java/lib/Compiler/Compiler.cpp
Message-ID: <200405230116.UAA08351@zion.cs.uiuc.edu>


Changes in directory llvm-java/lib/Compiler:

Compiler.cpp updated: 1.4 -> 1.5

---
Log message:

Rename initForCode() to methodCompileInit().
Create a bytecode to llvm::BasicBlock* map in a separate pass before
compilation.


---
Diffs of the changes:  (+130 -5)

Index: llvm-java/lib/Compiler/Compiler.cpp
diff -u llvm-java/lib/Compiler/Compiler.cpp:1.4 llvm-java/lib/Compiler/Compiler.cpp:1.5
--- llvm-java/lib/Compiler/Compiler.cpp:1.4	Sat May 22 19:18:36 2004
+++ llvm-java/lib/Compiler/Compiler.cpp	Sat May 22 20:15:53 2004
@@ -19,6 +19,7 @@
 #include <llvm/Instructions.h>
 #include <llvm/Value.h>
 #include <llvm/Type.h>
+#include <Support/StringExtras.h>
 
 using namespace llvm;
 using namespace llvm::Java;
@@ -60,27 +61,151 @@
         return readIntSigned(code, i);
     }
 
+    inline void skipPadBytes(const uint8_t* code, unsigned& i) {
+        while (((unsigned)&code[++i]) & 0XFF);
+    }
+
 } // namespace
 
-void Compiler::initForCode(const CodeAttribute& codeAttr)
+void Compiler::compileMethodInit(Function& function,
+                                 const CodeAttribute& codeAttr)
 {
     while (!opStack_.empty())
         opStack_.pop();
 
     locals_.clear();
     locals_.assign(codeAttr.getMaxLocals(), NULL);
+
+    bc2bbMap_.clear();
+    bc2bbMap_.assign(codeAttr.getCodeSize(), NULL);
+    bc2bbMap_[0] = new BasicBlock("entry", &function);
+
+    const uint8_t* code = codeAttr.getCode();
+    for (unsigned i = 0; i < codeAttr.getCodeSize(); ++i) {
+        using namespace llvm::Java::Opcode;
+
+        unsigned bcStart = i;
+        bool wide = code[i] == WIDE;
+        i += wide;
+        switch (code[i]) {
+        case BIPUSH:
+        case LDC:
+        case NEWARRAY:
+            ++i;
+            break;
+        case ILOAD:
+        case LLOAD:
+        case FLOAD:
+        case DLOAD:
+        case ALOAD:
+        case ISTORE:
+        case LSTORE:
+        case FSTORE:
+        case DSTORE:
+        case ASTORE:
+        case RET:
+            i += 1 + wide;
+            break;
+        case SIPUSH:
+        case LDC_W:
+        case LDC2_W:
+        case GOTO:
+        case JSR:
+        case GETSTATIC:
+        case PUTSTATIC:
+        case GETFIELD:
+        case PUTFIELD:
+        case INVOKEVIRTUAL:
+        case INVOKESPECIAL:
+        case INVOKESTATIC:
+        case INVOKEINTERFACE:
+        case NEW:
+        case ANEWARRAY:
+        case ARRAYLENGTH:
+        case ATHROW:
+        case CHECKCAST:
+        case INSTANCEOF:
+            i += 2;
+            break;
+        case IFEQ:
+        case IFNE:
+        case IFLT:
+        case IFGE:
+        case IFGT:
+        case IFLE:
+        case IF_ICMPEQ:
+        case IF_ICMPNE:
+        case IF_ICMPLT:
+        case IF_ICMPGE:
+        case IF_ICMPGT:
+        case IF_ICMPLE:
+        case IF_ICMPACMPEQ:
+        case IF_ICMPACMPNE:
+        case IFNULL:
+        case IFNONNULL: {
+            unsigned index = readShortUnsigned(code, i);
+            bc2bbMap_[bcStart] = new BasicBlock(
+                std::string("bb at bc") + utostr(bcStart), &function);
+            break;
+        }
+        case TABLESWITCH: {
+            skipPadBytes(code, i);
+            readIntSigned(code, i);
+            int low = readIntSigned(code, i);
+            int high = readIntSigned(code, i);
+            unsigned offsetCount = high - low + 1;
+            while (offsetCount--) {
+                unsigned bcIndex = bcStart + readIntSigned(code, i);
+                bc2bbMap_[bcIndex] = new BasicBlock(
+                    std::string("bb at bc") + utostr(bcIndex), &function);
+            }
+            break;
+        }
+        case LOOKUPSWITCH: {
+            skipPadBytes(code, i);
+            unsigned pairCount = readIntUnsigned(code, i);
+            while (pairCount--) {
+                readIntSigned(code, i);
+                unsigned bcIndex = bcStart + readIntSigned(code, i);
+                bc2bbMap_[bcIndex] = new BasicBlock(
+                    std::string("bb at bc") + utostr(bcIndex), &function);
+            }
+            break;
+        }
+        case XXXUNUSEDXXX:
+            throw "FIXME: create new exception class";
+        case MULTIANEWARRAY:
+            i += 3;
+            break;
+        case GOTO_W:
+        case JSR_W:
+            i+= 4;
+            break;
+        default:
+            break;
+        }
+    }
+
+    unsigned i = 0;
+    BasicBlock* bb = bc2bbMap_[i];
+    while (++i < codeAttr.getCodeSize()) {
+        if (bc2bbMap_[i])
+            bc2bbMap_[i] = bb;
+        else
+            bb = bc2bbMap_[i];
+    }
 }
 
 void Compiler::compileMethod(Module& module, const Java::Method& method) {
     using namespace llvm::Java::Opcode;
 
+    Function* function =
+        module.getOrInsertFunction(method.getName()->str(), Type::VoidTy);
+
     const Java::CodeAttribute* codeAttr =
         Java::getCodeAttribute(method.getAttributes());
 
-    initForCode(*codeAttr);
-
-    Function* function =
-        module.getOrInsertFunction(method.getName()->str(), Type::VoidTy);
+    compileMethodInit(*function, *codeAttr);
 
     const uint8_t* code = codeAttr->getCode();
     for (unsigned i = 0; i < codeAttr->getCodeSize(); ++i) {




From alkis at cs.uiuc.edu  Sat May 22 20:18:09 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sat May 22 20:18:09 2004
Subject: [llvm-commits] CVS: llvm-java/include/llvm/Java/Compiler.h
Message-ID: <200405230116.UAA08356@zion.cs.uiuc.edu>


Changes in directory llvm-java/include/llvm/Java:

Compiler.h updated: 1.2 -> 1.3

---
Log message:

Rename initForCode() to methodCompileInit().
Create a bytecode to llvm::BasicBlock* map in a separate pass before
compilation.


---
Diffs of the changes:  (+4 -1)

Index: llvm-java/include/llvm/Java/Compiler.h
diff -u llvm-java/include/llvm/Java/Compiler.h:1.2 llvm-java/include/llvm/Java/Compiler.h:1.3
--- llvm-java/include/llvm/Java/Compiler.h:1.2	Sat May 22 18:25:22 2004
+++ llvm-java/include/llvm/Java/Compiler.h	Sat May 22 20:15:53 2004
@@ -25,15 +25,18 @@
         Module* compile(const ClassFile& cf);
 
     private:
+        void compileMethodInit(Function& function,
+                               const CodeAttribute& codeAttr);
         void compileMethod(Module& module, const Method& method);
-        void initForCode(const CodeAttribute& codeAttr);
 
     private:
         typedef std::stack<Value*, std::vector<Value*> > OperandStack;
         typedef std::vector<Value*> Locals;
+        typedef std::vector<BasicBlock*> BC2BBMap;
 
         OperandStack opStack_;
         Locals locals_;
+        BC2BBMap bc2bbMap_;
     };
 
 } } // namespace llvm::Java




From alkis at cs.uiuc.edu  Sat May 22 20:26:05 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sat May 22 20:26:05 2004
Subject: [llvm-commits] CVS: llvm-java/lib/Compiler/Compiler.cpp
Message-ID: <200405230124.UAA08403@zion.cs.uiuc.edu>


Changes in directory llvm-java/lib/Compiler:

Compiler.cpp updated: 1.5 -> 1.6

---
Log message:

Make correct byte skips for local variable increament.


---
Diffs of the changes:  (+3 -0)

Index: llvm-java/lib/Compiler/Compiler.cpp
diff -u llvm-java/lib/Compiler/Compiler.cpp:1.5 llvm-java/lib/Compiler/Compiler.cpp:1.6
--- llvm-java/lib/Compiler/Compiler.cpp:1.5	Sat May 22 20:15:53 2004
+++ llvm-java/lib/Compiler/Compiler.cpp	Sat May 22 20:24:43 2004
@@ -127,6 +127,9 @@
         case INSTANCEOF:
             i += 2;
             break;
+        case IINC:
+            i += 2 * (1 + wide);
+            break;
         case IFEQ:
         case IFNE:
         case IFLT:




From gaeke at cs.uiuc.edu  Sat May 22 21:03:01 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Sat May 22 21:03:01 2004
Subject: [llvm-commits] CVS: reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp
Message-ID: <200405230202.VAA04409@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/LightWtProfiling:

UnpackTraceFunction.cpp updated: 1.65 -> 1.66

---
Log message:

Save live-in values that originate in MatrixFn's register on TraceFn's stack.
(Still need to reload them into TraceFn's registers from there.)
Remove some dead references from rewriteProlog().


---
Diffs of the changes:  (+25 -9)

Index: reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp
diff -u reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp:1.65 reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp:1.66
--- reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp:1.65	Sat May 22 19:09:59 2004
+++ reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp	Sat May 22 21:02:13 2004
@@ -270,11 +270,30 @@
   E.push_back (BuildMI (V9::ADDi, 3).addMReg (sp).addSImm (-Size).addMReg (sp,
     MachineOperand::Def));
 
+  std::set<unsigned> regsToSave (regsUsed);
+  // Save live-in values in registers onto the stack
+  LiveVariableSet &Si = TF->LiveInSet;
+  std::map<Value *, std::pair<AllocInfo, AllocInfo> > AllocStates;
+  for (LiveVariableSet::iterator SI = Si.begin (), SE = Si.end (); SI != SE;
+       ++SI) {
+    Value *V = *SI;
+    std::pair<Value *, std::pair<AllocInfo, AllocInfo> > as =
+      std::make_pair (V, getValueAllocState (TF, V));
+    AllocStates.insert (as);
+    AllocInfo &Source = as.second.first, &Target = as.second.second;
+    if (Source != Target)
+      if (Source.AllocState == AllocInfo::Allocated) {
+        DEBUG (std::cerr << "Need to save incoming live value in reg "
+               << Source.Placement << " onto the stack\n");
+        regsToSave.insert (Source.Placement);
+      }
+  }
+
   // 3. Save used registers onto the stack.
   std::vector<MachineInstr *> mvec;
   int RegType;
-  for (std::set<unsigned>::iterator i = regsUsed.begin (), e = regsUsed.end ();
-    i != e; ++i) {
+  for (std::set<unsigned>::iterator i = regsToSave.begin (),
+       e = regsToSave.end (); i != e; ++i) {
     mvec.clear ();
     unsigned R = *i;
     unsigned RegType = TRI.getRegType (R);
@@ -300,25 +319,22 @@
     }
   }
 
+
   // 4. Insert copies from each live-in variable's reg. in the matrix fn.
   // to its reg. in the trace.
-  Function *TraceF = TF->TraceFn, *MatrixF = TF->MatrixFn;
-  LiveVariableSet &Si = TF->LiveInSet;
+  Function *MatrixF = TF->MatrixFn;
   std::vector<CopyInfo> EntryCopies;
   for (LiveVariableSet::iterator SI = Si.begin (), SE = Si.end (); SI != SE;
        ++SI) {
     Value *V = *SI;
-    DEBUG(std::cerr
-          << "rewriteProlog: Getting alloc state for live-in value ";
-          WriteAsOperand (std::cerr, V, true, true, MatrixF->getParent ());
-          std::cerr << "\n");
-    std::pair<AllocInfo, AllocInfo> ai = getValueAllocState (TF, V);
+    std::pair<AllocInfo, AllocInfo> &ai = AllocStates[V];
     AllocInfo &Source = ai.first, &Target = ai.second;
     DEBUG(std::cerr << "rewriteProlog: Source in MatrixFn = " << Source << "\n"
                     << "rewriteProlog: Target in TraceFn = " << Target << "\n");
     if (Source != Target)
       EntryCopies.push_back (CopyInfo (Source, Target, &E, V->getType ()));
   }
+
   for (std::vector<CopyInfo>::iterator i = EntryCopies.begin (),
          e = EntryCopies.end (); i != e; ++i) {
     DEBUG(std::cerr << "rewriteProlog: Inserting copy " << *i << "\n");




From alkis at cs.uiuc.edu  Sat May 22 21:10:01 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sat May 22 21:10:01 2004
Subject: [llvm-commits] CVS: llvm-java/lib/Compiler/Compiler.cpp
Message-ID: <200405230208.VAA08527@zion.cs.uiuc.edu>


Changes in directory llvm-java/lib/Compiler:

Compiler.cpp updated: 1.6 -> 1.7

---
Log message:

Fix pop2 compilation.


---
Diffs of the changes:  (+5 -3)

Index: llvm-java/lib/Compiler/Compiler.cpp
diff -u llvm-java/lib/Compiler/Compiler.cpp:1.6 llvm-java/lib/Compiler/Compiler.cpp:1.7
--- llvm-java/lib/Compiler/Compiler.cpp:1.6	Sat May 22 20:24:43 2004
+++ llvm-java/lib/Compiler/Compiler.cpp	Sat May 22 21:08:01 2004
@@ -350,10 +350,12 @@
         case POP:
             opStack_.pop();
             break;
-        case POP2:
-            opStack_.pop();
-            opStack_.pop();
+        case POP2: {
+            Value* v1 = opStack_.top(); opStack_.pop();
+            if (isOneSlotValue(v1))
+                opStack_.pop();
             break;
+        }
         case DUP:
             opStack_.push(opStack_.top());
             break;




From gaeke at cs.uiuc.edu  Sun May 23 00:06:02 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Sun May 23 00:06:02 2004
Subject: [llvm-commits] CVS: llvm/include/llvm/CodeGen/MachineInstrBuilder.h
Message-ID: <200405230504.AAA09340@zion.cs.uiuc.edu>


Changes in directory llvm/include/llvm/CodeGen:

MachineInstrBuilder.h updated: 1.21 -> 1.22

---
Log message:

Fix a typo in the head-of-file comment.
Rewrite the comments about the different versions of BuildMI, to
better emphasize their differences.


---
Diffs of the changes:  (+21 -14)

Index: llvm/include/llvm/CodeGen/MachineInstrBuilder.h
diff -u llvm/include/llvm/CodeGen/MachineInstrBuilder.h:1.21 llvm/include/llvm/CodeGen/MachineInstrBuilder.h:1.22
--- llvm/include/llvm/CodeGen/MachineInstrBuilder.h:1.21	Wed Mar 31 22:03:10 2004
+++ llvm/include/llvm/CodeGen/MachineInstrBuilder.h	Sun May 23 00:04:00 2004
@@ -10,7 +10,7 @@
 // This file exposes a function named BuildMI, which is useful for dramatically
 // simplifying how MachineInstr's are created.  Instead of using code like this:
 //
-//   M = new MachineInstr(X86::ADDrr32);
+//   M = new MachineInstr(X86::ADDrr8);
 //   M->SetMachineOperandVal(0, MachineOperand::MO_VirtualRegister, argVal1);
 //   M->SetMachineOperandVal(1, MachineOperand::MO_VirtualRegister, argVal2);
 //
@@ -144,9 +144,9 @@
   return MachineInstrBuilder(new MachineInstr(Opcode, NumOperands, true, true));
 }
 
-/// BuildMI - This version of the builder also sets up the first "operand" as a
+/// BuildMI - This version of the builder sets up the first operand as a
 /// destination virtual register.  NumOperands is the number of additional add*
-/// calls that are expected, it does not include the destination register.
+/// calls that are expected, not including the destination register.
 ///
 inline MachineInstrBuilder BuildMI(
   int Opcode, unsigned NumOperands,
@@ -156,9 +156,12 @@
                                    true, true)).addReg(DestReg, useType);
 }
 
-
-/// BuildMI - Insert the instruction before a specified location in the basic
-/// block.
+/// BuildMI - This version of the builder inserts the newly-built
+/// instruction before the given position in the given MachineBasicBlock, and
+/// sets up the first operand as a destination virtual register.
+/// NumOperands is the number of additional add* calls that are expected,
+/// not including the destination register.
+///
 inline MachineInstrBuilder BuildMI(MachineBasicBlock &BB,
                                    MachineBasicBlock::iterator I,
                                    int Opcode, unsigned NumOperands,
@@ -168,8 +171,10 @@
   return MachineInstrBuilder(MI).addReg(DestReg, MachineOperand::Def);
 }
 
-/// BMI - A special BuildMI variant that takes an iterator to insert the
-/// instruction at as well as a basic block.
+/// BuildMI - This version of the builder inserts the newly-built
+/// instruction before the given position in the given MachineBasicBlock, and
+/// does NOT take a destination register.
+///
 inline MachineInstrBuilder BuildMI(MachineBasicBlock &BB,
                                    MachineBasicBlock::iterator I,
                                    int Opcode, unsigned NumOperands) {
@@ -178,18 +183,20 @@
   return MachineInstrBuilder(MI);
 }
 
-/// BuildMI - This version of the builder inserts the built MachineInstr into
-/// the specified MachineBasicBlock.
+/// BuildMI - This version of the builder inserts the newly-built
+/// instruction at the end of the given MachineBasicBlock, and does NOT take a
+/// destination register.
 ///
 inline MachineInstrBuilder BuildMI(MachineBasicBlock *BB, int Opcode,
                                    unsigned NumOperands) {
   return BuildMI(*BB, BB->end(), Opcode, NumOperands);
 }
 
-/// BuildMI - This version of the builder inserts the built MachineInstr into
-/// the specified MachineBasicBlock, and also sets up the first "operand" as a
-/// destination virtual register.  NumOperands is the number of additional add*
-/// calls that are expected, it does not include the destination register.
+/// BuildMI - This version of the builder inserts the newly-built
+/// instruction at the end of the given MachineBasicBlock, and sets up the first
+/// operand as a destination virtual register. NumOperands is the number of
+/// additional add* calls that are expected, not including the destination
+/// register.
 ///
 inline MachineInstrBuilder BuildMI(MachineBasicBlock *BB, int Opcode,
                                    unsigned NumOperands, unsigned DestReg) {




From gaeke at cs.uiuc.edu  Sun May 23 00:11:03 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Sun May 23 00:11:03 2004
Subject: [llvm-commits] CVS: llvm/lib/Target/SparcV9/SparcV9InstrSelection.cpp
Message-ID: <200405230510.AAA04993@seraph.cs.uiuc.edu>


Changes in directory llvm/lib/Target/SparcV9:

SparcV9InstrSelection.cpp updated: 1.140 -> 1.141

---
Log message:

Fix a typo in a comment.


---
Diffs of the changes:  (+1 -1)

Index: llvm/lib/Target/SparcV9/SparcV9InstrSelection.cpp
diff -u llvm/lib/Target/SparcV9/SparcV9InstrSelection.cpp:1.140 llvm/lib/Target/SparcV9/SparcV9InstrSelection.cpp:1.141
--- llvm/lib/Target/SparcV9/SparcV9InstrSelection.cpp:1.140	Sun Apr 25 02:04:49 2004
+++ llvm/lib/Target/SparcV9/SparcV9InstrSelection.cpp	Sun May 23 00:10:06 2004
@@ -2867,7 +2867,7 @@
           // Make sure not to substitute an upwards-exposed use -- that would
           // introduce a use of `tmpI' with no preceding def.  Therefore,
           // substitute a use or def-and-use operand only if a previous def
-          // operand has already been substituted (i.e., numSusbt > 0).
+          // operand has already been substituted (i.e., numSubst > 0).
           // 
           numSubst += mvec[i]->substituteValue(dest, tmpI,
                                                /*defsOnly*/ numSubst == 0,




From vadve at cs.uiuc.edu  Sun May 23 02:36:02 2004
From: vadve at cs.uiuc.edu (Vikram Adve)
Date: Sun May 23 02:36:02 2004
Subject: [llvm-commits] CVS: llvm/include/llvm/Analysis/DSNode.h
Message-ID: <200405230735.CAA31737@psmith.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Analysis:

DSNode.h updated: 1.41 -> 1.42

---
Log message:

Fix size/offset assertion to allow negative offsets and folded nodes.


---
Diffs of the changes:  (+8 -1)

Index: llvm/include/llvm/Analysis/DSNode.h
diff -u llvm/include/llvm/Analysis/DSNode.h:1.41 llvm/include/llvm/Analysis/DSNode.h:1.42
--- llvm/include/llvm/Analysis/DSNode.h:1.41	Thu Mar 11 17:08:20 2004
+++ llvm/include/llvm/Analysis/DSNode.h	Sun May 23 02:34:53 2004
@@ -386,7 +386,14 @@
 // Define inline DSNodeHandle functions that depend on the definition of DSNode
 //
 inline DSNode *DSNodeHandle::getNode() const {
-  assert((!N || Offset < N->Size || (N->Size == 0 && Offset == 0) ||
+  // Disabling this assertion because it is failing on a "magic" struct
+  // in named (from bind).  The fourth field is an array of length 0,
+  // presumably used to create struct instances of different sizes.
+  assert((!N ||
+          N->isNodeCompletelyFolded() ||
+          (N->Size == 0 && Offset == 0) ||
+          (int(Offset) >= 0 && Offset < N->Size) ||
+          (int(Offset) < 0 && -int(Offset) < int(N->Size)) ||
           N->isForwarding()) && "Node handle offset out of range!");
   if (N == 0 || !N->isForwarding())
     return N;




From gaeke at cs.uiuc.edu  Sun May 23 02:47:01 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Sun May 23 02:47:01 2004
Subject: [llvm-commits] CVS: reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp
Message-ID: <200405230746.CAA15917@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/LightWtProfiling:

UnpackTraceFunction.cpp updated: 1.66 -> 1.67

---
Log message:

Move the calculation of stack sizes and register sets out of rewriteProlog(),
and save the answers in the UnpackTraceFunction object for later use in
rewriteEpilog().

Split regsUsed out into RegsToRestore and RegsToSave.  findRegsUsedInFunction
has been renamed to findRegsToRestore(). I haven't decided the fates of these
sets yet.

Refactor the calculation of stack offsets for saved registers in the trace
function out into stackOffsetForReg().

Do not try to save and restore the stack pointer.

Perform live-in value copies by loading them off the stack where they were
saved.  Do not batch live-in value copies into the EntryCopies vector.
Ignore insertCopyMachineInstrs for now because it is going to screw up
copies from the MatrixFn's stack to the TraceFn's stack.

Other minor edits, primarily concerning debug messages.


---
Diffs of the changes:  (+105 -85)

Index: reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp
diff -u reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp:1.66 reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp:1.67
--- reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp:1.66	Sat May 22 21:02:13 2004
+++ reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp	Sun May 23 02:46:26 2004
@@ -40,12 +40,24 @@
   TargetMachine *TM;
   TraceFunction *TF;
 
+  /// Static stack size of the MachineFunction. Filled in by rewriteProlog ()
+  /// by calling getStaticStackSize (), and then used again by rewriteEpilog ().
+  ///
+  unsigned StaticStackSize;
+
+  /// Total stack size of the MachineFunction, including space for regs
+  /// modified in TraceFn to be saved.
+  ///
+  int TotalStackSize;
+
   /// Registers clobbered in the trace. Filled in by findRegsUsedInFunction ().
   ///
-  std::set<unsigned> regsUsed;
+  std::set<unsigned> RegsToRestore;
+  std::set<unsigned> RegsToSave;
 
   unsigned getStaticStackSize (MachineFunction &MF);
-  void findRegsUsedInFunction (MachineFunction &MF);
+  unsigned stackOffsetForReg (unsigned R);
+  void findRegsToRestore (MachineFunction &MF);
   void insertCopyMachineInstrs (AllocInfo &Source, AllocInfo &Target,
                                 MachineBasicBlock &B, const Type *Ty);
   void insertBranchMachineInstrs (uint64_t Target, MachineBasicBlock &B);
@@ -190,13 +202,13 @@
 static AllocInfo getValueAllocStateFromGlobal (Function *F, Value *V);
 
 /// Fill in the set of registers used in this function, which is kept in
-/// 'regsUsed' in the UnpackTraceFunction Pass object, and is used by
+/// 'RegsToRestore' in the UnpackTraceFunction Pass object, and is used by
 /// rewriteProlog() and rewriteEpilog(). Registers are
 /// represented by their 'unified register numbers' as used in the SPARCv9
 /// back-end.
 ///
 void
-UnpackTraceFunction::findRegsUsedInFunction (MachineFunction &MF) {
+UnpackTraceFunction::findRegsToRestore (MachineFunction &MF) {
   const TargetRegInfo &TRI = TM->getRegInfo ();
   bool intCCRegSeen = false, floatCCRegSeen = false;
   for (MachineFunction::iterator fi = MF.begin (), fe = MF.end ();
@@ -212,8 +224,9 @@
           } else if (TRI.getRegType(regNo) == SparcV9RegInfo::FloatCCRegType
                   || TRI.getRegType(regNo) == SparcV9RegInfo::SpecialRegType) {
             floatCCRegSeen = true;
-          } else if (regNo != SparcV9::g0) { // Ignore defs of the zero register
-            regsUsed.insert (regNo);
+          } else if (regNo != SparcV9::g0 && regNo != SparcV9::o6 /* sp */) {
+            // Defs of certain registers are ignored
+            RegsToRestore.insert (regNo);
           }
         }
       }
@@ -223,13 +236,19 @@
   // If the intcc regs are used, then we only put %ccr in the
   // set, not the individual intcc regs.
   if (floatCCRegSeen)
-    regsUsed.insert (SparcV9::fsr);
+    RegsToRestore.insert (SparcV9::fsr);
   if (intCCRegSeen)
-    regsUsed.insert (SparcV9::ccr);
+    RegsToRestore.insert (SparcV9::ccr);
 
   // Always put fp in the set because it is restored unconditionally.
   static const unsigned fp = SparcV9::i6;
-  regsUsed.insert (fp);
+  RegsToRestore.insert (fp);
+
+  DEBUG(std::cerr << "findRegsToRestore: RegsToRestore (size " 
+        << RegsToRestore.size () << ") contains: (";
+        for (std::set<unsigned>::iterator i = RegsToRestore.begin (),
+             e = RegsToRestore.end (); i != e; ++i) { std::cerr << *i << " "; }
+        std::cerr << " )\n");
 }
 
 static std::pair<AllocInfo, AllocInfo>
@@ -238,6 +257,10 @@
     getValueAllocStateFromGlobal (TF->TraceFn, TF->getCorrespondingValue (V)));
 }
 
+unsigned UnpackTraceFunction::stackOffsetForReg (unsigned R) {
+  return (2047 + StaticStackSize + 176 + (R) * 8);
+}
+
 void UnpackTraceFunction::rewriteProlog (MachineFunction &MF,
                                          MachineBasicBlock &E) {
   const TargetRegInfo &TRI = TM->getRegInfo ();
@@ -253,25 +276,14 @@
   E.push_back (BuildMI (V9::ORr, 3).addMReg (sp).addZImm (0).addMReg (g1,
     MachineOperand::UseAndDef));
 
-  // 1. Initialize regsUsed with the set of registers used in this function.
-  findRegsUsedInFunction (MF);
-
-  unsigned stackSize = getStaticStackSize (MF);
-  DEBUG(std::cerr << "rewriteProlog: Static stack size is "
-                  << stackSize << "\n"
-                  << "rewriteProlog: Reg set (size " << regsUsed.size ()
-                  << ") contains: (");
-  DEBUG(for (std::set<unsigned>::iterator i = regsUsed.begin (),
-             e = regsUsed.end (); i != e; ++i) { std::cerr << *i << " "; });
-  DEBUG(std::cerr << " )\n");
-
-  // 2. Get some stack space: (Stack Frame Size + Slop + Space for Regs).
-  int Size = (stackSize + 176 + 104 * 8);
-  E.push_back (BuildMI (V9::ADDi, 3).addMReg (sp).addSImm (-Size).addMReg (sp,
-    MachineOperand::Def));
-
-  std::set<unsigned> regsToSave (regsUsed);
-  // Save live-in values in registers onto the stack
+  // 1. Emit ADD instruction to allocate the right amount of stack space.
+  E.push_back (BuildMI (V9::ADDi, 3).addMReg (sp).addSImm (-TotalStackSize)
+    .addMReg (sp, MachineOperand::Def));
+
+  // 2. Get the saved register allocator state for all live-in variables.
+  // We are going to need to save live-in values which reside in registers
+  // on the stack, so we need to dig their register numbers out now.
+  RegsToSave = RegsToRestore;
   LiveVariableSet &Si = TF->LiveInSet;
   std::map<Value *, std::pair<AllocInfo, AllocInfo> > AllocStates;
   for (LiveVariableSet::iterator SI = Si.begin (), SE = Si.end (); SI != SE;
@@ -285,60 +297,57 @@
       if (Source.AllocState == AllocInfo::Allocated) {
         DEBUG (std::cerr << "Need to save incoming live value in reg "
                << Source.Placement << " onto the stack\n");
-        regsToSave.insert (Source.Placement);
+        RegsToSave.insert (Source.Placement);
       }
   }
 
   // 3. Save used registers onto the stack.
   std::vector<MachineInstr *> mvec;
   int RegType;
-  for (std::set<unsigned>::iterator i = regsToSave.begin (),
-       e = regsToSave.end (); i != e; ++i) {
+  for (std::set<unsigned>::iterator i = RegsToSave.begin (),
+       e = RegsToSave.end (); i != e; ++i) {
     mvec.clear ();
     unsigned R = *i;
     unsigned RegType = TRI.getRegType (R);
-    static const char *RegTypeStrings[] = { "IntRegType",
-      "FPSingleRegType", "FPDoubleRegType", "IntCCRegType",
-      "FloatCCRegType", "SpecialRegType" };
+
+    static const char *RegTypeStrings[] = { "IntRegType", "FPSingleRegType",
+      "FPDoubleRegType", "IntCCRegType", "FloatCCRegType", "SpecialRegType" };
     static const char *RegClassStrings[] = { "IntRegClassID",
       "FloatRegClassID", "IntCCRegClassID", "FloatCCRegClassID",
       "SpecialRegClassID" };
     DEBUG (std::cerr << "rewriteProlog: Saving reg#" << R << ", type = "
-                     << RegType << " (" << RegTypeStrings[RegType]
-                     << "), Class = " << TRI.getRegClassIDOfRegType(RegType)
-                     << " ("
-                     << RegClassStrings[TRI.getRegClassIDOfRegType(RegType)]
-                     << ")\n");
-    TRI.cpReg2MemMI (mvec, R, sp, 2047 + stackSize + 176 + R * 8, RegType, g2);
-    // Add whatever the TargetRegInfo gave us to the MachineBasicBlock we are
-    // working on.
+           << RegType << " (" << RegTypeStrings[RegType]
+           << "), ClassID = " << TRI.getRegClassIDOfRegType(RegType) << " ("
+           << RegClassStrings[TRI.getRegClassIDOfRegType(RegType)] << ")\n");
+    TRI.cpReg2MemMI (mvec, R, sp, stackOffsetForReg (R), RegType, g2);
     for (std::vector<MachineInstr *>::iterator vi = mvec.begin (),
-           ve = mvec.end (); vi != ve; ++vi) {
-      DEBUG (std::cerr << "rewriteProlog: Insns: " << **vi);
+           ve = mvec.end (); vi != ve; ++vi)
       E.push_back (*vi);
-    }
   }
 
-
-  // 4. Insert copies from each live-in variable's reg. in the matrix fn.
-  // to its reg. in the trace.
-  Function *MatrixF = TF->MatrixFn;
-  std::vector<CopyInfo> EntryCopies;
+  // 4. Insert a copy for each live-in variable to copy it from the stack
+  // to the register that was allocated for it in the TraceFn.
   for (LiveVariableSet::iterator SI = Si.begin (), SE = Si.end (); SI != SE;
        ++SI) {
     Value *V = *SI;
     std::pair<AllocInfo, AllocInfo> &ai = AllocStates[V];
     AllocInfo &Source = ai.first, &Target = ai.second;
-    DEBUG(std::cerr << "rewriteProlog: Source in MatrixFn = " << Source << "\n"
-                    << "rewriteProlog: Target in TraceFn = " << Target << "\n");
-    if (Source != Target)
-      EntryCopies.push_back (CopyInfo (Source, Target, &E, V->getType ()));
-  }
-
-  for (std::vector<CopyInfo>::iterator i = EntryCopies.begin (),
-         e = EntryCopies.end (); i != e; ++i) {
-    DEBUG(std::cerr << "rewriteProlog: Inserting copy " << *i << "\n");
-    insertCopyMachineInstrs (i->Src, i->Targ, *i->Blk, i->Ty);
+    if (Source != Target) {
+      assert (Target.AllocState == AllocInfo::Allocated
+              && "FIXME: can't do mem-->mem copy of live-ins yet");
+      // FIXME: The following should really be using insertCopyMachineInstrs or
+      // something like it.
+      mvec.clear ();
+      assert (TRI.getRegType (Target.Placement)
+              == TRI.getRegType (Source.Placement)
+              && "Live-in value changed reg type?!");
+      unsigned RegType = TRI.getRegType (Target.Placement);
+      TRI.cpMem2RegMI (mvec, sp, stackOffsetForReg (Source.Placement),
+                       Target.Placement, RegType, g2);
+      for (std::vector<MachineInstr *>::iterator vi = mvec.begin (),
+             ve = mvec.end (); vi != ve; ++vi)
+        E.push_back (*vi);
+    }
   }
 }
 
@@ -432,10 +441,10 @@
   // (InstructionKey, OperandKey, ...):
   for (unsigned i = 0; i < FAllocState->numTuples; ++i) {
 	OperandAllocState &T = FAllocState->tuples[i];
-	if (T.Instruction == InstructionKey && T.Operand == OperandKey)
-	  return AllocInfo (T.Instruction, T.Operand,
-						(AllocInfo::AllocStateTy) T.AllocState,
-						T.Placement);
+        if (T.Instruction == InstructionKey && T.Operand == OperandKey)
+          return AllocInfo (T.Instruction, T.Operand,
+                            (AllocInfo::AllocStateTy) T.AllocState,
+                            T.Placement);
   }
   // By this time we had better have found it, otherwise we are about to do bad
   // things.
@@ -493,8 +502,6 @@
   const TargetRegInfo &TRI = TM->getRegInfo ();
   static const unsigned fp = SparcV9::i6, sp = SparcV9::o6, g0 = SparcV9::g0,
     g1 = SparcV9::g1, g2 = SparcV9::g2;
-  Function *TraceF = TF->TraceFn, *MatrixF = TF->MatrixFn;
-  LiveVariableSet &So = TF->LiveOutSet;
 
   // UTF epilog: start out by clearing everything out of the exit basic block
   // (FIXME: may not work once we start doing optimizations!!! We will probably
@@ -503,16 +510,17 @@
 
   // Restore old FP.
   std::vector<MachineInstr *> mvec;
-  unsigned stackSize = getStaticStackSize (MF);
-  TRI.cpMem2RegMI (mvec, sp, 2047 + stackSize + 176 + fp * 8, fp,
-                   TRI.getRegType(fp), g2);
+  TRI.cpMem2RegMI (mvec, sp, stackOffsetForReg (fp), fp, TRI.getRegType(fp),
+                   g2);
   for (std::vector<MachineInstr *>::iterator vi = mvec.begin (),
        ve = mvec.end (); vi != ve; ++vi)
     MBB.push_back (*vi);
-  regsUsed.erase (fp);
+  RegsToRestore.erase (fp);
 
   // Insert copies from each live-out variable's reg. in the trace
   // to its reg. in the matrix function.
+  Function *TraceF = TF->TraceFn, *MatrixF = TF->MatrixFn;
+  LiveVariableSet &So = TF->LiveOutSet;
   for (LiveVariableSet::iterator SI = So.begin (), SE = So.end ();
 	   SI != SE; ++SI) {
 	Value *V = *SI;
@@ -526,41 +534,38 @@
   // rewriteProlog, earlier, and restore all used registers from stack
   // except SP.
   int RegType;
-  for (std::set<unsigned>::iterator i = regsUsed.begin (), e = regsUsed.end ();
-    i != e; ++i) {
+  for (std::set<unsigned>::iterator i = RegsToSave.begin (),
+         e = RegsToSave.end (); i != e; ++i) {
     mvec.clear ();
     unsigned R = *i;
     unsigned RegType = TRI.getRegType (R);
     DEBUG (std::cerr << "rewriteEpilog: Reloading reg#" << R << ", type = "
-                     << RegType << ", " << "Class = "
-                     << TRI.getRegClassIDOfRegType(RegType) << "\n");
-    TRI.cpMem2RegMI (mvec, sp, 2047 + stackSize + 176 + R * 8, R, RegType, g2);
-    // Add whatever the TargetRegInfo gave us to the MachineBasicBlock we are
-    // working on.
+           << RegType << ", " << "ClassID = "
+           << TRI.getRegClassIDOfRegType(RegType) << "\n");
+    TRI.cpMem2RegMI (mvec, sp, stackOffsetForReg (R), R, RegType, g2);
     for (std::vector<MachineInstr *>::iterator vi = mvec.begin (),
            ve = mvec.end (); vi != ve; ++vi)
       MBB.push_back (*vi);
   }
 
   // Restore stack pointer.
-  int Size = (stackSize + 176 + 104 * 8);
-  MBB.push_back (BuildMI (V9::ADDi, 3).addMReg (sp).addSImm (Size).addMReg (sp,
-    MachineOperand::Def));
+  MBB.push_back (BuildMI (V9::ADDi, 3).addMReg (sp).addSImm (TotalStackSize)
+    .addMReg (sp, MachineOperand::Def));
 
   // Let ReturnAddress be the address in memory of the compiled
   // code for the MachineBasicBlock in MatrixF that RI would have
   // returned to. Find it by using mapping info.
   const BasicBlock *RBB = MBB.getBasicBlock ();
   assert ((TF->ReturnBlockForTraceExit.find (RBB) !=
-		   TF->ReturnBlockForTraceExit.end ())
-		  && "Can't find matrix fn BB address to return to from trace");
+           TF->ReturnBlockForTraceExit.end ())
+          && "Can't find matrix fn BB address to return to from trace");
   std::pair<uint64_t, uint64_t> BlockAddrs =
     getBasicBlockInfo(TF->ReturnBlockForTraceExit[RBB]);
   uint64_t ReturnAddress = BlockAddrs.first;
-  DEBUG (std::cerr << "Return block for trace exit path is: "
+  DEBUG (std::cerr << "rewriteEpilog: Return block for trace exit path is:\n"
          << *TF->ReturnBlockForTraceExit[RBB] << "\n"
-         << "Mapping info says addresses are: return address = 0x"
-         << std::hex << BlockAddrs.first << ", end of block = 0x"
+         << "rewriteEpilog: Mapping info says addresses are: return address ="
+         << " 0x" << std::hex << BlockAddrs.first << ", end of block = 0x"
          << BlockAddrs.second << std::dec << "\n");
   // Insert a branch back to the return BasicBlock of the matrix fn.
   insertBranchMachineInstrs (ReturnAddress, MBB);
@@ -585,6 +590,21 @@
 
   DEBUG(std::cerr << "UnpackTraceFunction: unpacking "
           << MF.getFunction()->getName() << "()\n");
+
+  // Initialize RegsToRestore with the set of registers we'll need to restore
+  // in the epilog. This is a subset of the registers we'll need to save on the
+  // stack in the prolog!
+  findRegsToRestore (MF);
+
+  // Calculate the stack size.
+  // The actual amount we will allocate is (Static Stack Frame Size + Slop +
+  // Space for Regs).  The Slop = 176 number is from looking at the SparcV9
+  // ABI. The Space for Regs = 104 number is the number of 64-bit registers
+  // that are known to the SparcV9 backend.
+  StaticStackSize = getStaticStackSize (MF);
+  TotalStackSize = (StaticStackSize + 176 + 104 * 8);
+  DEBUG(std::cerr << "rewriteProlog: Stack sizes: static = " << StaticStackSize
+        << ", total = " << TotalStackSize << "\n");
 
   // Rewrite function prolog, found in the entry MachineBasicBlock of MF
   rewriteProlog (MF, MF.front ());




From gaeke at cs.uiuc.edu  Sun May 23 02:47:08 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Sun May 23 02:47:08 2004
Subject: [llvm-commits] CVS: reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp
Message-ID: <200405230746.CAA15910@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/LightWtProfiling:

RuntimeOptimizations.cpp updated: 1.37 -> 1.38

---
Log message:

Add another useful command to the gdb command printout in
TraceOptimizerDone(), which makes it easy to print the trace
disassembly.

Make optimizeTrace() keep track of its own initialization, and make
sure that it only runs it once (using the initDone flag). We should
really get around to making this into a class sometime soon.

By the time we get to optimizeTrace(), we have certainly called
initModules(), so just assert that, instead of trying to clean up
if we haven't.

Make the FunctionPassManager and TargetData pointer static.


---
Diffs of the changes:  (+28 -19)

Index: reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp
diff -u reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp:1.37 reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp:1.38
--- reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp:1.37	Thu May 20 13:09:30 2004
+++ reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp	Sun May 23 02:46:23 2004
@@ -37,8 +37,9 @@
 
 namespace llvm {
 
-static TargetMachine *Target = 0;
 static IntrinsicLowering *IL = 0;
+static TargetMachine *Target = 0;
+static const TargetData *TD = 0;
 static MachineCodeEmitter *MCE = 0;
 
 extern bool SaveStateToModule;
@@ -70,10 +71,15 @@
 
 extern void WriteTraceToFile (Trace &T);
 
-extern "C" void TraceOptimizerDone (uint64_t a, uint64_t traceStartAddr) {
+extern "C" void TraceOptimizerDone (uint64_t a, uint64_t traceStartAddr,
+                                    uint64_t traceEndAddr) {
   DEBUG (std::cerr << "\n*** About to return from optimizeTrace(). GDB"
-    << " breakpoints:\n" << std::hex << "display/i $pc\n" << "b *0x" << a
-    << "\n" << "b *0x" << traceStartAddr << std::dec << "\n");
+    << " commands:\n" << std::hex
+    << "display/i $pc\n"
+    << "b *0x" << a << "\n"
+    << "b *0x" << traceStartAddr << "\n"
+    << "disassemble 0x" << traceStartAddr << " 0x" << traceEndAddr << "\n"
+    << std::dec);
 }
 
 /// This method is called when we have finally constructed a
@@ -91,11 +97,20 @@
 
   // Initialization stuff: ensure module has been read in, and allocate a
   // target machine, if there isn't one already.
-  if (!MP) initModules ();
-  if (!IL) IL = new DefaultIntrinsicLowering ();
-  if (!Target) Target = allocateSparcV9TargetMachine (*MP->getModule (), IL);
-  const TargetData &TD = Target->getTargetData ();
-  if (!MCE) MCE = createTraceOptEmitter (TD);
+  static bool initDone = false;
+  if (!initDone) {
+    assert (MP);
+    if (!IL) IL = new DefaultIntrinsicLowering ();
+    if (!Target) Target = allocateSparcV9TargetMachine (*MP->getModule (), IL);
+    if (!TD) TD = &Target->getTargetData ();
+    if (!MCE) MCE = createTraceOptEmitter (*TD);
+    initDone = true;
+
+    // Force the SPARCv9 register allocator to save its state into a global
+    // variable
+    SaveRegAllocState = true;
+    SaveStateToModule = false;
+  }
 
   // Turn the vector of basic blocks into a Trace, and then turn the Trace into
   // a TraceFunction.
@@ -103,21 +118,15 @@
   DEBUG (WriteTraceToFile (T));
   TraceFunction *TF = TraceFunction::get (T);
 
-  // Force the SPARCv9 register allocator to save its state into a global
-  // variable
-  SaveRegAllocState = true;
-  SaveStateToModule = false;
-
   // Verify that the generated function's bytecode is good, then compile it
   // down to machine code. Then, "unpack" it back into its matrix function.
   // FIXME: Register allocation hints also need to be added to the PassManager,
   // so that we do not get clobbered by the overhead of adding copies in
   // UnpackTraceFunction.
-  FunctionPassManager PM (MP);
-  DEBUG(PM.add (new PrintFunctionPass ("Function created from trace for "
-                                 + T.getFunction ()->getName () + ":\n",
+  static FunctionPassManager PM (MP);
+  DEBUG(PM.add (new PrintFunctionPass ("Function created from trace:\n",
                                  &std::cerr)));
-  PM.add (new TargetData (TD));
+  PM.add (new TargetData (*TD));
   PM.add (createVerifierPass ());
   Target->getJITInfo ()->addPassesToJITCompile (PM);
   DEBUG(PM.add (createMachineFunctionPrinterPass (&std::cerr,
@@ -138,7 +147,7 @@
                   << " to point to 0x" << traceStartAddr << std::dec << "\n");
   vm->writeBranchInstruction(a, traceStartAddr);
   doFlush (a, a + 4);
-  TraceOptimizerDone(a, traceStartAddr);
+  TraceOptimizerDone(a, traceStartAddr, MCE->getCurrentPCValue ());
 }
 
 } // end namespace llvm




From gaeke at cs.uiuc.edu  Sun May 23 02:48:01 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Sun May 23 02:48:01 2004
Subject: [llvm-commits] CVS: reopt/lib/LightWtProfiling/SLI.cpp
Message-ID: <200405230747.CAA15922@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/LightWtProfiling:

SLI.cpp updated: 1.19 -> 1.20

---
Log message:

By the time we get to doInlining(), we have certainly called
initModules(), so just assert that, instead of trying to clean up
if we haven't.


---
Diffs of the changes:  (+1 -2)

Index: reopt/lib/LightWtProfiling/SLI.cpp
diff -u reopt/lib/LightWtProfiling/SLI.cpp:1.19 reopt/lib/LightWtProfiling/SLI.cpp:1.20
--- reopt/lib/LightWtProfiling/SLI.cpp:1.19	Sun May 16 00:12:14 2004
+++ reopt/lib/LightWtProfiling/SLI.cpp	Sun May 23 02:46:25 2004
@@ -451,8 +451,7 @@
 void doInlining(uint64_t addr1, std::vector<Function *> &stack, VirtualMem *vm){
   // Make sure we have already parsed the LLVM bytecode and have the
   // resulting LLVM Module handy.
-  if (!M)
-    initModules();
+  assert (M);
 
   Function *F = getRevFunction(M, addr1); 
 




From vadve at cs.uiuc.edu  Sun May 23 02:56:01 2004
From: vadve at cs.uiuc.edu (Vikram Adve)
Date: Sun May 23 02:56:01 2004
Subject: [llvm-commits] CVS: poolalloc/lib/PoolAllocate/PoolAllocate.cpp
Message-ID: <200405230755.CAA32700@psmith.cs.uiuc.edu>


Changes in directory poolalloc/lib/PoolAllocate:

PoolAllocate.cpp updated: 1.68 -> 1.69

---
Log message:

Pull out equivalance-class partitioning and graph merging into
a separate pass (class EquivClassGraphs), and fix it to inline
the resulting graphs bottom-up on the SCCs of the CBU call graph.


---
Diffs of the changes:  (+8 -182)

Index: poolalloc/lib/PoolAllocate/PoolAllocate.cpp
diff -u poolalloc/lib/PoolAllocate/PoolAllocate.cpp:1.68 poolalloc/lib/PoolAllocate/PoolAllocate.cpp:1.69
--- poolalloc/lib/PoolAllocate/PoolAllocate.cpp:1.68	Fri May 14 17:43:09 2004
+++ poolalloc/lib/PoolAllocate/PoolAllocate.cpp	Sun May 23 02:55:12 2004
@@ -6,6 +6,7 @@
 //===----------------------------------------------------------------------===//
 
 #define DEBUG_TYPE "PoolAllocator"
+#include "EquivClassGraphs.h"
 #include "poolalloc/PoolAllocate.h"
 #include "llvm/Constants.h"
 #include "llvm/DerivedTypes.h"
@@ -77,20 +78,18 @@
 
 void PoolAllocate::getAnalysisUsage(AnalysisUsage &AU) const {
   AU.addRequired<CompleteBUDataStructures>();
+  AU.addRequired<EquivClassGraphs>();
   AU.addRequired<TargetData>();
 }
 
 bool PoolAllocate::run(Module &M) {
   if (M.begin() == M.end()) return false;
   CurModule = &M;
-  BU = &getAnalysis<CompleteBUDataStructures>();
+  ECGraphs = &getAnalysis<EquivClassGraphs>();   // folded inlined CBU graphs
 
   // Add the pool* prototypes to the module
   AddPoolPrototypes();
 
-  // Figure out what the equivalence classes are for indirectly called functions
-  BuildIndirectFunctionSets(M);
-
   // Create the pools for memory objects reachable by global variables.
   if (SetupGlobalPools(M))
     return true;
@@ -223,186 +222,13 @@
   }
 }
 
-const PA::EquivClassInfo &PoolAllocate::getECIForIndirectCallSite(CallSite CS) {
-  Instruction *I = CS.getInstruction();
-  assert(I && "Not a call site?");
-  Function *thisFunc = I->getParent()->getParent();
-  DSNode *calleeNode = BU->getDSGraph(*thisFunc).getNodeForValue(CS.getCalledValue()).getNode();
-  if (!OneCalledFunction.count(calleeNode))
-    return ECInfoForLeadersMap[0];    // Special null function for empty graphs
-  Function *Called = OneCalledFunction[calleeNode];
-  Function *Leader = FuncECs.findClass(Called);
-  assert(Leader && "Leader not found for indirect call target!");
-  assert(ECInfoForLeadersMap.count(Leader) && "No ECI for indirect call site!");
-  return ECInfoForLeadersMap[Leader];
-}
-
-// BuildIndirectFunctionSets - Iterate over the module looking for indirect
-// calls to functions.  If a call site can invoke any functions [F1, F2... FN],
-// unify the N functions together in the FuncECs set.
-//
-void PoolAllocate::BuildIndirectFunctionSets(Module &M) {
-  const CompleteBUDataStructures::ActualCalleesTy AC = BU->getActualCallees();
-
-  // Loop over all of the indirect calls in the program.  If a call site can
-  // call multiple different functions, we need to unify all of the callees into
-  // the same equivalence class.
-  Instruction *LastInst = 0;
-  Function *FirstFunc = 0;
-  for (CompleteBUDataStructures::ActualCalleesTy::const_iterator I = AC.begin(),
-         E = AC.end(); I != E; ++I) {
-    CallSite CS = CallSite::get(I->first);
-    if (!CS.getCalledFunction() &&    // Ignore direct calls
-        !I->second->isExternal()) {   // Ignore functions we cannot modify
-      //std::cerr << "CALLEE: " << I->second->getName() << " from : " <<
-      //I->first;
-      if (I->first != LastInst) {
-        // This is the first callee from this call site.
-        LastInst = I->first;
-        FirstFunc = I->second;
-	//Instead of storing the lastInst For Indirection call Sites we store the
-	//DSNode for the function ptr arguemnt
-	Function *thisFunc = LastInst->getParent()->getParent();
-	DSNode *calleeNode = BU->getDSGraph(*thisFunc).getNodeForValue(CS.getCalledValue()).getNode();
-        OneCalledFunction[calleeNode] = FirstFunc;
-        FuncECs.addElement(I->second);
-      } else {
-        // This is not the first possible callee from a particular call site.
-        // Union the callee in with the other functions.
-        FuncECs.unionSetsWith(FirstFunc, I->second);
-      }
-    }
-  }
-
-  // Now that all of the equivalences have been built, turn the union-find data
-  // structure into a simple map from each function in the equiv class to the
-  // DSGraph used to represent the union of graphs.
-  //
-  std::map<Function*, Function*> &leaderMap = FuncECs.getLeaderMap();
-  DEBUG(std::cerr << "Indirect Function Equivalence Map:\n");
-  for (std::map<Function*, Function*>::iterator LI = leaderMap.begin(),
-	 LE = leaderMap.end(); LI != LE; ++LI) {
-    DEBUG(std::cerr << "  " << LI->first->getName() << ": leader is "
-                    << LI->second->getName() << "\n");
-
-    // Now the we have the equiv class info for this object, merge the DSGraph
-    // for this function into the composite DSGraph.
-    EquivClassInfo &ECI = ECInfoForLeadersMap[LI->second];
-
-    // If this is the first function in this equiv class, create the graph now.
-    if (ECI.G == 0)
-      ECI.G = new DSGraph(BU->getGlobalsGraph().getTargetData());
-
-    // Clone this member of the equivalence class into ECI.
-    DSGraph::NodeMapTy NodeMap;    
-    ECI.G->cloneInto(BU->getDSGraph(*LI->first), ECI.G->getScalarMap(),
-                     ECI.G->getReturnNodes(), NodeMap, 0);
-
-    // This is N^2 with the number of functions in this equiv class, but I don't
-    // really care right now.  FIXME!
-    for (unsigned i = 0, e = ECI.FuncsInClass.size(); i != e; ++i) {
-      Function *F = ECI.FuncsInClass[i];
-      // Merge the return nodes together.
-      ECI.G->getReturnNodes()[F].mergeWith(ECI.G->getReturnNodes()[LI->first]);
-
-      // Merge the arguments of the functions together.
-      Function::aiterator AI1 = F->abegin();
-      Function::aiterator AI2 = LI->first->abegin();
-      for (; AI1 != F->aend() && AI2 != LI->first->aend(); ++AI1,++AI2)
-        ECI.G->getNodeForValue(AI1).mergeWith(ECI.G->getNodeForValue(AI2));
-    }
-    ECI.FuncsInClass.push_back(LI->first);
-  }
-  DEBUG(std::cerr << "\n");
-
-  // Now that we have created the graphs for each of the equivalence sets, we
-  // have to figure out which pool descriptors to pass into the functions.  We
-  // must pass arguments for any pool descriptor that is needed by any function
-  // in the equiv. class.
-  for (ECInfoForLeadersMapTy::iterator ECII = ECInfoForLeadersMap.begin(),
-         E = ECInfoForLeadersMap.end(); ECII != E; ++ECII) {
-    EquivClassInfo &ECI = ECII->second;
-
-    // Traverse part of the graph to provoke most of the node forwardings to
-    // occur.
-    DSGraph::ScalarMapTy &SM = ECI.G->getScalarMap();
-    for (DSGraph::ScalarMapTy::iterator I = SM.begin(), E = SM.end(); I!=E; ++I)
-      I->second.getNode();   // Collapse forward references...
-
-    // Remove breadcrumbs from merging nodes.
-    ECI.G->removeTriviallyDeadNodes();
-
-    // Loop over all of the functions in this equiv class, figuring out which
-    // pools must be passed in for each function.
-    for (unsigned i = 0, e = ECI.FuncsInClass.size(); i != e; ++i) {
-      Function *F = ECI.FuncsInClass[i];
-      DSGraph &FG = BU->getDSGraph(*F);
-
-      // Figure out which nodes need to be passed in for this function (if any)
-      hash_set<DSNode*> &MarkedNodes = FunctionInfo[F].MarkedNodes;
-      MarkNodesWhichMustBePassedIn(MarkedNodes, *F, FG);
-
-      if (!MarkedNodes.empty()) {
-        // If any nodes need to be passed in, figure out which nodes these are
-        // in the unified graph for this equivalence class.
-        DSGraph::NodeMapTy NodeMapping;
-        for (Function::aiterator I = F->abegin(), E = F->aend(); I != E; ++I)
-          DSGraph::computeNodeMapping(FG.getNodeForValue(I),
-                                      ECI.G->getNodeForValue(I), NodeMapping);
-        DSGraph::computeNodeMapping(FG.getReturnNodeFor(*F),
-                                    ECI.G->getReturnNodeFor(*F), NodeMapping);
-        
-        // Loop through all of the nodes which must be passed through for this
-        // callee, and add them to the arguments list.
-        for (hash_set<DSNode*>::iterator I = MarkedNodes.begin(),
-               E = MarkedNodes.end(); I != E; ++I) {
-          DSNode *LocGraphNode = *I, *ECIGraphNode = NodeMapping[*I].getNode();
-
-          // Remember the mapping of this node
-          ECI.ECGraphToPrivateMap[std::make_pair(F,ECIGraphNode)] =LocGraphNode;
-
-          // Add this argument to be passed in.  Don't worry about duplicates,
-          // they will be eliminated soon.
-          ECI.ArgNodes.push_back(ECIGraphNode);
-        }
-      }
-    }
-
-    // Okay, all of the functions have had their required nodes added to the
-    // ECI.ArgNodes list, but there might be duplicates.  Eliminate the dups
-    // now.
-    std::sort(ECI.ArgNodes.begin(), ECI.ArgNodes.end());
-    ECI.ArgNodes.erase(std::unique(ECI.ArgNodes.begin(), ECI.ArgNodes.end()),
-                       ECI.ArgNodes.end());
-    
-    // Uncomment this if you want to see the aggregate graph result
-    //ECI.G->viewGraph();
-  }
-}
-
 
 void PoolAllocate::FindFunctionPoolArgs(Function &F) {
-  DSGraph &G = BU->getDSGraph(F);
+  DSGraph &G = ECGraphs->getDSGraph(F);
 
   FuncInfo &FI = FunctionInfo[&F];   // Create a new entry for F
   hash_set<DSNode*> &MarkedNodes = FI.MarkedNodes;
 
-  // If this function is part of an indirectly called function equivalence
-  // class, we have to handle it specially.
-  if (Function *Leader = FuncECs.findClass(&F)) {
-    EquivClassInfo &ECI = ECInfoForLeadersMap[Leader];
-
-    // The arguments passed in will be the ones specified by the ArgNodes list.
-    for (unsigned i = 0, e = ECI.ArgNodes.size(); i != e; ++i) {
-      DSNode *ArgNode =
-        ECI.ECGraphToPrivateMap[std::make_pair(&F, ECI.ArgNodes[i])];
-      FI.ArgNodes.push_back(ArgNode);
-      MarkedNodes.insert(ArgNode);
-    }
-
-    return;
-  }
-
   if (G.node_begin() == G.node_end())
     return;  // No memory activity, nothing is required
 
@@ -419,7 +245,7 @@
 // necessary, and return it.  If not, just return null.
 //
 Function *PoolAllocate::MakeFunctionClone(Function &F) {
-  DSGraph &G = BU->getDSGraph(F);
+  DSGraph &G = ECGraphs->getDSGraph(F);
   if (G.node_begin() == G.node_end()) return 0;
     
   FuncInfo &FI = FunctionInfo[&F];
@@ -540,7 +366,7 @@
 //
 bool PoolAllocate::SetupGlobalPools(Module &M) {
   // Get the globals graph for the program.
-  DSGraph &GG = BU->getGlobalsGraph();
+  DSGraph &GG = ECGraphs->getGlobalsGraph();
 
   // Get all of the nodes reachable from globals.
   hash_set<DSNode*> GlobalHeapNodes;
@@ -791,7 +617,7 @@
 // the specified function...
 //
 void PoolAllocate::ProcessFunctionBody(Function &F, Function &NewF) {
-  DSGraph &G = BU->getDSGraph(F);
+  DSGraph &G = ECGraphs->getDSGraph(F);
 
   if (G.node_begin() == G.node_end()) return;  // Quick exit if nothing to do...
   
@@ -801,7 +627,7 @@
   // Calculate which DSNodes are reachable from globals.  If a node is reachable
   // from a global, we will create a global pool for it, so no argument passage
   // is required.
-  DSGraph &GG = BU->getGlobalsGraph();
+  DSGraph &GG = ECGraphs->getGlobalsGraph();
   DSGraph::NodeMapTy GlobalsGraphNodeMapping;
   for (DSScalarMap::global_iterator I = G.getScalarMap().global_begin(),
          E = G.getScalarMap().global_end(); I != E; ++I) {




From vadve at cs.uiuc.edu  Sun May 23 02:56:08 2004
From: vadve at cs.uiuc.edu (Vikram Adve)
Date: Sun May 23 02:56:08 2004
Subject: [llvm-commits] CVS: poolalloc/lib/PoolAllocate/EquivClassGraphs.cpp EquivClassGraphs.h
Message-ID: <200405230755.CAA32692@psmith.cs.uiuc.edu>


Changes in directory poolalloc/lib/PoolAllocate:

EquivClassGraphs.cpp added (r1.1)
EquivClassGraphs.h added (r1.1)

---
Log message:

Complete rewrite of the code that merges DS graphs for equivalence classes
of functions called at a common call site.  The rewrite inlines the
resulting graphs bottom-up on the SCCs of the CBU call graph.  It also
simplifies the merging of equivalence classes by exploiting the fact that
functions in non-trivial SCCs are already merged.


---
Diffs of the changes:  (+604 -0)

Index: poolalloc/lib/PoolAllocate/EquivClassGraphs.cpp
diff -c /dev/null poolalloc/lib/PoolAllocate/EquivClassGraphs.cpp:1.1
*** /dev/null	Sun May 23 02:55:13 2004
--- poolalloc/lib/PoolAllocate/EquivClassGraphs.cpp	Sun May 23 02:54:02 2004
***************
*** 0 ****
--- 1,428 ----
+ //===- EquivClassGraphs.cpp - Merge equiv-class graphs & inline bottom-up -===//
+ // 
+ //                     The LLVM Compiler Infrastructure
+ //
+ // This file was developed by the LLVM research group and is distributed under
+ // the University of Illinois Open Source License. See LICENSE.TXT for details.
+ // 
+ //===----------------------------------------------------------------------===//
+ //
+ // This pass is the same as the complete bottom-up graphs, but
+ // with functions partitioned into equivalence classes and a single merged
+ // DS graph for all functions in an equivalence class.  After this merging,
+ // graphs are inlined bottom-up on the SCCs of the final (CBU) call graph.
+ //
+ //===----------------------------------------------------------------------===//
+ 
+ #define DEBUG_TYPE "ECGraphs"
+ #include "EquivClassGraphs.h"
+ #include "llvm/Analysis/DataStructure.h"
+ #include "llvm/Module.h"
+ #include "llvm/Pass.h"
+ #include "llvm/Analysis/DSGraph.h"
+ #include "llvm/Support/CallSite.h"
+ #include "Support/Debug.h"
+ #include "Support/SCCIterator.h"
+ #include "Support/Statistic.h"
+ #include "Support/EquivalenceClasses.h"
+ #include "Support/STLExtras.h"
+ using namespace llvm;
+ 
+ namespace llvm {
+   namespace PA {
+     Statistic<> NumFoldGraphInlines("Inline equiv-class graphs bottom up",
+                                     "Number of graphs inlined");
+     
+   } // End PA namespace
+ } // End llvm namespace
+ 
+ 
+ namespace {
+   RegisterAnalysis<llvm::PA::EquivClassGraphs> X("equivdatastructure",
+                     "Equivalence-class Bottom-up Data Structure Analysis");
+   Statistic<> NumEquivBUInlines("equivdatastructures", "Number of graphs inlined");
+ }
+ 
+ 
+ // getDSGraphForCallSite - Return the common data structure graph for
+ // callees at the specified call site.
+ // 
+ Function *llvm::PA::EquivClassGraphs::getSomeCalleeForCallSite(const CallSite &CS) const {
+   Function *thisFunc = CS.getCaller();
+   assert(thisFunc && "getDSGraphForCallSite(): Not a valid call site?");
+   DSNode *calleeNode = CBU->getDSGraph(*thisFunc).
+     getNodeForValue(CS.getCalledValue()).getNode();
+   std::map<DSNode*, Function *>::const_iterator I =
+     OneCalledFunction.find(calleeNode);
+   return (I == OneCalledFunction.end())? NULL : I->second;
+ }
+ 
+ // computeFoldedGraphs - Calculate the bottom up data structure
+ // graphs for each function in the program.
+ //
+ void llvm::PA::EquivClassGraphs::computeFoldedGraphs(Module &M) {
+   CBU = &getAnalysis<CompleteBUDataStructures>();
+ 
+   // Find equivalence classes of functions called from common call sites.
+   // Fold the CBU graphs for all functions in an equivalence class.
+   buildIndirectFunctionSets(M);
+ 
+   // Stack of functions used for Tarjan's SCC-finding algorithm.
+   std::vector<Function*> Stack;
+   hash_map<Function*, unsigned> ValMap;
+   unsigned NextID = 1;
+ 
+   if (Function *Main = M.getMainFunction()) {
+     if (!Main->isExternal())
+       processSCC(getOrCreateGraph(*Main), *Main, Stack, NextID, ValMap);
+   } else {
+     std::cerr << "Fold Graphs: No 'main' function found!\n";
+   }
+   
+   for (Module::iterator I = M.begin(), E = M.end(); I != E; ++I)
+     if (!I->isExternal() && !FoldedGraphsMap.count(I))
+       processSCC(getOrCreateGraph(*I), *I, Stack, NextID, ValMap);
+ 
+   getGlobalsGraph().removeTriviallyDeadNodes();
+ }
+ 
+ 
+ // buildIndirectFunctionSets - Iterate over the module looking for indirect
+ // calls to functions.  If a call site can invoke any functions [F1, F2... FN],
+ // unify the N functions together in the FuncECs set.
+ //
+ void llvm::PA::EquivClassGraphs::buildIndirectFunctionSets(Module &M) {
+   const ActualCalleesTy& AC = CBU->getActualCallees();
+ 
+   // Loop over all of the indirect calls in the program.  If a call site can
+   // call multiple different functions, we need to unify all of the callees into
+   // the same equivalence class.
+   Instruction *LastInst = 0;
+   Function *FirstFunc = 0;
+   for (ActualCalleesTy::const_iterator I=AC.begin(), E=AC.end(); I != E; ++I) {
+     if (I->second->isExternal())
+       continue;                         // Ignore functions we cannot modify
+ 
+     CallSite CS = CallSite::get(I->first);
+ 
+     if (CS.getCalledFunction()) {       // Direct call:
+       FuncECs.addElement(I->second);    // -- Make sure function has equiv class
+       FirstFunc = I->second;            // -- First callee at this site
+     } else {                            // Else indirect call
+       // DEBUG(std::cerr << "CALLEE: " << I->second->getName()
+       //       << " from : " << I->first);
+       if (I->first != LastInst) {
+         // This is the first callee from this call site.
+         LastInst = I->first;
+         FirstFunc = I->second;
+ 	// Instead of storing the lastInst For Indirection call Sites we store
+ 	// the DSNode for the function ptr arguemnt
+ 	Function *thisFunc = LastInst->getParent()->getParent();
+ 	DSNode *calleeNode = CBU->getDSGraph(*thisFunc).getNodeForValue(CS.getCalledValue()).getNode();
+         OneCalledFunction[calleeNode] = FirstFunc;
+         FuncECs.addElement(I->second);
+       } else {
+         // This is not the first possible callee from a particular call site.
+         // Union the callee in with the other functions.
+         FuncECs.unionSetsWith(FirstFunc, I->second);
+ #ifndef NDEBUG
+ 	Function *thisFunc = LastInst->getParent()->getParent();
+ 	DSNode *calleeNode = CBU->getDSGraph(*thisFunc).getNodeForValue(CS.getCalledValue()).getNode();
+         assert(OneCalledFunction.count(calleeNode) > 0 && "Missed a call?");
+ #endif
+       }
+     }
+ 
+     // Now include all functions that share a graph with any function in the
+     // equivalence class.  More precisely, if F is in the class, and G(F) is
+     // its graph, then we include all other functions that are also in G(F).
+     // Currently, that is just the functions in the same call-graph-SCC as F.
+     // 
+     DSGraph& funcDSGraph = CBU->getDSGraph(*I->second);
+     const DSGraph::ReturnNodesTy &RetNodes = funcDSGraph.getReturnNodes();
+     for (DSGraph::ReturnNodesTy::const_iterator RI=RetNodes.begin(),
+            RE=RetNodes.end(); RI != RE; ++RI)
+       FuncECs.unionSetsWith(FirstFunc, RI->first);
+   }
+ 
+   // Now that all of the equivalences have been built, merge the graphs for
+   // each equivalence class.
+   //
+   std::set<Function*> &leaderSet = FuncECs.getLeaderSet();
+   DEBUG(std::cerr << "\nIndirect Function Equivalence Sets:\n");
+   for (std::set<Function*>::iterator LI = leaderSet.begin(),
+ 	 LE = leaderSet.end(); LI != LE; ++LI) {
+ 
+     Function* LF = *LI;
+     const std::set<Function*>& EqClass = FuncECs.getEqClass(LF);
+ 
+ #ifndef NDEBUG
+     if (EqClass.size() > 1) {
+       DEBUG(std::cerr <<"  Equivalence set for leader " <<LF->getName()<<" = ");
+       for (std::set<Function*>::const_iterator EqI = EqClass.begin(),
+              EqEnd = EqClass.end(); EqI != EqEnd; ++EqI)
+         DEBUG(std::cerr << " " << (*EqI)->getName() << ",");
+       DEBUG(std::cerr << "\n");
+     }
+ #endif
+ 
+     if (EqClass.size() > 1) {
+       // This equiv class has multiple functions: merge their graphs.
+       // First, clone the CBU graph for the leader and make it the 
+       // common graph for the equivalence graph.
+       DSGraph* mergedG = cloneGraph(*LF);
+ 
+       // Record the argument nodes for use in merging later below
+       EquivClassGraphArgsInfo& GraphInfo = getECGraphInfo(mergedG);
+       for (Function::aiterator AI1 = LF->abegin(); AI1 != LF->aend(); ++AI1)
+         GraphInfo.argNodes.push_back(mergedG->getNodeForValue(AI1));
+       
+       // Merge in the graphs of all other functions in this equiv. class.
+       // Note that two or more functions may have the same graph, and it
+       // only needs to be merged in once.  Use a set to find repetitions.
+       std::set<DSGraph*> GraphsMerged;
+       for (std::set<Function*>::const_iterator EqI = EqClass.begin(),
+              EqEnd = EqClass.end(); EqI != EqEnd; ++EqI) {
+         Function* F = *EqI;
+         DSGraph*& FG = FoldedGraphsMap[F];
+         if (F == LF || FG == mergedG)
+           continue;
+         
+         // Record the "folded" graph for the function.
+         FG = mergedG;
+         
+         // Clone this member of the equivalence class into mergedG
+         DSGraph* CBUGraph = &CBU->getDSGraph(*F); 
+         if (GraphsMerged.count(CBUGraph) > 0)
+           continue;
+         
+         GraphsMerged.insert(CBUGraph);
+         DSGraph::NodeMapTy NodeMap;    
+         mergedG->cloneInto(*CBUGraph, mergedG->getScalarMap(),
+                            mergedG->getReturnNodes(), NodeMap, 0);
+ 
+         // Merge the return nodes of all functions together.
+         mergedG->getReturnNodes()[LF].mergeWith(mergedG->getReturnNodes()[F]);
+ 
+         // Merge the function arguments with all argument nodes found so far.
+         // If there are extra function args, add them to the vector of argNodes
+         Function::aiterator AI2 = F->abegin(), AI2end = F->aend();
+         for (unsigned arg=0, numArgs=GraphInfo.argNodes.size();
+              arg < numArgs && AI2 != AI2end; ++AI2, ++arg)
+           GraphInfo.argNodes[arg].mergeWith(mergedG->getNodeForValue(AI2));
+         for ( ; AI2 != AI2end; ++AI2)
+           GraphInfo.argNodes.push_back(mergedG->getNodeForValue(AI2));
+       }
+     }
+   }
+   DEBUG(std::cerr << "\n");
+ }
+ 
+ 
+ DSGraph &llvm::PA::EquivClassGraphs::getOrCreateGraph(Function &F) {
+   // Has the graph already been created?
+   DSGraph *&Graph = FoldedGraphsMap[&F];
+   if (Graph) return *Graph;
+ 
+   // Use the CBU graph directly without copying it.
+   // This automatically updates the FoldedGraphsMap via the reference.
+   Graph = &CBU->getDSGraph(F);
+ 
+   return *Graph;
+ }
+ 
+ DSGraph* llvm::PA::EquivClassGraphs::cloneGraph(Function &F) {
+   DSGraph *&Graph = FoldedGraphsMap[&F];
+   DSGraph &CBUGraph = CBU->getDSGraph(F);
+   assert(Graph == NULL || Graph == &CBUGraph && "Cloning a graph twice?");
+ 
+   // Copy the CBU graph...
+   Graph = new DSGraph(CBUGraph);           // updates the map via reference
+   Graph->setGlobalsGraph(&getGlobalsGraph());
+   Graph->setPrintAuxCalls();
+ 
+   // Make sure to update the FoldedGraphsMap map for all functions in the graph!
+   for (DSGraph::ReturnNodesTy::iterator I = Graph->getReturnNodes().begin();
+        I != Graph->getReturnNodes().end(); ++I)
+     if (I->first != &F) {
+       DSGraph*& FG = FoldedGraphsMap[I->first];
+       assert(FG == NULL || FG == &CBU->getDSGraph(*I->first) &&
+              "Merging function in SCC twice?");
+       FG = Graph;
+     }
+ 
+   return Graph;
+ }
+ 
+ 
+ unsigned llvm::PA::EquivClassGraphs::processSCC(DSGraph &FG, Function& F,
+                                         std::vector<Function*> &Stack,
+                                         unsigned &NextID, 
+                                         hash_map<Function*, unsigned> &ValMap) {
+   DEBUG(std::cerr << "    ProcessSCC for function " << F.getName() << "\n");
+ 
+   assert(!ValMap.count(&F) && "Shouldn't revisit functions!");
+   unsigned Min = NextID++, MyID = Min;
+   ValMap[&F] = Min;
+   Stack.push_back(&F);
+ 
+   // The edges out of the current node are the call site targets...
+   for (unsigned i = 0, e = FG.getFunctionCalls().size(); i != e; ++i) {
+     Instruction *Call = FG.getFunctionCalls()[i].getCallSite().getInstruction();
+ 
+     // Loop over all of the actually called functions...
+     ActualCalleesTy::const_iterator I, E;
+     for (tie(I, E) = getActualCallees().equal_range(Call); I != E; ++I)
+       if (!I->second->isExternal()) {
+         DSGraph &CalleeG = getOrCreateGraph(*I->second);
+ 
+         // Have we visited the destination function yet?
+         hash_map<Function*, unsigned>::iterator It = ValMap.find(I->second);
+         unsigned M = (It == ValMap.end())  // No, visit it now.
+           ? processSCC(CalleeG, *I->second, Stack, NextID, ValMap)
+           : It->second;                    // Yes, get it's number.
+         
+         if (M < Min) Min = M;
+       }
+   }
+ 
+   assert(ValMap[&F] == MyID && "SCC construction assumption wrong!");
+   if (Min != MyID)
+     return Min;         // This is part of a larger SCC!
+ 
+   // If this is a new SCC, process it now.
+   bool IsMultiNodeSCC = false;
+   while (Stack.back() != &F) {
+     DSGraph *NG = &getOrCreateGraph(* Stack.back());
+     ValMap[Stack.back()] = ~0U;
+ 
+     // Since all SCCs must be the same as those found in CBU, we do not need to
+     // do any merging.  Make sure all functions in the SCC share the same graph.
+     assert(NG == &FG &&
+            "FoldGraphs: Functions in the same SCC have different graphs?");
+     
+     Stack.pop_back();
+     IsMultiNodeSCC = true;
+   }
+ 
+   // Clean up the graph before we start inlining a bunch again...
+   if (IsMultiNodeSCC)
+     FG.removeTriviallyDeadNodes();
+ 
+   Stack.pop_back();
+   processGraph(FG, F);
+   ValMap[&F] = ~0U;
+   return MyID;
+ }
+ 
+ 
+ /// processGraph - Process the CBU graphs for the program in bottom-up order on
+ /// the SCC of the __ACTUAL__ call graph.  This builds final folded CBU graphs.
+ void llvm::PA::EquivClassGraphs::processGraph(DSGraph &G, Function& F) {
+   DEBUG(std::cerr << "    ProcessGraph for function " << F.getName() << "\n");
+ 
+   hash_set<Instruction*> calls;
+ 
+   DSGraph* CallerGraph = sameAsCBUGraph(F)? NULL : &getOrCreateGraph(F);
+ 
+   // If the function has not yet been cloned, let's check if any callees
+   // need to be inlined before cloning it.
+   // 
+   for (unsigned i=0, e=G.getFunctionCalls().size(); i!=e && !CallerGraph; ++i) {
+     const DSCallSite &CS = G.getFunctionCalls()[i];
+     Instruction *TheCall = CS.getCallSite().getInstruction();
+     
+     // Loop over all potential callees to find the first non-external callee.
+     // Some inlining is needed if there is such a callee and it has changed.
+     ActualCalleesTy::const_iterator I, E;
+     for (tie(I, E) = getActualCallees().equal_range(TheCall); I != E; ++I)
+       if (!I->second->isExternal() && !sameAsCBUGraph(*I->second)) {
+         // Ok, the caller does need to be cloned... go ahead and do it now.
+         // clone the CBU graph for F now because we have not cloned it so far
+         CallerGraph = cloneGraph(F);
+         break;
+       }
+   }
+ 
+   if (!CallerGraph) {                   // No inlining is needed.
+     DEBUG(std::cerr << "  --DONE ProcessGraph for function " << F.getName()
+           << " (NO INLINING NEEDED)\n");
+     return;
+   }
+ 
+   // Else we need to inline some callee graph.  Visit all call sites.
+   // The edges out of the current node are the call site targets...
+   for (unsigned i=0, e = CallerGraph->getFunctionCalls().size(); i != e; ++i) {
+     const DSCallSite &CS = CallerGraph->getFunctionCalls()[i];
+     Instruction *TheCall = CS.getCallSite().getInstruction();
+ 
+     assert(calls.insert(TheCall).second &&
+            "Call instruction occurs multiple times in graph??");
+     
+     // Inline the common callee graph into the current graph, if the callee
+     // graph has not changed.  Note that all callees should have the same
+     // graph so we only need to do this once.
+     // 
+     DSGraph* CalleeGraph = NULL;
+     ActualCalleesTy::const_iterator I, E;
+     tie(I, E) = getActualCallees().equal_range(TheCall);
+     unsigned TNum, Num;
+ 
+     // Loop over all potential callees to find the first non-external callee.
+     for (TNum = 0, Num = std::distance(I, E); I != E; ++I, ++TNum)
+       if (!I->second->isExternal())
+         break;
+ 
+     // Now check if the graph has changed and if so, clone and inline it.
+     if (I != E && !sameAsCBUGraph(*I->second)) {
+       Function *CalleeFunc = I->second;
+       
+       // Merge the callee's graph into this graph, if not already the same.
+       // Callees in the same equivalence class (which subsumes those
+       // in the same SCCs) have the same graph.  Note that all recursion
+       // including self-recursion have been folded in the equiv classes.
+       // 
+       CalleeGraph = &getOrCreateGraph(*CalleeFunc);
+       if (CalleeGraph != CallerGraph) {
+         ++NumFoldGraphInlines;
+         CallerGraph->mergeInGraph(CS, *CalleeFunc, *CalleeGraph,
+                                   DSGraph::KeepModRefBits |
+                                   DSGraph::StripAllocaBit |
+                                   DSGraph::DontCloneCallNodes |
+                                   DSGraph::DontCloneAuxCallNodes);
+         DEBUG(std::cerr << "    Inlining graph [" << i << "/" << e-1
+               << ":" << TNum << "/" << Num-1 << "] for "
+               << CalleeFunc->getName() << "["
+               << CalleeGraph->getGraphSize() << "+"
+               << CalleeGraph->getAuxFunctionCalls().size()
+               << "] into '" /*<< CallerGraph->getFunctionNames()*/ << "' ["
+               << CallerGraph->getGraphSize() << "+"
+               << CallerGraph->getAuxFunctionCalls().size()
+               << "]\n");
+       }
+     }
+ 
+ #ifndef NDEBUG
+     // Now loop over the rest of the callees and make sure they have the
+     // same graph as the one inlined above.
+     if (CalleeGraph)
+       for (++I, ++TNum; I != E; ++I, ++TNum)
+         if (!I->second->isExternal())
+           assert(CalleeGraph == &getOrCreateGraph(*I->second) &&
+                  "Callees at a call site have different graphs?");
+ #endif
+   }
+ 
+   // Recompute the Incomplete markers
+   if (CallerGraph != NULL) {
+     assert(CallerGraph->getInlinedGlobals().empty());
+     CallerGraph->maskIncompleteMarkers();
+     CallerGraph->markIncompleteNodes(DSGraph::MarkFormalArgs);
+ 
+     // Delete dead nodes.  Treat globals that are unreachable but that can
+     // reach live nodes as live.
+     CallerGraph->removeDeadNodes(DSGraph::KeepUnreachableGlobals);
+   }
+ 
+   DEBUG(std::cerr << "  --DONE ProcessGraph for function " << F.getName() << "\n");
+ }


Index: poolalloc/lib/PoolAllocate/EquivClassGraphs.h
diff -c /dev/null poolalloc/lib/PoolAllocate/EquivClassGraphs.h:1.1
*** /dev/null	Sun May 23 02:55:13 2004
--- poolalloc/lib/PoolAllocate/EquivClassGraphs.h	Sun May 23 02:54:02 2004
***************
*** 0 ****
--- 1,176 ----
+ //===-- EquivClassGraphs.h - Merge equiv-class graphs & inline bottom-up --===//
+ // 
+ //                     The LLVM Compiler Infrastructure
+ //
+ // This file was developed by the LLVM research group and is distributed under
+ // the University of Illinois Open Source License. See LICENSE.TXT for details.
+ // 
+ //===----------------------------------------------------------------------===//
+ //
+ // This pass is the same as the complete bottom-up graphs, but
+ // with functions partitioned into equivalence classes and a single merged
+ // DS graph for all functions in an equivalence class.  After this merging,
+ // graphs are inlined bottom-up on the SCCs of the final (CBU) call graph.
+ //
+ //===----------------------------------------------------------------------===//
+ 
+ 
+ #include "llvm/Analysis/DataStructure.h"
+ #include "llvm/Analysis/DSGraph.h"
+ #include "Support/EquivalenceClasses.h"
+ #include "Support/STLExtras.h"
+ #include <vector>
+ #include <map>
+ #include <ext/hash_map>
+ 
+ namespace llvm {
+ 
+ class Module;
+ class Function;
+ 
+ namespace PA {
+ 
+   /// EquivClassGraphArgsInfo - Information about the set of argument nodes 
+   /// in a DS graph (the number of argument nodes is the max of argument nodes
+   /// for all functions folded into the graph).
+   /// FIXME: This class is only used temporarily and could be eliminated.
+   ///  
+   struct EquivClassGraphArgsInfo {
+     const DSGraph* ECGraph;
+     std::vector<DSNodeHandle> argNodes;
+     EquivClassGraphArgsInfo() : ECGraph(NULL) {}
+   };
+ 
+   /// EquivClassGraphs - This is the same as the complete bottom-up graphs, but
+   /// with functions partitioned into equivalence classes and a single merged
+   /// DS graph for all functions in an equivalence class.  After this merging,
+   /// graphs are inlined bottom-up on the SCCs of the final (CBU) call graph.
+   ///
+   struct EquivClassGraphs : public Pass {
+     CompleteBUDataStructures *CBU;
+ 
+     // FoldedGraphsMap, one graph for each function
+     hash_map<const Function*, DSGraph*> FoldedGraphsMap;
+   
+     // Equivalence class where functions that can potentially be called via the
+     // same function pointer are in the same class.
+     EquivalenceClasses<Function*> FuncECs;
+ 
+     // Each equivalence class graph contains several functions.
+     // Remember their argument nodes (and return nodes?)
+     std::map<const DSGraph*, EquivClassGraphArgsInfo> ECGraphInfo;
+     
+     /// OneCalledFunction - For each indirect call, we keep track of one
+     /// target of the call.  This is used to find equivalence class called by
+     /// a call site.
+     std::map<DSNode*, Function *> OneCalledFunction;
+ 
+   public:
+     /// EquivClassGraphs - Computes the equivalence classes and then the
+     /// folded DS graphs for each class.
+     /// 
+     virtual bool run(Module &M) { computeFoldedGraphs(M); return true; }
+ 
+     /// getCBUDataStructures - Get the CompleteBUDataStructures object
+     /// 
+     CompleteBUDataStructures *getCBUDataStructures() { return CBU; }
+ 
+     /// getDSGraph - Return the data structure graph for the specified function.
+     /// This returns the folded graph.  The folded graph is the same as the CBU
+     /// graph iff the function is in a singleton equivalence class AND all its 
+     /// callees also have the same folded graph as the CBU graph.
+     /// 
+     DSGraph &getDSGraph(const Function &F) const {
+       hash_map<const Function*, DSGraph*>::const_iterator I =
+         FoldedGraphsMap.find(const_cast<Function*>(&F));
+       assert(I != FoldedGraphsMap.end() && "No folded graph for function!");
+       return *I->second;
+     }
+ 
+     /// getSomeCalleeForCallSite - Return any one callee function at
+     /// a call site.
+     /// 
+     Function *getSomeCalleeForCallSite(const CallSite &CS) const;
+ 
+     /// getDSGraphForCallSite - Return the common data structure graph for
+     /// callees at the specified call site.
+     /// 
+     DSGraph &getDSGraphForCallSite(const CallSite &CS) const {
+       return this->getDSGraph(*getSomeCalleeForCallSite(CS));
+     }
+ 
+     /// getEquivClassForCallSite - Get the set of functions in the equivalence
+     /// class for a given call site.
+     /// 
+     const std::set<Function*>& getEquivClassForCallSite(const CallSite& CS) {
+       Function* leaderF = FuncECs.findClass(getSomeCalleeForCallSite(CS));
+       return FuncECs.getEqClass(leaderF);
+     }
+ 
+     /// getECGraphInfo - Get the graph info object with arg nodes info
+     /// 
+     EquivClassGraphArgsInfo &getECGraphInfo(const DSGraph* G) {
+       assert(G != NULL && "getECGraphInfo: Null graph!");
+       EquivClassGraphArgsInfo& GraphInfo = ECGraphInfo[G];
+       if (GraphInfo.ECGraph == NULL)
+         GraphInfo.ECGraph = G;
+       return GraphInfo;
+     }
+ 
+     /// sameAsCBUGraph - Check if the folded graph for this function is
+     /// the same as the CBU graph.
+     bool sameAsCBUGraph(const Function &F) const {
+       DSGraph& foldedGraph = getDSGraph(F);
+       return (&foldedGraph == &CBU->getDSGraph(F));
+     }
+ 
+     DSGraph &getGlobalsGraph() const {
+       return CBU->getGlobalsGraph();
+     }
+     
+     typedef llvm::BUDataStructures::ActualCalleesTy ActualCalleesTy;
+     const ActualCalleesTy &getActualCallees() const {
+       return CBU->getActualCallees();
+     }
+ 
+     virtual void getAnalysisUsage(AnalysisUsage &AU) const {
+       AU.setPreservesAll();
+       AU.addRequired<CompleteBUDataStructures>();
+     }
+ 
+     /// print - Print out the analysis results...
+     ///
+     void print(std::ostream &O, const Module *M) const { CBU->print(O, M); }
+ 
+   private:
+     void computeFoldedGraphs(Module &M);
+ 
+     void buildIndirectFunctionSets(Module &M);
+ 
+     unsigned processSCC(DSGraph &FG, Function &F, std::vector<Function*> &Stack,
+                         unsigned &NextID, 
+                         hash_map<Function*, unsigned> &ValMap);
+ 
+     void processGraph(DSGraph &FG, Function &F);
+ 
+     DSGraph &getOrCreateGraph(Function &F);
+ 
+     DSGraph* cloneGraph(Function &F);
+ 
+     bool hasFoldedGraph(const Function& F) const {
+       hash_map<const Function*, DSGraph*>::const_iterator I =
+         FoldedGraphsMap.find(const_cast<Function*>(&F));
+       return (I != FoldedGraphsMap.end());
+     }
+ 
+     DSGraph* getOrCreateLeaderGraph(const Function& leader) {
+       DSGraph*& leaderGraph = FoldedGraphsMap[&leader];
+       if (leaderGraph == NULL)
+         leaderGraph = new DSGraph(CBU->getGlobalsGraph().getTargetData());
+       return leaderGraph;
+     }
+   };
+ 
+ }; // end PA namespace
+ 
+ }; // end llvm namespace




From vadve at cs.uiuc.edu  Sun May 23 02:57:03 2004
From: vadve at cs.uiuc.edu (Vikram Adve)
Date: Sun May 23 02:57:03 2004
Subject: [llvm-commits] CVS: poolalloc/include/poolalloc/PoolAllocate.h
Message-ID: <200405230756.CAA32715@psmith.cs.uiuc.edu>


Changes in directory poolalloc/include/poolalloc:

PoolAllocate.h updated: 1.22 -> 1.23

---
Log message:

Pull out equivalance-class partitioning and graph merging into
a separate pass (class EquivClassGraphs).


---
Diffs of the changes:  (+7 -40)

Index: poolalloc/include/poolalloc/PoolAllocate.h
diff -u poolalloc/include/poolalloc/PoolAllocate.h:1.22 poolalloc/include/poolalloc/PoolAllocate.h:1.23
--- poolalloc/include/poolalloc/PoolAllocate.h:1.22	Fri May 14 17:41:35 2004
+++ poolalloc/include/poolalloc/PoolAllocate.h	Sun May 23 02:56:05 2004
@@ -20,13 +20,15 @@
 
 namespace llvm {
 
-class CompleteBUDataStructures;
 class DSNode;
 class DSGraph;
 class Type;
 class AllocaInst;
 
 namespace PA {
+
+  class EquivClassGraphs;
+
   /// FuncInfo - Represent the pool allocation information for one function in
   /// the program.  Note that many functions must actually be cloned in order
   /// for pool allocation to add arguments to the function signature.  In this
@@ -71,27 +73,7 @@
     std::map<Value*, const Value*> NewToOldValueMap;
   };
 
-  struct EquivClassInfo {
-    // FuncsInClass - This is a list of all of the functions in this equiv
-    // class.
-    std::vector<Function*> FuncsInClass;
-
-    // G - The DSGraph which contains a union of all of the nodes in this
-    // equivalence class.
-    DSGraph *G;
-
-    // ECGraphToPrivateMap - This map, given a DSNode in the ECI DSGraph and a
-    // function, will tell you which DS node in the function DS graph this node
-    // corresponds to (or null if none).
-    std::map<std::pair<Function*,DSNode*>, DSNode*> ECGraphToPrivateMap;
-
-    // ArgNodes - The list of DSNodes which require arguments to be passed in
-    // for all members of this equivalence class.
-    std::vector<DSNode*> ArgNodes;
-
-    EquivClassInfo() : G(0) {}
-  };
-}
+}; // end PA namespace
 
 
 
@@ -99,41 +81,26 @@
 ///
 class PoolAllocate : public Pass {
   Module *CurModule;
-  CompleteBUDataStructures *BU;
+  PA::EquivClassGraphs *ECGraphs;
 
   std::map<Function*, PA::FuncInfo> FunctionInfo;
 
-  // Equivalence class where functions that can potentially be called via the
-  // same function pointer are in the same class.
-  EquivalenceClasses<Function*> FuncECs;
-
-  // Each equivalence class leader has an EquivClassInfo object.  This map holds
-  // them.
-  typedef std::map<Function*, PA::EquivClassInfo> ECInfoForLeadersMapTy;
-  ECInfoForLeadersMapTy ECInfoForLeadersMap;
-
-  /// OneCalledFunction - For each indirect function call, we keep track of one
-  /// target of the call.  This is used to find the equivalence class called by
-  /// a call site.
-  std::map<DSNode*, Function *> OneCalledFunction;
-
  public:
 
   Function *PoolInit, *PoolDestroy, *PoolAlloc, *PoolAllocArray, *PoolFree;
   static const Type *PoolDescPtrTy;
 
-  const PA::EquivClassInfo &getECIForIndirectCallSite(CallSite CS);
-
   /// GlobalNodes - For each node (with an H marker) in the globals graph, this
   /// map contains the global variable that holds the pool descriptor for the
   /// node.
   std::map<DSNode*, Value*> GlobalNodes;
+
  public:
   bool run(Module &M);
   
   virtual void getAnalysisUsage(AnalysisUsage &AU) const;
   
-  CompleteBUDataStructures &getBUDataStructures() const { return *BU; }
+  PA::EquivClassGraphs &getECGraphs() const { return *ECGraphs; }
   
   //Dinakar to get function info for all (cloned functions) 
   PA::FuncInfo *getFunctionInfo(Function *F) {




From vadve at cs.uiuc.edu  Sun May 23 02:57:09 2004
From: vadve at cs.uiuc.edu (Vikram Adve)
Date: Sun May 23 02:57:09 2004
Subject: [llvm-commits] CVS: poolalloc/lib/PoolAllocate/TransformFunctionBody.cpp
Message-ID: <200405230756.CAA32730@psmith.cs.uiuc.edu>


Changes in directory poolalloc/lib/PoolAllocate:

TransformFunctionBody.cpp updated: 1.23 -> 1.24

---
Log message:

Rewrite to use results of new merging pass, EquivClassGraphs, instead
of old ECI stuff.


---
Diffs of the changes:  (+50 -24)

Index: poolalloc/lib/PoolAllocate/TransformFunctionBody.cpp
diff -u poolalloc/lib/PoolAllocate/TransformFunctionBody.cpp:1.23 poolalloc/lib/PoolAllocate/TransformFunctionBody.cpp:1.24
--- poolalloc/lib/PoolAllocate/TransformFunctionBody.cpp:1.23	Fri Mar  5 00:54:55 2004
+++ poolalloc/lib/PoolAllocate/TransformFunctionBody.cpp	Sun May 23 02:56:48 2004
@@ -5,6 +5,7 @@
 //===----------------------------------------------------------------------===//
 
 #define DEBUG_TYPE "PoolAllocator"
+#include "EquivClassGraphs.h"
 #include "poolalloc/PoolAllocate.h"
 #include "llvm/Analysis/DataStructure.h"
 #include "llvm/Analysis/DSGraph.h"
@@ -75,15 +76,18 @@
       FI.NewToOldValueMap.erase(I);
     }
 
-    DSNodeHandle& getDSNodeHFor(Value *V) {
+    Value* getOldValueIfAvailable(Value* V) {
       if (!FI.NewToOldValueMap.empty()) {
         // If the NewToOldValueMap is in effect, use it.
         std::map<Value*,const Value*>::iterator I = FI.NewToOldValueMap.find(V);
         if (I != FI.NewToOldValueMap.end())
           V = (Value*)I->second;
       }
+      return V;
+    }
 
-      return G.getScalarMap()[V];
+    DSNodeHandle& getDSNodeHFor(Value *V) {
+      return G.getScalarMap()[getOldValueIfAvailable(V)];
     }
 
     Value *getPoolHandle(Value *V) {
@@ -337,13 +341,16 @@
   // between the graphs to figure out which pool descriptors need to be passed
   // in.  The roots of this mapping is found from arguments and return values.
   //
+  EquivClassGraphs& ECGraphs = PAInfo.getECGraphs();
   DSGraph::NodeMapTy NodeMapping;
   Instruction *NewCall;
   Value *NewCallee;
   std::vector<DSNode*> ArgNodes;
   DSGraph *CalleeGraph;  // The callee graph
 
+  // For indirect callees find any callee since all DS graphs have been merged.
   if (CF) {   // Direct calls are nice and simple.
+    DEBUG(std::cerr << "  Handling direct call: " << *TheCall);
     FuncInfo *CFI = PAInfo.getFuncInfo(*CF);
     if (CFI == 0 || CFI->Clone == 0) {   // Nothing to transform...
       visitInstruction(*TheCall);
@@ -352,30 +359,50 @@
     NewCallee = CFI->Clone;
     ArgNodes = CFI->ArgNodes;
     
-    DEBUG(std::cerr << "  Handling direct call: " << *TheCall);
-    CalleeGraph = &PAInfo.getBUDataStructures().getDSGraph(*CF);
+    CalleeGraph = &ECGraphs.getDSGraph(*CF);
   } else {
     DEBUG(std::cerr << "  Handling indirect call: " << *TheCall);
     
-    // Figure out which set of functions this call may invoke
-    Instruction *OrigInst = CS.getInstruction();
-
-    // If this call site is in a clone function, map it back to the original
-    if (!FI.NewToOldValueMap.empty())
-      OrigInst = cast<Instruction>((Value*)FI.NewToOldValueMap[OrigInst]);
-    const PA::EquivClassInfo &ECI =
-      PAInfo.getECIForIndirectCallSite(CallSite::get(OrigInst));
-
-    if (ECI.ArgNodes.empty())
-      return;   // No arguments to add?  Transformation is a noop!
-
     // Here we fill in CF with one of the possible called functions.  Because we
     // merged together all of the arguments to all of the functions in the
     // equivalence set, it doesn't really matter which one we pick.
-    CalleeGraph = ECI.G;
-    CF = ECI.FuncsInClass.back();
-    NewCallee = CS.getCalledValue();
-    ArgNodes = ECI.ArgNodes;
+    // (If the function was cloned, we have to map the cloned call instruction
+    // in CS back to the original call instruction.)
+    Instruction *OrigInst =
+      cast<Instruction>(getOldValueIfAvailable(CS.getInstruction()));
+    CF = isa<CallInst>(OrigInst)?
+      ECGraphs.getSomeCalleeForCallSite(cast<CallInst>(OrigInst)) :
+      ECGraphs.getSomeCalleeForCallSite(cast<InvokeInst>(OrigInst));
+
+    if (!CF) {
+      // FIXME: Unknown callees for a call-site. Warn and ignore.
+      std::cerr << "\n***\n*** WARNING (FuncTransform::visitCallSite): "
+                << "Unknown callees for call-site in function "
+                << CS.getCaller()->getName() << "\n***\n";
+      return;
+    }
+
+    // Get the common graph for the set of functions this call may invoke.
+    CalleeGraph = &ECGraphs.getDSGraph(*CF);
+    
+#ifndef NDEBUG
+    // Verify that all potential callees at call site have the same DS graph.
+    const BUDataStructures::ActualCalleesTy& ActualCallees =
+      ECGraphs.getCBUDataStructures()->getActualCallees();
+    BUDataStructures::ActualCalleesTy::const_iterator I, E;
+    for (tie(I, E) = ActualCallees.equal_range(OrigInst); I != E; ++I)
+      if (!I->second->isExternal())
+        assert(CalleeGraph == &ECGraphs.getDSGraph(*I->second) &&
+               "Callees at call site do not have a common graph!");
+#endif    
+
+    // Find the DS nodes for the arguments that need to be added, if any.
+    FuncInfo *CFI = PAInfo.getFuncInfo(*CF);
+    assert(CFI && "No function info for callee at indirect call?");
+    ArgNodes = CFI->ArgNodes;
+
+    if (ArgNodes.empty())
+      return;           // No arguments to add?  Transformation is a noop!
 
     // Cast the function pointer to an appropriate type!
     std::vector<const Type*> ArgTys(ArgNodes.size(),
@@ -383,13 +410,12 @@
     for (CallSite::arg_iterator I = CS.arg_begin(), E = CS.arg_end();
          I != E; ++I)
       ArgTys.push_back((*I)->getType());
-      
+    
     FunctionType *FTy = FunctionType::get(TheCall->getType(), ArgTys, false);
     PointerType *PFTy = PointerType::get(FTy);
     
-    // If there are any pool arguments cast the function pointer to the right
-    // type.
-    NewCallee = new CastInst(NewCallee, PFTy, "tmp", TheCall);
+    // If there are any pool arguments cast the func ptr to the right type.
+    NewCallee = new CastInst(CS.getCalledValue(), PFTy, "tmp", TheCall);
   }
 
   Function::aiterator FAI = CF->abegin(), E = CF->aend();




From gaeke at cs.uiuc.edu  Sun May 23 02:58:01 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Sun May 23 02:58:01 2004
Subject: [llvm-commits] CVS: reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp
Message-ID: <200405230757.CAA15944@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/LightWtProfiling:

UnpackTraceFunction.cpp updated: 1.67 -> 1.68

---
Log message:

Untabify.


---
Diffs of the changes:  (+3 -3)

Index: reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp
diff -u reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp:1.67 reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp:1.68
--- reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp:1.67	Sun May 23 02:46:26 2004
+++ reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp	Sun May 23 02:57:26 2004
@@ -523,11 +523,11 @@
   LiveVariableSet &So = TF->LiveOutSet;
   for (LiveVariableSet::iterator SI = So.begin (), SE = So.end ();
 	   SI != SE; ++SI) {
-	Value *V = *SI;
+    Value *V = *SI;
     std::pair<AllocInfo, AllocInfo> ai = getValueAllocState (TF, V);
     AllocInfo &Target = ai.first, &Source = ai.second;
-	if (Source != Target)
-	  insertCopyMachineInstrs (Source, Target, MBB, V->getType ());
+    if (Source != Target)
+      insertCopyMachineInstrs (Source, Target, MBB, V->getType ());
   }
 
   // Get the set of registers used in this function which we saved in




From vadve at cs.uiuc.edu  Sun May 23 03:01:01 2004
From: vadve at cs.uiuc.edu (Vikram Adve)
Date: Sun May 23 03:01:01 2004
Subject: [llvm-commits] CVS: llvm/lib/Analysis/DataStructure/CompleteBottomUp.cpp
Message-ID: <200405230800.DAA32767@psmith.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis/DataStructure:

CompleteBottomUp.cpp updated: 1.10 -> 1.11

---
Log message:

Inline both direct and indirect callees in the CBU phase because
a direct callee may have indirect callees and so may have changed.


---
Diffs of the changes:  (+34 -34)

Index: llvm/lib/Analysis/DataStructure/CompleteBottomUp.cpp
diff -u llvm/lib/Analysis/DataStructure/CompleteBottomUp.cpp:1.10 llvm/lib/Analysis/DataStructure/CompleteBottomUp.cpp:1.11
--- llvm/lib/Analysis/DataStructure/CompleteBottomUp.cpp:1.10	Fri Mar  5 16:04:07 2004
+++ llvm/lib/Analysis/DataStructure/CompleteBottomUp.cpp	Sun May 23 03:00:34 2004
@@ -37,14 +37,13 @@
   GlobalsGraph = new DSGraph(BU.getGlobalsGraph());
   GlobalsGraph->setPrintAuxCalls();
 
-  // Our call graph is the same as the BU data structures call graph
-  ActualCallees = BU.getActualCallees();
-
 #if 1   // REMOVE ME EVENTUALLY
   // FIXME: TEMPORARY (remove once finalization of indirect call sites in the
   // globals graph has been implemented in the BU pass)
   TDDataStructures &TD = getAnalysis<TDDataStructures>();
 
+  ActualCallees.clear();
+
   // The call graph extractable from the TD pass is _much more complete_ and
   // trustable than that generated by the BU pass so far.  Until this is fixed,
   // we hack it like this:
@@ -53,17 +52,23 @@
     const std::vector<DSCallSite> &CSs = TD.getDSGraph(*MI).getFunctionCalls();
 
     for (unsigned CSi = 0, e = CSs.size(); CSi != e; ++CSi) {
-      if (CSs[CSi].isIndirectCall()) {
-        Instruction *TheCall = CSs[CSi].getCallSite().getInstruction();
+      Instruction *TheCall = CSs[CSi].getCallSite().getInstruction();
 
+      if (CSs[CSi].isIndirectCall()) { // indirect call: insert all callees
         const std::vector<GlobalValue*> &Callees =
           CSs[CSi].getCalleeNode()->getGlobals();
         for (unsigned i = 0, e = Callees.size(); i != e; ++i)
           if (Function *F = dyn_cast<Function>(Callees[i]))
             ActualCallees.insert(std::make_pair(TheCall, F));
+      } else {        // direct call: insert the single callee directly
+        ActualCallees.insert(std::make_pair(TheCall,
+                                            CSs[CSi].getCalleeFunc()));
       }
     }
   }
+#else
+  // Our call graph is the same as the BU data structures call graph
+  ActualCallees = BU.getActualCallees();
 #endif
 
   std::vector<DSGraph*> Stack;
@@ -183,35 +188,30 @@
            "Call instruction occurs multiple times in graph??");
       
 
-    // The Normal BU pass will have taken care of direct calls well already,
-    // don't worry about them.
-
-    // FIXME: if a direct callee had indirect callees, it seems like they could
-    // be updated and we would have to reinline even direct calls!
-
-    if (!CS.getCallSite().getCalledFunction()) {
-      // Loop over all of the actually called functions...
-      ActualCalleesTy::iterator I, E;
-      tie(I, E) = ActualCallees.equal_range(TheCall);
-      unsigned TNum = 0, Num = std::distance(I, E);
-      for (; I != E; ++I, ++TNum) {
-        Function *CalleeFunc = I->second;
-        if (!CalleeFunc->isExternal()) {
-          // Merge the callee's graph into this graph.  This works for normal
-          // calls or for self recursion within an SCC.
-          DSGraph &GI = getOrCreateGraph(*CalleeFunc);
-          ++NumCBUInlines;
-          G.mergeInGraph(CS, *CalleeFunc, GI, DSGraph::KeepModRefBits |
-                         DSGraph::StripAllocaBit | DSGraph::DontCloneCallNodes |
-                         DSGraph::DontCloneAuxCallNodes);
-          DEBUG(std::cerr << "    Inlining graph [" << i << "/" << e-1
-                << ":" << TNum << "/" << Num-1 << "] for "
-                << CalleeFunc->getName() << "["
-                << GI.getGraphSize() << "+" << GI.getAuxFunctionCalls().size()
-                << "] into '" /*<< G.getFunctionNames()*/ << "' ["
-                << G.getGraphSize() << "+" << G.getAuxFunctionCalls().size()
-                << "]\n");
-        }
+    // Loop over all of the potentially called functions...
+    // Inline direct calls as well as indirect calls because the direct
+    // callee may have indirect callees and so may have changed.
+    // 
+    ActualCalleesTy::iterator I, E;
+    tie(I, E) = ActualCallees.equal_range(TheCall);
+    unsigned TNum = 0, Num = std::distance(I, E);
+    for (; I != E; ++I, ++TNum) {
+      Function *CalleeFunc = I->second;
+      if (!CalleeFunc->isExternal()) {
+        // Merge the callee's graph into this graph.  This works for normal
+        // calls or for self recursion within an SCC.
+        DSGraph &GI = getOrCreateGraph(*CalleeFunc);
+        ++NumCBUInlines;
+        G.mergeInGraph(CS, *CalleeFunc, GI, DSGraph::KeepModRefBits |
+                       DSGraph::StripAllocaBit | DSGraph::DontCloneCallNodes |
+                       DSGraph::DontCloneAuxCallNodes);
+        DEBUG(std::cerr << "    Inlining graph [" << i << "/" << e-1
+              << ":" << TNum << "/" << Num-1 << "] for "
+              << CalleeFunc->getName() << "["
+              << GI.getGraphSize() << "+" << GI.getAuxFunctionCalls().size()
+              << "] into '" /*<< G.getFunctionNames()*/ << "' ["
+              << G.getGraphSize() << "+" << G.getAuxFunctionCalls().size()
+              << "]\n");
       }
     }
   }




From vadve at cs.uiuc.edu  Sun May 23 03:03:00 2004
From: vadve at cs.uiuc.edu (Vikram Adve)
Date: Sun May 23 03:03:00 2004
Subject: [llvm-commits] CVS: llvm/include/llvm/Support/CallSite.h
Message-ID: <200405230802.DAA00326@psmith.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Support:

CallSite.h updated: 1.13 -> 1.14

---
Log message:

Add getCaller() method.


---
Diffs of the changes:  (+5 -0)

Index: llvm/include/llvm/Support/CallSite.h
diff -u llvm/include/llvm/Support/CallSite.h:1.13 llvm/include/llvm/Support/CallSite.h:1.14
--- llvm/include/llvm/Support/CallSite.h:1.13	Tue Nov 11 16:41:31 2003
+++ llvm/include/llvm/Support/CallSite.h	Sun May 23 03:02:45 2004
@@ -22,6 +22,7 @@
 #define LLVM_SUPPORT_CALLSITE_H
 
 #include "llvm/Instruction.h"
+#include "llvm/BasicBlock.h"
 
 namespace llvm {
 
@@ -59,6 +60,10 @@
   /// getInstruction - Return the instruction this call site corresponds to
   ///
   Instruction *getInstruction() const { return I; }
+
+  /// getCaller - Return the caller function for this call site
+  ///
+  Function *getCaller() const { return I->getParent()->getParent(); }
 
   /// getCalledValue - Return the pointer to function that is being called...
   ///




From vadve at cs.uiuc.edu  Sun May 23 03:06:03 2004
From: vadve at cs.uiuc.edu (Vikram Adve)
Date: Sun May 23 03:06:03 2004
Subject: [llvm-commits] CVS: llvm/include/Support/EquivalenceClasses.h
Message-ID: <200405230805.DAA00348@psmith.cs.uiuc.edu>


Changes in directory llvm/include/Support:

EquivalenceClasses.h updated: 1.6 -> 1.7

---
Log message:

Remember the set of leaders.  Also compute on demand and cache the equiv
class for each leader.   Finally, rename Elem2ECLeaderMap to Elem2LeaderMap
(most of the changed lines are only due to the latter).


---
Diffs of the changes:  (+52 -32)

Index: llvm/include/Support/EquivalenceClasses.h
diff -u llvm/include/Support/EquivalenceClasses.h:1.6 llvm/include/Support/EquivalenceClasses.h:1.7
--- llvm/include/Support/EquivalenceClasses.h:1.6	Tue Nov 11 16:41:29 2003
+++ llvm/include/Support/EquivalenceClasses.h	Sun May 23 03:05:14 2004
@@ -18,6 +18,7 @@
 #define SUPPORT_EQUIVALENCECLASSES_H
 
 #include <map>
+#include <set>
 #include <vector>
 
 namespace llvm {
@@ -26,69 +27,88 @@
 class EquivalenceClasses {
   // Maps each element to the element that is the leader of its 
   // equivalence class.
-  std::map<ElemTy, ElemTy> Elem2ECLeaderMap;
+  std::map<ElemTy, ElemTy> Elem2LeaderMap;
   
+  // Maintains the set of leaders
+  std::set<ElemTy> LeaderSet;
+
+  // Caches the equivalence class for each leader
+  std::map<ElemTy, std::set<ElemTy> > LeaderToEqClassMap;
+
   // Make Element2 the leader of the union of classes Element1 and Element2
   // Element1 and Element2 are presumed to be leaders of their respective
   // equivalence classes.
   void attach(ElemTy Element1, ElemTy Element2) {
     for (typename std::map<ElemTy, ElemTy>::iterator ElemI = 
-	   Elem2ECLeaderMap.begin(), ElemE = Elem2ECLeaderMap.end(); 
+	   Elem2LeaderMap.begin(), ElemE = Elem2LeaderMap.end(); 
 	 ElemI != ElemE; ++ElemI) {
       if (ElemI->second == Element1)
-	Elem2ECLeaderMap[ElemI->first] = Element2;
+	Elem2LeaderMap[ElemI->first] = Element2;
     }
   }
 
 public:
-  
-  void addElement (ElemTy NewElement) {
-    if (Elem2ECLeaderMap.find(NewElement) == Elem2ECLeaderMap.end())
-      Elem2ECLeaderMap[NewElement] = NewElement;
+  // If an element has not yet in any class, make it a separate new class.
+  // Return the leader of the class containing the element.
+  ElemTy addElement (ElemTy NewElement) {
+    typename std::map<ElemTy, ElemTy>::iterator ElemI = 
+      Elem2LeaderMap.find(NewElement);
+    if (ElemI == Elem2LeaderMap.end()) {
+      Elem2LeaderMap[NewElement] = NewElement;
+      LeaderSet.insert(NewElement);
+      return NewElement;
+    }
+    else
+      return ElemI->second;
   }
   
-  ElemTy findClass(ElemTy Element) {
-    if (Elem2ECLeaderMap.find(Element) == Elem2ECLeaderMap.end())
-      return 0;
-    else 
-      return Elem2ECLeaderMap[Element];
+  ElemTy findClass(ElemTy Element) const {
+    typename std::map<ElemTy, ElemTy>::const_iterator I =
+      Elem2LeaderMap.find(Element);
+    return (I == Elem2LeaderMap.end())? (ElemTy) 0 : I->second;
   }
 
   /// Attach the set with Element1 to the set with Element2 adding Element1 and
   /// Element2 to the set of equivalence classes if they are not there already.
   /// Implication: Make Element1 the element in the smaller set.
+  /// Take Leader[Element1] out of the set of leaders.
   void unionSetsWith(ElemTy Element1, ElemTy Element2) {
     // If either Element1 or Element2 does not already exist, include it
-    if (Elem2ECLeaderMap.find(Element1) == Elem2ECLeaderMap.end())
-      Elem2ECLeaderMap[Element1] = Element1;
-    if (Elem2ECLeaderMap.find(Element2) == Elem2ECLeaderMap.end())
-      Elem2ECLeaderMap[Element2] = Element2;
-
-    attach(Elem2ECLeaderMap[Element1], Elem2ECLeaderMap[Element2]);
+    const ElemTy& leader1 = addElement(Element1);
+    const ElemTy& leader2 = addElement(Element2);
+    assert(leader1 != (ElemTy) 0 && leader2 != (ElemTy) 0);
+    if (leader1 != leader2) {
+      attach(leader1, leader2);
+      LeaderSet.erase(leader1);
+    }
   }
   
-  // Returns a vector containing all the elements in the equivalent class
+  // Returns a vector containing all the elements in the equivalence class
   // including Element1
-  std::vector<ElemTy> getEqClass(ElemTy Element1) {
-    std::vector<ElemTy> EqClass;
+  const std::set<ElemTy> & getEqClass(ElemTy Element1) {
+    assert(Elem2LeaderMap.find(Element1) != Elem2LeaderMap.end());
+    const ElemTy classLeader = Elem2LeaderMap[Element1];
     
-    if (Elem2ECLeaderMap.find(EqClass) == Elem2ECLeaderMap.end())
-      return EqClass;
+    std::set<ElemTy> & EqClass = LeaderToEqClassMap[classLeader];
     
-    ElemTy classLeader = Elem2ECLeaderMap[Element1];
-    for (typename std::map<ElemTy, ElemTy>::iterator ElemI = 
-	   Elem2ECLeaderMap.begin(), ElemE = Elem2ECLeaderMap.end(); 
-	 ElemI != ElemE; ++ElemI) {
-      if (ElemI->second == classLeader)
-	EqClass.push_back(ElemI->first);
+    // If the EqClass vector is empty, it has not been computed yet: do it now
+    if (EqClass.empty()) {
+      for (typename std::map<ElemTy, ElemTy>::iterator
+             ElemI = Elem2LeaderMap.begin(), ElemE = Elem2LeaderMap.end(); 
+           ElemI != ElemE; ++ElemI)
+        if (ElemI->second == classLeader)
+          EqClass.insert(ElemI->first);
+      assert(! EqClass.empty());        // must at least include the leader
     }
     
     return EqClass;
   }
 
-  std::map<ElemTy, ElemTy>& getLeaderMap() {
-    return Elem2ECLeaderMap ;
-  }
+        std::set<ElemTy>& getLeaderSet()       { return LeaderSet; }
+  const std::set<ElemTy>& getLeaderSet() const { return LeaderSet; }
+
+        std::map<ElemTy, ElemTy>& getLeaderMap()       { return Elem2LeaderMap;}
+  const std::map<ElemTy, ElemTy>& getLeaderMap() const { return Elem2LeaderMap;}
 };
 
 } // End llvm namespace




From gaeke at cs.uiuc.edu  Sun May 23 05:06:01 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Sun May 23 05:06:01 2004
Subject: [llvm-commits] CVS: reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp
Message-ID: <200405231005.FAA00467@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/LightWtProfiling:

UnpackTraceFunction.cpp updated: 1.68 -> 1.69

---
Log message:

Move declaration of UnpackTraceFunction to reopt/UnpackTraceFunction.h.
Include it. Prune other excess #includes.

createUnpackTraceFunctionPass() is history.

Give UnpackTraceFunction::runOnMachineFunction() a new doxygen comment.


---
Diffs of the changes:  (+8 -63)

Index: reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp
diff -u reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp:1.68 reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp:1.69
--- reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp:1.68	Sun May 23 02:57:26 2004
+++ reopt/lib/LightWtProfiling/UnpackTraceFunction.cpp	Sun May 23 05:05:02 2004
@@ -13,16 +13,15 @@
 //
 //===----------------------------------------------------------------------===//
 
+#include "reopt/UnpackTraceFunction.h"
 #include "reopt/TraceToFunction.h"
-#include "llvm/CodeGen/MachineFunctionPass.h"
+#include "reopt/MappingInfo.h"
 #include "llvm/CodeGen/MachineFunctionInfo.h"
-#include "llvm/CodeGen/MachineInstr.h"
 #include "llvm/Module.h"
 #include "llvm/Argument.h"
 #include "llvm/Assembly/Writer.h"
 #include "llvm/Support/InstIterator.h"
 #include "Support/Debug.h"
-#include "reopt/MappingInfo.h"
 #include "../../../../lib/Target/SparcV9/RegAlloc/AllocInfo.h"
 #include "../../../../lib/Target/SparcV9/RegAlloc/PhyRegAlloc.h"
 #include "../../../../lib/Target/SparcV9/SparcV9RegInfo.h"
@@ -30,47 +29,6 @@
 
 namespace llvm {
 
-/// This pass inserts copies in two places in the machine
-/// code for TF->TraceFn: first, at the entry basic block to copy the values in
-/// TF->LiveInSet from TF->MatrixFn's registers to TF->TraceFn's registers, and
-/// second, at each exit basic block to copy the values in TF->LiveOutSet from
-/// TF->TraceFn's registers to TF->MatrixFn's registers.
-///
-class UnpackTraceFunction : public MachineFunctionPass {
-  TargetMachine *TM;
-  TraceFunction *TF;
-
-  /// Static stack size of the MachineFunction. Filled in by rewriteProlog ()
-  /// by calling getStaticStackSize (), and then used again by rewriteEpilog ().
-  ///
-  unsigned StaticStackSize;
-
-  /// Total stack size of the MachineFunction, including space for regs
-  /// modified in TraceFn to be saved.
-  ///
-  int TotalStackSize;
-
-  /// Registers clobbered in the trace. Filled in by findRegsUsedInFunction ().
-  ///
-  std::set<unsigned> RegsToRestore;
-  std::set<unsigned> RegsToSave;
-
-  unsigned getStaticStackSize (MachineFunction &MF);
-  unsigned stackOffsetForReg (unsigned R);
-  void findRegsToRestore (MachineFunction &MF);
-  void insertCopyMachineInstrs (AllocInfo &Source, AllocInfo &Target,
-                                MachineBasicBlock &B, const Type *Ty);
-  void insertBranchMachineInstrs (uint64_t Target, MachineBasicBlock &B);
-  const MachineInstr *containsReturnInstr (MachineBasicBlock &B);
-  void rewriteProlog (MachineFunction &MF, MachineBasicBlock &MBB);
-  void rewriteEpilog (MachineFunction &MF, MachineBasicBlock &MBB);
-public:
-  UnpackTraceFunction (TargetMachine *_TM, TraceFunction *_TF) :
-    TM (_TM), TF (_TF) { }
-  const char *getPassName () const { return "Unpack trace function"; }
-  virtual bool runOnMachineFunction (MachineFunction &MF);
-};
-
 /// Structure describing a single live variable copy.
 ///
 struct CopyInfo {
@@ -91,15 +49,6 @@
   return OS;
 }
 
-/// Create a new UnpackTraceFunction pass that will unpack a given
-/// TraceFunction, for which machine code has been generated, into
-/// its matrix function.
-///
-FunctionPass *createUnpackTraceFunctionPass (TargetMachine *TM,
-                                             TraceFunction *TF) {
-  return new UnpackTraceFunction (TM, TF);
-}
-
 // Ripped off from SparcV9PrologEpilogInserter
 unsigned UnpackTraceFunction::getStaticStackSize (MachineFunction &MF) {
   const TargetFrameInfo& frameInfo = MF.getTarget().getFrameInfo();
@@ -571,16 +520,12 @@
   insertBranchMachineInstrs (ReturnAddress, MBB);
 }
 
-/// This method is provided with MF, which is the
-/// machine code for TF->TraceFn.  We modify it (according to the description
-/// of the pass, above) using the information provided along with TF when this
-/// Pass object was created.
-///
-/// Admittedly, that is a pretty lame design, but for now, it means that we can
-/// only deal with one TraceFn and one MatrixFn in any given
-/// UnpackTraceFunction pass. If this stuff gets popular, we can have
-/// TraceFunctionPasses and TraceMachineFunctionPasses, or the Trace stuff can
-/// be an AnalysisPass, or something.
+/// runOnMachineFunction - Prepare MF, which is the machine code for
+/// TF->TraceFn, to be executed using a low-overhead calling convention for
+/// traces. We use the live-variable and return-block information in TF and the
+/// mapping information provided by the SparcV9 backend.  We assume that
+/// someone has called setTraceFunction(TraceFunction *) to set TF to point to
+/// the TraceFunction we are interested in.
 ///
 bool UnpackTraceFunction::runOnMachineFunction (MachineFunction &MF) {
   // Make sure we are looking at the MachineFunction corresponding to the




From gaeke at cs.uiuc.edu  Sun May 23 05:06:08 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Sun May 23 05:06:08 2004
Subject: [llvm-commits] CVS: reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp
Message-ID: <200405231005.FAA00460@seraph.cs.uiuc.edu>


Changes in directory reopt/lib/LightWtProfiling:

RuntimeOptimizations.cpp updated: 1.38 -> 1.39

---
Log message:

Include UnpackTraceFunction.h.

Explicitly save the UnpackTraceFunction object so that we can call
setTraceFunction() on it as necessary.

Move the FunctionPassManager out to global scope.

Move the adding of passes into addPassesToReoptimizeTrace().

In optimizeTrace()'s initialization section, don't check each pointer
individually. Also, set up the FunctionPassManager here and add passes to it.
This means that the only thing we need to do on second and subsequent calls to
optimizeTrace() is set the TraceFunction and run the PassManager.


---
Diffs of the changes:  (+36 -23)

Index: reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp
diff -u reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp:1.38 reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp:1.39
--- reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp:1.38	Sun May 23 02:46:23 2004
+++ reopt/lib/LightWtProfiling/RuntimeOptimizations.cpp	Sun May 23 05:05:01 2004
@@ -14,6 +14,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "ReoptimizerInternal.h"
+#include "reopt/UnpackTraceFunction.h"
 #include "reopt/TraceToFunction.h"
 #include "reopt/VirtualMem.h"
 #include "reopt/InstrUtils.h"
@@ -41,6 +42,8 @@
 static TargetMachine *Target = 0;
 static const TargetData *TD = 0;
 static MachineCodeEmitter *MCE = 0;
+static FunctionPassManager *PM = 0;
+static UnpackTraceFunction *UTF = 0;
 
 extern bool SaveStateToModule;
 extern bool SaveRegAllocState;
@@ -82,6 +85,29 @@
     << std::dec);
 }
 
+static void addPassesToReoptimizeTrace (FunctionPassManager &PM,
+                                        const TargetData *TD,
+                                        TargetMachine *Target,
+                                        MachineCodeEmitter *MCE) {
+  // Verify that the generated function's bytecode is good, then compile it
+  // down to machine code. Then, "unpack" it back into its matrix function.
+  // FIXME: Register allocation hints also need to be added to the PassManager,
+  // so that we do not get clobbered by the overhead of adding copies in
+  // UnpackTraceFunction.
+  DEBUG(PM.add (new PrintFunctionPass ("Function created from trace:\n",
+                                 &std::cerr)));
+  PM.add (new TargetData (*TD));
+  DEBUG (PM.add (createVerifierPass ()));
+  Target->getJITInfo ()->addPassesToJITCompile (PM);
+  DEBUG(PM.add (createMachineFunctionPrinterPass (&std::cerr,
+                                                  "Before unpacking:\n")));
+  UTF = new UnpackTraceFunction (Target);
+  PM.add (UTF);
+  DEBUG(PM.add (createMachineFunctionPrinterPass (&std::cerr,
+                                                  "After unpacking:\n")));
+  Target->addPassesToEmitMachineCode (PM, *MCE);
+}
+
 /// This method is called when we have finally constructed a
 /// trace. The first parameter is the vector of basic blocks that form
 /// the trace; the second parameter is the starting
@@ -100,10 +126,13 @@
   static bool initDone = false;
   if (!initDone) {
     assert (MP);
-    if (!IL) IL = new DefaultIntrinsicLowering ();
-    if (!Target) Target = allocateSparcV9TargetMachine (*MP->getModule (), IL);
-    if (!TD) TD = &Target->getTargetData ();
-    if (!MCE) MCE = createTraceOptEmitter (*TD);
+    IL = new DefaultIntrinsicLowering ();
+    Target = allocateSparcV9TargetMachine (*MP->getModule (), IL);
+    TD = &Target->getTargetData ();
+    MCE = createTraceOptEmitter (*TD);
+    // Set up the pass manager and trace function unpacker
+    PM = new FunctionPassManager (MP);
+    addPassesToReoptimizeTrace (*PM, TD, Target, MCE);
     initDone = true;
 
     // Force the SPARCv9 register allocator to save its state into a global
@@ -118,31 +147,15 @@
   DEBUG (WriteTraceToFile (T));
   TraceFunction *TF = TraceFunction::get (T);
 
-  // Verify that the generated function's bytecode is good, then compile it
-  // down to machine code. Then, "unpack" it back into its matrix function.
-  // FIXME: Register allocation hints also need to be added to the PassManager,
-  // so that we do not get clobbered by the overhead of adding copies in
-  // UnpackTraceFunction.
-  static FunctionPassManager PM (MP);
-  DEBUG(PM.add (new PrintFunctionPass ("Function created from trace:\n",
-                                 &std::cerr)));
-  PM.add (new TargetData (*TD));
-  PM.add (createVerifierPass ());
-  Target->getJITInfo ()->addPassesToJITCompile (PM);
-  DEBUG(PM.add (createMachineFunctionPrinterPass (&std::cerr,
-                                                  "Before unpacking:\n")));
-  PM.add (createUnpackTraceFunctionPass (Target, TF));
-  DEBUG(PM.add (createMachineFunctionPrinterPass (&std::cerr,
-                                                  "After unpacking:\n")));
-  Target->addPassesToEmitMachineCode (PM, *MCE);
-  PM.run (*TF->TraceFn);
+  // Run optimization passes, then unpack the trace function.
+  UTF->setTraceFunction (TF); // Pass out-of-band information to UTF...
+  PM->run (*TF->TraceFn);
 
   // Add a branch from address A (the parameter to this method) to the 
   // entry basic block of the unpacked TraceFn. Future executions of the trace
   // will proceed from the optimized version of the code.
   uint64_t traceStartAddr = getPointerToGlobalIfAvailable (TF->TraceFn);
   assert (traceStartAddr && "Address of code for TraceFn was NULL after JIT?!");
-  
   DEBUG(std::cerr << "Writing branch at 0x" << std::hex << a
                   << " to point to 0x" << traceStartAddr << std::dec << "\n");
   vm->writeBranchInstruction(a, traceStartAddr);




From gaeke at cs.uiuc.edu  Sun May 23 05:06:14 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Sun May 23 05:06:14 2004
Subject: [llvm-commits] CVS: reopt/include/reopt/TraceToFunction.h
Message-ID: <200405231005.FAA00453@seraph.cs.uiuc.edu>


Changes in directory reopt/include/reopt:

TraceToFunction.h updated: 1.11 -> 1.12

---
Log message:

Remove declaration of createUnpackTraceFunctionPass().


---
Diffs of the changes:  (+0 -6)

Index: reopt/include/reopt/TraceToFunction.h
diff -u reopt/include/reopt/TraceToFunction.h:1.11 reopt/include/reopt/TraceToFunction.h:1.12
--- reopt/include/reopt/TraceToFunction.h:1.11	Mon May 10 19:20:00 2004
+++ reopt/include/reopt/TraceToFunction.h	Sun May 23 05:05:00 2004
@@ -89,12 +89,6 @@
   }
 };
 
-class FunctionPass;
-class TargetMachine;
-
-FunctionPass *createUnpackTraceFunctionPass (TargetMachine *TM,
-                                             TraceFunction *TF);
-
 } // end namespace llvm
 
 #endif // TRACETOFUNCTION_H




From gaeke at cs.uiuc.edu  Sun May 23 05:09:01 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Sun May 23 05:09:01 2004
Subject: [llvm-commits] CVS: reopt/include/reopt/UnpackTraceFunction.h
Message-ID: <200405231008.FAA00501@seraph.cs.uiuc.edu>


Changes in directory reopt/include/reopt:

UnpackTraceFunction.h added (r1.1)

---
Log message:

Definition of class UnpackTraceFunction moved into this public header file.


---
Diffs of the changes:  (+64 -0)

Index: reopt/include/reopt/UnpackTraceFunction.h
diff -c /dev/null reopt/include/reopt/UnpackTraceFunction.h:1.1
*** /dev/null	Sun May 23 05:08:48 2004
--- reopt/include/reopt/UnpackTraceFunction.h	Sun May 23 05:08:38 2004
***************
*** 0 ****
--- 1,64 ----
+ //===- reopt/UnpackTraceFunction.h -------------------------------*- C++ -*--=//
+ //
+ // Unpackage functions previously formed from traces by TraceToFunction.
+ //
+ //===----------------------------------------------------------------------===//
+ 
+ #ifndef REOPT_UNPACKTRACEFUNCTION_H
+ #define REOPT_UNPACKTRACEFUNCTION_H
+ 
+ #include "llvm/CodeGen/MachineFunctionPass.h"
+ #include <set>
+ 
+ namespace llvm {
+ 
+ class TargetMachine;
+ class TraceFunction;
+ class AllocInfo;
+ class Type;
+ 
+ /// This pass inserts copies in two places in the machine
+ /// code for TF->TraceFn: first, at the entry basic block to copy the values in
+ /// TF->LiveInSet from TF->MatrixFn's registers to TF->TraceFn's registers, and
+ /// second, at each exit basic block to copy the values in TF->LiveOutSet from
+ /// TF->TraceFn's registers to TF->MatrixFn's registers.
+ ///
+ class UnpackTraceFunction : public MachineFunctionPass {
+   TargetMachine *TM;
+   TraceFunction *TF;
+ 
+   /// Static stack size of the MachineFunction. Filled in by rewriteProlog ()
+   /// by calling getStaticStackSize (), and then used again by rewriteEpilog ().
+   ///
+   unsigned StaticStackSize;
+ 
+   /// Total stack size of the MachineFunction, including space for regs
+   /// modified in TraceFn to be saved.
+   ///
+   int TotalStackSize;
+ 
+   /// Registers clobbered in the trace. Filled in by findRegsUsedInFunction ().
+   ///
+   std::set<unsigned> RegsToRestore;
+   std::set<unsigned> RegsToSave;
+ 
+   unsigned getStaticStackSize (MachineFunction &MF);
+   unsigned stackOffsetForReg (unsigned R);
+   void findRegsToRestore (MachineFunction &MF);
+   void insertCopyMachineInstrs (AllocInfo &Source, AllocInfo &Target,
+                                 MachineBasicBlock &B, const Type *Ty);
+   void insertBranchMachineInstrs (uint64_t Target, MachineBasicBlock &B);
+   const MachineInstr *containsReturnInstr (MachineBasicBlock &B);
+   void rewriteProlog (MachineFunction &MF, MachineBasicBlock &MBB);
+   void rewriteEpilog (MachineFunction &MF, MachineBasicBlock &MBB);
+ 
+ public:
+   UnpackTraceFunction (TargetMachine *TM_) : TM (TM_), TF (0) { }
+   void setTraceFunction (TraceFunction *TF_) { TF = TF_; }
+   const char *getPassName () const { return "Unpack trace function"; }
+   virtual bool runOnMachineFunction (MachineFunction &MF);
+ };
+ 
+ } // end namespace llvm
+ 
+ #endif // REOPT_UNPACKTRACEFUNCTION_H




From gaeke at cs.uiuc.edu  Sun May 23 05:10:01 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Sun May 23 05:10:01 2004
Subject: [llvm-commits] CVS: reopt/include/reopt/GetTraceTime.h CFG.h
Message-ID: <200405231009.FAA00541@seraph.cs.uiuc.edu>


Changes in directory reopt/include/reopt:

GetTraceTime.h updated: 1.6 -> 1.7
CFG.h updated: 1.3 -> 1.4

---
Log message:

Regularize #include guards.


---
Diffs of the changes:  (+4 -4)

Index: reopt/include/reopt/GetTraceTime.h
diff -u reopt/include/reopt/GetTraceTime.h:1.6 reopt/include/reopt/GetTraceTime.h:1.7
--- reopt/include/reopt/GetTraceTime.h:1.6	Wed Nov 19 16:51:41 2003
+++ reopt/include/reopt/GetTraceTime.h	Sun May 23 05:09:42 2004
@@ -7,8 +7,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef LLVM_REOPTIMIZER_TRACECACHE_GETTRACETIME_H
-#define LLVM_REOPTIMIZER_TRACECACHE_GETTRACETIME_H
+#ifndef REOPT_GETTRACETIME_H
+#define REOPT_GETTRACETIME_H
 
 namespace llvm {
 


Index: reopt/include/reopt/CFG.h
diff -u reopt/include/reopt/CFG.h:1.3 reopt/include/reopt/CFG.h:1.4
--- reopt/include/reopt/CFG.h:1.3	Wed Nov 19 16:51:41 2003
+++ reopt/include/reopt/CFG.h	Sun May 23 05:09:42 2004
@@ -4,8 +4,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef LLVM_REOPTIMIZER_CFG_H
-#define LLVM_REOPTIMIZER_CFG_H
+#ifndef REOPT_CFG_H
+#define REOPT_CFG_H
 
 #include <list>
 #include <vector>




From gaeke at cs.uiuc.edu  Sun May 23 05:10:08 2004
From: gaeke at cs.uiuc.edu (Brian Gaeke)
Date: Sun May 23 05:10:08 2004
Subject: [llvm-commits] CVS: reopt/include/reopt/TraceToFunction.h
Message-ID: <200405231009.FAA00548@seraph.cs.uiuc.edu>


Changes in directory reopt/include/reopt:

TraceToFunction.h updated: 1.12 -> 1.13

---
Log message:

Regularize #include guards and fix head-of-file comment.


---
Diffs of the changes:  (+4 -4)

Index: reopt/include/reopt/TraceToFunction.h
diff -u reopt/include/reopt/TraceToFunction.h:1.12 reopt/include/reopt/TraceToFunction.h:1.13
--- reopt/include/reopt/TraceToFunction.h:1.12	Sun May 23 05:05:00 2004
+++ reopt/include/reopt/TraceToFunction.h	Sun May 23 05:09:43 2004
@@ -1,12 +1,12 @@
-//===- LightWtProfiling/TraceToFunction.h ------------------------*- C++ -*--=//
+//===- reopt/TraceToFunction.h -----------------------------------*- C++ -*--=//
 //
 // Repackage traces as functions, so that global (function-level) optimizations
 // can be applied to traces.
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef TRACETOFUNCTION_H
-#define TRACETOFUNCTION_H
+#ifndef REOPT_TRACETOFUNCTION_H
+#define REOPT_TRACETOFUNCTION_H
 
 #include "llvm/Analysis/Trace.h"
 #include "llvm/Function.h"
@@ -91,4 +91,4 @@
 
 } // end namespace llvm
 
-#endif // TRACETOFUNCTION_H
+#endif // REOPT_TRACETOFUNCTION_H




From alkis at cs.uiuc.edu  Sun May 23 05:14:02 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sun May 23 05:14:02 2004
Subject: [llvm-commits] CVS: llvm-java/lib/Compiler/Compiler.cpp
Message-ID: <200405231012.FAA12392@zion.cs.uiuc.edu>


Changes in directory llvm-java/lib/Compiler:

Compiler.cpp updated: 1.7 -> 1.8

---
Log message:

Correct java opcode name.


---
Diffs of the changes:  (+4 -4)

Index: llvm-java/lib/Compiler/Compiler.cpp
diff -u llvm-java/lib/Compiler/Compiler.cpp:1.7 llvm-java/lib/Compiler/Compiler.cpp:1.8
--- llvm-java/lib/Compiler/Compiler.cpp:1.7	Sat May 22 21:08:01 2004
+++ llvm-java/lib/Compiler/Compiler.cpp	Sun May 23 05:11:59 2004
@@ -142,8 +142,8 @@
         case IF_ICMPGE:
         case IF_ICMPGT:
         case IF_ICMPLE:
-        case IF_ICMPACMPEQ:
-        case IF_ICMPACMPNE:
+        case IF_IACMPEQ:
+        case IF_IACMPNE:
         case IFNULL:
         case IFNONNULL: {
             unsigned index = readShortUnsigned(code, i);
@@ -531,8 +531,8 @@
         case IF_ICMPGE:
         case IF_ICMPGT:
         case IF_ICMPLE:
-        case IF_ICMPACMPEQ:
-        case IF_ICMPACMPNE:
+        case IF_IACMPEQ:
+        case IF_IACMPNE:
         case GOTO:
         case JSR:
         case RET:




From alkis at cs.uiuc.edu  Sun May 23 05:14:09 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sun May 23 05:14:09 2004
Subject: [llvm-commits] CVS: llvm-java/include/llvm/Java/Bytecode.h
Message-ID: <200405231012.FAA12393@zion.cs.uiuc.edu>


Changes in directory llvm-java/include/llvm/Java:

Bytecode.h updated: 1.2 -> 1.3

---
Log message:

Correct java opcode name.


---
Diffs of the changes:  (+2 -2)

Index: llvm-java/include/llvm/Java/Bytecode.h
diff -u llvm-java/include/llvm/Java/Bytecode.h:1.2 llvm-java/include/llvm/Java/Bytecode.h:1.3
--- llvm-java/include/llvm/Java/Bytecode.h:1.2	Fri May 21 16:35:56 2004
+++ llvm-java/include/llvm/Java/Bytecode.h	Sun May 23 05:12:00 2004
@@ -184,8 +184,8 @@
         IF_ICMPGE = 0xA2,
         IF_ICMPGT = 0xA3,
         IF_ICMPLE = 0xA4,
-        IF_ICMPACMPEQ = 0xA5,
-        IF_ICMPACMPNE = 0xA6,
+        IF_IACMPEQ = 0xA5,
+        IF_IACMPNE = 0xA6,
         GOTO = 0xA7,
         JSR = 0xA8,
         RET = 0xA9,




From alkis at cs.uiuc.edu  Sun May 23 05:19:01 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sun May 23 05:19:01 2004
Subject: [llvm-commits] CVS: llvm-java/include/llvm/Java/Bytecode.h
Message-ID: <200405231017.FAA13159@zion.cs.uiuc.edu>


Changes in directory llvm-java/include/llvm/Java:

Bytecode.h updated: 1.3 -> 1.4

---
Log message:

Move immediate reading functions to bytecode header.


---
Diffs of the changes:  (+244 -209)

Index: llvm-java/include/llvm/Java/Bytecode.h
diff -u llvm-java/include/llvm/Java/Bytecode.h:1.3 llvm-java/include/llvm/Java/Bytecode.h:1.4
--- llvm-java/include/llvm/Java/Bytecode.h:1.3	Sun May 23 05:12:00 2004
+++ llvm-java/include/llvm/Java/Bytecode.h	Sun May 23 05:17:31 2004
@@ -16,214 +16,249 @@
 
 #include <stdint.h>
 
-namespace llvm { namespace Java { namespace Opcode {
+namespace llvm { namespace Java {
 
-    enum {
-        NOP = 0x00,
-        ACONST_NULL = 0x01,
-        ICONST_M1 = 0x02,
-        ICONST_0 = 0x03,
-        ICONST_1 = 0x04,
-        ICONST_2 = 0x05,
-        ICONST_3 = 0x06,
-        ICONST_4 = 0x07,
-        ICONST_5 = 0x08,
-        LCONST_0 = 0x09,
-        LCONST_1 = 0x0A,
-        FCONST_0 = 0x0B,
-        FCONST_1 = 0x0C,
-        FCONST_2 = 0x0D,
-        DCONST_0 = 0x0E,
-        DCONST_1 = 0x0F,
-        BIPUSH = 0x10,
-        SIPUSH = 0x11,
-        LDC = 0x12,
-        LDC_W = 0x13,
-        LDC2_W = 0x14,
-        ILOAD = 0x15,
-        LLOAD = 0x16,
-        FLOAD = 0x17,
-        DLOAD = 0x18,
-        ALOAD = 0x19,
-        ILOAD_0 = 0x1A,
-        ILOAD_1 = 0x1B,
-        ILOAD_2 = 0x1C,
-        ILOAD_3 = 0x1D,
-        LLOAD_0 = 0x1E,
-        LLOAD_1 = 0x1F,
-        LLOAD_2 = 0x20,
-        LLOAD_3 = 0x21,
-        FLOAD_0 = 0x22,
-        FLOAD_1 = 0x23,
-        FLOAD_2 = 0x24,
-        FLOAD_3 = 0x25,
-        DLOAD_0 = 0x26,
-        DLOAD_1 = 0x27,
-        DLOAD_2 = 0x28,
-        DLOAD_3 = 0x29,
-        ALOAD_0 = 0x2A,
-        ALOAD_1 = 0x2B,
-        ALOAD_2 = 0x2C,
-        ALOAD_3 = 0x2D,
-        IALOAD = 0x2E,
-        LALOAD = 0x2F,
-        FALOAD = 0x30,
-        DALOAD = 0x31,
-        AALOAD = 0x32,
-        BALOAD = 0x33,
-        CALOAD = 0x34,
-        SALOAD = 0x35,
-        ISTORE = 0x36,
-        LSTORE = 0x37,
-        FSTORE = 0x38,
-        DSTORE = 0x39,
-        ASTORE = 0x3A,
-        ISTORE_0 = 0x3B,
-        ISTORE_1 = 0x3C,
-        ISTORE_2 = 0x3D,
-        ISTORE_3 = 0x3E,
-        LSTORE_0 = 0x3F,
-        LSTORE_1 = 0x40,
-        LSTORE_2 = 0x41,
-        LSTORE_3 = 0x42,
-        FSTORE_0 = 0x43,
-        FSTORE_1 = 0x44,
-        FSTORE_2 = 0x45,
-        FSTORE_3 = 0x46,
-        DSTORE_0 = 0x47,
-        DSTORE_1 = 0x48,
-        DSTORE_2 = 0x49,
-        DSTORE_3 = 0x4A,
-        ASTORE_0 = 0x4B,
-        ASTORE_1 = 0x4C,
-        ASTORE_2 = 0x4D,
-        ASTORE_3 = 0x4E,
-        IASTORE = 0x4F,
-        LASTORE = 0x50,
-        FASTORE = 0x51,
-        DASTORE = 0x52,
-        AASTORE = 0x53,
-        BASTORE = 0x54,
-        CASTORE = 0x55,
-        SASTORE = 0x56,
-        POP = 0x57,
-        POP2 = 0x58,
-        DUP = 0x59,
-        DUP_X1 = 0x5A,
-        DUP_X2 = 0x5B,
-        DUP2 = 0x5C,
-        DUP2_X1 = 0x5D,
-        DUP2_X2 = 0x5E,
-        SWAP = 0x5F,
-        IADD = 0x60,
-        LADD = 0x61,
-        FADD = 0x62,
-        DADD = 0x63,
-        ISUB = 0x64,
-        LSUB = 0x65,
-        FSUB = 0x66,
-        DSUB = 0x67,
-        IMUL = 0x68,
-        LMUL = 0x69,
-        FMUL = 0x6A,
-        DMUL = 0x6B,
-        IDIV = 0x6C,
-        LDIV = 0x6D,
-        FDIV = 0x6E,
-        DDIV = 0x6F,
-        IREM = 0x70,
-        LREM = 0x71,
-        FREM = 0x72,
-        DREM = 0x73,
-        INEG = 0x74,
-        LNEG = 0x75,
-        FNEG = 0x76,
-        DNEG = 0x77,
-        ISHL = 0x78,
-        LSHL = 0x79,
-        ISHR = 0x7A,
-        LSHR = 0x7B,
-        IUSHR = 0x7C,
-        LUSHR = 0x7D,
-        IAND = 0x7E,
-        LAND = 0x7F,
-        IOR = 0x80,
-        LOR = 0x81,
-        IXOR = 0x82,
-        LXOR = 0x83,
-        IINC = 0x84,
-        I2L = 0x85,
-        I2F = 0x86,
-        I2D = 0x87,
-        L2I = 0x88,
-        L2F = 0x89,
-        L2D = 0x8A,
-        F2I = 0x8B,
-        F2L = 0x8C,
-        F2D = 0x8D,
-        D2I = 0x8E,
-        D2L = 0x8F,
-        D2F = 0x90,
-        I2B = 0x91,
-        I2C = 0x92,
-        I2S = 0x93,
-        LCMP = 0x94,
-        FCMPL = 0x95,
-        FCMPG = 0x96,
-        DCMPL = 0x97,
-        DCMPG = 0x98,
-        IFEQ = 0x99,
-        IFNE = 0x9A,
-        IFLT = 0x9B,
-        IFGE = 0x9C,
-        IFGT = 0x9D,
-        IFLE = 0x9E,
-        IF_ICMPEQ = 0x9F,
-        IF_ICMPNE = 0xA0,
-        IF_ICMPLT = 0xA1,
-        IF_ICMPGE = 0xA2,
-        IF_ICMPGT = 0xA3,
-        IF_ICMPLE = 0xA4,
-        IF_IACMPEQ = 0xA5,
-        IF_IACMPNE = 0xA6,
-        GOTO = 0xA7,
-        JSR = 0xA8,
-        RET = 0xA9,
-        TABLESWITCH = 0xAA,
-        LOOKUPSWITCH = 0xAB,
-        IRETURN = 0xAC,
-        LRETURN = 0xAD,
-        FRETURN = 0xAE,
-        DRETURN = 0xAF,
-        ARETURN = 0xB0,
-        RETURN = 0xB1,
-        GETSTATIC = 0xB2,
-        PUTSTATIC = 0xB3,
-        GETFIELD = 0xB4,
-        PUTFIELD = 0xB5,
-        INVOKEVIRTUAL = 0xB6,
-        INVOKESPECIAL = 0xB7,
-        INVOKESTATIC = 0xB8,
-        INVOKEINTERFACE = 0xB9,
-        XXXUNUSEDXXX = 0xBA,
-        NEW = 0xBB,
-        NEWARRAY = 0xBC,
-        ANEWARRAY = 0xBD,
-        ARRAYLENGTH = 0xBE,
-        ATHROW = 0xBF,
-        CHECKCAST = 0xC0,
-        INSTANCEOF = 0xC1,
-        MONITORENTER = 0xC2,
-        MONITOREXIT = 0xC3,
-        WIDE = 0xC4,
-        MULTIANEWARRAY = 0xC5,
-        IFNULL = 0xC6,
-        IFNONNULL = 0xC7,
-        GOTO_W = 0xC8,
-        JSR_W = 0xC9,
-        BREAKPOINT = 0xCA,
-        IMPDEP1 = 0xFE,
-        IMPDEP2 = 0xFF,
-    };
+    namespace Opcode {
 
-} } } // namespace llvm::Java::BC
+        enum {
+            NOP = 0x00,
+            ACONST_NULL = 0x01,
+            ICONST_M1 = 0x02,
+            ICONST_0 = 0x03,
+            ICONST_1 = 0x04,
+            ICONST_2 = 0x05,
+            ICONST_3 = 0x06,
+            ICONST_4 = 0x07,
+            ICONST_5 = 0x08,
+            LCONST_0 = 0x09,
+            LCONST_1 = 0x0A,
+            FCONST_0 = 0x0B,
+            FCONST_1 = 0x0C,
+            FCONST_2 = 0x0D,
+            DCONST_0 = 0x0E,
+            DCONST_1 = 0x0F,
+            BIPUSH = 0x10,
+            SIPUSH = 0x11,
+            LDC = 0x12,
+            LDC_W = 0x13,
+            LDC2_W = 0x14,
+            ILOAD = 0x15,
+            LLOAD = 0x16,
+            FLOAD = 0x17,
+            DLOAD = 0x18,
+            ALOAD = 0x19,
+            ILOAD_0 = 0x1A,
+            ILOAD_1 = 0x1B,
+            ILOAD_2 = 0x1C,
+            ILOAD_3 = 0x1D,
+            LLOAD_0 = 0x1E,
+            LLOAD_1 = 0x1F,
+            LLOAD_2 = 0x20,
+            LLOAD_3 = 0x21,
+            FLOAD_0 = 0x22,
+            FLOAD_1 = 0x23,
+            FLOAD_2 = 0x24,
+            FLOAD_3 = 0x25,
+            DLOAD_0 = 0x26,
+            DLOAD_1 = 0x27,
+            DLOAD_2 = 0x28,
+            DLOAD_3 = 0x29,
+            ALOAD_0 = 0x2A,
+            ALOAD_1 = 0x2B,
+            ALOAD_2 = 0x2C,
+            ALOAD_3 = 0x2D,
+            IALOAD = 0x2E,
+            LALOAD = 0x2F,
+            FALOAD = 0x30,
+            DALOAD = 0x31,
+            AALOAD = 0x32,
+            BALOAD = 0x33,
+            CALOAD = 0x34,
+            SALOAD = 0x35,
+            ISTORE = 0x36,
+            LSTORE = 0x37,
+            FSTORE = 0x38,
+            DSTORE = 0x39,
+            ASTORE = 0x3A,
+            ISTORE_0 = 0x3B,
+            ISTORE_1 = 0x3C,
+            ISTORE_2 = 0x3D,
+            ISTORE_3 = 0x3E,
+            LSTORE_0 = 0x3F,
+            LSTORE_1 = 0x40,
+            LSTORE_2 = 0x41,
+            LSTORE_3 = 0x42,
+            FSTORE_0 = 0x43,
+            FSTORE_1 = 0x44,
+            FSTORE_2 = 0x45,
+            FSTORE_3 = 0x46,
+            DSTORE_0 = 0x47,
+            DSTORE_1 = 0x48,
+            DSTORE_2 = 0x49,
+            DSTORE_3 = 0x4A,
+            ASTORE_0 = 0x4B,
+            ASTORE_1 = 0x4C,
+            ASTORE_2 = 0x4D,
+            ASTORE_3 = 0x4E,
+            IASTORE = 0x4F,
+            LASTORE = 0x50,
+            FASTORE = 0x51,
+            DASTORE = 0x52,
+            AASTORE = 0x53,
+            BASTORE = 0x54,
+            CASTORE = 0x55,
+            SASTORE = 0x56,
+            POP = 0x57,
+            POP2 = 0x58,
+            DUP = 0x59,
+            DUP_X1 = 0x5A,
+            DUP_X2 = 0x5B,
+            DUP2 = 0x5C,
+            DUP2_X1 = 0x5D,
+            DUP2_X2 = 0x5E,
+            SWAP = 0x5F,
+            IADD = 0x60,
+            LADD = 0x61,
+            FADD = 0x62,
+            DADD = 0x63,
+            ISUB = 0x64,
+            LSUB = 0x65,
+            FSUB = 0x66,
+            DSUB = 0x67,
+            IMUL = 0x68,
+            LMUL = 0x69,
+            FMUL = 0x6A,
+            DMUL = 0x6B,
+            IDIV = 0x6C,
+            LDIV = 0x6D,
+            FDIV = 0x6E,
+            DDIV = 0x6F,
+            IREM = 0x70,
+            LREM = 0x71,
+            FREM = 0x72,
+            DREM = 0x73,
+            INEG = 0x74,
+            LNEG = 0x75,
+            FNEG = 0x76,
+            DNEG = 0x77,
+            ISHL = 0x78,
+            LSHL = 0x79,
+            ISHR = 0x7A,
+            LSHR = 0x7B,
+            IUSHR = 0x7C,
+            LUSHR = 0x7D,
+            IAND = 0x7E,
+            LAND = 0x7F,
+            IOR = 0x80,
+            LOR = 0x81,
+            IXOR = 0x82,
+            LXOR = 0x83,
+            IINC = 0x84,
+            I2L = 0x85,
+            I2F = 0x86,
+            I2D = 0x87,
+            L2I = 0x88,
+            L2F = 0x89,
+            L2D = 0x8A,
+            F2I = 0x8B,
+            F2L = 0x8C,
+            F2D = 0x8D,
+            D2I = 0x8E,
+            D2L = 0x8F,
+            D2F = 0x90,
+            I2B = 0x91,
+            I2C = 0x92,
+            I2S = 0x93,
+            LCMP = 0x94,
+            FCMPL = 0x95,
+            FCMPG = 0x96,
+            DCMPL = 0x97,
+            DCMPG = 0x98,
+            IFEQ = 0x99,
+            IFNE = 0x9A,
+            IFLT = 0x9B,
+            IFGE = 0x9C,
+            IFGT = 0x9D,
+            IFLE = 0x9E,
+            IF_ICMPEQ = 0x9F,
+            IF_ICMPNE = 0xA0,
+            IF_ICMPLT = 0xA1,
+            IF_ICMPGE = 0xA2,
+            IF_ICMPGT = 0xA3,
+            IF_ICMPLE = 0xA4,
+            IF_IACMPEQ = 0xA5,
+            IF_IACMPNE = 0xA6,
+            GOTO = 0xA7,
+            JSR = 0xA8,
+            RET = 0xA9,
+            TABLESWITCH = 0xAA,
+            LOOKUPSWITCH = 0xAB,
+            IRETURN = 0xAC,
+            LRETURN = 0xAD,
+            FRETURN = 0xAE,
+            DRETURN = 0xAF,
+            ARETURN = 0xB0,
+            RETURN = 0xB1,
+            GETSTATIC = 0xB2,
+            PUTSTATIC = 0xB3,
+            GETFIELD = 0xB4,
+            PUTFIELD = 0xB5,
+            INVOKEVIRTUAL = 0xB6,
+            INVOKESPECIAL = 0xB7,
+            INVOKESTATIC = 0xB8,
+            INVOKEINTERFACE = 0xB9,
+            XXXUNUSEDXXX = 0xBA,
+            NEW = 0xBB,
+            NEWARRAY = 0xBC,
+            ANEWARRAY = 0xBD,
+            ARRAYLENGTH = 0xBE,
+            ATHROW = 0xBF,
+            CHECKCAST = 0xC0,
+            INSTANCEOF = 0xC1,
+            MONITORENTER = 0xC2,
+            MONITOREXIT = 0xC3,
+            WIDE = 0xC4,
+            MULTIANEWARRAY = 0xC5,
+            IFNULL = 0xC6,
+            IFNONNULL = 0xC7,
+            GOTO_W = 0xC8,
+            JSR_W = 0xC9,
+            BREAKPOINT = 0xCA,
+            IMPDEP1 = 0xFE,
+            IMPDEP2 = 0xFF,
+        };
+
+    } // namespace Opcode
+
+    inline int readSByte(const uint8_t* code, unsigned& i) {
+        return code[++i];
+    }
+
+    inline unsigned readUByte(const uint8_t* code, unsigned& i) {
+        return code[++i];
+    }
+
+    inline int readSShort(const uint8_t* code, unsigned& i) {
+        return (readSByte(code, i) << 8) | readUByte(code, i);
+    }
+
+    inline unsigned readUShort(const uint8_t* code, unsigned& i) {
+        return (readUByte(code, i) << 8) | readUByte(code, i);
+    }
+
+    inline int readSInt(const uint8_t* code, unsigned& i) {
+        return ((readUByte(code, i) << 24) |
+                (readUByte(code, i) << 16) |
+                (readUByte(code, i) << 8) |
+                readUByte(code, i));
+    }
+
+    inline unsigned readUInt(const uint8_t* code, unsigned& i) {
+        return readSInt(code, i);
+    }
+
+    inline void skipPadBytes(const uint8_t* code, unsigned& i) {
+        while (((unsigned)&code[++i]) & 0XFF);
+    }
+
+} } // namespace llvm::Java




From alkis at cs.uiuc.edu  Sun May 23 05:19:07 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sun May 23 05:19:07 2004
Subject: [llvm-commits] CVS: llvm-java/lib/Compiler/Compiler.cpp
Message-ID: <200405231017.FAA13152@zion.cs.uiuc.edu>


Changes in directory llvm-java/lib/Compiler:

Compiler.cpp updated: 1.8 -> 1.9

---
Log message:

Move immediate reading functions to bytecode header.


---
Diffs of the changes:  (+14 -45)

Index: llvm-java/lib/Compiler/Compiler.cpp
diff -u llvm-java/lib/Compiler/Compiler.cpp:1.8 llvm-java/lib/Compiler/Compiler.cpp:1.9
--- llvm-java/lib/Compiler/Compiler.cpp:1.8	Sun May 23 05:11:59 2004
+++ llvm-java/lib/Compiler/Compiler.cpp	Sun May 23 05:17:31 2004
@@ -34,37 +34,6 @@
         return !isTwoSlotValue(v);
     }
 
-    inline int readByteSigned(const uint8_t* code, unsigned& i) {
-        return code[++i];
-    }
-
-    inline unsigned readByteUnsigned(const uint8_t* code, unsigned& i) {
-        return code[++i];
-    }
-
-    inline int readShortSigned(const uint8_t* code, unsigned& i) {
-        return (readByteSigned(code, i) << 8) | readByteUnsigned(code, i);
-    }
-
-    inline unsigned readShortUnsigned(const uint8_t* code, unsigned& i) {
-        return (readByteUnsigned(code, i) << 8) | readByteUnsigned(code, i);
-    }
-
-    inline int readIntSigned(const uint8_t* code, unsigned& i) {
-        return ((readByteUnsigned(code, i) << 24) |
-                (readByteUnsigned(code, i) << 16) |
-                (readByteUnsigned(code, i) << 8) |
-                readByteUnsigned(code, i));
-    }
-
-    inline unsigned readIntUnsigned(const uint8_t* code, unsigned& i) {
-        return readIntSigned(code, i);
-    }
-
-    inline void skipPadBytes(const uint8_t* code, unsigned& i) {
-        while (((unsigned)&code[++i]) & 0XFF);
-    }
-
 } // namespace
 
 void Compiler::compileMethodInit(Function& function,
@@ -146,19 +115,19 @@
         case IF_IACMPNE:
         case IFNULL:
         case IFNONNULL: {
-            unsigned index = readShortUnsigned(code, i);
+            unsigned index = readUShort(code, i);
             bc2bbMap_[bcStart] = new BasicBlock(
                 std::string("bb at bc") + utostr(bcStart), &function);
             break;
         }
         case TABLESWITCH: {
             skipPadBytes(code, i);
-            readIntSigned(code, i);
-            int low = readIntSigned(code, i);
-            int high = readIntSigned(code, i);
+            readSInt(code, i);
+            int low = readSInt(code, i);
+            int high = readSInt(code, i);
             unsigned offsetCount = high - low + 1;
             while (offsetCount--) {
-                unsigned bcIndex = bcStart + readIntSigned(code, i);
+                unsigned bcIndex = bcStart + readSInt(code, i);
                 bc2bbMap_[bcIndex] = new BasicBlock(
                     std::string("bb at bc") + utostr(bcIndex), &function);
             }
@@ -166,10 +135,10 @@
         }
         case LOOKUPSWITCH: {
             skipPadBytes(code, i);
-            unsigned pairCount = readIntUnsigned(code, i);
+            unsigned pairCount = readUInt(code, i);
             while (pairCount--) {
-                readIntSigned(code, i);
-                unsigned bcIndex = bcStart + readIntSigned(code, i);
+                readSInt(code, i);
+                unsigned bcIndex = bcStart + readSInt(code, i);
                 bc2bbMap_[bcIndex] = new BasicBlock(
                     std::string("bb at bc") + utostr(bcIndex), &function);
             }
@@ -244,25 +213,25 @@
             opStack_.push(ConstantFP::get(Type::DoubleTy, code[i]-DCONST_0));
             break;
         case BIPUSH: {
-            int imm = readByteSigned(code, i);
+            int imm = readSByte(code, i);
             opStack_.push(ConstantInt::get(Type::IntTy, imm));
             break;
         }
         case SIPUSH: {
-            int imm = readShortSigned(code, i);
+            int imm = readSShort(code, i);
             opStack_.push(ConstantInt::get(Type::IntTy, imm));
             break;
         }
         case LDC: {
-            unsigned index = readByteUnsigned(code, i);
+            unsigned index = readUByte(code, i);
             // FIXME: load constant from constant pool
         }
         case LDC_W: {
-            unsigned index = readShortUnsigned(code, i);
+            unsigned index = readUShort(code, i);
             // FIXME: load constant from constant pool
         }
         case LDC2_W: {
-            unsigned index = readShortUnsigned(code, i);
+            unsigned index = readUShort(code, i);
             // FIXME: load constant from constant pool
         }
         case ILOAD:
@@ -271,7 +240,7 @@
         case DLOAD:
         case ALOAD: {
             // FIXME: use opcodes to perform type checking
-            unsigned index = readByteUnsigned(code, i);
+            unsigned index = readUByte(code, i);
             opStack_.push(locals_[index]);
             break;
         }




From alkis at cs.uiuc.edu  Sun May 23 05:35:01 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sun May 23 05:35:01 2004
Subject: [llvm-commits] CVS: llvm-java/lib/Compiler/Compiler.cpp
Message-ID: <200405231033.FAA14633@zion.cs.uiuc.edu>


Changes in directory llvm-java/lib/Compiler:

Compiler.cpp updated: 1.9 -> 1.10

---
Log message:

Remove one variable from tableswitch skipping code.
Fix lookupswitch skipping code.


---
Diffs of the changes:  (+2 -2)

Index: llvm-java/lib/Compiler/Compiler.cpp
diff -u llvm-java/lib/Compiler/Compiler.cpp:1.9 llvm-java/lib/Compiler/Compiler.cpp:1.10
--- llvm-java/lib/Compiler/Compiler.cpp:1.9	Sun May 23 05:17:31 2004
+++ llvm-java/lib/Compiler/Compiler.cpp	Sun May 23 05:33:37 2004
@@ -125,8 +125,7 @@
             readSInt(code, i);
             int low = readSInt(code, i);
             int high = readSInt(code, i);
-            unsigned offsetCount = high - low + 1;
-            while (offsetCount--) {
+            while (low++ <= high) {
                 unsigned bcIndex = bcStart + readSInt(code, i);
                 bc2bbMap_[bcIndex] = new BasicBlock(
                     std::string("bb at bc") + utostr(bcIndex), &function);
@@ -135,6 +134,7 @@
         }
         case LOOKUPSWITCH: {
             skipPadBytes(code, i);
+            readSInt(code, i);
             unsigned pairCount = readUInt(code, i);
             while (pairCount--) {
                 readSInt(code, i);




From alkis at cs.uiuc.edu  Sun May 23 06:13:02 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sun May 23 06:13:02 2004
Subject: [llvm-commits] CVS: llvm/docs/OpenProjects.html
Message-ID: <200405231111.GAA17416@zion.cs.uiuc.edu>


Changes in directory llvm/docs:

OpenProjects.html updated: 1.25 -> 1.26

---
Log message:

Put back the inlined bug list. Use the OBJECT tag to make it HTML-4.01
(Strict)-compliant.


---
Diffs of the changes:  (+3 -2)

Index: llvm/docs/OpenProjects.html
diff -u llvm/docs/OpenProjects.html:1.25 llvm/docs/OpenProjects.html:1.26
--- llvm/docs/OpenProjects.html:1.25	Wed May 12 13:57:54 2004
+++ llvm/docs/OpenProjects.html	Sun May 23 06:10:55 2004
@@ -57,7 +57,8 @@
 
 <p>The projects in this page are open-ended. More specific projects are
 filed as unassigned enhancements in the <a href="http://llvm.cs.uiuc.edu/bugs/">
-LLVM bug tracker</a>. See the <a href="http://llvm.cs.uiuc.edu/bugs/buglist.cgi?keywords_type=allwords&amp;keywords=&amp;bug_status=NEW&amp;bug_status=ASSIGNED&amp;bug_status=REOPENED&amp;bug_severity=enhancement&amp;emailassigned_to1=1&amp;emailtype1=substring&amp;email1=unassigned">list of currently outstanding issues</a> if you wish to help improve LLVM.</p>
+LLVM bug tracker</a>. Here is the current list:
+<object width="100%" height="400" data="http://llvm.cs.uiuc.edu/bugs/buglist.cgi?keywords_type=allwords&amp;keywords=&amp;bug_status=NEW&amp;bug_status=ASSIGNED&amp;bug_status=REOPENED&amp;bug_severity=enhancement&amp;emailassigned_to1=1&amp;emailtype1=substring&amp;email1=unassigned" /></p>
 
 </div>
 
@@ -333,7 +334,7 @@
 
   <a href="mailto:sabre at nondot.org">Chris Lattner</a><br>
   <a href="http://llvm.cs.uiuc.edu">LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/05/12 18:57:54 $
+  Last modified: $Date: 2004/05/23 11:10:55 $
 </address>
 
 </body>




From alkis at cs.uiuc.edu  Sun May 23 06:16:01 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sun May 23 06:16:01 2004
Subject: [llvm-commits] CVS: llvm-java/lib/Compiler/Compiler.cpp
Message-ID: <200405231114.GAA17929@zion.cs.uiuc.edu>


Changes in directory llvm-java/lib/Compiler:

Compiler.cpp updated: 1.10 -> 1.11

---
Log message:

Use ConstantSInt::get instead of ConstantInt::get so that all possible
int values can be handled.


---
Diffs of the changes:  (+4 -4)

Index: llvm-java/lib/Compiler/Compiler.cpp
diff -u llvm-java/lib/Compiler/Compiler.cpp:1.10 llvm-java/lib/Compiler/Compiler.cpp:1.11
--- llvm-java/lib/Compiler/Compiler.cpp:1.10	Sun May 23 05:33:37 2004
+++ llvm-java/lib/Compiler/Compiler.cpp	Sun May 23 06:14:42 2004
@@ -197,11 +197,11 @@
         case ICONST_3:
         case ICONST_4:
         case ICONST_5:
-            opStack_.push(ConstantInt::get(Type::IntTy, code[i]-ICONST_0));
+            opStack_.push(ConstantSInt::get(Type::IntTy, code[i]-ICONST_0));
             break;
         case LCONST_0:
         case LCONST_1:
-            opStack_.push(ConstantInt::get(Type::LongTy, code[i]-LCONST_0));
+            opStack_.push(ConstantSInt::get(Type::LongTy, code[i]-LCONST_0));
             break;
         case FCONST_0:
         case FCONST_1:
@@ -214,12 +214,12 @@
             break;
         case BIPUSH: {
             int imm = readSByte(code, i);
-            opStack_.push(ConstantInt::get(Type::IntTy, imm));
+            opStack_.push(ConstantSInt::get(Type::IntTy, imm));
             break;
         }
         case SIPUSH: {
             int imm = readSShort(code, i);
-            opStack_.push(ConstantInt::get(Type::IntTy, imm));
+            opStack_.push(ConstantSInt::get(Type::IntTy, imm));
             break;
         }
         case LDC: {




From alkis at cs.uiuc.edu  Sun May 23 06:25:02 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sun May 23 06:25:02 2004
Subject: [llvm-commits] CVS: llvm-java/lib/Compiler/Compiler.cpp
Message-ID: <200405231123.GAA18529@zion.cs.uiuc.edu>


Changes in directory llvm-java/lib/Compiler:

Compiler.cpp updated: 1.11 -> 1.12

---
Log message:

Add compilation for add, sub, mul, div, rem, shl, shr, ushr, and, or,
xor, if*, tableswitch, lookupswitch, and *return opcodes.


---
Diffs of the changes:  (+182 -19)

Index: llvm-java/lib/Compiler/Compiler.cpp
diff -u llvm-java/lib/Compiler/Compiler.cpp:1.11 llvm-java/lib/Compiler/Compiler.cpp:1.12
--- llvm-java/lib/Compiler/Compiler.cpp:1.11	Sun May 23 06:14:42 2004
+++ llvm-java/lib/Compiler/Compiler.cpp	Sun May 23 06:23:24 2004
@@ -434,39 +434,108 @@
         case IADD:
         case LADD:
         case FADD:
-        case DADD:
+        case DADD: {
+            Value* v1 = opStack_.top(); opStack_.pop();
+            Value* v2 = opStack_.top(); opStack_.pop();
+            Instruction* in = BinaryOperator::create(Instruction::Add, v1, v2);
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
+            break;
+        }
         case ISUB:
         case LSUB:
         case FSUB:
-        case DSUB:
+        case DSUB: {
+            Value* v1 = opStack_.top(); opStack_.pop();
+            Value* v2 = opStack_.top(); opStack_.pop();
+            Instruction* in = BinaryOperator::create(Instruction::Sub, v1, v2);
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
+            break;
+        }
         case IMUL:
         case LMUL:
         case FMUL:
-        case DMUL:
+        case DMUL: {
+            Value* v1 = opStack_.top(); opStack_.pop();
+            Value* v2 = opStack_.top(); opStack_.pop();
+            Instruction* in = BinaryOperator::create(Instruction::Mul, v1, v2);
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
+            break;
+        }
         case IDIV:
         case LDIV:
         case FDIV:
-        case DDIV:
+        case DDIV: {
+            Value* v1 = opStack_.top(); opStack_.pop();
+            Value* v2 = opStack_.top(); opStack_.pop();
+            Instruction* in = BinaryOperator::create(Instruction::Div, v1, v2);
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
+            break;
+        }
         case IREM:
         case LREM:
         case FREM:
-        case DREM:
+        case DREM: {
+            Value* v1 = opStack_.top(); opStack_.pop();
+            Value* v2 = opStack_.top(); opStack_.pop();
+            Instruction* in = BinaryOperator::create(Instruction::Rem, v1, v2);
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
+            break;
+        }
         case INEG:
         case LNEG:
         case FNEG:
         case DNEG:
         case ISHL:
-        case LSHL:
+        case LSHL: {
+            Value* v1 = opStack_.top(); opStack_.pop();
+            Value* v2 = opStack_.top(); opStack_.pop();
+            Instruction* in = new ShiftInst(Instruction::Shl, v1, v2);
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
+            break;
+        }
         case ISHR:
-        case LSHR:
+        case LSHR: {
+            Value* v1 = opStack_.top(); opStack_.pop();
+            Value* v2 = opStack_.top(); opStack_.pop();
+            Instruction* in = new ShiftInst(Instruction::Shr, v1, v2);
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
+            break;
+        }
         case IUSHR:
-        case LUSHR:
+        case LUSHR: {
+            Value* v1 = opStack_.top(); opStack_.pop();
+            Instruction* in =
+                new CastInst(v1, v1->getType()->getUnsignedVersion());
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
+            Value* v2 = opStack_.top(); opStack_.pop();
+            in = new ShiftInst(Instruction::Shr, in, v2);
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
+            break;
+        }
         case IAND:
-        case LAND:
+        case LAND: {
+            Value* v1 = opStack_.top(); opStack_.pop();
+            Value* v2 = opStack_.top(); opStack_.pop();
+            Instruction* in = BinaryOperator::create(Instruction::And, v1, v2);
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
+            break;
+        }
         case IOR:
-        case LOR:
+        case LOR: {
+            Value* v1 = opStack_.top(); opStack_.pop();
+            Value* v2 = opStack_.top(); opStack_.pop();
+            Instruction* in = BinaryOperator::create(Instruction::Or, v1, v2);
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
+            break;
+        }
         case IXOR:
-        case LXOR:
+        case LXOR: {
+            Value* v1 = opStack_.top(); opStack_.pop();
+            Value* v2 = opStack_.top(); opStack_.pop();
+            Instruction* in = BinaryOperator::create(Instruction::Xor, v1, v2);
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
+            break;
+        }
         case IINC:
         case I2L:
         case I2F:
@@ -493,26 +562,104 @@
         case IFLT:
         case IFGE:
         case IFGT:
-        case IFLE:
+        case IFLE: {
+            static Instruction::BinaryOps java2llvm[] = {
+                Instruction::SetEQ,
+                Instruction::SetNE,
+                Instruction::SetLT,
+                Instruction::SetGE,
+                Instruction::SetGT,
+                Instruction::SetLE,
+            };
+            Value* v1 = opStack_.top(); opStack_.pop();
+            Value* v2 = ConstantSInt::get(Type::IntTy, 0);
+            Instruction* in = new SetCondInst(java2llvm[i-IFEQ], v1, v2);
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
+            new BranchInst(bc2bbMap_[bcStart + readSShort(code, i)],
+                           bc2bbMap_[i + 1],
+                           bc2bbMap_[bcStart]);
+            break;
+        }
         case IF_ICMPEQ:
         case IF_ICMPNE:
         case IF_ICMPLT:
         case IF_ICMPGE:
         case IF_ICMPGT:
-        case IF_ICMPLE:
+        case IF_ICMPLE: {
+            static Instruction::BinaryOps java2llvm[] = {
+                Instruction::SetEQ,
+                Instruction::SetNE,
+                Instruction::SetLT,
+                Instruction::SetGE,
+                Instruction::SetGT,
+                Instruction::SetLE,
+            };
+            Value* v1 = opStack_.top(); opStack_.pop();
+            Value* v2 = opStack_.top(); opStack_.pop();
+            Instruction* in = new SetCondInst(java2llvm[i-IF_ICMPEQ], v1, v2);
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
+            new BranchInst(bc2bbMap_[bcStart + readSShort(code, i)],
+                           bc2bbMap_[i + 1],
+                           bc2bbMap_[bcStart]);
+            break;
+        }
         case IF_IACMPEQ:
-        case IF_IACMPNE:
+        case IF_IACMPNE: {
+            static Instruction::BinaryOps java2llvm[] = {
+                Instruction::SetEQ,
+                Instruction::SetNE,
+            };
+            Value* v1 = opStack_.top(); opStack_.pop();
+            Value* v2 = opStack_.top(); opStack_.pop();
+            Instruction* in = new SetCondInst(java2llvm[i-IF_IACMPEQ], v1, v2);
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
+            new BranchInst(bc2bbMap_[bcStart + readSShort(code, i)],
+                           bc2bbMap_[i + 1],
+                           bc2bbMap_[bcStart]);
+            break;
+        }
         case GOTO:
+            new BranchInst(bc2bbMap_[bcStart + readSShort(code, i)]);
+            break;
         case JSR:
         case RET:
-        case TABLESWITCH:
-        case LOOKUPSWITCH:
+        case TABLESWITCH: {
+            Value* v1 = opStack_.top(); opStack_.pop();
+            skipPadBytes(code, i);
+            int def = readSInt(code, i);
+            int low = readSInt(code, i);
+            int high = readSInt(code, i);
+            SwitchInst* in =
+                new SwitchInst(v1, bc2bbMap_[bcStart + def],bc2bbMap_[bcStart]);
+            while (low <= high)
+                in->addCase(ConstantSInt::get(Type::IntTy, low++),
+                            bc2bbMap_[bcStart + readSInt(code, i)]);
+            break;
+        }
+        case LOOKUPSWITCH: {
+            Value* v1 = opStack_.top(); opStack_.pop();
+            skipPadBytes(code, i);
+            int def = readSInt(code, i);
+            unsigned pairCount = readUInt(code, i);
+            SwitchInst* in =
+                new SwitchInst(v1, bc2bbMap_[bcStart + def],bc2bbMap_[bcStart]);
+            while (pairCount--)
+                in->addCase(ConstantSInt::get(Type::IntTy, readSInt(code, i)),
+                           bc2bbMap_[bcStart + readSInt(code, i)]);
+            break;
+        }
         case IRETURN:
         case LRETURN:
         case FRETURN:
         case DRETURN:
-        case ARETURN:
+        case ARETURN: {
+            Value* v1 = opStack_.top(); opStack_.pop();
+            new ReturnInst(v1, bc2bbMap_[bcStart]);
+            break;
+        }
         case RETURN:
+            new ReturnInst(NULL, bc2bbMap_[bcStart]);
+            break;
         case GETSTATIC:
         case PUTSTATIC:
         case GETFIELD:
@@ -531,11 +678,27 @@
         case INSTANCEOF:
         case MONITORENTER:
         case MONITOREXIT:
-        case WIDE:
+//      case WIDE:
         case MULTIANEWARRAY:
         case IFNULL:
-        case IFNONNULL:
+        case IFNONNULL: {
+            static Instruction::BinaryOps java2llvm[] = {
+                Instruction::SetEQ,
+                Instruction::SetNE,
+            };
+            Value* v1 = opStack_.top(); opStack_.pop();
+            // FIXME: should compare to a null pointer of type Object*
+            Value* v2 =ConstantPointerNull::get(PointerType::get(Type::VoidTy));
+            Instruction* in = new SetCondInst(java2llvm[i-IFNULL], v1, v2);
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
+            new BranchInst(bc2bbMap_[bcStart + readSShort(code, i)],
+                           bc2bbMap_[i + 1],
+                           bc2bbMap_[bcStart]);
+            break;
+        }
         case GOTO_W:
+            new BranchInst(bc2bbMap_[bcStart + readSInt(code, i)]);
+            break;
         case JSR_W:
         case BREAKPOINT:
         case IMPDEP1:




From llvm at cs.uiuc.edu  Sun May 23 12:07:01 2004
From: llvm at cs.uiuc.edu (LLVM)
Date: Sun May 23 12:07:01 2004
Subject: [llvm-commits] CVS: llvm/docs/BytecodeFormat.html
Message-ID: <200405231705.MAA27293@zion.cs.uiuc.edu>


Changes in directory llvm/docs:

BytecodeFormat.html updated: 1.2 -> 1.3

---
Log message:

Additional content describing variable bit rate encoding, correcting the
definition of the signature block and module block, and adding definitions
of the global type pool and symbol table. More still to come.


---
Diffs of the changes:  (+238 -48)

Index: llvm/docs/BytecodeFormat.html
diff -u llvm/docs/BytecodeFormat.html:1.2 llvm/docs/BytecodeFormat.html:1.3
--- llvm/docs/BytecodeFormat.html:1.2	Sat May 22 00:56:41 2004
+++ llvm/docs/BytecodeFormat.html	Sun May 23 12:05:09 2004
@@ -6,7 +6,7 @@
   <link rel="stylesheet" href="llvm.css" type="text/css">
   <style>
     table, tr, td { border: 2px solid gray }
-    th { border: 2px sold gray; font-weight: bold; }
+    th { border: 2px solid gray; font-weight: bold; }
     table { border-collapse: collapse; margin-top: 1em margin-bottom: 1em }
   </style>
 </head>
@@ -19,13 +19,15 @@
       <li><a href="#blocks">Blocks</a></li>
       <li><a href="#lists">Lists</a></li>
       <li><a href="#fields">Fields</a></li>
+      <li><a href="#encoding">Encoding Rules</a></li>
       <li><a href="#align">Alignment</a></li>
     </ol>
   <li><a href="#details">Detailed Layout</a>
     <ol>
       <li><a href="#notation">Notation</a></li>
       <li><a href="#blocktypes">Blocks Types</a></li>
-      <li><a href="#header">Header Block</a></li>
+      <li><a href="#signature">Signature Block</a></li>
+      <li><a href="#module">Module Block</a></li>
       <li><a href="#typeool">Global Type Pool</a></li>
       <li><a href="#modinfo">Module Info Block</a></li>
       <li><a href="#constants">Global Constant Pool</a></li>
@@ -63,7 +65,7 @@
 <p>LLVM bytecode files consist simply of a sequence of blocks of bytes. 
 Each block begins with an identification value that determines the type of 
 the next block.  The possible types of blocks are described below in the section 
-<a href="#blockstypes">Block Types</a>. The block identifier is used because
+<a href="#blocktypes">Block Types</a>. The block identifier is used because
 it is possible for entire blocks to be omitted from the file if they are
 empty. The block identifier helps the reader determine which kind of block is
 next in the file.</p>
@@ -80,8 +82,9 @@
   <li><b>InstructionList (0x32)</b>.</li>
   <li><b>CompactionTable (0x33)</b>.</li>
 </ol>
-<p> Except for the <a href="#header">Header Block</a> all blocks are variable
-length. They consume just enough bytes to express their contents.</p>
+<p> All blocks are variable length. They consume just enough bytes to express 
+their contents.  Each block begins with an integer identifier and the length 
+of the block.</p>
 </div>
 <!-- _______________________________________________________________________ -->
 <div class="doc_subsection"><a name="lists">Lists</a> </div>
@@ -112,6 +115,88 @@
 written and how the bits are to be interpreted.</p>
 </div>
 <!-- _______________________________________________________________________ -->
+<div class="doc_subsection"><a name="encoding">Encoding Primitives</a> </div>
+<div class="doc_text">
+<p>Each field that can be put out is encoded into the file using a small set 
+of primitives. The rules for these primitives are described below.</p>
+<h3>Variable Bit Rate Encoding</h3>
+<p>To minimize the number of bytes written for small quantities, an encoding
+scheme similar to UTF-8 is used to write integer data. The scheme is known as
+variable bit rate (vbr) encoding.  In this encoding, the high bit of each 
+byte is used to indicate if more bytes follow. If (byte &amp; 0x80) is non-zero 
+in any given byte, it means there is another byte immediately following that 
+also contributes to the value. For the final byte (byte &amp; 0x80) is false 
+(the high bit is not set). In each byte only the low seven bits contribute to 
+the value. Consequently 32-bit quantities can take from one to <em>five</em> 
+bytes to encode. In general, smaller quantities will encode in fewer bytes, 
+as follows:</p>
+<table class="doc_table_nw">
+  <tr>
+    <th>Byte #</th>
+    <th>Significant Bits</th>
+    <th>Maximum Value</th>
+  </tr>
+  <tr><td>1</td><td>0-6</td><td>127</td></tr>
+  <tr><td>2</td><td>7-13</td><td>16,383</td></tr>
+  <tr><td>3</td><td>14-20</td><td>2,097,151</td></tr>
+  <tr><td>4</td><td>21-27</td><td>268,435,455</td></tr>
+  <tr><td>5</td><td>28-34</td><td>34,359,738,367</td></tr>
+  <tr><td>6</td><td>35-41</td><td>4,398,046,511,103</td></tr>
+  <tr><td>7</td><td>42-48</td><td>562,949,953,421,311</td></tr>
+  <tr><td>8</td><td>49-55</td><td>72,057,594,037,927,935</td></tr>
+  <tr><td>9</td><td>56-62</td><td>9,223,372,036,854,775,807</td></tr>
+  <tr><td>10</td><td>63-69</td><td>1,180,591,620,717,411,303,423</td></tr>
+</table>
+<p>Note that in practice, the tenth byte could only encode bits 63 and 64
+since the maximum quantity to use this encoding is a 64-bit integer.</p>
+<p>The table below defines the encoding rules for type names used in the
+descriptions of blocks and fields in the next section. Any type name with
+the suffix <em>_vbr</em> indicate a quantity that is encoded using 
+variable bit rate encoding as described above.</p>
+<table class="doc_table" >
+  <tr>
+    <th><b>Type</b></th>
+    <th align="left"><b>Rule</b></th>
+  </tr>
+  <tr>
+    <td>unsigned</td>
+    <td align="left">A 32-bit unsigned integer that always occupies four 
+      consecutive bytes. The unsigned integer is encoded using LSB first 
+      ordering. That is bits 2<sup>0</sup> through 2<sup>7</sup> are in the 
+      byte with the lowest file offset (little endian).</td>
+  </tr><tr>
+    <td>uint_vbr</td>
+    <td align="left">A 32-bit unsigned integer that occupies from one to five 
+    bytes using variable bit rate encoding.</td>
+  </tr><tr>
+    <td>uint64_vbr</td>
+    <td align="left">A 64-bit unsigned integer that occupies from one to ten 
+    bytes using variable bit rate encoding.</td>
+  </tr><tr>
+    <td>int64_vbr</td>
+    <td align="left">A 64-bit signed integer that occupies from one to ten 
+    bytes using variable bit rate encoding.</td>
+  </tr><tr>
+    <td>char</td>
+    <td align="left">A single unsigned character encoded into one byte</td>
+  </tr><tr>
+    <td>bit</td>
+    <td align="left">A single bit within a byte.</td>
+  </tr><tr>
+    <td>string</td>
+    <td align="left">A uint_vbr indicating the length of the character string 
+    immediately followed by the characters of the string. There is no 
+    terminating null byte in the string. Characters are interpreted as unsigned 
+    char and are generally US-ASCII encoded.</td>
+  </tr><tr>
+    <td>data</td>
+    <td align="left">An arbitrarily long segment of data to which no 
+    interpretation is implied. This is used for float, double, and constant 
+    initializers.</td>
+  </tr>
+</table>
+</div>
+<!-- _______________________________________________________________________ -->
 <div class="doc_subsection"><a name="align">Alignment</a> </div>
 <div class="doc_text">
 <p>To support cross-platform differences, the bytecode file is aligned on 
@@ -123,8 +208,8 @@
 <div class="doc_section"> <a name="details">Detailed Layout</a> </div>
 <!-- *********************************************************************** -->
 <div class="doc_text">
-  <p>This section provides the detailed layout of the LLVM bytecode file format.
-  bit and byte level specifics.</p>
+<p>This section provides the detailed layout of the LLVM bytecode file format.
+bit and byte level specifics.</p>
 </div>
 <!-- _______________________________________________________________________ -->
 <div class="doc_subsection"><a name="notation">Notation</a></div>
@@ -153,9 +238,10 @@
   of bytes known as blocks. The blocks are written sequentially to the file in
   the following order:</p>
 <ol>
-  <li><a href="#header">Header</a>. This block contains the file signature 
-  (magic number), machine description and file format version (not LLVM 
-  version).</li>
+  <li><a href="#signature">Signature</a>. This block contains the file signature 
+  (magic number) that identifies the file as LLVM bytecode.</li>
+  <li><a href="#module">Module Block</a>. This is the top level block in a
+  bytecode file. It contains all the other blocks.</li>
   <li><a href="#gtypepool">Global Type Pool</a>. This block contains all the
   global (module) level types.</li>
   <li><a href="#modinfo">Module Info</a>. This block contains the types of the
@@ -171,77 +257,181 @@
 </ol>
 </div>
 <!-- _______________________________________________________________________ -->
-<div class="doc_subsection"><a name="header">Header Block</a> </div>
+<div class="doc_subsection"><a name="signature">Signature Block</a> </div>
 <div class="doc_text">
-<p>The Header Block occurs in every LLVM bytecode file and is always first. It
-simply provides a few bytes of data to identify the file, its format, and the
-bytecode version. This block is fixed length and always eight bytes, as follows:
-<table class="doc_table" width="90%">
+<p>The signature block occurs in every LLVM bytecode file and is always first. 
+It simply provides a few bytes of data to identify the file as being an LLVM
+bytecode file. This block is always four bytes in length and differs from the
+other blocks because there is no identifier and no block length at the start
+of the block. Essentially, this block is just the "magic number" for the file.
+<table class="doc_table_nw" >
   <tr>
     <th><b>Byte(s)</b></th>
     <th><b>Bit(s)</b></th>
     <th><b>Align?</b></th>
     <th><b>Type</b></th>
     <th align="left"><b>Field Description</b></th>
-  </tr>
-  <tr>
-    <td>00</td><td>00-07</td><td>No</td><td>Char</td>
+  </tr><tr>
+    <td>00</td><td>-</td><td>No</td><td>char</td>
     <td align="left">Constant "l" (0x6C)</td>
-  </tr>
-  <tr>
-    <td>01</td><td>00-07</td><td>No</td><td>Char</td>
+  </tr><tr>
+    <td>01</td><td>-</td><td>No</td><td>char</td>
     <td align="left">Constant "l" (0x6C)</td>
-  </tr>
-  <tr>
-    <td>02</td><td>00-07</td><td>No</td><td>Char</td>
+  </tr><tr>
+    <td>02</td><td>-</td><td>No</td><td>char</td>
     <td align="left">Constant "v" (0x76)</td>
-  </tr>
-  <tr>
-    <td>03</td><td>00-07</td><td>No</td><td>Char</td>
+  </tr><tr>
+    <td>03</td><td>-</td><td>No</td><td>char</td>
     <td align="left">Constant "m" (0x6D)</td>
   </tr>
+</table>
+</div>
+<!-- _______________________________________________________________________ -->
+<div class="doc_subsection"><a name="module">Module Block</a> </div>
+<div class="doc_text">
+<p>The module block contains a small pre-amble and all the other blocks in
+the file. Of particular note, the bytecode format number is simply a 28-bit
+monotonically increase integer that identifiers the version of the bytecode
+format. While the bytecode format version is not related to the LLVM release
+(it doesn't automatically get increased with each new LLVM release), there is
+a definite correspondence between the bytecode format version and the LLVM
+release.</p>
+<p>The table below shows the format of the module block header. The blocks it
+contains are detailed in other sections.</p>
+<table class="doc_table_nw" >
   <tr>
-    <td>04-07</td><td>00</td><td>No</td><td>Bool</td>
-    <td align="left">Target is big endian?</td>
-  </tr>
-  <tr>
-    <td>04-07</td><td>01</td><td>No</td><td>Bool</td>
-    <td align="left">Target has long pointers?</td>
-  </tr>
-  <tr>
-    <td>04-07</td><td>02</td><td>No</td><td>Bool</td>
-    <td align="left">Target has no endianess?</td>
-  </tr>
-  <tr>
-    <td>04-07</td><td>03</td><td>No</td><td>Bool</td>
-    <td align="left">Target has no pointer size?</td>
-  </tr>
-  <tr>
-    <td>04-07</td><td>04-31</td><td>Yes</td><td>Unsigned</td>
-    <td align="left">The LLVM bytecode format version number</td>
+    <th><b>Byte(s)</b></th>
+    <th><b>Bit(s)</b></th>
+    <th><b>Align?</b></th>
+    <th><b>Type</b></th>
+    <th align="left"><b>Field Description</b></th>
+  </tr><tr>
+    <td>04-07</td><td>-</td><td>No</td><td>unsigned</td>
+    <td align="left">Module Identifier (0x01)</td>
+  </tr><tr>
+    <td>08-11</td><td>-</td><td>No</td><td>unsigned</td>
+    <td align="left">Size of the module block in bytes</td>
+  </tr><tr>
+    <td>12-15</td><td>00</td><td>Yes</td><td>uint32_vbr</td>
+    <td align="left">Format Information</td>
+  </tr><tr>
+    <td>''</td><td>0</td><td>-</td><td>bit</td>
+    <td align="left">Big Endian?</td>
+  </tr><tr>
+    <td>''</td><td>1</td><td>-</td><td>bit</td>
+    <td align="left">Pointers Are 64-bit?</td>
+  </tr><tr>
+    <td>''</td><td>2</td><td>-</td><td>bit</td>
+    <td align="left">Has No Endianess?</td>
+  </tr><tr>
+    <td>''</td><td>3</td><td>-</td><td>bit</td>
+    <td align="left">Has No Pointer Size?</td>
+  </tr><tr>
+    <td>''</td><td>4-31</td><td>-</td><td>bit</td>
+    <td align="left">Bytecode Format Version</td>
+  </tr><tr>
+    <td>16-end</td><td>-</td><td>-</td><td>blocks</td>
+    <td align="left">The remaining bytes in the block consist
+    solely of other block types in sequence.</td>
   </tr>
 </table>
 </div>
 <!-- _______________________________________________________________________ -->
 <div class="doc_subsection"><a name="gtypepool">Global Type Pool</a> </div>
 <div class="doc_text">
+<p>The global type pool consists of type definitions. Their order of appearnce
+in the file determines their slot number (0 based). Slot numbers are used to 
+replace pointers in the intermediate representation. Each slot number uniquely
+identifies one entry in a type plane (a collection of values of the same type).
+Since all values have types and are associated with the order in which the type
+pool is written, the global type pool <em>must</em> be written as the first 
+block of a module. If it is not, attempts to read the file will fail because
+both forward and backward type resolution will not be possible.</p>
+<p>The type pool is simply a list of types definitions, as shown in the table 
+below.</p>
+<table class="doc_table_nw" >
+  <tr>
+    <th><b>Byte(s)</b></th>
+    <th><b>Bit(s)</b></th>
+    <th><b>Align?</b></th>
+    <th><b>Type</b></th>
+    <th align="left"><b>Field Description</b></th>
+  </tr><tr>
+    <td>00-03</td><td>-</td><td>No</td><td>unsigned</td>
+    <td align="left">Type Pool Identifier (0x13)</td>
+  </tr><tr>
+    <td>04-07</td><td>-</td><td>No</td><td>unsigned</td>
+    <td align="left">Size in bytes of the symbol table block.</td>
+  </tr><tr>
+    <td>08-11<sup>1</sup></td><td>-</td><td>No</td><td>uint32_vbr</td>
+    <td align="left">Number of entries in type plane</td>
+  </tr><tr>
+    <td>12-15<sup>1</sup></td><td>-</td><td>No</td><td>uint32_vbr</td>
+    <td align="left">Type plane index for following entries</td>
+  </tr><tr>
+    <td>16-end<sup>1,2</sup></td><td>-</td><td>No</td><td>type</td>
+    <td align="left">Each of the type definitions.</td>
+  </tr><tr>
+    <td align="left" colspan="5"><sup>1</sup>Maximum length shown, 
+      may be smaller<br><sup>2</sup>Repeated field.
+  </tr>
+</table>
 </div>
 <!-- _______________________________________________________________________ -->
 <div class="doc_subsection"><a name="modinfo">Module Info</a> </div>
 <div class="doc_text">
+  <p>To be determined.</p>
 </div>
 <!-- _______________________________________________________________________ -->
 <div class="doc_subsection"><a name="constants">Constants</a> </div>
 <div class="doc_text">
+  <p>To be determined.</p>
 </div>
 <!-- _______________________________________________________________________ -->
 <div class="doc_subsection"><a name="functions">Functions</a> </div>
 <div class="doc_text">
+  <p>To be determined.</p>
 </div>
 <!-- _______________________________________________________________________ -->
-<div class="doc_subsection"><a name="symtab">Module Symbol Table</a> </div>
+<div class="doc_subsection"><a name="symtab">Symbol Table</a> </div>
 <div class="doc_text">
-<p>The module symbol table is a list of
+<p>A symbol table can be put out in conjunction with a module or a function.
+A symbol table is a list of type planes. Each type plane starts with the number
+of entries in the plane and the type plane's slot number (so the type can be 
+looked up in the global type pool). For each entry in a type plane, the slot 
+number of the value and the name associated with that value are written.  The 
+format is given in the table below. </p>
+<table class="doc_table_nw" >
+  <tr>
+    <th><b>Byte(s)</b></th>
+    <th><b>Bit(s)</b></th>
+    <th><b>Align?</b></th>
+    <th><b>Type</b></th>
+    <th align="left"><b>Field Description</b></th>
+  </tr><tr>
+    <td>00-03</td><td>-</td><td>No</td><td>unsigned</td>
+    <td align="left">Symbol Table Identifier (0x13)</td>
+  </tr><tr>
+    <td>04-07</td><td>-</td><td>No</td><td>unsigned</td>
+    <td align="left">Size in bytes of the symbol table block.</td>
+  </tr><tr>
+    <td>08-11<sup>1</sup></td><td>-</td><td>No</td><td>uint32_vbr</td>
+    <td align="left">Number of entries in type plane</td>
+  </tr><tr>
+    <td>12-15<sup>1</sup></td><td>-</td><td>No</td><td>uint32_vbr</td>
+    <td align="left">Type plane index for following entries</td>
+  </tr><tr>
+    <td>16-19<sup>1,2</sup></td><td>-</td><td>No</td><td>uint32_vbr</td>
+    <td align="left">Slot number of a value.</td>
+  </tr><tr>
+    <td>variable<sup>1,2</sup></td><td>-</td><td>No</td><td>string</td>
+    <td align="left">Name of the value in the symbol table.</td>
+  </tr>
+  <tr>
+    <td align="left" colspan="5"><sup>1</sup>Maximum length shown, 
+      may be smaller<br><sup>2</sup>Repeated field.
+  </tr>
+</table>
 </div>
 
 <!-- *********************************************************************** -->
@@ -255,7 +445,7 @@
   <a href="mailto:rspencer at x10sys.com">Reid Spencer</a> and 
   <a href="mailto:sabre at nondot.org">Chris Lattner</a><br>
   <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/05/22 05:56:41 $
+  Last modified: $Date: 2004/05/23 17:05:09 $
 </address>
 </body>
 </html>




From llvm at cs.uiuc.edu  Sun May 23 12:15:08 2004
From: llvm at cs.uiuc.edu (LLVM)
Date: Sun May 23 12:15:08 2004
Subject: [llvm-commits] CVS: llvm/docs/BytecodeFormat.html
Message-ID: <200405231712.MAA27360@zion.cs.uiuc.edu>


Changes in directory llvm/docs:

BytecodeFormat.html updated: 1.3 -> 1.4

---
Log message:

Make the file HTML 4.01 (Strict) clean.


---
Diffs of the changes:  (+7 -5)

Index: llvm/docs/BytecodeFormat.html
diff -u llvm/docs/BytecodeFormat.html:1.3 llvm/docs/BytecodeFormat.html:1.4
--- llvm/docs/BytecodeFormat.html:1.3	Sun May 23 12:05:09 2004
+++ llvm/docs/BytecodeFormat.html	Sun May 23 12:12:45 2004
@@ -4,7 +4,7 @@
 <head>
   <title>LLVM Bytecode File Format</title>
   <link rel="stylesheet" href="llvm.css" type="text/css">
-  <style>
+  <style type="css">
     table, tr, td { border: 2px solid gray }
     th { border: 2px solid gray; font-weight: bold; }
     table { border-collapse: collapse; margin-top: 1em margin-bottom: 1em }
@@ -14,7 +14,7 @@
   <div class="doc_title"> LLVM Bytecode File Format </div>
 <ol>
   <li><a href="#abstract">Abstract</a></li>
-  <li><a href="#general">General Concepts</a></li>
+  <li><a href="#general">General Concepts</a>
     <ol>
       <li><a href="#blocks">Blocks</a></li>
       <li><a href="#lists">Lists</a></li>
@@ -22,6 +22,7 @@
       <li><a href="#encoding">Encoding Rules</a></li>
       <li><a href="#align">Alignment</a></li>
     </ol>
+  </li>
   <li><a href="#details">Detailed Layout</a>
     <ol>
       <li><a href="#notation">Notation</a></li>
@@ -219,10 +220,11 @@
   has four columns that specify:</p>
   <ol>
     <li><b>Byte(s)</b>. The offset in bytes of the field from the start of
-    its container (block, list, other field).<li>
+    its container (block, list, other field).</li>
     <li><b>Bit(s)</b>. The offset in bits of the field from the start of
     the byte field. Bits are always little endian. That is, bit addresses with
-    smaller values have smaller address (i.e. 2^0 is at bit 0, 2^1 at 1, etc.)
+    smaller values have smaller address (i.e. 2<sup>0</sup> is at bit 0, 
+    2<sup>1</sup> at 1, etc.)
     </li>
     <li><b>Align?</b> Indicates if this field is aligned to 32 bits or not.
     This indicates where the <em>next</em> field starts, always on a 32 bit
@@ -445,7 +447,7 @@
   <a href="mailto:rspencer at x10sys.com">Reid Spencer</a> and 
   <a href="mailto:sabre at nondot.org">Chris Lattner</a><br>
   <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/05/23 17:05:09 $
+  Last modified: $Date: 2004/05/23 17:12:45 $
 </address>
 </body>
 </html>




From llvm at cs.uiuc.edu  Sun May 23 12:25:01 2004
From: llvm at cs.uiuc.edu (LLVM)
Date: Sun May 23 12:25:01 2004
Subject: [llvm-commits] CVS: llvm/docs/ReleaseNotes.html
Message-ID: <200405231723.MAA27418@zion.cs.uiuc.edu>


Changes in directory llvm/docs:

ReleaseNotes.html updated: 1.180 -> 1.181

---
Log message:

Added a "Miscellaneous Improvements" section to cover things like the 
Bytecode File Format documentation, getting all documents in HTML 4.01 
(Strict) compliance, and any other minor improvement that shoud be noted.


---
Diffs of the changes:  (+16 -2)

Index: llvm/docs/ReleaseNotes.html
diff -u llvm/docs/ReleaseNotes.html:1.180 llvm/docs/ReleaseNotes.html:1.181
--- llvm/docs/ReleaseNotes.html:1.180	Fri May 21 11:48:20 2004
+++ llvm/docs/ReleaseNotes.html	Sun May 23 12:23:13 2004
@@ -150,7 +150,6 @@
 <div class="doc_subsubsection">
 <a name="codequality">This release includes the following Code Quality
 improvements:</a>
-</div>
 
 <ol>
 <li>Fixed: <a href="http://llvm.cs.uiuc.edu/PR309">[vmcore] Code quality problem
@@ -189,6 +188,7 @@
 functions like sqrt/sin/cos/etc.</li>
 </ol>
 
+</div>
 
 <!--=========================================================================-->
 <div class="doc_subsubsection">
@@ -224,9 +224,23 @@
 <li><a href="http://llvm.cs.uiuc.edu/PR298">[llvmgcc] Variable length array
 indexing miscompiled</a></li>
 </ol>
+</div>
+
+
+<!--=========================================================================-->
+<div class="doc_subsubsection">
+<a name="misc">This release includes the following miscellaneous
+improvements:</a>
 
+<p>Documentation Improvements</p>
+<ol>
+<li><a href="BytecodeFormat.html">Documentation of the LLVM Bytecode file format
+was added.</a></li>
+<li>All documentation was brought up to HTML 4.01 (Strict) level.</li>
+</ol>
 </div>
 
+
 <!-- *********************************************************************** -->
 <div class="doc_section">
   <a name="portability">Portability and Supported Platforms</a>
@@ -646,7 +660,7 @@
   src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
 
   <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/05/21 16:48:20 $
+  Last modified: $Date: 2004/05/23 17:23:13 $
 </address>
 
 </body>




From tbrethou at cs.uiuc.edu  Sun May 23 14:36:01 2004
From: tbrethou at cs.uiuc.edu (Tanya Brethour)
Date: Sun May 23 14:36:01 2004
Subject: [llvm-commits] CVS: llvm/include/llvm/CodeGen/MachineInstr.h
Message-ID: <200405231935.OAA24049@seraph.cs.uiuc.edu>


Changes in directory llvm/include/llvm/CodeGen:

MachineInstr.h updated: 1.146 -> 1.147

---
Log message:

Adding support to clone MachineInstr



---
Diffs of the changes:  (+13 -1)

Index: llvm/include/llvm/CodeGen/MachineInstr.h
diff -u llvm/include/llvm/CodeGen/MachineInstr.h:1.146 llvm/include/llvm/CodeGen/MachineInstr.h:1.147
--- llvm/include/llvm/CodeGen/MachineInstr.h:1.146	Fri Apr 16 10:57:14 2004
+++ llvm/include/llvm/CodeGen/MachineInstr.h	Sun May 23 14:35:12 2004
@@ -172,6 +172,7 @@
       contents.SymbolName = new std::string(M.getSymbolName());
   }
 
+ 
   ~MachineOperand() {
     if (isExternalSymbol())
       delete contents.SymbolName;
@@ -367,7 +368,9 @@
   // OperandComplete - Return true if it's illegal to add a new operand
   bool OperandsComplete() const;
 
-  MachineInstr(const MachineInstr &);  // DO NOT IMPLEMENT
+  //Constructor used by clone() method
+  MachineInstr(const MachineInstr&);
+
   void operator=(const MachineInstr&); // DO NOT IMPLEMENT
 
   // Intrusive list support
@@ -395,6 +398,9 @@
   const MachineBasicBlock* getParent() const { return parent; }
   MachineBasicBlock* getParent() { return parent; }
 
+  //void setParent(const MachineBasicBlock *MBB) { parent = MBB; } 
+  void setParent(MachineBasicBlock *MBB) { parent = MBB; } 
+
   /// getOpcode - Returns the opcode of this MachineInstr.
   ///
   const int getOpcode() const { return Opcode; }
@@ -454,6 +460,12 @@
     SetMachineOperandVal(i + getNumOperands(),
                          MachineOperand::MO_VirtualRegister, V);
   }
+
+  //Clone Instruction 
+  //Create a copy of 'this' instruction that is
+  //identical in all ways except the following: The instruction has no
+  //parent The instruction has no name
+  MachineInstr* clone();
 
   //
   // Debugging support




From tbrethou at cs.uiuc.edu  Sun May 23 14:36:08 2004
From: tbrethou at cs.uiuc.edu (Tanya Brethour)
Date: Sun May 23 14:36:08 2004
Subject: [llvm-commits] CVS: llvm/lib/CodeGen/MachineInstr.cpp
Message-ID: <200405231935.OAA24042@seraph.cs.uiuc.edu>


Changes in directory llvm/lib/CodeGen:

MachineInstr.cpp updated: 1.96 -> 1.97

---
Log message:

Adding support to clone MachineInstr



---
Diffs of the changes:  (+19 -0)

Index: llvm/lib/CodeGen/MachineInstr.cpp
diff -u llvm/lib/CodeGen/MachineInstr.cpp:1.96 llvm/lib/CodeGen/MachineInstr.cpp:1.97
--- llvm/lib/CodeGen/MachineInstr.cpp:1.96	Wed Mar  3 13:07:27 2004
+++ llvm/lib/CodeGen/MachineInstr.cpp	Sun May 23 14:35:04 2004
@@ -69,11 +69,29 @@
   MBB->push_back(this);  // Add instruction to end of basic block!
 }
 
+///MachineInstr ctor - Copies MachineInstr arg exactly
+MachineInstr::MachineInstr(const MachineInstr &MI) {
+  Opcode = MI.getOpcode();
+  numImplicitRefs = MI.getNumImplicitRefs();
+ 
+  //Add operands
+  for(unsigned i=0; i < MI.getNumOperands(); ++i)
+    operands.push_back(MachineOperand(MI.getOperand(i)));
+}
+
+
 MachineInstr::~MachineInstr()
 {
   LeakDetector::removeGarbageObject(this);
 }
 
+///clone - Create a copy of 'this' instruction that is identical in
+///all ways except the following: The instruction has no parent The
+///instruction has no name
+MachineInstr* MachineInstr::clone() {
+  MachineInstr* newInst = new MachineInstr(*this);
+}
+
 /// OperandComplete - Return true if it's illegal to add a new operand
 ///
 bool MachineInstr::OperandsComplete() const {
@@ -93,6 +111,7 @@
   Opcode = opcode;
   operands.clear();
   operands.resize(numOperands, MachineOperand());
+
 }
 
 void MachineInstr::SetMachineOperandVal(unsigned i,




From lattner at cs.uiuc.edu  Sun May 23 15:31:01 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 15:31:01 2004
Subject: [llvm-commits] CVS: llvm/docs/OpenProjects.html
Message-ID: <200405232029.PAA29176@zion.cs.uiuc.edu>


Changes in directory llvm/docs:

OpenProjects.html updated: 1.26 -> 1.27

---
Log message:

Revert back to the previous version, as it doesn't work with mozilla, and people REALLY CAN just click the link. :)


---
Diffs of the changes:  (+2 -3)

Index: llvm/docs/OpenProjects.html
diff -u llvm/docs/OpenProjects.html:1.26 llvm/docs/OpenProjects.html:1.27
--- llvm/docs/OpenProjects.html:1.26	Sun May 23 06:10:55 2004
+++ llvm/docs/OpenProjects.html	Sun May 23 15:29:35 2004
@@ -57,8 +57,7 @@
 
 <p>The projects in this page are open-ended. More specific projects are
 filed as unassigned enhancements in the <a href="http://llvm.cs.uiuc.edu/bugs/">
-LLVM bug tracker</a>. Here is the current list:
-<object width="100%" height="400" data="http://llvm.cs.uiuc.edu/bugs/buglist.cgi?keywords_type=allwords&amp;keywords=&amp;bug_status=NEW&amp;bug_status=ASSIGNED&amp;bug_status=REOPENED&amp;bug_severity=enhancement&amp;emailassigned_to1=1&amp;emailtype1=substring&amp;email1=unassigned" /></p>
+LLVM bug tracker</a>. See the <a href="http://llvm.cs.uiuc.edu/bugs/buglist.cgi?keywords_type=allwords&amp;keywords=&amp;bug_status=NEW&amp;bug_status=ASSIGNED&amp;bug_status=REOPENED&amp;bug_severity=enhancement&amp;emailassigned_to1=1&amp;emailtype1=substring&amp;email1=unassigned">list of currently outstanding issues</a> if you wish to help improve LLVM.</p>
 
 </div>
 
@@ -334,7 +333,7 @@
 
   <a href="mailto:sabre at nondot.org">Chris Lattner</a><br>
   <a href="http://llvm.cs.uiuc.edu">LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/05/23 11:10:55 $
+  Last modified: $Date: 2004/05/23 20:29:35 $
 </address>
 
 </body>




From lattner at cs.uiuc.edu  Sun May 23 15:44:01 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 15:44:01 2004
Subject: [llvm-commits] CVS: llvm/docs/ReleaseNotes.html
Message-ID: <200405232042.PAA30022@zion.cs.uiuc.edu>


Changes in directory llvm/docs:

ReleaseNotes.html updated: 1.181 -> 1.182

---
Log message:

Fix a couple misplaced div tags


---
Diffs of the changes:  (+3 -3)

Index: llvm/docs/ReleaseNotes.html
diff -u llvm/docs/ReleaseNotes.html:1.181 llvm/docs/ReleaseNotes.html:1.182
--- llvm/docs/ReleaseNotes.html:1.181	Sun May 23 12:23:13 2004
+++ llvm/docs/ReleaseNotes.html	Sun May 23 15:42:08 2004
@@ -150,6 +150,7 @@
 <div class="doc_subsubsection">
 <a name="codequality">This release includes the following Code Quality
 improvements:</a>
+</div>
 
 <ol>
 <li>Fixed: <a href="http://llvm.cs.uiuc.edu/PR309">[vmcore] Code quality problem
@@ -224,13 +225,13 @@
 <li><a href="http://llvm.cs.uiuc.edu/PR298">[llvmgcc] Variable length array
 indexing miscompiled</a></li>
 </ol>
-</div>
 
 
 <!--=========================================================================-->
 <div class="doc_subsubsection">
 <a name="misc">This release includes the following miscellaneous
 improvements:</a>
+</div>
 
 <p>Documentation Improvements</p>
 <ol>
@@ -238,7 +239,6 @@
 was added.</a></li>
 <li>All documentation was brought up to HTML 4.01 (Strict) level.</li>
 </ol>
-</div>
 
 
 <!-- *********************************************************************** -->
@@ -660,7 +660,7 @@
   src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
 
   <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/05/23 17:23:13 $
+  Last modified: $Date: 2004/05/23 20:42:08 $
 </address>
 
 </body>




From alkis at cs.uiuc.edu  Sun May 23 15:57:01 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sun May 23 15:57:01 2004
Subject: [llvm-commits] CVS: llvm/include/llvm/Support/InstVisitor.h
Message-ID: <200405232054.PAA30114@zion.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Support:

InstVisitor.h updated: 1.30 -> 1.31

---
Log message:

Remove virtual destructor from InstVisitor. This class should never be
used through a base pointer/reference so inproper destruction should
never be an issue. Removing this last virtual function also saves 4
bytes off each InstVisitor instance.


---
Diffs of the changes:  (+0 -2)

Index: llvm/include/llvm/Support/InstVisitor.h
diff -u llvm/include/llvm/Support/InstVisitor.h:1.30 llvm/include/llvm/Support/InstVisitor.h:1.31
--- llvm/include/llvm/Support/InstVisitor.h:1.30	Wed May  5 21:07:42 2004
+++ llvm/include/llvm/Support/InstVisitor.h	Sun May 23 15:54:39 2004
@@ -71,8 +71,6 @@
 
 template<typename SubClass, typename RetTy=void>
 struct InstVisitor {
-  virtual ~InstVisitor() {}           // We are meant to be derived from
-
   //===--------------------------------------------------------------------===//
   // Interface code - This is the public interface of the InstVisitor that you
   // use to visit instructions...




From tbrethou at cs.uiuc.edu  Sun May 23 15:59:01 2004
From: tbrethou at cs.uiuc.edu (Tanya Brethour)
Date: Sun May 23 15:59:01 2004
Subject: [llvm-commits] CVS: llvm/include/llvm/CodeGen/MachineInstr.h
Message-ID: <200405232058.PAA27533@seraph.cs.uiuc.edu>


Changes in directory llvm/include/llvm/CodeGen:

MachineInstr.h updated: 1.147 -> 1.148

---
Log message:

Fixed up my changes to add support for cloning Machine Instructions.


---
Diffs of the changes:  (+2 -7)

Index: llvm/include/llvm/CodeGen/MachineInstr.h
diff -u llvm/include/llvm/CodeGen/MachineInstr.h:1.147 llvm/include/llvm/CodeGen/MachineInstr.h:1.148
--- llvm/include/llvm/CodeGen/MachineInstr.h:1.147	Sun May 23 14:35:12 2004
+++ llvm/include/llvm/CodeGen/MachineInstr.h	Sun May 23 15:58:02 2004
@@ -398,9 +398,6 @@
   const MachineBasicBlock* getParent() const { return parent; }
   MachineBasicBlock* getParent() { return parent; }
 
-  //void setParent(const MachineBasicBlock *MBB) { parent = MBB; } 
-  void setParent(MachineBasicBlock *MBB) { parent = MBB; } 
-
   /// getOpcode - Returns the opcode of this MachineInstr.
   ///
   const int getOpcode() const { return Opcode; }
@@ -461,10 +458,8 @@
                          MachineOperand::MO_VirtualRegister, V);
   }
 
-  //Clone Instruction 
-  //Create a copy of 'this' instruction that is
-  //identical in all ways except the following: The instruction has no
-  //parent The instruction has no name
+  /// clone - Create a copy of 'this' instruction that is identical in
+  /// all ways except the the instruction has no parent, prev, or next.
   MachineInstr* clone();
 
   //




From tbrethou at cs.uiuc.edu  Sun May 23 15:59:09 2004
From: tbrethou at cs.uiuc.edu (Tanya Brethour)
Date: Sun May 23 15:59:09 2004
Subject: [llvm-commits] CVS: llvm/lib/CodeGen/MachineInstr.cpp
Message-ID: <200405232058.PAA27420@seraph.cs.uiuc.edu>


Changes in directory llvm/lib/CodeGen:

MachineInstr.cpp updated: 1.97 -> 1.98

---
Log message:

Fixed up my changes to add support for cloning Machine Instructions.


---
Diffs of the changes:  (+3 -2)

Index: llvm/lib/CodeGen/MachineInstr.cpp
diff -u llvm/lib/CodeGen/MachineInstr.cpp:1.97 llvm/lib/CodeGen/MachineInstr.cpp:1.98
--- llvm/lib/CodeGen/MachineInstr.cpp:1.97	Sun May 23 14:35:04 2004
+++ llvm/lib/CodeGen/MachineInstr.cpp	Sun May 23 15:57:57 2004
@@ -73,7 +73,8 @@
 MachineInstr::MachineInstr(const MachineInstr &MI) {
   Opcode = MI.getOpcode();
   numImplicitRefs = MI.getNumImplicitRefs();
- 
+  operands.reserve(MI.getNumOperands());
+
   //Add operands
   for(unsigned i=0; i < MI.getNumOperands(); ++i)
     operands.push_back(MachineOperand(MI.getOperand(i)));
@@ -89,7 +90,7 @@
 ///all ways except the following: The instruction has no parent The
 ///instruction has no name
 MachineInstr* MachineInstr::clone() {
-  MachineInstr* newInst = new MachineInstr(*this);
+  return new MachineInstr(*this);
 }
 
 /// OperandComplete - Return true if it's illegal to add a new operand




From lattner at cs.uiuc.edu  Sun May 23 16:03:02 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:03:02 2004
Subject: [llvm-commits] CVS: llvm/lib/Analysis/IPA/Andersens.cpp
Message-ID: <200405232100.QAA30241@zion.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis/IPA:

Andersens.cpp added (r1.1)

---
Log message:

Add a simple implementation of Andersen's interprocedural pointer analysis


---
Diffs of the changes:  (+1029 -0)

Index: llvm/lib/Analysis/IPA/Andersens.cpp
diff -c /dev/null llvm/lib/Analysis/IPA/Andersens.cpp:1.1
*** /dev/null	Sun May 23 16:00:59 2004
--- llvm/lib/Analysis/IPA/Andersens.cpp	Sun May 23 16:00:47 2004
***************
*** 0 ****
--- 1,1029 ----
+ //===- Andersens.cpp - Andersen's Interprocedural Alias Analysis -----------==//
+ // 
+ //                     The LLVM Compiler Infrastructure
+ //
+ // This file was developed by the LLVM research group and is distributed under
+ // the University of Illinois Open Source License. See LICENSE.TXT for details.
+ // 
+ //===----------------------------------------------------------------------===//
+ //
+ // This file defines a very simple implementation of Andersen's interprocedural
+ // alias analysis.  This implementation does not include any of the fancy
+ // features that make Andersen's reasonably efficient (like cycle elimination or
+ // variable substitution), but it should be useful for getting precision
+ // numbers and can be extended in the future.
+ //
+ // In pointer analysis terms, this is a subset-based, flow-insensitive,
+ // field-insensitive, and context-insensitive algorithm pointer algorithm.
+ //
+ // This algorithm is implemented as three stages:
+ //   1. Object identification.
+ //   2. Inclusion constraint identification.
+ //   3. Inclusion constraint solving.
+ //
+ // The object identification stage identifies all of the memory objects in the
+ // program, which includes globals, heap allocated objects, and stack allocated
+ // objects.
+ //
+ // The inclusion constraint identification stage finds all inclusion constraints
+ // in the program by scanning the program, looking for pointer assignments and
+ // other statements that effect the points-to graph.  For a statement like "A =
+ // B", this statement is processed to indicate that A can point to anything that
+ // B can point to.  Constraints can handle copies, loads, and stores.
+ //
+ // The inclusion constraint solving phase iteratively propagates the inclusion
+ // constraints until a fixed point is reached.  This is an O(N^3) algorithm.
+ //
+ // In the initial pass, all indirect function calls are completely ignored.  As
+ // the analysis discovers new targets of function pointers, it iteratively
+ // resolves a precise (and conservative) call graph.  Also related, this
+ // analysis initially assumes that all internal functions have known incoming
+ // pointers.  If we find that an internal function's address escapes outside of
+ // the program, we update this assumption.
+ //
+ //===----------------------------------------------------------------------===//
+ 
+ #define DEBUG_TYPE "anders-aa"
+ #include "llvm/Constants.h"
+ #include "llvm/DerivedTypes.h"
+ #include "llvm/Instructions.h"
+ #include "llvm/Module.h"
+ #include "llvm/Pass.h"
+ #include "llvm/Support/InstIterator.h"
+ #include "llvm/Support/InstVisitor.h"
+ #include "llvm/Analysis/AliasAnalysis.h"
+ #include "Support/Debug.h"
+ #include "Support/Statistic.h"
+ #include <set>
+ using namespace llvm;
+ 
+ namespace {
+   Statistic<>
+   NumIters("anders-aa", "Number of iterations to reach convergence");
+   Statistic<>
+   NumConstraints("anders-aa", "Number of constraints");
+   Statistic<>
+   NumNodes("anders-aa", "Number of nodes");
+   Statistic<>
+   NumEscapingFunctions("anders-aa", "Number of internal functions that escape");
+   Statistic<>
+   NumIndirectCallees("anders-aa", "Number of indirect callees found");
+ 
+   class Andersens : public Pass, public AliasAnalysis,
+                     private InstVisitor<Andersens> {
+     /// Node class - This class is used to represent a memory object in the
+     /// program, and is the primitive used to build the points-to graph.
+     class Node {
+       std::vector<Node*> Pointees;
+       Value *Val;
+     public:
+       Node() : Val(0) {}
+       Node *setValue(Value *V) {
+         assert(Val == 0 && "Value already set for this node!");
+         Val = V;
+         return this;
+       }
+ 
+       /// getValue - Return the LLVM value corresponding to this node.
+       Value *getValue() const { return Val; }
+ 
+       typedef std::vector<Node*>::const_iterator iterator;
+       iterator begin() const { return Pointees.begin(); }
+       iterator end() const { return Pointees.end(); }
+ 
+       /// addPointerTo - Add a pointer to the list of pointees of this node,
+       /// returning true if this caused a new pointer to be added, or false if
+       /// we already knew about the points-to relation.
+       bool addPointerTo(Node *N) {
+         std::vector<Node*>::iterator I = std::lower_bound(Pointees.begin(),
+                                                           Pointees.end(),
+                                                           N);
+         if (I != Pointees.end() && *I == N)
+           return false;
+         Pointees.insert(I, N);
+         return true;
+       }
+ 
+       /// intersects - Return true if the points-to set of this node intersects
+       /// with the points-to set of the specified node.
+       bool intersects(Node *N) const;
+ 
+       /// intersectsIgnoring - Return true if the points-to set of this node
+       /// intersects with the points-to set of the specified node on any nodes
+       /// except for the specified node to ignore.
+       bool intersectsIgnoring(Node *N, Node *Ignoring) const;
+ 
+       // Constraint application methods.
+       bool copyFrom(Node *N);
+       bool loadFrom(Node *N);
+       bool storeThrough(Node *N);
+     };
+ 
+     /// GraphNodes - This vector is populated as part of the object
+     /// identification stage of the analysis, which populates this vector with a
+     /// node for each memory object and fills in the ValueNodes map.
+     std::vector<Node> GraphNodes;
+ 
+     /// ValueNodes - This map indicates the Node that a particular Value* is
+     /// represented by.  This contains entries for all pointers.
+     std::map<Value*, unsigned> ValueNodes;
+ 
+     /// ObjectNodes - This map contains entries for each memory object in the
+     /// program: globals, alloca's and mallocs.  
+     std::map<Value*, unsigned> ObjectNodes;
+ 
+     /// ReturnNodes - This map contains an entry for each function in the
+     /// program that returns a value.
+     std::map<Function*, unsigned> ReturnNodes;
+ 
+     /// VarargNodes - This map contains the entry used to represent all pointers
+     /// passed through the varargs portion of a function call for a particular
+     /// function.  An entry is not present in this map for functions that do not
+     /// take variable arguments.
+     std::map<Function*, unsigned> VarargNodes;
+ 
+     /// Constraint - Objects of this structure are used to represent the various
+     /// constraints identified by the algorithm.  The constraints are 'copy',
+     /// for statements like "A = B", 'load' for statements like "A = *B", and
+     /// 'store' for statements like "*A = B".
+     struct Constraint {
+       enum ConstraintType { Copy, Load, Store } Type;
+       Node *Dest, *Src;
+ 
+       Constraint(ConstraintType Ty, Node *D, Node *S)
+         : Type(Ty), Dest(D), Src(S) {}
+     };
+     
+     /// Constraints - This vector contains a list of all of the constraints
+     /// identified by the program.
+     std::vector<Constraint> Constraints;
+ 
+     /// EscapingInternalFunctions - This set contains all of the internal
+     /// functions that are found to escape from the program.  If the address of
+     /// an internal function is passed to an external function or otherwise
+     /// escapes from the analyzed portion of the program, we must assume that
+     /// any pointer arguments can alias the universal node.  This set keeps
+     /// track of those functions we are assuming to escape so far.
+     std::set<Function*> EscapingInternalFunctions;
+ 
+     /// IndirectCalls - This contains a list of all of the indirect call sites
+     /// in the program.  Since the call graph is iteratively discovered, we may
+     /// need to add constraints to our graph as we find new targets of function
+     /// pointers.
+     std::vector<CallSite> IndirectCalls;
+ 
+     /// IndirectCallees - For each call site in the indirect calls list, keep
+     /// track of the callees that we have discovered so far.  As the analysis
+     /// proceeds, more callees are discovered, until the call graph finally
+     /// stabilizes.
+     std::map<CallSite, std::vector<Function*> > IndirectCallees;
+ 
+     /// This enum defines the GraphNodes indices that correspond to important
+     /// fixed sets.
+     enum {
+       UniversalSet = 0,
+       NullPtr      = 1,
+       NullObject   = 2,
+     };
+     
+   public:
+     bool run(Module &M) {
+       InitializeAliasAnalysis(this);
+       IdentifyObjects(M);
+       CollectConstraints(M);
+       DEBUG(PrintConstraints());
+       SolveConstraints();
+       DEBUG(PrintPointsToGraph());
+ 
+       // Free the constraints list, as we don't need it to respond to alias
+       // requests.
+       ObjectNodes.clear();
+       ReturnNodes.clear();
+       VarargNodes.clear();
+       EscapingInternalFunctions.clear();
+       std::vector<Constraint>().swap(Constraints);      
+       return false;
+     }
+ 
+     void releaseMemory() {
+       // FIXME: Until we have transitively required passes working correctly,
+       // this cannot be enabled!  Otherwise, using -count-aa with the pass
+       // causes memory to be freed too early. :(
+ #if 0
+       // The memory objects and ValueNodes data structures at the only ones that
+       // are still live after construction.
+       std::vector<Node>().swap(GraphNodes);
+       ValueNodes.clear();
+ #endif
+     }
+ 
+     virtual void getAnalysisUsage(AnalysisUsage &AU) const {
+       AliasAnalysis::getAnalysisUsage(AU);
+       AU.setPreservesAll();                         // Does not transform code
+     }
+ 
+     //------------------------------------------------
+     // Implement the AliasAnalysis API
+     //  
+     AliasResult alias(const Value *V1, unsigned V1Size,
+                       const Value *V2, unsigned V2Size);
+     void getMustAliases(Value *P, std::vector<Value*> &RetVals);
+     bool pointsToConstantMemory(const Value *P);
+ 
+     virtual void deleteValue(Value *V) {
+       ValueNodes.erase(V);
+       getAnalysis<AliasAnalysis>().deleteValue(V);
+     }
+ 
+     virtual void copyValue(Value *From, Value *To) {
+       ValueNodes[To] = ValueNodes[From];
+       getAnalysis<AliasAnalysis>().copyValue(From, To);
+     }
+ 
+   private:
+     /// getNode - Return the node corresponding to the specified pointer scalar.
+     ///
+     Node *getNode(Value *V) {
+       if (Constant *C = dyn_cast<Constant>(V))
+         return getNodeForConstantPointer(C);
+ 
+       std::map<Value*, unsigned>::iterator I = ValueNodes.find(V);
+       if (I == ValueNodes.end()) {
+         V->dump();
+         assert(I != ValueNodes.end() &&
+                "Value does not have a node in the points-to graph!");
+       }
+       return &GraphNodes[I->second];
+     }
+     
+     /// getObject - Return the node corresponding to the memory object for the
+     /// specified global or allocation instruction.
+     Node *getObject(Value *V) {
+       std::map<Value*, unsigned>::iterator I = ObjectNodes.find(V);
+       assert(I != ObjectNodes.end() &&
+              "Value does not have an object in the points-to graph!");
+       return &GraphNodes[I->second];
+     }
+ 
+     /// getReturnNode - Return the node representing the return value for the
+     /// specified function.
+     Node *getReturnNode(Function *F) {
+       std::map<Function*, unsigned>::iterator I = ReturnNodes.find(F);
+       assert(I != ReturnNodes.end() && "Function does not return a value!");
+       return &GraphNodes[I->second];
+     }
+ 
+     /// getVarargNode - Return the node representing the variable arguments
+     /// formal for the specified function.
+     Node *getVarargNode(Function *F) {
+       std::map<Function*, unsigned>::iterator I = VarargNodes.find(F);
+       assert(I != VarargNodes.end() && "Function does not take var args!");
+       return &GraphNodes[I->second];
+     }
+ 
+     /// getNodeValue - Get the node for the specified LLVM value and set the
+     /// value for it to be the specified value.
+     Node *getNodeValue(Value &V) {
+       return getNode(&V)->setValue(&V);
+     }
+ 
+     void IdentifyObjects(Module &M);
+     void CollectConstraints(Module &M);
+     void SolveConstraints();
+ 
+     Node *getNodeForConstantPointer(Constant *C);
+     Node *getNodeForConstantPointerTarget(Constant *C);
+     void AddGlobalInitializerConstraints(Node *N, Constant *C);
+     void AddConstraintsForNonInternalLinkage(Function *F);
+     void AddConstraintsForCall(CallSite CS, Function *F);
+ 
+ 
+     void PrintNode(Node *N);
+     void PrintConstraints();
+     void PrintPointsToGraph();
+ 
+     //===------------------------------------------------------------------===//
+     // Instruction visitation methods for adding constraints
+     //
+     friend class InstVisitor<Andersens>;
+     void visitReturnInst(ReturnInst &RI);
+     void visitInvokeInst(InvokeInst &II) { visitCallSite(CallSite(&II)); }
+     void visitCallInst(CallInst &CI) { visitCallSite(CallSite(&CI)); }
+     void visitCallSite(CallSite CS);
+     void visitAllocationInst(AllocationInst &AI);
+     void visitLoadInst(LoadInst &LI);
+     void visitStoreInst(StoreInst &SI);
+     void visitGetElementPtrInst(GetElementPtrInst &GEP);
+     void visitPHINode(PHINode &PN);
+     void visitCastInst(CastInst &CI);
+     void visitSelectInst(SelectInst &SI);
+     void visitVANext(VANextInst &I);
+     void visitVAArg(VAArgInst &I);
+     void visitInstruction(Instruction &I);
+   };
+ 
+   RegisterOpt<Andersens> X("anders-aa",
+                            "Andersen's Interprocedural Alias Analysis");
+   RegisterAnalysisGroup<AliasAnalysis, Andersens> Y;
+ }
+ 
+ //===----------------------------------------------------------------------===//
+ //                  AliasAnalysis Interface Implementation
+ //===----------------------------------------------------------------------===//
+ 
+ AliasAnalysis::AliasResult Andersens::alias(const Value *V1, unsigned V1Size,
+                                             const Value *V2, unsigned V2Size) {
+   Node *N1 = getNode((Value*)V1);
+   Node *N2 = getNode((Value*)V2);
+ 
+   // Check to see if the two pointers are known to not alias.  They don't alias
+   // if their points-to sets do not intersect.
+   if (!N1->intersectsIgnoring(N2, &GraphNodes[NullObject]))
+     return NoAlias;
+ 
+   return AliasAnalysis::alias(V1, V1Size, V2, V2Size);
+ }
+ 
+ /// getMustAlias - We can provide must alias information if we know that a
+ /// pointer can only point to a specific function or the null pointer.
+ /// Unfortunately we cannot determine must-alias information for global
+ /// variables or any other memory memory objects because we do not track whether
+ /// a pointer points to the beginning of an object or a field of it.
+ void Andersens::getMustAliases(Value *P, std::vector<Value*> &RetVals) {
+   Node *N = getNode(P);
+   Node::iterator I = N->begin();
+   if (I != N->end()) {
+     // If there is exactly one element in the points-to set for the object...
+     ++I;
+     if (I == N->end()) {
+       Node *Pointee = *N->begin();
+ 
+       // If a function is the only object in the points-to set, then it must be
+       // the destination.  Note that we can't handle global variables here,
+       // because we don't know if the pointer is actually pointing to a field of
+       // the global or to the beginning of it.
+       if (Value *V = Pointee->getValue()) {
+         if (Function *F = dyn_cast<Function>(V))
+           RetVals.push_back(F);
+       } else {
+         // If the object in the points-to set is the null object, then the null
+         // pointer is a must alias.
+         if (Pointee == &GraphNodes[NullObject])
+           RetVals.push_back(Constant::getNullValue(P->getType()));
+       }
+     }
+   }
+   
+   AliasAnalysis::getMustAliases(P, RetVals);
+ }
+ 
+ /// pointsToConstantMemory - If we can determine that this pointer only points
+ /// to constant memory, return true.  In practice, this means that if the
+ /// pointer can only point to constant globals, functions, or the null pointer,
+ /// return true.
+ ///
+ bool Andersens::pointsToConstantMemory(const Value *P) {
+   Node *N = getNode((Value*)P);
+   for (Node::iterator I = N->begin(), E = N->end(); I != E; ++I) {
+     if (Value *V = (*I)->getValue()) {
+       if (!isa<GlobalValue>(V) || (isa<GlobalVariable>(V) &&
+                                    !cast<GlobalVariable>(V)->isConstant()))
+         return AliasAnalysis::pointsToConstantMemory(P);
+     } else {
+       if (*I != &GraphNodes[NullObject])
+         return AliasAnalysis::pointsToConstantMemory(P);
+     }
+   }
+ 
+   return true;
+ }
+ 
+ //===----------------------------------------------------------------------===//
+ //                       Object Identification Phase
+ //===----------------------------------------------------------------------===//
+ 
+ /// IdentifyObjects - This stage scans the program, adding an entry to the
+ /// GraphNodes list for each memory object in the program (global stack or
+ /// heap), and populates the ValueNodes and ObjectNodes maps for these objects.
+ ///
+ void Andersens::IdentifyObjects(Module &M) {
+   unsigned NumObjects = 0;
+ 
+   // Object #0 is always the universal set: the object that we don't know
+   // anything about.
+   assert(NumObjects == UniversalSet && "Something changed!");
+   ++NumObjects;
+ 
+   // Object #1 always represents the null pointer.
+   assert(NumObjects == NullPtr && "Something changed!");
+   ++NumObjects;
+ 
+   // Object #2 always represents the null object (the object pointed to by null)
+   assert(NumObjects == NullObject && "Something changed!");
+   ++NumObjects;
+ 
+   // Add all the globals first.
+   for (Module::giterator I = M.gbegin(), E = M.gend(); I != E; ++I) {
+     ObjectNodes[I] = NumObjects++;
+     ValueNodes[I] = NumObjects++;
+   }
+ 
+   // Add nodes for all of the functions and the instructions inside of them.
+   for (Module::iterator F = M.begin(), E = M.end(); F != E; ++F) {
+     // The function itself is a memory object.
+     ValueNodes[F] = NumObjects++;
+     ObjectNodes[F] = NumObjects++;
+     if (isa<PointerType>(F->getFunctionType()->getReturnType()))
+       ReturnNodes[F] = NumObjects++;
+     if (F->getFunctionType()->isVarArg())
+       VarargNodes[F] = NumObjects++;
+ 
+     // Add nodes for all of the incoming pointer arguments.
+     for (Function::aiterator I = F->abegin(), E = F->aend(); I != E; ++I)
+       if (isa<PointerType>(I->getType()))
+         ValueNodes[I] = NumObjects++;
+ 
+     // Scan the function body, creating a memory object for each heap/stack
+     // allocation in the body of the function and a node to represent all
+     // pointer values defined by instructions and used as operands.
+     for (inst_iterator II = inst_begin(F), E = inst_end(F); II != E; ++II) {
+       // If this is an heap or stack allocation, create a node for the memory
+       // object.
+       if (isa<PointerType>(II->getType())) {
+         ValueNodes[&*II] = NumObjects++;
+         if (AllocationInst *AI = dyn_cast<AllocationInst>(&*II))
+           ObjectNodes[AI] = NumObjects++;
+       }
+     }
+   }
+ 
+   // Now that we know how many objects to create, make them all now!
+   GraphNodes.resize(NumObjects);
+   NumNodes += NumObjects;
+ }
+ 
+ //===----------------------------------------------------------------------===//
+ //                     Constraint Identification Phase
+ //===----------------------------------------------------------------------===//
+ 
+ /// getNodeForConstantPointer - Return the node corresponding to the constant
+ /// pointer itself.
+ Andersens::Node *Andersens::getNodeForConstantPointer(Constant *C) {
+   assert(isa<PointerType>(C->getType()) && "Not a constant pointer!");
+ 
+   if (isa<ConstantPointerNull>(C))
+     return &GraphNodes[NullPtr];
+   else if (ConstantPointerRef *CPR = dyn_cast<ConstantPointerRef>(C))
+     return getNode(CPR->getValue());
+   else if (ConstantExpr *CE = dyn_cast<ConstantExpr>(C)) {
+     switch (CE->getOpcode()) {
+     case Instruction::GetElementPtr:
+       return getNodeForConstantPointer(CE->getOperand(0));
+     case Instruction::Cast:
+       if (isa<PointerType>(CE->getOperand(0)->getType()))
+         return getNodeForConstantPointer(CE->getOperand(0));
+       else
+         return &GraphNodes[UniversalSet];
+     default:
+       std::cerr << "Constant Expr not yet handled: " << *CE << "\n";
+       assert(0);
+     }
+   } else {
+     assert(0 && "Unknown constant pointer!");
+   }
+ }
+ 
+ /// getNodeForConstantPointerTarget - Return the node POINTED TO by the
+ /// specified constant pointer.
+ Andersens::Node *Andersens::getNodeForConstantPointerTarget(Constant *C) {
+   assert(isa<PointerType>(C->getType()) && "Not a constant pointer!");
+ 
+   if (isa<ConstantPointerNull>(C))
+     return &GraphNodes[NullObject];
+   else if (ConstantPointerRef *CPR = dyn_cast<ConstantPointerRef>(C))
+     return getObject(CPR->getValue());
+   else if (ConstantExpr *CE = dyn_cast<ConstantExpr>(C)) {
+     switch (CE->getOpcode()) {
+     case Instruction::GetElementPtr:
+       return getNodeForConstantPointerTarget(CE->getOperand(0));
+     case Instruction::Cast:
+       if (isa<PointerType>(CE->getOperand(0)->getType()))
+         return getNodeForConstantPointerTarget(CE->getOperand(0));
+       else
+         return &GraphNodes[UniversalSet];
+     default:
+       std::cerr << "Constant Expr not yet handled: " << *CE << "\n";
+       assert(0);
+     }
+   } else {
+     assert(0 && "Unknown constant pointer!");
+   }
+ }
+ 
+ /// AddGlobalInitializerConstraints - Add inclusion constraints for the memory
+ /// object N, which contains values indicated by C.
+ void Andersens::AddGlobalInitializerConstraints(Node *N, Constant *C) {
+   if (C->getType()->isFirstClassType()) {
+     if (isa<PointerType>(C->getType()))
+       N->addPointerTo(getNodeForConstantPointer(C));
+   } else if (C->isNullValue()) {
+     N->addPointerTo(&GraphNodes[NullObject]);
+     return;
+   } else {
+     // If this is an array or struct, include constraints for each element.
+     assert(isa<ConstantArray>(C) || isa<ConstantStruct>(C));
+     for (unsigned i = 0, e = C->getNumOperands(); i != e; ++i)
+       AddGlobalInitializerConstraints(N, cast<Constant>(C->getOperand(i)));
+   }
+ }
+ 
+ void Andersens::AddConstraintsForNonInternalLinkage(Function *F) {
+   for (Function::aiterator I = F->abegin(), E = F->aend(); I != E; ++I)
+     if (isa<PointerType>(I->getType()))
+       // If this is an argument of an externally accessible function, the
+       // incoming pointer might point to anything.
+       Constraints.push_back(Constraint(Constraint::Copy, getNode(I),
+                                        &GraphNodes[UniversalSet]));
+ }
+ 
+ 
+ /// CollectConstraints - This stage scans the program, adding a constraint to
+ /// the Constraints list for each instruction in the program that induces a
+ /// constraint, and setting up the initial points-to graph.
+ ///
+ void Andersens::CollectConstraints(Module &M) {
+   // First, the universal set points to itself.
+   GraphNodes[UniversalSet].addPointerTo(&GraphNodes[UniversalSet]);
+ 
+   // Next, the null pointer points to the null object.
+   GraphNodes[NullPtr].addPointerTo(&GraphNodes[NullObject]);
+ 
+   // Next, add any constraints on global variables and their initializers.
+   for (Module::giterator I = M.gbegin(), E = M.gend(); I != E; ++I) {
+     // Associate the address of the global object as pointing to the memory for
+     // the global: &G = <G memory>
+     Node *Object = getObject(I);
+     Object->setValue(I);
+     getNodeValue(*I)->addPointerTo(Object);
+ 
+     if (I->hasInitializer()) {
+       AddGlobalInitializerConstraints(Object, I->getInitializer());
+     } else {
+       // If it doesn't have an initializer (i.e. it's defined in another
+       // translation unit), it points to the universal set.
+       Constraints.push_back(Constraint(Constraint::Copy, Object,
+                                        &GraphNodes[UniversalSet]));
+     }
+   }
+   
+   for (Module::iterator F = M.begin(), E = M.end(); F != E; ++F) {
+     // Make the function address point to the function object.
+     getNodeValue(*F)->addPointerTo(getObject(F)->setValue(F));
+ 
+     // Set up the return value node.
+     if (isa<PointerType>(F->getFunctionType()->getReturnType()))
+       getReturnNode(F)->setValue(F);
+     if (F->getFunctionType()->isVarArg())
+       getVarargNode(F)->setValue(F);
+ 
+     // Set up incoming argument nodes.
+     for (Function::aiterator I = F->abegin(), E = F->aend(); I != E; ++I)
+       if (isa<PointerType>(I->getType()))
+         getNodeValue(*I);
+ 
+     if (!F->hasInternalLinkage())
+       AddConstraintsForNonInternalLinkage(F);
+ 
+     if (!F->isExternal()) {
+       // Scan the function body, creating a memory object for each heap/stack
+       // allocation in the body of the function and a node to represent all
+       // pointer values defined by instructions and used as operands.
+       visit(F);
+     } else {
+       // External functions that return pointers return the universal set.
+       if (isa<PointerType>(F->getFunctionType()->getReturnType()))
+         Constraints.push_back(Constraint(Constraint::Copy,
+                                          getReturnNode(F),
+                                          &GraphNodes[UniversalSet]));
+ 
+       // Any pointers that are passed into the function have the universal set
+       // stored into them.
+       for (Function::aiterator I = F->abegin(), E = F->aend(); I != E; ++I)
+         if (isa<PointerType>(I->getType())) {
+           // Pointers passed into external functions could have anything stored
+           // through them.
+           Constraints.push_back(Constraint(Constraint::Store, getNode(I),
+                                            &GraphNodes[UniversalSet]));
+           // Memory objects passed into external function calls can have the
+           // universal set point to them.
+           Constraints.push_back(Constraint(Constraint::Copy,
+                                            &GraphNodes[UniversalSet],
+                                            getNode(I)));
+         }
+ 
+       // If this is an external varargs function, it can also store pointers
+       // into any pointers passed through the varargs section.
+       if (F->getFunctionType()->isVarArg())
+         Constraints.push_back(Constraint(Constraint::Store, getVarargNode(F),
+                                          &GraphNodes[UniversalSet]));
+     }
+   }
+   NumConstraints += Constraints.size();
+ }
+ 
+ 
+ void Andersens::visitInstruction(Instruction &I) {
+ #ifdef NDEBUG
+   return;          // This function is just a big assert.
+ #endif
+   if (isa<BinaryOperator>(I))
+     return;
+   // Most instructions don't have any effect on pointer values.
+   switch (I.getOpcode()) {
+   case Instruction::Br:
+   case Instruction::Switch:
+   case Instruction::Unwind:
+   case Instruction::Free:
+   case Instruction::Shl:
+   case Instruction::Shr:
+     return;
+   default:
+     // Is this something we aren't handling yet?
+     std::cerr << "Unknown instruction: " << I;
+     abort();
+   }
+ }
+ 
+ void Andersens::visitAllocationInst(AllocationInst &AI) {
+   getNodeValue(AI)->addPointerTo(getObject(&AI)->setValue(&AI));
+ }
+ 
+ void Andersens::visitReturnInst(ReturnInst &RI) {
+   if (RI.getNumOperands() && isa<PointerType>(RI.getOperand(0)->getType()))
+     // return V   -->   <Copy/retval{F}/v>
+     Constraints.push_back(Constraint(Constraint::Copy,
+                                      getReturnNode(RI.getParent()->getParent()),
+                                      getNode(RI.getOperand(0))));
+ }
+ 
+ void Andersens::visitLoadInst(LoadInst &LI) {
+   if (isa<PointerType>(LI.getType()))
+     // P1 = load P2  -->  <Load/P1/P2>
+     Constraints.push_back(Constraint(Constraint::Load, getNodeValue(LI),
+                                      getNode(LI.getOperand(0))));
+ }
+ 
+ void Andersens::visitStoreInst(StoreInst &SI) {
+   if (isa<PointerType>(SI.getOperand(0)->getType()))
+     // store P1, P2  -->  <Store/P2/P1>
+     Constraints.push_back(Constraint(Constraint::Store,
+                                      getNode(SI.getOperand(1)),
+                                      getNode(SI.getOperand(0))));
+ }
+ 
+ void Andersens::visitGetElementPtrInst(GetElementPtrInst &GEP) {
+   // P1 = getelementptr P2, ... --> <Copy/P1/P2>
+   Constraints.push_back(Constraint(Constraint::Copy, getNodeValue(GEP),
+                                    getNode(GEP.getOperand(0))));
+ }
+ 
+ void Andersens::visitPHINode(PHINode &PN) {
+   if (isa<PointerType>(PN.getType())) {
+     Node *PNN = getNodeValue(PN);
+     for (unsigned i = 0, e = PN.getNumIncomingValues(); i != e; ++i)
+       // P1 = phi P2, P3  -->  <Copy/P1/P2>, <Copy/P1/P3>, ...
+       Constraints.push_back(Constraint(Constraint::Copy, PNN,
+                                        getNode(PN.getIncomingValue(i))));
+   }
+ }
+ 
+ void Andersens::visitCastInst(CastInst &CI) {
+   Value *Op = CI.getOperand(0);
+   if (isa<PointerType>(CI.getType())) {
+     if (isa<PointerType>(Op->getType())) {
+       // P1 = cast P2  --> <Copy/P1/P2>
+       Constraints.push_back(Constraint(Constraint::Copy, getNodeValue(CI),
+                                        getNode(CI.getOperand(0))));
+     } else {
+       // P1 = cast int --> <Copy/P1/Univ>
+       Constraints.push_back(Constraint(Constraint::Copy, getNodeValue(CI),
+                                        &GraphNodes[UniversalSet]));
+     }
+   } else if (isa<PointerType>(Op->getType())) {
+     // int = cast P1 --> <Copy/Univ/P1>
+     Constraints.push_back(Constraint(Constraint::Copy,
+                                      &GraphNodes[UniversalSet],
+                                      getNode(CI.getOperand(0))));
+   }
+ }
+ 
+ void Andersens::visitSelectInst(SelectInst &SI) {
+   if (isa<PointerType>(SI.getType())) {
+     Node *SIN = getNodeValue(SI);
+     // P1 = select C, P2, P3   ---> <Copy/P1/P2>, <Copy/P1/P3>
+     Constraints.push_back(Constraint(Constraint::Copy, SIN,
+                                      getNode(SI.getOperand(1))));
+     Constraints.push_back(Constraint(Constraint::Copy, SIN,
+                                      getNode(SI.getOperand(2))));
+   }
+ }
+ 
+ void Andersens::visitVANext(VANextInst &I) {
+   // FIXME: Implement
+   assert(0 && "vanext not handled yet!");
+ }
+ void Andersens::visitVAArg(VAArgInst &I) {
+   assert(0 && "vaarg not handled yet!");
+ }
+ 
+ /// AddConstraintsForCall - Add constraints for a call with actual arguments
+ /// specified by CS to the function specified by F.  Note that the types of
+ /// arguments might not match up in the case where this is an indirect call and
+ /// the function pointer has been casted.  If this is the case, do something
+ /// reasonable.
+ void Andersens::AddConstraintsForCall(CallSite CS, Function *F) {
+   if (isa<PointerType>(CS.getType())) {
+     Node *CSN = getNode(CS.getInstruction());
+     if (isa<PointerType>(F->getFunctionType()->getReturnType())) {
+       Constraints.push_back(Constraint(Constraint::Copy, CSN,
+                                        getReturnNode(F)));
+     } else {
+       // If the function returns a non-pointer value, handle this just like we
+       // treat a nonpointer cast to pointer.
+       Constraints.push_back(Constraint(Constraint::Copy, CSN,
+                                        &GraphNodes[UniversalSet]));
+     }
+   } else if (isa<PointerType>(F->getFunctionType()->getReturnType())) {
+     Constraints.push_back(Constraint(Constraint::Copy,
+                                      &GraphNodes[UniversalSet],
+                                      getReturnNode(F)));
+   }
+   
+   Function::aiterator AI = F->abegin(), AE = F->aend();
+   CallSite::arg_iterator ArgI = CS.arg_begin(), ArgE = CS.arg_end();
+   for (; AI != AE && ArgI != ArgE; ++AI, ++ArgI)
+     if (isa<PointerType>(AI->getType())) {
+       if (isa<PointerType>((*ArgI)->getType())) {
+         // Copy the actual argument into the formal argument.
+         Constraints.push_back(Constraint(Constraint::Copy, getNode(AI),
+                                          getNode(*ArgI)));
+       } else {
+         Constraints.push_back(Constraint(Constraint::Copy, getNode(AI),
+                                          &GraphNodes[UniversalSet]));
+       }
+     } else if (isa<PointerType>((*ArgI)->getType())) {
+       Constraints.push_back(Constraint(Constraint::Copy,
+                                        &GraphNodes[UniversalSet],
+                                        getNode(*ArgI)));
+     }
+   
+   // Copy all pointers passed through the varargs section to the varargs node.
+   if (F->getFunctionType()->isVarArg())
+     for (; ArgI != ArgE; ++ArgI)
+       if (isa<PointerType>((*ArgI)->getType()))
+         Constraints.push_back(Constraint(Constraint::Copy, getVarargNode(F),
+                                          getNode(*ArgI)));
+   // If more arguments are passed in than we track, just drop them on the floor.
+ }
+ 
+ void Andersens::visitCallSite(CallSite CS) {
+   if (isa<PointerType>(CS.getType()))
+     getNodeValue(*CS.getInstruction());
+ 
+   if (Function *F = CS.getCalledFunction()) {
+     AddConstraintsForCall(CS, F);
+   } else {
+     // We don't handle indirect call sites yet.  Keep track of them for when we
+     // discover the call graph incrementally.
+     IndirectCalls.push_back(CS);
+   }
+ }
+ 
+ //===----------------------------------------------------------------------===//
+ //                         Constraint Solving Phase
+ //===----------------------------------------------------------------------===//
+ 
+ /// intersects - Return true if the points-to set of this node intersects
+ /// with the points-to set of the specified node.
+ bool Andersens::Node::intersects(Node *N) const {
+   iterator I1 = begin(), I2 = N->begin(), E1 = end(), E2 = N->end();
+   while (I1 != E1 && I2 != E2) {
+     if (*I1 == *I2) return true;
+     if (*I1 < *I2)
+       ++I1;
+     else
+       ++I2;
+   }
+   return false;
+ }
+ 
+ /// intersectsIgnoring - Return true if the points-to set of this node
+ /// intersects with the points-to set of the specified node on any nodes
+ /// except for the specified node to ignore.
+ bool Andersens::Node::intersectsIgnoring(Node *N, Node *Ignoring) const {
+   iterator I1 = begin(), I2 = N->begin(), E1 = end(), E2 = N->end();
+   while (I1 != E1 && I2 != E2) {
+     if (*I1 == *I2) {
+       if (*I1 != Ignoring) return true;
+       ++I1; ++I2;
+     } else if (*I1 < *I2)
+       ++I1;
+     else
+       ++I2;
+   }
+   return false;
+ }
+ 
+ // Copy constraint: all edges out of the source node get copied to the
+ // destination node.  This returns true if a change is made.
+ bool Andersens::Node::copyFrom(Node *N) {
+   // Use a mostly linear-time merge since both of the lists are sorted.
+   bool Changed = false;
+   iterator I = N->begin(), E = N->end();
+   unsigned i = 0;
+   while (I != E && i != Pointees.size()) {
+     if (Pointees[i] < *I) {
+       ++i;
+     } else if (Pointees[i] == *I) {
+       ++i; ++I;
+     } else {
+       // We found a new element to copy over.
+       Changed = true;
+       Pointees.insert(Pointees.begin()+i, *I);
+        ++i; ++I;
+     }
+   }
+ 
+   if (I != E) {
+     Pointees.insert(Pointees.end(), I, E);
+     Changed = true;
+   }
+ 
+   return Changed;
+ }
+ 
+ bool Andersens::Node::loadFrom(Node *N) {
+   bool Changed = false;
+   for (iterator I = N->begin(), E = N->end(); I != E; ++I)
+     Changed |= copyFrom(*I);
+   return Changed;
+ }
+ 
+ bool Andersens::Node::storeThrough(Node *N) {
+   bool Changed = false;
+   for (iterator I = begin(), E = end(); I != E; ++I)
+     Changed |= (*I)->copyFrom(N);
+   return Changed;
+ }
+ 
+ 
+ /// SolveConstraints - This stage iteratively processes the constraints list
+ /// propagating constraints (adding edges to the Nodes in the points-to graph)
+ /// until a fixed point is reached.
+ ///
+ void Andersens::SolveConstraints() {
+   bool Changed = true;
+   unsigned Iteration = 0;
+   while (Changed) {
+     Changed = false;
+     ++NumIters;
+     DEBUG(std::cerr << "Starting iteration #" << Iteration++ << "!\n");
+ 
+     // Loop over all of the constraints, applying them in turn.
+     for (unsigned i = 0, e = Constraints.size(); i != e; ++i) {
+       Constraint &C = Constraints[i];
+       switch (C.Type) {
+       case Constraint::Copy:
+         Changed |= C.Dest->copyFrom(C.Src);
+         break;
+       case Constraint::Load:
+         Changed |= C.Dest->loadFrom(C.Src);
+         break;
+       case Constraint::Store:
+         Changed |= C.Dest->storeThrough(C.Src);
+         break;
+       default:
+         assert(0 && "Unknown constraint!");
+       }
+     }
+ 
+     if (Changed) {
+       // Check to see if any internal function's addresses have been passed to
+       // external functions.  If so, we have to assume that their incoming
+       // arguments could be anything.  If there are any internal functions in
+       // the universal node that we don't know about, we must iterate.
+       for (Node::iterator I = GraphNodes[UniversalSet].begin(),
+              E = GraphNodes[UniversalSet].end(); I != E; ++I)
+         if (Function *F = dyn_cast_or_null<Function>((*I)->getValue()))
+           if (F->hasInternalLinkage() &&
+               EscapingInternalFunctions.insert(F).second) {
+             // We found a function that is just now escaping.  Mark it as if it
+             // didn't have internal linkage.
+             AddConstraintsForNonInternalLinkage(F);
+             DEBUG(std::cerr << "Found escaping internal function: "
+                             << F->getName() << "\n");
+             ++NumEscapingFunctions;
+           }
+ 
+       // Check to see if we have discovered any new callees of the indirect call
+       // sites.  If so, add constraints to the analysis.
+       for (unsigned i = 0, e = IndirectCalls.size(); i != e; ++i) {
+         CallSite CS = IndirectCalls[i];
+         std::vector<Function*> &KnownCallees = IndirectCallees[CS];
+         Node *CN = getNode(CS.getCalledValue());
+ 
+         for (Node::iterator NI = CN->begin(), E = CN->end(); NI != E; ++NI)
+           if (Function *F = dyn_cast_or_null<Function>((*NI)->getValue())) {
+             std::vector<Function*>::iterator IP =
+               std::lower_bound(KnownCallees.begin(), KnownCallees.end(), F);
+             if (IP == KnownCallees.end() || *IP != F) {
+               // Add the constraints for the call now.
+               AddConstraintsForCall(CS, F);
+               DEBUG(std::cerr << "Found actual callee '"
+                               << F->getName() << "' for call: "
+                               << *CS.getInstruction() << "\n");
+               ++NumIndirectCallees;
+               KnownCallees.insert(IP, F);
+             }
+           }
+       }
+     }
+   }
+ }
+ 
+ 
+ 
+ //===----------------------------------------------------------------------===//
+ //                               Debugging Output
+ //===----------------------------------------------------------------------===//
+ 
+ void Andersens::PrintNode(Node *N) {
+   if (N == &GraphNodes[UniversalSet]) {
+     std::cerr << "<universal>";
+     return;
+   } else if (N == &GraphNodes[NullPtr]) {
+     std::cerr << "<nullptr>";
+     return;
+   } else if (N == &GraphNodes[NullObject]) {
+     std::cerr << "<null>";
+     return;
+   }
+ 
+   assert(N->getValue() != 0 && "Never set node label!");
+   Value *V = N->getValue();
+   if (Function *F = dyn_cast<Function>(V)) {
+     if (isa<PointerType>(F->getFunctionType()->getReturnType()) &&
+         N == getReturnNode(F)) {
+       std::cerr << F->getName() << ":retval";
+       return;
+     } else if (F->getFunctionType()->isVarArg() && N == getVarargNode(F)) {
+       std::cerr << F->getName() << ":vararg";
+       return;
+     }
+   }
+ 
+   if (Instruction *I = dyn_cast<Instruction>(V))
+     std::cerr << I->getParent()->getParent()->getName() << ":";
+   else if (Argument *Arg = dyn_cast<Argument>(V))
+     std::cerr << Arg->getParent()->getName() << ":";
+ 
+   if (V->hasName())
+     std::cerr << V->getName();
+   else
+     std::cerr << "(unnamed)";
+ 
+   if (isa<GlobalValue>(V) || isa<AllocationInst>(V))
+     if (N == getObject(V))
+       std::cerr << "<mem>";
+ }
+ 
+ void Andersens::PrintConstraints() {
+   std::cerr << "Constraints:\n";
+   for (unsigned i = 0, e = Constraints.size(); i != e; ++i) {
+     std::cerr << "  #" << i << ":  ";
+     Constraint &C = Constraints[i];
+     if (C.Type == Constraint::Store)
+       std::cerr << "*";
+     PrintNode(C.Dest);
+     std::cerr << " = ";
+     if (C.Type == Constraint::Load)
+       std::cerr << "*";
+     PrintNode(C.Src);
+     std::cerr << "\n";
+   }
+ }
+ 
+ void Andersens::PrintPointsToGraph() {
+   std::cerr << "Points-to graph:\n";
+   for (unsigned i = 0, e = GraphNodes.size(); i != e; ++i) {
+     Node *N = &GraphNodes[i];
+     std::cerr << "[" << (N->end() - N->begin()) << "] ";
+     PrintNode(N);
+     std::cerr << "\t--> ";
+     for (Node::iterator I = N->begin(), E = N->end(); I != E; ++I) {
+       if (I != N->begin()) std::cerr << ", ";
+       PrintNode(*I);
+     }
+     std::cerr << "\n";
+   }
+ }




From lattner at cs.uiuc.edu  Sun May 23 16:04:01 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:04:01 2004
Subject: [llvm-commits] CVS: llvm/docs/GarbageCollection.html
Message-ID: <200405232102.QAA31123@zion.cs.uiuc.edu>


Changes in directory llvm/docs:

GarbageCollection.html added (r1.1)

---
Log message:

Add a new document describing the LLVM accurate garbage collection support.


---
Diffs of the changes:  (+418 -0)

Index: llvm/docs/GarbageCollection.html
diff -c /dev/null llvm/docs/GarbageCollection.html:1.1
*** /dev/null	Sun May 23 16:02:30 2004
--- llvm/docs/GarbageCollection.html	Sun May 23 16:02:20 2004
***************
*** 0 ****
--- 1,418 ----
+ <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
+                       "http://www.w3.org/TR/html4/strict.dtd">
+ <html>
+ <head>
+   <title>Accurate Garbage Collection with LLVM</title>
+   <link rel="stylesheet" href="llvm.css" type="text/css">
+ </head>
+ <body>
+ 
+ <div class="doc_title">
+   Accurate Garbage Collection with LLVM
+ </div>
+ 
+ <ol>
+   <li><a href="#introduction">Introduction</a>
+     <ul>
+     <li><a href="#feature">GC features provided and algorithms supported</a></li>
+     </ul>
+   </li>
+ 
+   <li><a href="#interfaces">Interfaces for user programs</a>
+     <ul>
+     <li><a href="#roots">Identifying GC roots on the stack: <tt>llvm.gcroot</tt></a></li>
+     <li><a href="#gcdescriptors">GC descriptor format for heap objects</a></li>
+     <li><a href="#allocate">Allocating memory from the GC</a></li>
+     <li><a href="#barriers">Reading and writing references to the heap</a></li>
+     <li><a href="#explicit">Explicit invocation of the garbage collector</a></li>
+     </ul>
+   </li>
+ 
+   <li><a href="#gcimpl">Implementing a garbage collector</a>
+     <ul>
+     <li><a href="#llvm_gc_readwrite">Implementing <tt>llvm_gc_read</tt> and <tt>llvm_gc_write</tt></a></li>
+     <li><a href="#traceroots">Tracing the GC roots from the program stack</a></li>
+     <li><a href="#gcimpls">GC implementations available</a></li>
+     </ul>
+   </li>
+ 
+ <!--
+   <li><a href="#codegen">Implementing GC support in a code generator</a></li>
+ -->
+ </ol>
+ 
+ <div class="doc_author">
+   <p>Written by <a href="mailto:sabre at nondot.org">Chris Lattner</a></p>
+ </div>
+ 
+ <!-- *********************************************************************** -->
+ <div class="doc_section">
+   <a name="introduction">Introduction</a>
+ </div>
+ <!-- *********************************************************************** -->
+ 
+ <div class="doc_text">
+ 
+ <p>Garbage collection is a widely used technique that frees the programmer from
+ having to know the life-times of heap objects, making software easier to produce
+ and maintain.  Many programming languages rely on garbage collection for
+ automatic memory management.  There are two primary forms of garbage collection:
+ conservative and accurate.</p>
+ 
+ <p>Conservative garbage collection often does not require any special support
+ from either the language or the compiler: it can handle non-type-safe
+ programming languages (such as C/C++) and does not require any special
+ information from the compiler.  The [LINK] Boehm collector is an example of a
+ state-of-the-art conservative collector.</p>
+ 
+ <p>Accurate garbage collection requires the ability to identify all pointers in
+ the program at run-time (which requires that the source-language be type-safe in
+ most cases).  Identifying pointers at run-time requires compiler support to
+ locate all places that hold live pointer variables at run-time, including the
+ <a href="#roots">processor stack and registers</a>.</p>
+ 
+ <p>
+ Conservative garbage collection is attractive because it does not require any
+ special compiler support, but it does have problems.  In particular, because the
+ conservative garbage collector cannot <i>know</i> that a particular word in the
+ machine is a pointer, it cannot move live objects in the heap (preventing the
+ use of compacting and generational GC algorithms) and it can occasionally suffer
+ from memory leaks due to integer values that happen to point to objects in the
+ program.  In addition, some aggressive compiler transformations can break
+ conservative garbage collectors (though these seem rare in practice).
+ </p>
+ 
+ <p>
+ Accurate garbage collectors do not suffer from any of these problems, but they
+ can suffer from degraded scalar optimization of the program.  In particular,
+ because the runtime must be able to identify and update all pointers active in
+ the program, some optimizations are less effective.  In practice, however, the
+ locality and performance benefits of using aggressive garbage allocation
+ techniques dominates any low-level losses.
+ </p>
+ 
+ <p>
+ This document describes the mechanisms and interfaces provided by LLVM to
+ support accurate garbage collection.
+ </p>
+ 
+ </div>
+ 
+ <!-- ======================================================================= -->
+ <div class="doc_subsection">
+   <a name="feature">GC features provided and algorithms supported</a>
+ </div>
+ 
+ <div class="doc_text">
+ 
+ <p>
+ LLVM provides support for a broad class of garbage collection algorithms,
+ including compacting semi-space collectors, mark-sweep collectors, generational
+ collectors, and even reference counting implementations.  It includes support
+ for <a href="#barriers">read and write barriers</a>, and associating <a
+ href="#roots">meta-data with stack objects</a> (used for tagless garbage
+ collection).  All LLVM code generators support garbage collection, including the
+ C backend.
+ </p>
+ 
+ <p>
+ We hope that the primitive support built into LLVM is sufficient to support a
+ broad class of garbage collected languages, including Scheme, ML, scripting
+ languages, Java, C#, etc.  That said, the implemented garbage collectors may
+ need to be extended to support language-specific features such as finalization,
+ weak references, or other features.  As these needs are identified and
+ implemented, they should be added to this specification.
+ </p>
+ 
+ <p>
+ LLVM does not currently support garbage collection of multi-threaded programs or
+ GC-safe points other than function calls, but these will be added in the future
+ as there is interest.
+ </p>
+ 
+ </div>
+ 
+ <!-- *********************************************************************** -->
+ <div class="doc_section">
+   <a name="interfaces">Interfaces for user programs</a>
+ </div>
+ <!-- *********************************************************************** -->
+ 
+ <div class="doc_text">
+ 
+ <p>This section describes the interfaces provided by LLVM and by the garbage
+ collector run-time that should be used by user programs.  As such, this is the
+ interface that front-end authors should generate code for.
+ </p>
+ 
+ </div>
+ 
+ <!-- ======================================================================= -->
+ <div class="doc_subsection">
+   <a name="roots">Identifying GC roots on the stack: <tt>llvm.gcroot</tt></a>
+ </div>
+ 
+ <div class="doc_text">
+ 
+ <div class="doc_code"><tt>
+   void %llvm.gcroot(&lt;ty&gt;** %ptrloc, &lt;ty2&gt;* %metadata)
+ </tt></div>
+ 
+ <p>
+ The <tt>llvm.gcroot</tt> intrinsic is used to inform LLVM of a pointer variable
+ on the stack.  The first argument contains the address of the variable on the
+ stack, and the second contains a pointer to metadata that should be associated
+ with the pointer (which <b>must</b> be a constant or global value address).  At
+ runtime, the <tt>llvm.gcroot</tt> intrinsic stores a null pointer into the
+ specified location to initialize the pointer.</p>
+ 
+ <p>
+ Consider the following fragment of Java code:
+ </p>
+ 
+ <pre>
+        {
+          Object X;   // A null-initialized reference to an object
+          ...
+        }
+ </pre>
+ 
+ <p>
+ This block (which may be located in the middle of a function or in a loop nest),
+ could be compiled to this LLVM code:
+ </p>
+ 
+ <pre>
+ Entry:
+    ;; In the entry block for the function, allocate the
+    ;; stack space for X, which is an LLVM pointer.
+    %X = alloca %Object*
+    ...
+ 
+    ;; "CodeBlock" is the block corresponding to the start
+    ;;  of the scope scope above.
+ CodeBlock:
+    ;; Initialize the object, telling LLVM that it is now live.
+    ;; Java has type-tags on objects, so it doesn't need any
+    ;; metadata.
+    call void %llvm.gcroot(%Object** %X, sbyte* null)
+    ...
+ 
+    ;; As the pointer goes out of scope, store a null value into
+    ;; it, to indicate that the value is no longer live.
+    store %Object* null, %Object** %X
+    ...
+ </pre>
+ 
+ </div>
+ 
+ <!-- ======================================================================= -->
+ <div class="doc_subsection">
+   <a name="gcdescriptors">GC descriptor format for heap objects</a>
+ </div>
+ 
+ <div class="doc_text">
+ 
+ <p>
+ Either from root meta data, or from object headers.  Front-end can provide a
+ call-back to get descriptor from object without meta-data.
+ </p>
+ 
+ </div>
+ 
+ <!-- ======================================================================= -->
+ <div class="doc_subsection">
+   <a name="allocate">Allocating memory from the GC</a>
+ </div>
+ 
+ <div class="doc_text">
+ 
+ <div class="doc_code"><tt>
+   sbyte *%llvm_gc_allocate(unsigned %Size)
+ </tt></div>
+ 
+ <p>The <tt>llvm_gc_allocate</tt> function is a global function defined by the
+ garbage collector implementation to allocate memory.  It should return a
+ zeroed-out block of memory of the appropriate size.</p>
+ 
+ </div>
+ 
+ <!-- ======================================================================= -->
+ <div class="doc_subsection">
+   <a name="barriers">Reading and writing references to the heap</a>
+ </div>
+ 
+ <div class="doc_text">
+ 
+ <div class="doc_code"><tt>
+   sbyte *%llvm.gcread(sbyte **)<br>
+   void %llvm.gcwrite(sbyte*, sbyte**)
+ </tt></div>
+ 
+ <p>Several of the more interesting garbage collectors (e.g., generational
+ collectors) need to be informed when the mutator (the program that needs garbage
+ collection) reads or writes object references into the heap.  In the case of a
+ generational collector, it needs to keep track of which "old" generation objects
+ have references stored into them.  The amount of code that typically needs to be
+ executed is usually quite small, so the overall performance impact of the
+ inserted code is tolerable.</p>
+ 
+ <p>To support garbage collectors that use read or write barriers, LLVM provides
+ the <tt>llvm.gcread</tt> and <tt>llvm.gcwrite</tt> intrinsics.  The first
+ intrinsic has exactly the same semantics as a non-volatile LLVM load and the
+ second has the same semantics as a non-volatile LLVM store.  At code generation
+ time, these intrinsics are replaced with calls into the garbage collector
+ (<tt><a href="#llvm_gc_readwrite">llvm_gc_read</a></tt> and <tt><a
+ href="#llvm_gc_readwrite">llvm_gc_write</a></tt> respectively), which are then
+ inlined into the code.
+ </p>
+ 
+ <p>
+ If you are writing a front-end for a garbage collected language, every load or
+ store of a reference from or to the heap should use these intrinsics instead of
+ normal LLVM loads/stores.</p>
+ 
+ </div>
+ 
+ <!-- ======================================================================= -->
+ <div class="doc_subsection">
+   <a name="initialize">Garbage collector startup and initialization</a>
+ </div>
+ 
+ <div class="doc_text">
+ 
+ <div class="doc_code"><tt>
+   void %llvm_gc_initialize()
+ </tt></div>
+ 
+ <p>
+ The <tt>llvm_gc_initialize</tt> function should be called once before any other
+ garbage collection functions are called.  This gives the garbage collector the
+ chance to initialize itself and allocate the heap spaces.
+ </p>
+ 
+ </div>
+ 
+ <!-- ======================================================================= -->
+ <div class="doc_subsection">
+   <a name="explicit">Explicit invocation of the garbage collector</a>
+ </div>
+ 
+ <div class="doc_text">
+ 
+ <div class="doc_code"><tt>
+   void %llvm_gc_collect()
+ </tt></div>
+ 
+ <p>
+ The <tt>llvm_gc_collect</tt> function is exported by the garbage collector
+ implementations to provide a full collection, even when the heap is not
+ exhausted.  This can be used by end-user code as a hint, and may be ignored by
+ the garbage collector.
+ </p>
+ 
+ </div>
+ 
+ 
+ <!-- *********************************************************************** -->
+ <div class="doc_section">
+   <a name="gcimpl">Implementing a garbage collector</a>
+ </div>
+ <!-- *********************************************************************** -->
+ 
+ <div class="doc_text">
+ 
+ <p>
+ Implementing a garbage collector for LLVM is fairly straight-forward.  The
+ implementation must include the <a
+ href="#allocate"><tt>llvm_gc_allocate</tt></a> and <a
+ href="#explicit"><tt>llvm_gc_collect</tt></a> functions, and it must implement
+ the <a href="#llvm_gc_readwrite">read/write barrier</a> functions as well.  To
+ do this, it will probably have to <a href="#traceroots">trace through the roots
+ from the stack</a> and understand the <a href="#gcdescriptors">GC descriptors
+ for heap objects</a>.  Luckily, there are some <a href="#gcimpls">example
+ implementations</a> available.
+ </p>
+ </div>
+ 
+ 
+ <!-- ======================================================================= -->
+ <div class="doc_subsection">
+   <a name="llvm_gc_readwrite">Implementing <tt>llvm_gc_read</tt> and <tt>llvm_gc_write</tt></a>
+ </div>
+ 
+ <div class="doc_text">
+   <div class="doc_code"><tt>
+     void *llvm_gc_read(void **)<br>
+     void llvm_gc_write(void*, void**)
+  </tt></div>
+ 
+ <p>
+ These functions <i>must</i> be implemented in every garbage collector, even if
+ they do not need read/write barriers.  In this case, just load or store the
+ pointer, then return.
+ </p>
+ 
+ <p>
+ If an actual read or write barrier is needed, it should be straight-forward to
+ implement it.  Note that we may add a pointer to the start of the memory object
+ as a parameter in the future, if needed.
+ </p>
+ 
+ </div>
+ 
+ <!-- ======================================================================= -->
+ <div class="doc_subsection">
+   <a name="traceroots">Tracing the GC roots from the program stack</a>
+ </div>
+ 
+ <div class="doc_text">
+   <div class="doc_code"><tt>
+      void llvm_cg_walk_gcroots(void (*FP)(void **Root, void *Meta));
+   </tt></div>
+ 
+ <p>
+ The <tt>llvm_cg_walk_gcroots</tt> function is a function provided by the code
+ generator that iterates through all of the GC roots on the stack, calling the
+ specified function pointer with each record.  For each GC root, the address of
+ the pointer and the meta-data (from the <a
+ href="#gcroot"><tt>llvm.gcroot</tt></a> intrinsic) are provided.
+ </p>
+ </div>
+ 
+ 
+ <!-- ======================================================================= -->
+ <div class="doc_subsection">
+   <a name="gcimpls">GC implementations available</a>
+ </div>
+ 
+ <div class="doc_text">
+ 
+ <p>
+ To make this more concrete, the currently implemented LLVM garbage collectors
+ all live in the llvm/runtime/GC directory in the LLVM source-base.
+ </p>
+ 
+ <p>
+ TODO: Brief overview of each.
+ </p>
+ 
+ </div>
+ 
+ 
+ <!-- *********************************************************************** -->
+ 
+ <hr>
+ <address>
+   <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
+   src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
+   <a href="http://validator.w3.org/check/referer"><img
+   src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
+ 
+   <a href="mailto:sabre at nondot.org">Chris Lattner</a><br>
+   <a href="http://llvm.cs.uiuc.edu">LLVM Compiler Infrastructure</a><br>
+   Last modified: $Date: 2004/05/23 21:02:20 $
+ </address>
+ 
+ </body>
+ </html>




From lattner at cs.uiuc.edu  Sun May 23 16:05:01 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:05:01 2004
Subject: [llvm-commits] CVS: llvm/docs/llvm.css
Message-ID: <200405232103.QAA31733@zion.cs.uiuc.edu>


Changes in directory llvm/docs:

llvm.css updated: 1.16 -> 1.17

---
Log message:

Various changes to make the docs look nicer.  Add a new doc_author class.

The one thing that someone (misha?) should look at is fixing doc_code so that it sizes to fit the code instead of going to the right margin all of the way.  Fixing it at a particular number of pixels was just wrong. :)


---
Diffs of the changes:  (+10 -9)

Index: llvm/docs/llvm.css
diff -u llvm/docs/llvm.css:1.16 llvm/docs/llvm.css:1.17
--- llvm/docs/llvm.css:1.16	Thu May 13 11:37:52 2004
+++ llvm/docs/llvm.css	Sun May 23 16:03:30 2004
@@ -28,27 +28,28 @@
 }
 
 .doc_title      { text-align: left;   font-size: 25pt }
-.doc_section    { text-align: center; font-size: 22pt; }
-.doc_subsection { width: 50%;
+.doc_section    { text-align: center; font-size: 22pt;
+                  margin: 20pt 0pt 5pt 0pt; }
+.doc_subsection { width: 75%;
                   text-align: left;  font-size: 12pt; padding: 4pt 4pt 4pt 4pt;
-                  margin: 1.5em 0.5em 1.5em 0.5em }
+                  margin: 1.5em 0.5em 0.5em 0.5em }
 
-.doc_subsubsection { margin: 1.5em 0.5em 1.5em 0.5em;
-                     padding: 20pt 4pt 4pt 4pt;
+.doc_subsubsection { margin: 2.0em 0.5em 0.5em 0.5em;
                      font-weight: bold; font-style: oblique;
-                     border-bottom: 1px solid #999999; font-size: 12pt }
-
+                     border-bottom: 1px solid #999999; font-size: 12pt;
+                     width: 75%; }
+.doc_author     { text-align: left; font-weight: bold; padding-left: 20pt }
 .doc_text       { text-align: left; padding-left: 20pt }
 
 .doc_footer     { text-align: left; padding: 0 0 0 0 }
 
 .doc_red        { color: red }
 
-.doc_table	    { text-align: center; width: 90%; 
+.doc_table      { text-align: center; width: 90%; 
                   padding: 1px 1px 1px 1px; border: 1px; }
 
 .doc_warning    { color: red; font-weight: bold }
 
 .doc_code       { border: solid 1px gray; background: #eeeeee;
-                  width: 700px; margin: 0 0 0 1em; 
+                  margin: 0 0 0 1em; 
                   padding: 0 0 0 1em }




From lattner at cs.uiuc.edu  Sun May 23 16:06:03 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:06:03 2004
Subject: [llvm-commits] CVS: llvm/docs/AliasAnalysis.html
Message-ID: <200405232104.QAA32333@zion.cs.uiuc.edu>


Changes in directory llvm/docs:

AliasAnalysis.html updated: 1.18 -> 1.19

---
Log message:

Substantially expand and update the alias analysis documentation, including
adding blurbs about all of the implementations we have


---
Diffs of the changes:  (+521 -118)

Index: llvm/docs/AliasAnalysis.html
diff -u llvm/docs/AliasAnalysis.html:1.18 llvm/docs/AliasAnalysis.html:1.19
--- llvm/docs/AliasAnalysis.html:1.18	Wed May 12 13:02:00 2004
+++ llvm/docs/AliasAnalysis.html	Sun May 23 16:04:01 2004
@@ -2,52 +2,57 @@
                       "http://www.w3.org/TR/html4/strict.dtd">
 <html>
 <head>
-  <title>Alias Analysis Infrastructure in LLVM</title>
+  <title>The LLVM Alias Analysis Infrastructure</title>
   <link rel="stylesheet" href="llvm.css" type="text/css">
 </head>
 <body>
 
 <div class="doc_title">
-  Alias Analysis Infrastructure in LLVM
+  The LLVM Alias Analysis Infrastructure
 </div>
 
 <ol>
   <li><a href="#introduction">Introduction</a></li>
 
-  <li><a href="#overview">AliasAnalysis Overview</a>
+  <li><a href="#overview"><tt>AliasAnalysis</tt> Class Overview</a>
     <ul>
     <li><a href="#pointers">Representation of Pointers</a></li>
-    <li><a href="#MustMayNo">Must, May, and No Alias Responses</a></li>
+    <li><a href="#alias">The <tt>alias</tt> method</a></li>
     <li><a href="#ModRefInfo">The <tt>getModRefInfo</tt> methods</a></li>
-    </ul></li>
+    <li><a href="#OtherItfs">Other useful <tt>AliasAnalysis</tt> methods</a></li>
+    </ul>
+  </li>
 
-  <li><a href="#writingnew">Writing a new AliasAnalysis Implementation</a>
+  <li><a href="#writingnew">Writing a new <tt>AliasAnalysis</tt> Implementation</a>
     <ul>
     <li><a href="#passsubclasses">Different Pass styles</a></li>
     <li><a href="#requiredcalls">Required initialization calls</a></li>
     <li><a href="#interfaces">Interfaces which may be specified</a></li>
-    <li><a href="#chaining">The AliasAnalysis chaining behavior</a></li>
+    <li><a href="#chaining"><tt>AliasAnalysis</tt> chaining behavior</a></li>
+    <li><a href="#updating">Updating analysis results for transformations</a></li>
     <li><a href="#implefficiency">Efficiency Issues</a></li>
-    </ul></li>
+    </ul>
+  </li>
 
-  <li><a href="#using">Using AliasAnalysis results</a>
+  <li><a href="#using">Using alias analysis results</a>
     <ul>
     <li><a href="#loadvn">Using the <tt>-load-vn</tt> Pass</a></li>
     <li><a href="#ast">Using the <tt>AliasSetTracker</tt> class</a></li>
-    <li><a href="#direct">Using the AliasAnalysis interface directly</a></li>
-    </ul></li>
+    <li><a href="#direct">Using the <tt>AliasAnalysis</tt> interface directly</a></li>
+    </ul>
+  </li>
 
-  <li><a href="#tools">Helpful alias analysis related tools</a>
+  <li><a href="#exist">Existing alias analysis implementations and clients</a>
     <ul>
-    <li><a href="#no-aa">The <tt>-no-aa</tt> pass</a></li>
-    <li><a href="#print-alias-sets">The <tt>-print-alias-sets</tt> pass</a></li>
-    <li><a href="#count-aa">The <tt>-count-aa</tt> pass</a></li>
-    <li><a href="#aa-eval">The <tt>-aa-eval</tt> pass</a></li>
-    </ul></li>
+    <li><a href="#impls">Available <tt>AliasAnalysis</tt> implementations</a></li>
+    <li><a href="#aliasanalysis-xforms">Alias analysis driven transformations</a></li>
+    <li><a href="#aliasanalysis-debug">Clients for debugging and evaluation of implementations</a></li>
+    </ul>
+  </li>
 </ol>
 
-<div class="doc_text">    
-  <p><b>Written by <a href="mailto:sabre at nondot.org">Chris Lattner</a></b></p>
+<div class="doc_author">
+  <p>Written by <a href="mailto:sabre at nondot.org">Chris Lattner</a></p>
 </div>
 
 <!-- *********************************************************************** -->
@@ -58,20 +63,26 @@
 
 <div class="doc_text">
 
-<p>Alias Analysis (or Pointer Analysis) is a technique which attempts to
-determine whether or not two pointers ever can point to the same object in
-memory.  Traditionally, Alias Analyses respond to a query with either a <a
-href="#MustNoMay">Must, May, or No</a> alias response, indicating that two
-pointers do point to the same object, might point to the same object, or are
-known not to point to the same object.</p>
-
-<p>The <a href="/doxygen/classllvm_1_1AliasAnalysis.html">AliasAnalysis</a> class is the
-centerpiece of the LLVM Alias Analysis related infrastructure.  This class is
-the common interface between clients of alias analysis information and the
-implementations providing it.  In addition to simple alias analysis information,
-this class exposes Mod/Ref information from those implementations which can
-provide it, allowing for powerful analyses and transformations to work well
-together.</p>
+<p>Alias Analysis (aka Pointer Analysis) is a class of techniques which attempt
+to determine whether or not two pointers ever can point to the same object in
+memory.  There are many different algorithms for alias analysis and many
+different ways of classifying them: flow-sensitive vs flow-insensitive,
+context-sensitive vs context-insensitive, field-sensitive vs field-insensitive,
+unification-based vs subset-based, etc.  Traditionally, alias analyses respond
+to a query with a <a href="#MustNoMay">Must, May, or No</a> alias response,
+indicating that two pointers always point to the same object, might point to the
+same object, or are known to never point to the same object.</p>
+
+<p>The LLVM <a
+href="http://llvm.cs.uiuc.edu/doxygen/classllvm_1_1AliasAnalysis.html"><tt>AliasAnalysis</tt></a>
+class is the primary interface used by clients and implementations of alias
+analyses in the LLVM system.  This class is the common interface between clients
+of alias analysis information and the implementations providing it, and is
+designed to support a wide range of implementations and clients (but currently
+all clients are assumed to be flow-insensitive).  In addition to simple alias
+analysis information, this class exposes Mod/Ref information from those
+implementations which can provide it, allowing for powerful analyses and
+transformations to work well together.</p>
 
 <p>This document contains information necessary to successfully implement this
 interface, use it, and to test both sides.  It also explains some of the finer
@@ -83,24 +94,25 @@
 
 <!-- *********************************************************************** -->
 <div class="doc_section">
-  <a name="overview">AliasAnalysis Overview</a>
+  <a name="overview"><tt>AliasAnalysis</tt> Class Overview</a>
 </div>
 <!-- *********************************************************************** -->
 
 <div class="doc_text">
 
-<p>The <a href="/doxygen/classllvm_1_1AliasAnalysis.html">AliasAnalysis</a> class
-defines the interface that Alias Analysis implementations should support.  This
-class exports two important enums: <tt>AliasResult</tt> and
-<tt>ModRefResult</tt> which represent the result of an alias query or a mod/ref
-query, respectively.</p>
+<p>The <a
+href="http://llvm.cs.uiuc.edu/doxygen/classllvm_1_1AliasAnalysis.html"><tt>AliasAnalysis</tt></a>
+class defines the interface that the various alias analysis implementations
+should support.  This class exports two important enums: <tt>AliasResult</tt>
+and <tt>ModRefResult</tt> which represent the result of an alias query or a
+mod/ref query, respectively.</p>
 
-<p>The AliasAnalysis interface exposes information about memory, represented in
-several different ways.  In particular, memory objects are represented as a
-starting address and size, and function calls are represented as the actual
-<tt>call</tt> or <tt>invoke</tt> instructions that performs the call.  The
-AliasAnalysis interface also exposes some helper methods which allow you to get
-mod/ref information for arbitrary instructions.</p>
+<p>The <tt>AliasAnalysis</tt> interface exposes information about memory,
+represented in several different ways.  In particular, memory objects are
+represented as a starting address and size, and function calls are represented
+as the actual <tt>call</tt> or <tt>invoke</tt> instructions that performs the
+call.  The <tt>AliasAnalysis</tt> interface also exposes some helper methods
+which allow you to get mod/ref information for arbitrary instructions.</p>
 
 </div>
 
@@ -111,13 +123,15 @@
 
 <div class="doc_text">
 
-<p>Most importantly, the AliasAnalysis class provides several methods which are
-used to query whether or not pointers alias, whether function calls can modify
-or read memory, etc.</p>
+<p>Most importantly, the <tt>AliasAnalysis</tt> class provides several methods
+which are used to query whether or not two memory objects alias, whether
+function calls can modify or read a memory object, etc.  For all of these
+queries, memory objects are represented as a pair of their starting address (a
+symbolic LLVM <tt>Value*</tt>) and a static size.</p>
 
 <p>Representing memory objects as a starting address and a size is critically
-important for precise Alias Analyses.  For example, consider this (silly) C
-code:</p>
+important for correct Alias Analyses.  For example, consider this (silly, but
+possible) C code:</p>
 
 <pre>
   int i;
@@ -156,6 +170,17 @@
 
 <!-- ======================================================================= -->
 <div class="doc_subsection">
+  <a name="alias">The <tt>alias</tt> method</a>
+</div>
+  
+<div class="doc_text">
+The <tt>alias</tt> method is the primary interface used to determine whether or
+not two memory objects alias each other.  It takes two memory objects as input
+and returns MustAlias, MayAlias, or NoAlias as appropriate.
+</div>
+
+<!-- _______________________________________________________________________ -->
+<div class="doc_subsubsection">
   <a name="MustMayNo">Must, May, and No Alias Responses</a>
 </div>
 
@@ -163,13 +188,13 @@
 
 <p>An Alias Analysis implementation can return one of three responses:
 MustAlias, MayAlias, and NoAlias.  The No and May alias results are obvious: if
-the two pointers may never equal each other, return NoAlias, if they might,
+the two pointers can never equal each other, return NoAlias, if they might,
 return MayAlias.</p>
 
-<p>The Must Alias response is trickier though.  In LLVM, the Must Alias response
+<p>The MustAlias response is trickier though.  In LLVM, the Must Alias response
 may only be returned if the two memory objects are guaranteed to always start at
 exactly the same location.  If two memory objects overlap, but do not start at
-the same location, MayAlias must be returned.</p>
+the same location, return MayAlias.</p>
 
 </div>
 
@@ -182,14 +207,103 @@
 
 <p>The <tt>getModRefInfo</tt> methods return information about whether the
 execution of an instruction can read or modify a memory location.  Mod/Ref
-information is always conservative: if an action <b>may</b> read a location, Ref
-is returned.</p>
+information is always conservative: if an instruction <b>might</b> read or write
+a location, ModRef is returned.</p>
+
+<p>The <tt>AliasAnalysis</tt> class also provides a <tt>getModRefInfo</tt>
+method for testing dependencies between function calls.  This method takes two
+call sites (CS1 &amp; CS2), returns NoModRef if the two calls refer to disjoint
+memory locations, Ref if CS1 reads memory written by CS2, Mod if CS1 writes to
+memory read or written by CS2, or ModRef if CS1 might read or write memory
+accessed by CS2.  Note that this relation is not commutative.  Clients that use
+this method should be predicated on the <tt>hasNoModRefInfoForCalls()</tt>
+method, which indicates whether or not an analysis can provide mod/ref
+information for function call pairs (most can not).  If this predicate is false,
+the client shouldn't waste analysis time querying the <tt>getModRefInfo</tt>
+method many times.</p>
 
 </div>
 
+
+<!-- ======================================================================= -->
+<div class="doc_subsection">
+  <a name="OtherItfs">Other useful <tt>AliasAnalysis</tt> methods</a>
+</div>
+
+<div class="doc_text">
+
+<p>
+Several other tidbits of information are often collected by various alias
+analysis implementations and can be put to good use by various clients.
+</p>
+
+</div>
+
+<!-- _______________________________________________________________________ -->
+<div class="doc_subsubsection">
+  The <tt>getMustAliases</tt> method
+</div>
+
+<div class="doc_text">
+
+<p>The <tt>getMustAliases</tt> method returns all values that are known to
+always must alias a pointer.  This information can be provided in some cases for
+important objects like the null pointer and global values.  Knowing that a
+pointer always points to a particular function allows indirect calls to be
+turned into direct calls, for example.</p>
+
+</div>
+
+<!-- _______________________________________________________________________ -->
+<div class="doc_subsubsection">
+  The <tt>pointsToConstantMemory</tt> method
+</div>
+
+<div class="doc_text">
+
+<p>The <tt>pointsToConstantMemory</tt> method returns true if and only if the
+analysis can prove that the pointer only points to unchanging memory locations
+(functions, constant global variables, and the null pointer).  This information
+can be used to refine mod/ref information: it is impossible for an unchanging
+memory location to be modified.</p>
+
+</div>
+
+
+<!-- _______________________________________________________________________ -->
+<div class="doc_subsubsection">
+  <a name="simplemodref">The <tt>doesNotAccessMemory</tt> and
+  <tt>onlyReadsMemory</tt> methods</a>
+</div>
+
+<div class="doc_text">
+
+<p>These methods are used to provide very simple mod/ref information for
+function calls.  The <tt>doesNotAccessMemory</tt> method returns true for a
+function if the analysis can prove that the function never reads or writes to
+memory, or if the function only reads from constant memory.  Functions with this
+property are side-effect free and only depend on their input arguments, allowing
+them to be eliminated if they form common subexpressions or be hoisted out of
+loops.  Many common functions behave this way (e.g., <tt>sin</tt> and
+<tt>cos</tt>) but many others do not (e.g., <tt>acos</tt>, which modifies the
+<tt>errno</tt> variable).</p>
+
+<p>The <tt>onlyReadsMemory</tt> method returns true for a function if analysis
+can prove that (at most) the function only reads from non-volatile memory.
+Functions with this property are side-effect free, only depending on their input
+arguments and the state of memory when they are called.  This property allows
+calls to these functions to be eliminated and moved around, as long as there is
+no store instruction that changes the contents of memory.  Note that all
+functions that satisfy the <tt>doesNotAccessMemory</tt> method also satisfies
+<tt>onlyReadsMemory</tt>.</p>
+
+</div>
+
+
+
 <!-- *********************************************************************** -->
 <div class="doc_section">
-  <a name="writingnew">Writing a new AliasAnalysis Implementation</a>
+  <a name="writingnew">Writing a new <tt>AliasAnalysis</tt> Implementation</a>
 </div>
 <!-- *********************************************************************** -->
 
@@ -198,8 +312,8 @@
 <p>Writing a new alias analysis implementation for LLVM is quite
 straight-forward.  There are already several implementations that you can use
 for examples, and the following information should help fill in any details.
-For a minimal example, take a look at the <a
-href="/doxygen/structllvm_1_1NoAA.html"><tt>no-aa</tt></a> implementation.</p>
+For a examples, take a look at the <a href="#impls">various alias analysis
+implementations</a> included with LLVM.</p>
 
 </div>
 
@@ -219,8 +333,7 @@
 <ol>
   <li>If you require interprocedural analysis, it should be a
       <tt>Pass</tt>.</li>
-  <li>If you are a global analysis, subclass <tt>FunctionPass</tt>.</li>
-  <li>If you are a local pass, subclass <tt>BasicBlockPass</tt>.</li>
+  <li>If you are a function-local analysis, subclass <tt>FunctionPass</tt>.</li>
   <li>If you don't need to look at the program at all, subclass 
       <tt>ImmutablePass</tt>.</li>
 </ol>
@@ -239,8 +352,8 @@
 
 <div class="doc_text">
 
-<p>Your subclass of AliasAnalysis is required to invoke two methods on the
-AliasAnalysis base class: <tt>getAnalysisUsage</tt> and
+<p>Your subclass of <tt>AliasAnalysis</tt> is required to invoke two methods on
+the <tt>AliasAnalysis</tt> base class: <tt>getAnalysisUsage</tt> and
 <tt>InitializeAliasAnalysis</tt>.  In particular, your implementation of
 <tt>getAnalysisUsage</tt> should explicitly call into the
 <tt>AliasAnalysis::getAnalysisUsage</tt> method in addition to doing any
@@ -256,9 +369,8 @@
 
 <p>Additionally, your must invoke the <tt>InitializeAliasAnalysis</tt> method
 from your analysis run method (<tt>run</tt> for a <tt>Pass</tt>,
-<tt>runOnFunction</tt> for a <tt>FunctionPass</tt>, <tt>runOnBasicBlock</tt> for
-a <tt>BasicBlockPass</tt>, or <tt>InitializeAliasAnalysis</tt> for an
-<tt>ImmutablePass</tt>).  For example (as part of a <tt>Pass</tt>):</p>
+<tt>runOnFunction</tt> for a <tt>FunctionPass</tt>, or <tt>InitializePass</tt>
+for an <tt>ImmutablePass</tt>).  For example (as part of a <tt>Pass</tt>):</p>
 
 <pre>
     bool run(Module &amp;M) {
@@ -277,29 +389,110 @@
 
 <div class="doc_text">
 
-<p>All of the <a href="/doxygen/classllvm_1_1AliasAnalysis.html">AliasAnalysis</a>
-virtual methods default to providing conservatively correct information
-(returning "May" Alias and "Mod/Ref" for alias and mod/ref queries
+<p>All of the <a
+href="/doxygen/classllvm_1_1AliasAnalysis.html"><tt>AliasAnalysis</tt></a>
+virtual methods default to providing <a href="#chaining">chaining</a> to another
+alias analysis implementation, which ends up returning conservatively correct
+information (returning "May" Alias and "Mod/Ref" for alias and mod/ref queries
 respectively).  Depending on the capabilities of the analysis you are
 implementing, you just override the interfaces you can improve.</p>
 
 </div>
 
+
+
+<!-- ======================================================================= -->
+<div class="doc_subsection">
+  <a name="chaining"><tt>AliasAnalysis</tt> chaining behavior</a>
+</div>
+
+<div class="doc_text">
+
+<p>With only two special exceptions (the <tt><a
+href="#basic-aa">basicaa</a></tt> and <a href="#no-aa"><tt>no-aa</tt></a>
+passes) every alias analysis pass chains to another alias analysis
+implementation (for example, the user can specify "<tt>-basicaa -ds-aa
+-anders-aa -licm</tt>" to get the maximum benefit from the three alias
+analyses).  The alias analysis class automatically takes care of most of this
+for methods that you don't override.  For methods that you do override, in code
+paths that return a conservative MayAlias or Mod/Ref result, simply return
+whatever the superclass computes.  For example:</p>
+
+<pre>
+  AliasAnalysis::AliasResult alias(const Value *V1, unsigned V1Size,
+                                   const Value *V2, unsigned V2Size) {
+    if (...)
+      return NoAlias;
+    ...
+
+    <i>// Couldn't determine a must or no-alias result.</i>
+    return AliasAnalysis::alias(V1, V1Size, V2, V2Size);
+  }
+</pre>
+
+<p>In addition to analysis queries, you must make sure to unconditionally pass
+LLVM <a href="#updating">update notification</a> methods to the superclass as
+well if you override them, which allows all alias analyses in a change to be
+updated.</p>
+
+</div>
+
+
 <!-- ======================================================================= -->
 <div class="doc_subsection">
-  <a name="chaining">The AliasAnalysis chaining behavior</a>
+  <a name="updating">Updating analysis results for transformations</a>
 </div>
 
 <div class="doc_text">
+<p>
+Alias analysis information is initially computed for a static snapshot of the
+program, but clients will use this information to make transformations to the
+code.  All but the most trivial forms of alias analysis will need to have their
+analysis results updated to reflect the changes made by these transformations.
+</p>
 
-<p>With only two special exceptions (the <tt>basicaa</tt> and <a
-href="#no-aa"><tt>no-aa</tt></a> passes) every alias analysis pass should chain
-to another alias analysis implementation (for example, you could specify
-"<tt>-basic-aa -ds-aa -andersens-aa -licm</tt>" to get the maximum benefit from
-the three alias analyses).  To do this, simply "Require" AliasAnalysis in your
-<tt>getAnalysisUsage</tt> method, and if you need to return a conservative
-MayAlias or Mod/Ref result, simply chain to a lower analysis.</p>
+<p>
+The <tt>AliasAnalysis</tt> interface exposes two methods which are used to
+communicate program changes from the clients to the analysis implementations.
+Various alias analysis implementations should use these methods to ensure that
+their internal data structures are kept up-to-date as the program changes (for
+example, when an instruction is deleted), and clients of alias analysis must be
+sure to call these interfaces appropriately.
+</p>
+</div>
+
+<!-- _______________________________________________________________________ -->
+<div class="doc_subsubsection">The <tt>deleteValue</tt> method</div>
 
+<div class="doc_text">
+The <tt>deleteValue</tt> method is called by transformations when they remove an
+instruction or any other value from the program (including values that do not
+use pointers).  Typically alias analyses keep data structures that have entries
+for each value in the program.  When this method is called, they should remove
+any entries for the specified value, if they exist.
+</div>
+
+
+<!-- _______________________________________________________________________ -->
+<div class="doc_subsubsection">The <tt>copyValue</tt> method</div>
+
+<div class="doc_text">
+The <tt>copyValue</tt> method is used when a new value is introduced into the
+program.  There is no way to introduce a value into the program that did not
+exist before (this doesn't make sense for a safe compiler transformation), so
+this is the only way to introduce a new value.  This method indicates that the
+new value has exactly the same properties as the value being copied.
+</div>
+
+
+<!-- _______________________________________________________________________ -->
+<div class="doc_subsubsection">The <tt>replaceWithNewValue</tt> method</div>
+
+<div class="doc_text">
+This method is a simple helper method that is provided to make clients easier to
+use.  It is implemented by copying the old analysis information to the new
+value, then deleting the old value.  This method cannot be overridden by alias
+analysis implementations.
 </div>
 
 <!-- ======================================================================= -->
@@ -320,7 +513,7 @@
 
 <!-- *********************************************************************** -->
 <div class="doc_section">
-  <a name="using">Using AliasAnalysis results</a>
+  <a name="using">Using alias analysis results</a>
 </div>
 <!-- *********************************************************************** -->
 
@@ -339,10 +532,10 @@
 <div class="doc_text">
 
 <p>The <tt>load-vn</tt> pass uses alias analysis to provide value numbering
-information for <tt>load</tt> instructions.  If your analysis or transformation
-can be modelled in a form that uses value numbering information, you don't have
-to do anything special to handle load instructions: just use the
-<tt>load-vn</tt> pass, which uses alias analysis.</p>
+information for <tt>load</tt> instructions and pointer values.  If your analysis
+or transformation can be modeled in a form that uses value numbering
+information, you don't have to do anything special to handle load instructions:
+just use the <tt>load-vn</tt> pass, which uses alias analysis.</p>
 
 </div>
 
@@ -357,13 +550,13 @@
 in some scope, rather than information about pairwise aliasing.  The <tt><a
 href="/doxygen/classllvm_1_1AliasSetTracker.html">AliasSetTracker</a></tt> class is used
 to efficiently build these Alias Sets from the pairwise alias analysis
-information provided by the AliasAnalysis interface.</p>
+information provided by the <tt>AliasAnalysis</tt> interface.</p>
 
-<p>First you initialize the AliasSetTracker by use the "<tt>add</tt>" methods to
-add information about various potentially aliasing instructions in the scope you
-are interested in.  Once all of the alias sets are completed, your pass should
-simply iterate through the constructed alias sets, using the AliasSetTracker
-<tt>begin()</tt>/<tt>end()</tt> methods.</p>
+<p>First you initialize the AliasSetTracker by using the "<tt>add</tt>" methods
+to add information about various potentially aliasing instructions in the scope
+you are interested in.  Once all of the alias sets are completed, your pass
+should simply iterate through the constructed alias sets, using the
+<tt>AliasSetTracker</tt> <tt>begin()</tt>/<tt>end()</tt> methods.</p>
 
 <p>The <tt>AliasSet</tt>s formed by the <tt>AliasSetTracker</tt> are guaranteed
 to be disjoint, calculate mod/ref information and volatility for the set, and
@@ -372,16 +565,17 @@
 instructions, and can provide a list of pointers in each set.</p>
 
 <p>As an example user of this, the <a href="/doxygen/structLICM.html">Loop
-Invariant Code Motion</a> pass uses AliasSetTrackers to build alias information
-about each loop nest.  If an AliasSet in a loop is not modified, then all load
-instructions from that set may be hoisted out of the loop.  If any alias sets
-are stored <b>and</b> are must alias sets, then the stores may be sunk to
-outside of the loop, promoting the memory location to a register for the
-duration of the loop nest.  Both of these transformations obviously only apply
-if the pointer argument is loop-invariant.</p>
+Invariant Code Motion</a> pass uses <tt>AliasSetTracker</tt>s to calculate alias
+sets for each loop nest.  If an <tt>AliasSet</tt> in a loop is not modified,
+then all load instructions from that set may be hoisted out of the loop.  If any
+alias sets are stored to <b>and</b> are must alias sets, then the stores may be
+sunk to outside of the loop, promoting the memory location to a register for the
+duration of the loop nest.  Both of these transformations only apply if the
+pointer argument is loop-invariant.</p>
 
 </div>
 
+<!-- _______________________________________________________________________ -->
 <div class="doc_subsubsection">
   The AliasSetTracker implementation
 </div>
@@ -410,50 +604,257 @@
 
 <!-- ======================================================================= -->
 <div class="doc_subsection">
-  <a name="direct">Using the AliasAnalysis interface directly</a>
+  <a name="direct">Using the <tt>AliasAnalysis</tt> interface directly</a>
 </div>
 
 <div class="doc_text">
 
-<p>As a last resort, your pass could use the AliasAnalysis interface directly to
-service your pass.  If you find the need to do this, please <a
-href="mailto:sabre at nondot.org">let me know</a> so I can see if something new
-needs to be added to LLVM.</p>
+<p>If neither of these utility class are what your pass needs, you should use
+the interfaces exposed by the <tt>AliasAnalysis</tt> class directly.  Try to use
+the higher-level methods when possible (e.g., use mod/ref information instead of
+the <a href="#alias"><tt>alias</tt></a> method directly if possible) to get the
+best precision and efficiency.</p>
 
 </div>
 
 <!-- *********************************************************************** -->
 <div class="doc_section">
-  <a name="tools">Helpful alias-analysis-related tools</a>
+  <a name="exist">Existing alias analysis implementations and clients</a>
 </div>
 <!-- *********************************************************************** -->
 
 <div class="doc_text">
 
-<p>If you're going to be working with the AliasAnalysis infrastructure, there
-are several nice tools that may be useful for you and are worth knowing
-about...</p>
+<p>If you're going to be working with the LLVM alias analysis infrastructure,
+you should know what clients and implementations of alias analysis are
+available.  In particular, if you are implementing an alias analysis, you should
+be aware of the <a href="#aliasanalysis-debug">the clients</a> that are useful
+for monitoring and evaluating different implementations.</p>
 
 </div>
 
 <!-- ======================================================================= -->
 <div class="doc_subsection">
+  <a name="impls">Available <tt>AliasAnalysis</tt> implementations</a>
+</div>
+
+<div class="doc_text">
+
+<p>This section lists the various implementations of the <tt>AliasAnalysis</tt>
+interface.  With the exception of the <a href="#no-aa"><tt>-no-aa</tt></a> and
+<a href="#basic-aa"><tt>-basicaa</tt></a> implementations, all of these <a
+href="chaining">chain</a> to other alias analysis implementations.</p>
+
+</div>
+
+<!-- _______________________________________________________________________ -->
+<div class="doc_subsubsection">
   <a name="no-aa">The <tt>-no-aa</tt> pass</a>
 </div>
 
 <div class="doc_text">
 
-<p>The <tt>-no-aa</tt> analysis is just like what it sounds: an alias analysis
-that never returns any useful information.  This pass can be useful if you think
-that alias analysis is doing something wrong and are trying to narrow down a
-problem.  If you don't specify an alias analysis, the default will be to use the
-<tt>basicaa</tt> pass which does quite a bit of disambiguation on its own.</p>
+<p>The <tt>-no-aa</tt> pass is just like what it sounds: an alias analysis that
+never returns any useful information.  This pass can be useful if you think that
+alias analysis is doing something wrong and are trying to narrow down a
+problem.</p>
+
+</div>
+
+
+<!-- _______________________________________________________________________ -->
+<div class="doc_subsubsection">
+  <a name="basic-aa">The <tt>-basicaa</tt> pass</a>
+</div>
+
+<div class="doc_text">
+
+<p>The <tt>-basicaa</tt> pass is the default LLVM alias analysis.  It is an
+aggressive local analysis that "knows" many important facts:</p>
+
+<ul>
+<li>Distinct globals, stack allocations, and heap allocations can never
+    alias.</li>
+<li>Globals, stack allocations, and heap allocations never alias the null
+    pointer.</li>
+<li>Different fields of a structure do not alias.</li>
+<li>Indexes into arrays with statically differing subscripts cannot alias.</li>
+<li>Many common standard C library functions <a
+    href="#simplemodref">never access memory or only read memory</a>.</li>
+<li>Pointers that obviously point to constant globals
+    "<tt>pointToConstantMemory</tt>".</li>
+<li>Function calls can not modify or references stack allocations if they never
+    escape from the function that allocates them (a common case for automatic
+    arrays).</li>
+</ul>
+
+</div>
+
+
+<!-- _______________________________________________________________________ -->
+<div class="doc_subsubsection">
+  <a name="anders-aa">The <tt>-anders-aa</tt> pass</a>
+</div>
+
+<div class="doc_text">
+
+<p>The <tt>-anders-aa</tt> pass implements the well-known "Andersen's algorithm"
+for interprocedural alias analysis.  This algorithm is a subset-based,
+flow-insensitive, context-insensitive, and field-insensitive alias analysis that
+is widely believed to be fairly precise.  Unfortunately, this algorithm is also
+O(N<sup>3</sup>).  The LLVM implementation currently does not implement any of
+the refinements (such as "online cycle elimination" or "offline variable
+substitution") to improve its efficiency, so it can be quite slow in common
+cases.
+</p>
+
+</div>
+
+<!-- _______________________________________________________________________ -->
+<div class="doc_subsubsection">
+  <a name="steens-aa">The <tt>-steens-aa</tt> pass</a>
+</div>
+
+<div class="doc_text">
+
+<p>The <tt>-steens-aa</tt> pass implements a variation on the well-known
+"Steensgaard's algorithm" for interprocedural alias analysis.  Steensgaard's
+algorithm is a unification-based, flow-insensitive, context-insensitive, and
+field-insensitive alias analysis that is also very scalable (effectively linear
+time).</p>
+
+<p>The LLVM <tt>-steens-aa</tt> pass implements a "speculatively
+field-<b>sensitive</b>" version of Steensgaard's algorithm using the Data
+Structure Analysis framework.  This gives it substantially more precision than
+the standard algorithm while maintaining excellent analysis scalability.</p>
+
+</div>
+
+<!-- _______________________________________________________________________ -->
+<div class="doc_subsubsection">
+  <a name="ds-aa">The <tt>-ds-aa</tt> pass</a>
+</div>
+
+<div class="doc_text">
+
+<p>The <tt>-ds-aa</tt> pass implements the full Data Structure Analysis
+algorithm.  Data Structure Analysis is a modular unification-based,
+flow-insensitive, context-<b>sensitive</b>, and speculatively
+field-<b>sensitive</b> alias analysis that is also quite scalable, usually at
+O(n*log(n)).</p>
+
+<p>This algorithm is capable of responding to a full variety of alias analysis
+queries, and can provide context-sensitive mod/ref information as well.  The
+only major facility not implemented so far is support for must-alias
+information.</p>
 
 </div>
 
 
 <!-- ======================================================================= -->
 <div class="doc_subsection">
+  <a name="aliasanalysis-xforms">Alias analysis driven transformations</a>
+</div>
+
+<div class="doc_text">
+LLVM includes several alias-analysis driven transformations which can be used
+with any of the implementations above.
+</div>
+
+<!-- _______________________________________________________________________ -->
+<div class="doc_subsubsection">
+  <a name="adce">The <tt>-adce</tt> pass</a>
+</div>
+
+<div class="doc_text">
+
+<p>The <tt>-adce</tt> pass, which implements Aggressive Dead Code Elimination
+uses the <tt>AliasAnalysis</tt> interface to delete calls to functions that do
+not have side-effects and are not used.</p>
+
+</div>
+
+
+<!-- _______________________________________________________________________ -->
+<div class="doc_subsubsection">
+  <a name="licm">The <tt>-licm</tt> pass</a>
+</div>
+
+<div class="doc_text">
+
+<p>The <tt>-licm</tt> pass implements various Loop Invariant Code Motion related
+transformations.  It uses the <tt>AliasAnalysis</tt> interface for several
+different transformations:</p>
+
+<ul>
+<li>It uses mod/ref information to hoist or sink load instructions out of loops
+if there are no instructions in the loop that modifies the memory loaded.</li>
+
+<li>It uses mod/ref information to hoist function calls out of loops that do not
+write to memory and are loop-invariant.</li>
+
+<li>If uses alias information to promote memory objects that are loaded and
+stored to in loops to live in a register instead.  It can do this if there are
+no may aliases to the loaded/stored memory location.</li>
+</ul>
+
+</div>
+
+<!-- _______________________________________________________________________ -->
+<div class="doc_subsubsection">
+  <a name="argpromotion">The <tt>-argpromotion</tt> pass</a>
+</div>
+
+<div class="doc_text">
+<p>
+The <tt>-argpromotion</tt> pass promotes by-reference arguments to be passed in
+by-value instead.  In particular, if pointer arguments are only loaded from it
+passes in the value loaded instead of the address to the function.  This pass
+uses alias information to make sure that the value loaded from the argument
+pointer is not modified between the entry of the function and any load of the
+pointer.</p>
+</div>
+
+<!-- _______________________________________________________________________ -->
+<div class="doc_subsubsection">
+  <a name="gcseloadvn">The <tt>-load-vn</tt> &amp; <tt>-gcse</tt> passes</a>
+</div>
+
+<div class="doc_text">
+<p>
+The <tt>-load-vn</tt> pass uses alias analysis to "<a href="#loadvn">value
+number</a>" loads and pointers values, which is used by the GCSE pass to
+eliminate instructions.  The <tt>-load-vn</tt> pass relies on alias information
+and must-alias information.  This combination of passes can make the following
+transformations:</p>
+
+<ul>
+<li>Redundant load instructions are eliminated.</li>
+<li>Load instructions that follow a store to the same location are replaced with
+the stored value ("store forwarding").</li>
+<li>Pointers values (e.g. formal arguments) that must-alias simpler expressions
+(e.g. global variables or the null pointer) are replaced.  Note that this
+implements transformations like "virtual method resolution", turning indirect
+calls into direct calls.</li>
+</ul>
+
+</div>
+
+
+
+
+<!-- ======================================================================= -->
+<div class="doc_subsection">
+  <a name="aliasanalysis-debug">Clients for debugging and evaluation of implementations</a>
+</div>
+
+These passes are useful for evaluating the various alias analysis
+implementations.  You can use them with commands like '<tt>opt -anders-aa -ds-aa
+-aa-eval foo.bc -disable-output -stats</tt>'.
+
+
+<!-- _______________________________________________________________________ -->
+<div class="doc_subsubsection">
   <a name="print-alias-sets">The <tt>-print-alias-sets</tt> pass</a>
 </div>
 
@@ -462,20 +863,21 @@
 <p>The <tt>-print-alias-sets</tt> pass is exposed as part of the
 <tt>analyze</tt> tool to print out the Alias Sets formed by the <a
 href="#ast"><tt>AliasSetTracker</tt></a> class.  This is useful if you're using
-the <tt>AliasSetTracker</tt>.</p>
+the <tt>AliasSetTracker</tt> class.</p>
 
 </div>
 
-<!-- ======================================================================= -->
-<div class="doc_subsection">
+
+<!-- _______________________________________________________________________ -->
+<div class="doc_subsubsection">
   <a name="count-aa">The <tt>-count-aa</tt> pass</a>
 </div>
 
 <div class="doc_text">
 
 <p>The <tt>-count-aa</tt> pass is useful to see how many queries a particular
-pass is making and what kinds of responses are returned by the alias analysis.
-An example usage is:</p>
+pass is making and what responses are returned by the alias analysis.  An
+example usage is:</p>
 
 <pre>
   $ opt -basicaa -count-aa -ds-aa -count-aa -licm
@@ -484,12 +886,12 @@
 <p>Which will print out how many queries (and what responses are returned) by
 the <tt>-licm</tt> pass (of the <tt>-ds-aa</tt> pass) and how many queries are
 made of the <tt>-basicaa</tt> pass by the <tt>-ds-aa</tt> pass.  This can be
-useful when evaluating an alias analysis for precision.</p>
+useful when debugging a transformation or an alias analysis implementation.</p>
 
 </div>
 
-<!-- ======================================================================= -->
-<div class="doc_subsection">
+<!-- _______________________________________________________________________ -->
+<div class="doc_subsubsection">
   <a name="aa-eval">The <tt>-aa-eval</tt> pass</a>
 </div>
 
@@ -498,7 +900,8 @@
 <p>The <tt>-aa-eval</tt> pass simply iterates through all pairs of pointers in a
 function and asks an alias analysis whether or not the pointers alias.  This
 gives an indication of the precision of the alias analysis.  Statistics are
-printed.</p>
+printed indicating the percent of no/may/must aliases found (a more precise
+algorithm will have a lower number of may aliases).</p>
 
 </div>
 
@@ -513,7 +916,7 @@
 
   <a href="mailto:sabre at nondot.org">Chris Lattner</a><br>
   <a href="http://llvm.cs.uiuc.edu">LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/05/12 18:02:00 $
+  Last modified: $Date: 2004/05/23 21:04:01 $
 </address>
 
 </body>




From lattner at cs.uiuc.edu  Sun May 23 16:07:03 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:07:03 2004
Subject: [llvm-commits] CVS: llvm/docs/HowToSubmitABug.html
Message-ID: <200405232105.QAA01200@zion.cs.uiuc.edu>


Changes in directory llvm/docs:

HowToSubmitABug.html updated: 1.21 -> 1.22

---
Log message:

Lots of minor fixes and formatting improvements


---
Diffs of the changes:  (+28 -28)

Index: llvm/docs/HowToSubmitABug.html
diff -u llvm/docs/HowToSubmitABug.html:1.21 llvm/docs/HowToSubmitABug.html:1.22
--- llvm/docs/HowToSubmitABug.html:1.21	Fri May 14 14:04:07 2004
+++ llvm/docs/HowToSubmitABug.html	Sun May 23 16:05:39 2004
@@ -29,9 +29,9 @@
 
 </ol>
 
-<div class="doc_text">
-  <p><b>Written by <a href="mailto:sabre at nondot.org">Chris Lattner</a> and
-                   <a href="http://misha.brukman.net">Misha Brukman</a></b></p>
+<div class="doc_author">
+  <p>Written by <a href="mailto:sabre at nondot.org">Chris Lattner</a> and
+                <a href="http://misha.brukman.net">Misha Brukman</a></p>
 </div>
 
 </td>
@@ -148,7 +148,7 @@
 <tt>-save-temps</tt> option to <tt><b>llvm-gcc</b></tt>. Then run:</p>
 
 <div class="doc_code">
-<pre><b>gccas</b> -debug-pass=Arguments &lt; /dev/null -o - &gt; /dev/null</pre>
+<p><tt><b>gccas</b> -debug-pass=Arguments &lt; /dev/null -o - &gt; /dev/null</tt></p>
 </div>
 
 <p>... which will print a list of arguments, indicating the list of passes that
@@ -171,9 +171,9 @@
 the full list of objects linked).  Then run:</p>
 
 <div class="doc_code">
-<pre>
-<b>llvm-as</b> &lt; /dev/null &gt; null.bc
-<b>gccld</b> -debug-pass=Arguments null.bc</tt></p>
+<p><tt><b>llvm-as</b> &lt; /dev/null &gt; null.bc<br>
+<b>gccld</b> -debug-pass=Arguments null.bc</tt>
+</p>
 </div>
 
 <p>... which will print a list of arguments, indicating the list of passes that
@@ -196,19 +196,19 @@
 something tractable, use the <tt><b>bugpoint</b></tt> tool as follows:</p>
 
 <div class="doc_code">
-<pre><b>bugpoint</b> &lt;input files&gt; &lt;list of passes&gt;</pre>
+<p><tt><b>bugpoint</b> &lt;input files&gt; &lt;list of passes&gt;</tt></p>
 </div>
 
 <p><tt><b>bugpoint</b></tt> will print a bunch of output as it reduces the
 test-case, but it should eventually print something like this:</p>
 
 <div class="doc_code">
-<pre>
-...
-Emitted bytecode to 'bugpoint-reduced-simplified.bc'
-
-*** You can reproduce the problem with: opt bugpoint-reduced-simplified.bc -licm
-</pre>
+<p><tt>
+...<br>
+Emitted bytecode to 'bugpoint-reduced-simplified.bc'<br>
+<br>
+*** You can reproduce the problem with: opt bugpoint-reduced-simplified.bc -licm<br>
+</tt></p>
 </div>
 
 <p>Once you complete this, please send the LLVM bytecode file and the command
@@ -242,8 +242,8 @@
 JIT, or LLC) and optionally a series of LLVM passes to run.  For example:</p>
 
 <div class="doc_code">
-<pre>
-<b>bugpoint</b> -run-cbe [... optimization passes ...] file-to-test.bc --args -- [program arguments]</pre>
+<p><tt>
+<b>bugpoint</b> -run-cbe [... optzn passes ...] file-to-test.bc --args -- [program arguments]</tt></p>
 </div>
 
 <p><tt>bugpoint</tt> will try to narrow down your list of passes to the one pass
@@ -295,10 +295,10 @@
 will pass the program options specified in the Makefiles:</p>
 
 <div class="doc_code">
-<pre>
-cd llvm/test/../../program
+<p><tt>
+cd llvm/test/../../program<br>
 make bugpoint-jit
-</pre>
+</tt></p>
 </div>
 
 <p>At the end of a successful <tt>bugpoint</tt> run, you will be presented
@@ -314,28 +314,28 @@
 <li><p>Regenerate the shared object from the safe bytecode file:</p>
 
 <div class="doc_code">
-<pre>
-<b>llc</b> -march=c safe.bc -o safe.c
+<p><tt>
+<b>llc</b> -march=c safe.bc -o safe.c<br>
 <b>gcc</b> -shared safe.c -o safe.so
-</pre>
+</tt></p>
 </div></li>
 
 <li><p>If debugging LLC, compile test bytecode native and link with the shared
     object:</p>
 
 <div class="doc_code">
-<pre>
-<b>llc</b> test.bc -o test.s -f
-<b>gcc</b> test.s safe.so -o test.llc
+<p><tt>
+<b>llc</b> test.bc -o test.s -f<br>
+<b>gcc</b> test.s safe.so -o test.llc<br>
 ./test.llc [program options]
-</pre>
+</tt></p>
 </div></li>
     
 <li><p>If debugging the JIT, load the shared object and supply the test
     bytecode:</p>
 
 <div class="doc_code">
-<pre><b>lli</b> -load=safe.so test.bc [program options]</pre>
+<p><tt><b>lli</b> -load=safe.so test.bc [program options]</tt></p>
 </div></li>  
 
 </ol>
@@ -353,7 +353,7 @@
   <a href="mailto:sabre at nondot.org">Chris Lattner</a><br>
   <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a>
   <br>
-  Last modified: $Date: 2004/05/14 19:04:07 $
+  Last modified: $Date: 2004/05/23 21:05:39 $
 </address>
 
 </body>




From lattner at cs.uiuc.edu  Sun May 23 16:07:12 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:07:12 2004
Subject: [llvm-commits] CVS: llvm/docs/CodingStandards.html
Message-ID: <200405232105.QAA00547@zion.cs.uiuc.edu>


Changes in directory llvm/docs:

CodingStandards.html updated: 1.15 -> 1.16

---
Log message:

Update the coding standards, removign the bit about avoiding exceptions and
rtti (they shouldn't be used in the main compiler, but that's because 
optimizations aren't allowed to fail!)

Remove the big section about iterators, because people who do not know what
they are doing shouldn't be writing them.  :)


---
Diffs of the changes:  (+32 -381)

Index: llvm/docs/CodingStandards.html
diff -u llvm/docs/CodingStandards.html:1.15 llvm/docs/CodingStandards.html:1.16
--- llvm/docs/CodingStandards.html:1.15	Wed May 12 13:37:22 2004
+++ llvm/docs/CodingStandards.html	Sun May 23 16:05:07 2004
@@ -28,7 +28,6 @@
         <ol>
           <li><a href="#ci_warningerrors">Treat Compiler Warnings Like
               Errors</a></li>
-          <li><a href="#ci_cpp_features">Which C++ features can I use?</a></li>
           <li><a href="#ci_portable_code">Write Portable Code</a></li>
         </ol></li>
     </ol></li>
@@ -46,14 +45,17 @@
         <ol>
           <li><a href="#hl_assert">Assert Liberally</a></li>
           <li><a href="#hl_preincrement">Prefer Preincrement</a></li>
-          <li><a href="#hl_avoidendl">Avoid endl</a></li>
+          <li><a href="#hl_avoidendl">Avoid std::endl</a></li>
           <li><a href="#hl_exploitcpp">Exploit C++ to its Fullest</a></li>
         </ol></li>
-      <li><a href="#iterators">Writing Iterators</a></li>
     </ol></li>
   <li><a href="#seealso">See Also</a></li>
 </ol>
 
+<div class="doc_author">
+  <p>Written by <a href="mailto:sabre at nondot.org">Chris Lattner</a></p>
+</div>
+
 
 <!-- *********************************************************************** -->
 <div class="doc_section">
@@ -112,8 +114,7 @@
 comment our code more than we do, there are a few very critical places that
 documentation is very useful:</p>
 
-<ol>
-<li><h4>File Headers</h4> 
+<b>File Headers</b>
 
 <p>Every source file should have a header on it that
 describes the basic purpose of the file.  If a file does not have a header, it
@@ -123,28 +124,38 @@
 
 <pre>
 //===-- llvm/Instruction.h - Instruction class definition -------*- C++ -*-===//
+// 
+//                     The LLVM Compiler Infrastructure
+//
+// This file was developed by the LLVM research group and is distributed under
+// the University of Illinois Open Source License. See LICENSE.TXT for details.
+// 
+//===----------------------------------------------------------------------===//
 //
 // This file contains the declaration of the Instruction class, which is the
 // base class for all of the VM instructions.
 //
 //===----------------------------------------------------------------------===//
+
 </pre>
 
-<p>A few things to note about this particular format.  The "<tt>-*- C++
+<p>A few things to note about this particular format:  The "<tt>-*- C++
 -*-</tt>" string on the first line is there to tell Emacs that the source file
 is a C++ file, not a C file (Emacs assumes .h files are C files by default [Note
 that tag this is not necessary in .cpp files]).  The name of the file is also on
 the first line, along with a very short description of the purpose of the file.
 This is important when printing out code and flipping though lots of pages.</p>
 
+<p>The next section in the file is a concise note that defines the license that
+the file is released under.  This makes it perfectly clear what terms the source
+code can be distributed under.</p>
+
 <p>The main body of the description does not have to be very long in most cases.
 Here it's only two lines.  If an algorithm is being implemented or something
 tricky is going on, a reference to the paper where it is published should be
 included, as well as any notes or "gotchas" in the code to watch out for.</p>
 
-</li>
-
-<li><h4>Class overviews</h4>
+<b>Class overviews</b>
 
 <p>Classes are one fundemental part of a good object oriented design.  As such,
 a class definition should have a comment block that explains what the class is
@@ -152,9 +163,8 @@
 could figure it out, it's probably safe to leave it out.  Naming classes
 something sane goes a long ways towards avoiding writing documentation. :)</p>
 
-</li>
 
-<li><h4>Method information</h4>
+<b>Method information</b>
 
 <p>Methods defined in a class (as well as any global functions) should also be
 documented properly.  A quick note about what it does any a description of the
@@ -166,9 +176,6 @@
 <p>Good things to talk about here are what happens when something unexpected
 happens: does the method return null?  Abort?  Format your hard disk?</p>
 
-</li>
-</ol>
-
 </div>
 
 <!-- _______________________________________________________________________ -->
@@ -335,40 +342,6 @@
 
 <!-- _______________________________________________________________________ -->
 <div class="doc_subsubsection">
-  <a name="ci_cpp_features">Which C++ features can I use?</a>
-</div>
-
-<div class="doc_text">
-
-<p>Compilers are finally catching up to the C++ standard.  Most compilers
-implement most features, so you can use just about any features that you would
-like.  In the LLVM source tree, I have chosen to not use these features:</p>
-
-<ol>
-<li><p>Exceptions: Exceptions are very useful for error reporting and handling
-exceptional conditions.  I do not use them in LLVM because they do have an
-associated performance impact (by restricting restructuring of code), and parts
-of LLVM are designed for performance critical purposes.</p>
-
-<p>Just like most of the rules in this document, this isn't a hard and fast
-requirement.  Exceptions are used in the Parser, because it simplifies error
-reporting <b>significantly</b>, and the LLVM parser is not at all in the
-critical path.</p>
-</li>
-
-<li>RTTI: RTTI has a large cost in terms of executable size, and compilers are
-not yet very good at stomping out "dead" class information blocks.  Because of
-this, typeinfo and dynamic cast are not used.</li>
-</ol>
-
-<p>Other features, such as templates (without partial specialization) can be
-used freely.  The general goal is to have clear, consise, performant code... if
-a technique assists with that then use it.</p>
-
-</div>
-
-<!-- _______________________________________________________________________ -->
-<div class="doc_subsubsection">
   <a name="ci_portable_code">Write Portable Code</a>
 </div>
 
@@ -480,7 +453,7 @@
 </div>
 
 <!-- ======================================================================= -->
-<div class="doc_text">
+<div class="doc_subsection">
   <a name="micro">The Low Level Issues</a>
 </div>
 
@@ -553,18 +526,18 @@
 
 <!-- _______________________________________________________________________ -->
 <div class="doc_subsubsection">
-  <a name="hl_avoidendl">Avoid endl</a>
+  <a name="hl_avoidendl">Avoid std::endl</a>
 </div>
 
 <div class="doc_text">
 
-<p>The <tt>endl</tt> modifier, when used with iostreams outputs a newline to the
-output stream specified.  In addition to doing this, however, it also flushes
-the output stream.  In other words, these are equivalent:</p>
+<p>The <tt>std::endl</tt> modifier, when used with iostreams outputs a newline
+to the output stream specified.  In addition to doing this, however, it also
+flushes the output stream.  In other words, these are equivalent:</p>
 
 <pre>
-  cout &lt;&lt; endl;
-  cout &lt;&lt; "\n" &lt;&lt; flush;
+  std::cout &lt;&lt; std::endl;
+  std::cout &lt;&lt; "\n" &lt;&lt; std::flush;
 </pre>
 
 <p>Most of the time, you probably have no reason to flush the output stream, so
@@ -591,331 +564,6 @@
 
 </div>
 
-<!-- ======================================================================= -->
-<div class="doc_subsection">
-  <a name="iterators">Writing Iterators</a>
-</div>
-
-<div class="doc_text">
-
-<p>Here's a pretty good summary of how to write your own data structure iterators
-in a way that is compatible with the STL, and with a lot of other code out there
-(slightly edited by Chris):</p>
-
-<pre>
-From: Ross Smith &lt;ross.s at ihug.co.nz&gt;
-Newsgroups: comp.lang.c++.moderated
-Subject: Writing iterators (was: Re: Non-template functions that take iterators)
-Date: 28 Jun 2001 12:07:10 -0400
-
-Andre Majorel wrote:
-&gt; Any pointers handy on "writing STL-compatible iterators for
-&gt; dummies ?"
-
-I'll give it a try...
-
-The usual situation requiring user-defined iterators is that you have
-a type that bears some resemblance to an STL container, and you want
-to provide iterators so it can be used with STL algorithms. You need
-to ask three questions:
-
-First, is this simply a wrapper for an underlying collection of
-objects that's held somewhere as a real STL container, or is it a
-"virtual container" for which iteration is (under the hood) more
-complicated than simply incrementing some underlying iterator (or
-pointer or index or whatever)? In the former case you can frequently
-get away with making your container's iterators simply typedefs for
-those of the underlying container; your begin() function would call
-member_container.begin(), and so on.
-
-Second, do you only need read-only iterators, or do you need separate
-read-only (const) and read-write (non-const) iterators?
-
-Third, which kind of iterator (input, output, forward, bidirectional,
-or random access) is appropriate? If you're familiar with the
-properties of the iterator types (if not, visit
-<a href="http://www.sgi.com/tech/stl/">http://www.sgi.com/tech/stl/</a>), the appropriate choice should be
-obvious from the semantics of the container.
-
-I'll start with forward iterators, as the simplest case that's likely
-to come up in normal code. Input and output iterators have some odd
-properties and rarely need to be implemented in user code; I'll leave
-them out of discussion. Bidirectional and random access iterators are
-covered below.
-
-The exact behaviour of a forward iterator is spelled out in the
-Standard in terms of a set of expressions with specified behaviour,
-rather than a set of member functions, which leaves some leeway in how
-you actually implement it. Typically it looks something like this
-(I'll start with the const-iterator-only situation):
-
-  #include &lt;iterator&gt;
-
-  class container {
-    public:
-      typedef something_or_other value_type;
-      class const_iterator:
-        public std::iterator&lt;std::forward_iterator_tag, value_type&gt; {
-          friend class container;
-          public:
-            const value_type&amp; operator*() const;
-            const value_type* operator-&gt;() const;
-            const_iterator&amp; operator++();
-            const_iterator operator++(int);
-            friend bool operator==(const_iterator lhs,
-                                   const_iterator rhs);
-            friend bool operator!=(const_iterator lhs,
-                                   const_iterator rhs);
-          private:
-            //...
-        };
-      //...
-  };
-
-An iterator should always be derived from an instantiation of the
-std::iterator template. The iterator's life cycle functions
-(constructors, destructor, and assignment operator) aren't declared
-here; in most cases the compiler-generated ones are sufficient. The
-container needs to be a friend of the iterator so that the container's
-begin() and end() functions can fill in the iterator's private members
-with the appropriate values.
-
-<i>[Chris's Note: I prefer to not make my iterators friends.  Instead, two
-ctor's are provided for the iterator class: one to start at the end of the
-container, and one at the beginning.  Typically this is done by providing
-two constructors with different signatures.]</i>
-
-There are normally only three member functions that need nontrivial
-implementations; the rest are just boilerplate.
-
-  const container::value_type&amp;
-    container::const_iterator::operator*() const {
-      // find the element and return a reference to it
-    }
-
-  const container::value_type*
-    container::const_iterator::operator-&gt;() const {
-      return &amp;**this;
-    }
-
-If there's an underlying real container, operator*() can just return a
-reference to the appropriate element. If there's no actual container
-and the elements need to be generated on the fly -- what I think of as
-a "virtual container" -- things get a bit more complicated; you'll
-probably need to give the iterator a value_type member object, and
-fill it in when you need to. This might be done as part of the
-increment operator (below), or if the operation is nontrivial, you
-might choose the "lazy" approach and only generate the actual value
-when one of the dereferencing operators is called.
-
-The operator-&gt;() function is just boilerplate around a call to
-operator*().
-
-  container::const_iterator&amp;
-    container::const_iterator::operator++() {
-      // the incrementing logic goes here
-      return *this;
-    }
-
-  container::const_iterator
-    container::const_iterator::operator++(int) {
-      const_iterator old(*this);
-      ++*this;
-      return old;
-    }
-
-Again, the incrementing logic will usually be trivial if there's a
-real container involved, more complicated if you're working with a
-virtual container. In particular, watch out for what happens when you
-increment past the last valid item -- this needs to produce an
-iterator that will compare equal to container.end(), and making this
-work is often nontrivial for virtual containers.
-
-The post-increment function is just boilerplate again (and
-incidentally makes it obvious why all the experts recommend using
-pre-increment wherever possible).
-
-  bool operator==(container::const_iterator lhs,
-                  container::const_iterator rhs) {
-    // equality comparison goes here
-  }
-
-  bool operator!=(container::const_iterator lhs,
-                  container::const_iterator rhs) {
-    return !(lhs == rhs);
-  }
-
-For a real container, the equality comparison will usually just
-compare the underlying iterators (or pointers or indices or whatever).
-The semantics of comparisons for virtual container iterators are often
-tricky. Remember that iterator comparison only needs to be defined for
-iterators into the same container, so you can often simplify things by
-taking for granted that lhs and rhs both point into the same container
-object. Again, the second function is just boilerplate.
-
-It's a matter of taste whether iterator arguments are passed by value
-or reference; I've shown tham passed by value to reduce clutter, but
-if the iterator contains several data members, passing by reference
-may be better.
-
-That convers the const-iterator-only situation. When we need separate
-const and mutable iterators, one small complication is added beyond
-the simple addition of a second class.
-
-  class container {
-    public:
-      typedef something_or_other value_type;
-      class const_iterator;
-      class iterator:
-        public std::iterator&lt;std::forward_iterator_tag, value_type&gt; {
-          friend class container;
-          friend class container::const_iterator;
-          public:
-            value_type&amp; operator*() const;
-            value_type* operator-&gt;() const;
-            iterator&amp; operator++();
-            iterator operator++(int);
-            friend bool operator==(iterator lhs, iterator rhs);
-            friend bool operator!=(iterator lhs, iterator rhs);
-          private:
-            //...
-        };
-      class const_iterator:
-        public std::iterator&lt;std::forward_iterator_tag, value_type&gt; {
-          friend class container;
-          public:
-            const_iterator();
-            const_iterator(const iterator&amp; i);
-            const value_type&amp; operator*() const;
-            const value_type* operator-&gt;() const;
-            const_iterator&amp; operator++();
-            const_iterator operator++(int);
-            friend bool operator==(const_iterator lhs,
-                                   const_iterator rhs);
-            friend bool operator!=(const_iterator lhs,
-                                   const_iterator rhs);
-          private:
-            //...
-        };
-      //...
-  };
-
-There needs to be a conversion from iterator to const_iterator (so
-that mixed-type operations, such as comparison between an iterator and
-a const_iterator, will work). This is done here by giving
-const_iterator a conversion constructor from iterator (equivalently,
-we could have given iterator an operator const_iterator()), which
-requires const_iterator to be a friend of iterator, so it can copy its
-data members. (It also requires the addition of an explicit default
-constructor to const_iterator, since the existence of another
-user-defined constructor inhibits the compiler-defined one.)
-
-Bidirectional iterators add just two member functions to forward
-iterators:
-
-  class iterator:
-    public std::iterator&lt;std::bidirectional_iterator_tag, value_type&gt; {
-      public:
-        //...
-        iterator&amp; operator--();
-        iterator operator--(int);
-        //...
-    };
-
-I won't detail the implementations, they're obvious variations on
-operator++().
-
-Random access iterators add several more member and friend functions:
-
-  class iterator:
-    public std::iterator&lt;std::random_access_iterator_tag, value_type&gt; {
-      public:
-        //...
-        iterator&amp; operator+=(difference_type rhs);
-        iterator&amp; operator-=(difference_type rhs);
-        friend iterator operator+(iterator lhs, difference_type rhs);
-        friend iterator operator+(difference_type lhs, iterator rhs);
-        friend iterator operator-(iterator lhs, difference_type rhs);
-        friend difference_type operator-(iterator lhs, iterator rhs);
-        friend bool operator&lt;(iterator lhs, iterator rhs);
-        friend bool operator&gt;(iterator lhs, iterator rhs);
-        friend bool operator&lt;=(iterator lhs, iterator rhs);
-        friend bool operator&gt;=(iterator lhs, iterator rhs);
-        //...
-    };
-
-  container::iterator&amp;
-    container::iterator::operator+=(container::difference_type rhs) {
-      // add rhs to iterator position
-      return *this;
-    }
-
-  container::iterator&amp;
-    container::iterator::operator-=(container::difference_type rhs) {
-      // subtract rhs from iterator position
-      return *this;
-    }
-
-  container::iterator operator+(container::iterator lhs,
-                                container::difference_type rhs) {
-    return iterator(lhs) += rhs;
-  }
-
-  container::iterator operator+(container::difference_type lhs,
-                                container::iterator rhs) {
-    return iterator(rhs) += lhs;
-  }
-
-  container::iterator operator-(container::iterator lhs,
-                                container::difference_type rhs) {
-    return iterator(lhs) -= rhs;
-  }
-
-  container::difference_type operator-(container::iterator lhs,
-                                       container::iterator rhs) {
-    // calculate distance between iterators
-  }
-
-  bool operator&lt;(container::iterator lhs, container::iterator rhs) {
-    // perform less-than comparison
-  }
-
-  bool operator&gt;(container::iterator lhs, container::iterator rhs) {
-    return rhs &lt; lhs;
-  }
-
-  bool operator&lt;=(container::iterator lhs, container::iterator rhs) {
-    return !(rhs &lt; lhs);
-  }
-
-  bool operator&gt;=(container::iterator lhs, container::iterator rhs) {
-    return !(lhs &lt; rhs);
-  }
-
-Four of the functions (operator+=(), operator-=(), the second
-operator-(), and operator&lt;()) are nontrivial; the rest are
-boilerplate.
-
-One feature of the above code that some experts may disapprove of is
-the declaration of all the free functions as friends, when in fact
-only a few of them need direct access to the iterator's private data.
-I originally got into the habit of doing this simply to keep the
-declarations together; declaring some functions inside the class and
-some outside seemed awkward. Since then, though, I've been told that
-there's a subtle difference in the way name lookup works for functions
-declared inside a class (as friends) and outside, so keeping them
-together in the class is probably a good idea for practical as well as
-aesthetic reasons.
-
-I hope all this is some help to anyone who needs to write their own
-STL-like containers and iterators.
-
--- 
-Ross Smith &lt;ross.s at ihug.co.nz&gt; The Internet Group, Auckland, New Zealand
-</pre>
-
-</div>
-
 <!-- *********************************************************************** -->
 <div class="doc_section">
   <a name="seealso">See Also</a>
@@ -932,7 +580,9 @@
 <li><a href="http://www.aw.com/product/0,2627,0201924889,00.html">Effective
 C++</a> by Scott Meyers.  There is an online version of the book (only some
 chapters though) <a
-href="http://www.awlonline.com/cseng/meyerscddemo/">available as well</a>.</li>
+href="http://www.awlonline.com/cseng/meyerscddemo/">available as well</a>.  Also
+interesting and useful are "More Effective C++" and "Effective STL" by the same
+author.</li>
 
 <li><a href="http://cseng.aw.com/book/0,3828,0201633620,00.html">Large-Scale C++
 Software Design</a> by John Lakos</li>
@@ -955,7 +605,7 @@
 
   <a href="mailto:sabre at nondot.org">Chris Lattner</a><br>
   <a href="http://llvm.cs.uiuc.edu">LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/05/12 18:37:22 $
+  Last modified: $Date: 2004/05/23 21:05:07 $
 </address>
 
 </body>




From lattner at cs.uiuc.edu  Sun May 23 16:08:00 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:08:00 2004
Subject: [llvm-commits] CVS: llvm/docs/OpenProjects.html
Message-ID: <200405232106.QAA02418@zion.cs.uiuc.edu>


Changes in directory llvm/docs:

OpenProjects.html updated: 1.27 -> 1.28

---
Log message:

Remove implemented stuff


---
Diffs of the changes:  (+6 -6)

Index: llvm/docs/OpenProjects.html
diff -u llvm/docs/OpenProjects.html:1.27 llvm/docs/OpenProjects.html:1.28
--- llvm/docs/OpenProjects.html:1.27	Sun May 23 15:29:35 2004
+++ llvm/docs/OpenProjects.html	Sun May 23 16:06:21 2004
@@ -32,6 +32,11 @@
   </ol></li>
 </ul>
 
+<div class="doc_author">
+  <p>Written by the <a href="http://llvm.cs.uiuc.edu/">LLVM Team</a></p>
+</div>
+
+
 <!-- *********************************************************************** -->
 <div class="doc_section">
   <a name="what">What is this?</a>
@@ -209,9 +214,6 @@
 <li>Implement a flow-sensitive context-insensitive alias analysis algorithm<br>
   - Just an efficient local algorithm perhaps?</li>
 
-<li>Implement an interface to update analyses in response to common code motion
-    transformations</li>
-
 <li>Implement alias-analysis-based optimizations:
   <ul>
   <li>Dead store elimination</li>
@@ -275,8 +277,6 @@
      - Design some way to represent and query dep analysis</li>
 <li>Implement a strength reduction pass</li>
 <li>Value range propagation pass</li>
-<li>Implement an unswitching pass</li>
-<li>Write a loop unroller, with a simple heuristic for when to unroll</li>
 </ol>
 
 </div>
@@ -333,7 +333,7 @@
 
   <a href="mailto:sabre at nondot.org">Chris Lattner</a><br>
   <a href="http://llvm.cs.uiuc.edu">LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/05/23 20:29:35 $
+  Last modified: $Date: 2004/05/23 21:06:21 $
 </address>
 
 </body>




From lattner at cs.uiuc.edu  Sun May 23 16:08:09 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:08:09 2004
Subject: [llvm-commits] CVS: llvm/docs/LangRef.html
Message-ID: <200405232106.QAA01817@zion.cs.uiuc.edu>


Changes in directory llvm/docs:

LangRef.html updated: 1.60 -> 1.61

---
Log message:

Describe the new garbage collector intrinsics


---
Diffs of the changes:  (+177 -14)

Index: llvm/docs/LangRef.html
diff -u llvm/docs/LangRef.html:1.60 llvm/docs/LangRef.html:1.61
--- llvm/docs/LangRef.html:1.60	Mon Apr 12 11:33:19 2004
+++ llvm/docs/LangRef.html	Sun May 23 16:06:01 2004
@@ -5,7 +5,9 @@
   <title>LLVM Assembly Language Reference Manual</title>
   <link rel="stylesheet" href="llvm.css" type="text/css">
 </head>
+
 <body>
+
 <div class="doc_title"> LLVM Language Reference Manual </div>
 <ol>
   <li><a href="#abstract">Abstract</a></li>
@@ -97,6 +99,13 @@
           <li><a href="#i_va_copy">'<tt>llvm.va_copy</tt>'  Intrinsic</a></li>
         </ol>
       </li>
+      <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
+        <ol>
+          <li><a href="#i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
+          <li><a href="#i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
+          <li><a href="#i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
+        </ol>
+      </li>
       <li><a href="#int_codegen">Code Generator Intrinsics</a>
         <ol>
           <li><a href="#i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
@@ -117,18 +126,20 @@
           <li><a href="#i_memset">'<tt>llvm.memset</tt>' Intrinsic</a></li>
         </ol>
       </li>
-      <li><a href="#int_debugger">Debugger intrinsics</a>
+      <li><a href="#int_debugger">Debugger intrinsics</a></li>
     </ol>
   </li>
 </ol>
-<div class="doc_text">
-<p><b>Written by <a href="mailto:sabre at nondot.org">Chris Lattner</a>
-and <a href="mailto:vadve at cs.uiuc.edu">Vikram Adve</a></b></p>
-<p> </p>
+
+<div class="doc_author">
+  <p>Written by <a href="mailto:sabre at nondot.org">Chris Lattner</a>
+            and <a href="mailto:vadve at cs.uiuc.edu">Vikram Adve</a></p>
 </div>
+
 <!-- *********************************************************************** -->
 <div class="doc_section"> <a name="abstract">Abstract </a></div>
 <!-- *********************************************************************** -->
+
 <div class="doc_text">
 <p>This document is a reference manual for the LLVM assembly language. 
 LLVM is an SSA based representation that provides type safety,
@@ -137,10 +148,13 @@
 representation used throughout all phases of the LLVM compilation
 strategy.</p>
 </div>
+
 <!-- *********************************************************************** -->
 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
 <!-- *********************************************************************** -->
+
 <div class="doc_text">
+
 <p>The LLVM code representation is designed to be used in three
 different forms: as an in-memory compiler IR, as an on-disk bytecode
 representation (suitable for fast loading by a Just-In-Time compiler),
@@ -150,6 +164,7 @@
 to debug and visualize the transformations.  The three different forms
 of LLVM are all equivalent.  This document describes the human readable
 representation and notation.</p>
+
 <p>The LLVM representation aims to be a light-weight and low-level
 while being expressive, typed, and extensible at the same time.  It
 aims to be a "universal IR" of sorts, by being at a low enough level
@@ -160,15 +175,23 @@
 can be proven that a C automatic variable is never accessed outside of
 the current function... allowing it to be promoted to a simple SSA
 value instead of a memory location.</p>
+
 </div>
+
 <!-- _______________________________________________________________________ -->
 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
+
 <div class="doc_text">
+
 <p>It is important to note that this document describes 'well formed'
 LLVM assembly language.  There is a difference between what the parser
 accepts and what is considered 'well formed'.  For example, the
 following instruction is syntactically okay, but not well formed:</p>
-<pre>  %x = <a href="#i_add">add</a> int 1, %x<br></pre>
+
+<pre>
+  %x = <a href="#i_add">add</a> int 1, %x
+</pre>
+
 <p>...because the definition of <tt>%x</tt> does not dominate all of
 its uses. The LLVM infrastructure provides a verification pass that may
 be used to verify that an LLVM module is well formed.  This pass is
@@ -176,13 +199,18 @@
 the optimizer before it outputs bytecode.  The violations pointed out
 by the verifier pass indicate bugs in transformation passes or input to
 the parser.</p>
+
 <!-- Describe the typesetting conventions here. --> </div>
+
 <!-- *********************************************************************** -->
 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
 <!-- *********************************************************************** -->
+
 <div class="doc_text">
+
 <p>LLVM uses three different forms of identifiers, for different
 purposes:</p>
+
 <ol>
   <li>Numeric constants are represented as you would expect: 12, -3
 123.421,   etc.  Floating point constants have an optional hexadecimal
@@ -1803,18 +1831,22 @@
 </div>
 
 <div class="doc_text">
+
 <p>Variable argument support is defined in LLVM with the <a
  href="#i_vanext"><tt>vanext</tt></a> instruction and these three
 intrinsic functions.  These functions are related to the similarly
 named macros defined in the <tt>&lt;stdarg.h&gt;</tt> header file.</p>
+
 <p>All of these functions operate on arguments that use a
 target-specific value type "<tt>va_list</tt>".  The LLVM assembly
 language reference manual does not define what this type is, so all
 transformations should be prepared to handle intrinsics with any type
 used.</p>
+
 <p>This example shows how the <a href="#i_vanext"><tt>vanext</tt></a>
 instruction and the variable argument handling intrinsic functions are
 used.</p>
+
 <pre>
 int %test(int %X, ...) {
   ; Initialize variable argument processing
@@ -1888,21 +1920,152 @@
 </div>
 
 <div class="doc_text">
+
 <h5>Syntax:</h5>
-<pre>  call va_list (va_list)* %llvm.va_copy(va_list &lt;destarglist&gt;)<br></pre>
+
+<pre>
+  call va_list (va_list)* %llvm.va_copy(va_list &lt;destarglist&gt;)
+</pre>
+
 <h5>Overview:</h5>
-<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument
-position from the source argument list to the destination argument list.</p>
+
+<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
+from the source argument list to the destination argument list.</p>
+
 <h5>Arguments:</h5>
+
 <p>The argument is the <tt>va_list</tt> to copy.</p>
+
 <h5>Semantics:</h5>
+
 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
-macro available in C.  In a target-dependent way, it copies the source <tt>va_list</tt>
-element into the returned list.  This intrinsic is necessary because the <tt><a
- href="i_va_start">llvm.va_start</a></tt> intrinsic may be arbitrarily
-complex and require memory allocation, for example.</p>
+macro available in C.  In a target-dependent way, it copies the source
+<tt>va_list</tt> element into the returned list.  This intrinsic is necessary
+because the <tt><a href="i_va_start">llvm.va_start</a></tt> intrinsic may be
+arbitrarily complex and require memory allocation, for example.</p>
+
+</div>
+
+<!-- ======================================================================= -->
+<div class="doc_subsection">
+  <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
 </div>
 
+<div class="doc_text">
+
+<p>
+LLVM support for <a href="GarbageCollection.html">Accurate Garbage
+Collection</a> requires the implementation and generation of these intrinsics.
+These intrinsics allow identification of <a href="#i_gcroot">GC roots on the
+stack</a>, as well as garbage collector implementations that require <a
+href="#i_gcread">read</a> and <a href="#i_gcwrite">write</a> barriers.
+Front-ends for type-safe garbage collected languages should generate these
+intrinsics to make use of the LLVM garbage collectors.  For more details, see <a
+href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
+</p>
+</div>
+
+<!-- _______________________________________________________________________ -->
+<div class="doc_subsubsection">
+  <a name="i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
+</div>
+
+<div class="doc_text">
+
+<h5>Syntax:</h5>
+
+<pre>
+  call void (&lt;ty&gt;**, &lt;ty2&gt;*)* %llvm.gcroot(&lt;ty&gt;** %ptrloc, &lt;ty2&gt;* %metadata)
+</pre>
+
+<h5>Overview:</h5>
+
+<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existance of a GC root to
+the code generator, and allows some metadata to be associated with it.</p>
+
+<h5>Arguments:</h5>
+
+<p>The first argument specifies the address of a stack object that contains the
+root pointer.  The second pointer (which must be either a constant or a global
+value address) contains the meta-data to be associated with the root.</p>
+
+<h5>Semantics:</h5>
+
+<p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
+location.  At compile-time, the code generator generates information to allow
+the runtime to find the pointer at GC safe points.
+</p>
+
+</div>
+
+
+<!-- _______________________________________________________________________ -->
+<div class="doc_subsubsection">
+  <a name="i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
+</div>
+
+<div class="doc_text">
+
+<h5>Syntax:</h5>
+
+<pre>
+  call sbyte* (sbyte**)* %llvm.gcread(sbyte** %Ptr)
+</pre>
+
+<h5>Overview:</h5>
+
+<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
+locations, allowing garbage collector implementations that require read
+barriers.</p>
+
+<h5>Arguments:</h5>
+
+<p>The argument is the address to read from, which should be an address
+allocated from the garbage collector.</p>
+
+<h5>Semantics:</h5>
+
+<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
+instruction, but may be replaced with substantially more complex code by the
+garbage collector runtime, as needed.</p>
+
+</div>
+
+
+<!-- _______________________________________________________________________ -->
+<div class="doc_subsubsection">
+  <a name="i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
+</div>
+
+<div class="doc_text">
+
+<h5>Syntax:</h5>
+
+<pre>
+  call void (sbyte*, sbyte**)* %llvm.gcwrite(sbyte* %P1, sbyte** %P2)
+</pre>
+
+<h5>Overview:</h5>
+
+<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
+locations, allowing garbage collector implementations that require write
+barriers (such as generational or reference counting collectors).</p>
+
+<h5>Arguments:</h5>
+
+<p>The first argument is the reference to store, and the second is the heap
+location to store to.</p>
+
+<h5>Semantics:</h5>
+
+<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
+instruction, but may be replaced with substantially more complex code by the
+garbage collector runtime, as needed.</p>
+
+</div>
+
+
+
 <!-- ======================================================================= -->
 <div class="doc_subsection">
   <a name="int_codegen">Code Generator Intrinsics</a>
@@ -2390,7 +2553,7 @@
 
   <a href="mailto:sabre at nondot.org">Chris Lattner</a><br>
   <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/04/12 16:33:19 $
+  Last modified: $Date: 2004/05/23 21:06:01 $
 </address>
 </body>
 </html>




From lattner at cs.uiuc.edu  Sun May 23 16:09:01 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:09:01 2004
Subject: [llvm-commits] CVS: llvm/docs/ProgrammersManual.html
Message-ID: <200405232107.QAA03862@zion.cs.uiuc.edu>


Changes in directory llvm/docs:

ProgrammersManual.html updated: 1.55 -> 1.56

---
Log message:

Lots of minor cleanups and clarifications


---
Diffs of the changes:  (+49 -78)

Index: llvm/docs/ProgrammersManual.html
diff -u llvm/docs/ProgrammersManual.html:1.55 llvm/docs/ProgrammersManual.html:1.56
--- llvm/docs/ProgrammersManual.html:1.55	Thu Mar 18 08:58:55 2004
+++ llvm/docs/ProgrammersManual.html	Sun May 23 16:06:58 2004
@@ -98,14 +98,6 @@
             </ul></li>
           <li><a href="#Module">The <tt>Module</tt> class</a></li>
           <li><a href="#Constant">The <tt>Constant</tt> class</a>
-            <ul>
-                  <li> <br>
-                  </li>
-                  <li> <br>
-                  </li>
-                </ul>
-              </li>
-            </ul>
           </li>
           <li><a href="#Type">The <tt>Type</tt> class</a> </li>
           <li><a href="#Argument">The <tt>Argument</tt> class</a> </li>
@@ -117,15 +109,15 @@
           <li>Creating, inserting, moving and deleting from LLVM lists </li>
         </ul>
       </li>
-      <li>Important iterator invalidation semantics to be aware of </li>
+      <li>Important iterator invalidation semantics to be aware of.</li>
     </ul>
   </li>
 </ol>
 
-<div class="doc_text">    
-  <p><b>Written by <a href="mailto:sabre at nondot.org">Chris Lattner</a>, 
-  <a href="mailto:dhurjati at cs.uiuc.edu">Dinakar Dhurjati</a>, and <a
-  href="mailto:jstanley at cs.uiuc.edu">Joel Stanley</a></b></p>
+<div class="doc_author">    
+  <p>Written by <a href="mailto:sabre at nondot.org">Chris Lattner</a>, 
+                <a href="mailto:dhurjati at cs.uiuc.edu">Dinakar Dhurjati</a>, and
+                <a href="mailto:jstanley at cs.uiuc.edu">Joel Stanley</a></p>
 </div>
 
 <!-- *********************************************************************** -->
@@ -283,13 +275,14 @@
   that something is of the right type.  An example of the <tt>isa&lt;&gt;</tt>
   and <tt>cast&lt;&gt;</tt> template is:
 
-  <pre>static bool isLoopInvariant(const <a href="#Value">Value</a> *V, const
-  Loop *L) {<br> if (isa&lt;<a href="#Constant">Constant</a>&gt;(V) || isa&lt;<a
-  href="#Argument">Argument</a>&gt;(V) || isa&lt;<a
-  href="#GlobalValue">GlobalValue</a>&gt;(V))<br> return true;<br><br> <i>//
-  Otherwise, it must be an instruction...</i><br> return
-  !L-&gt;contains(cast&lt;<a
-  href="#Instruction">Instruction</a>&gt;(V)-&gt;getParent());<br></pre>
+  <pre>
+  static bool isLoopInvariant(const <a href="#Value">Value</a> *V, const Loop *L) {
+    if (isa&lt;<a href="#Constant">Constant</a>&gt;(V) || isa&lt;<a href="#Argument">Argument</a>&gt;(V) || isa&lt;<a href="#GlobalValue">GlobalValue</a>&gt;(V))
+      return true;
+
+  <i>// Otherwise, it must be an instruction...</i>
+  return !L-&gt;contains(cast&lt;<a href="#Instruction">Instruction</a>&gt;(V)-&gt;getParent());
+  </pre>
 
   <p>Note that you should <b>not</b> use an <tt>isa&lt;&gt;</tt> test followed
   by a <tt>cast&lt;&gt;</tt>, for that use the <tt>dyn_cast&lt;&gt;</tt>
@@ -307,8 +300,11 @@
   same circumstances.  Typically, the <tt>dyn_cast&lt;&gt;</tt> operator is used
   in an <tt>if</tt> statement or some other flow control statement like this:
 
-   <pre> if (<a href="#AllocationInst">AllocationInst</a> *AI = dyn_cast&lt;<a
-   href="#AllocationInst">AllocationInst</a>&gt;(Val)) {<br> ...<br> }<br></pre>
+   <pre>
+     if (<a href="#AllocationInst">AllocationInst</a> *AI = dyn_cast&lt;<a href="#AllocationInst">AllocationInst</a>&gt;(Val)) {
+       ...
+     }
+   </pre>
    
    <p> This form of the <tt>if</tt> statement effectively combines together a
    call to <tt>isa&lt;&gt;</tt> and a call to <tt>cast&lt;&gt;</tt> into one
@@ -316,10 +312,12 @@
 
    <p> Another common example is:</p>
 
-   <pre> <i>// Loop over all of the phi nodes in a basic block</i><br>
-   BasicBlock::iterator BBI = BB-&gt;begin();<br> for (; <a
-   href="#PhiNode">PHINode</a> *PN = dyn_cast&lt;<a
-   href="#PHINode">PHINode</a>&gt;(BBI); ++BBI)<br> cerr &lt;&lt; *PN;<br></pre>
+   <pre>
+     <i>// Loop over all of the phi nodes in a basic block</i>
+     BasicBlock::iterator BBI = BB-&gt;begin();
+     for (; <a href="#PhiNode">PHINode</a> *PN = dyn_cast&lt;<a href="#PHINode">PHINode</a>&gt;(BBI); ++BBI)
+       std::cerr &lt;&lt; *PN;
+   </pre>
 
    <p>Note that the <tt>dyn_cast&lt;&gt;</tt> operator, like C++'s
    <tt>dynamic_cast</tt> or Java's <tt>instanceof</tt> operator, can be abused.
@@ -525,7 +523,7 @@
 </div>
 
 <!-- _______________________________________________________________________ -->
-<div class="subsubsection">
+<div class="doc_subsubsection">
   <a name="iterate_function">Iterating over the </a><a
   href="#BasicBlock"><tt>BasicBlock</tt></a>s in a <a
   href="#Function"><tt>Function</tt></a>
@@ -551,7 +549,7 @@
 </div>
 
 <!-- _______________________________________________________________________ -->
-<div class="subsubsection">
+<div class="doc_subsubsection">
   <a name="iterate_basicblock">Iterating over the </a><a
   href="#Instruction"><tt>Instruction</tt></a>s in a <a
   href="#BasicBlock"><tt>BasicBlock</tt></a>
@@ -580,7 +578,7 @@
 </div>
 
 <!-- _______________________________________________________________________ -->
-<div class="subsubsection">
+<div class="doc_subsubsection">
   <a name="iterate_institer">Iterating over the </a><a
   href="#Instruction"><tt>Instruction</tt></a>s in a <a
   href="#Function"><tt>Function</tt></a>
@@ -593,11 +591,9 @@
 <tt>InstIterator</tt> should be used instead. You'll need to include <a
 href="/doxygen/InstIterator_8h-source.html"><tt>llvm/Support/InstIterator.h</tt></a>,
 and then instantiate <tt>InstIterator</tt>s explicitly in your code.  Here's a
-small example that shows how to dump all instructions in a function to stderr
-(<b>Note:</b> Dereferencing an <tt>InstIterator</tt> yields an
-<tt>Instruction*</tt>, <i>not</i> an <tt>Instruction&amp;</tt>!):</p>
+small example that shows how to dump all instructions in a function to the standard error stream:<p>
 
-  <pre>#include "<a href="/doxygen/InstIterator_8h-source.html">llvm/Support/InstIterator.h</a>"<br>...<br>// Suppose F is a ptr to a function<br>for (inst_iterator i = inst_begin(F), e = inst_end(F); i != e; ++i)<br>  cerr &lt;&lt; **i &lt;&lt; "\n";<br></pre>
+  <pre>#include "<a href="/doxygen/InstIterator_8h-source.html">llvm/Support/InstIterator.h</a>"<br>...<br>// Suppose F is a ptr to a function<br>for (inst_iterator i = inst_begin(F), e = inst_end(F); i != e; ++i)<br>  cerr &lt;&lt; *i &lt;&lt; "\n";<br></pre>
 Easy, isn't it?  You can also use <tt>InstIterator</tt>s to fill a
 worklist with its initial contents.  For example, if you wanted to
 initialize a worklist to contain all instructions in a <tt>Function</tt>
@@ -619,7 +615,7 @@
 
 <p>Sometimes, it'll be useful to grab a reference (or pointer) to a class
 instance when all you've got at hand is an iterator.  Well, extracting
-a reference or a pointer from an iterator is very straightforward.
+a reference or a pointer from an iterator is very straight-forward.
 Assuming that <tt>i</tt> is a <tt>BasicBlock::iterator</tt> and <tt>j</tt>
 is a <tt>BasicBlock::const_iterator</tt>:</p>
 
@@ -639,17 +635,14 @@
 
   <pre>Instruction* pinst = i;</pre>
 
-<p>It's also possible to turn a class pointer into the corresponding iterator.
-Usually, this conversion is quite inexpensive.  The following code snippet
-illustrates use of the conversion constructors provided by LLVM iterators.  By
-using these, you can explicitly grab the iterator of something without actually
-obtaining it via iteration over some structure:</p>
+<p>It's also possible to turn a class pointer into the corresponding iterator,
+and this is a constant time operation (very efficient).  The following code
+snippet illustrates use of the conversion constructors provided by LLVM
+iterators.  By using these, you can explicitly grab the iterator of something
+without actually obtaining it via iteration over some structure:</p>
 
   <pre>void printNextInstruction(Instruction* inst) {<br>    BasicBlock::iterator it(inst);<br>    ++it; // after this line, it refers to the instruction after *inst.<br>    if (it != inst-&gt;getParent()-&gt;end()) cerr &lt;&lt; *it &lt;&lt; "\n";<br>}<br></pre>
 
-<p>Of course, this example is strictly pedagogical, because it'd be much
-better to explicitly grab the next instruction directly from inst.</p>
-
 </div>
 
 <!--_______________________________________________________________________-->
@@ -664,7 +657,7 @@
 locations in the entire module (that is, across every <tt>Function</tt>) where a
 certain function (i.e., some <tt>Function</tt>*) is already in scope.  As you'll
 learn later, you may want to use an <tt>InstVisitor</tt> to accomplish this in a
-much more straightforward manner, but this example will allow us to explore how
+much more straight-forward manner, but this example will allow us to explore how
 you'd do it if you didn't have <tt>InstVisitor</tt> around. In pseudocode, this
 is what we want to do:</p>
 
@@ -694,17 +687,16 @@
 most-specific common base class is <tt>Instruction</tt>, which includes lots of
 less closely-related things. For these cases, LLVM provides a handy wrapper
 class called <a
-href="http://llvm.cs.uiuc.edu/doxygen/classCallSite.html"><tt>CallSite
-</tt></a>.  It is essentially a wrapper around an <tt>Instruction</tt> pointer,
-with some methods that provide functionality common to <tt>CallInst</tt>s and
+href="http://llvm.cs.uiuc.edu/doxygen/classCallSite.html"><tt>CallSite</tt></a>.
+It is essentially a wrapper around an <tt>Instruction</tt> pointer, with some
+methods that provide functionality common to <tt>CallInst</tt>s and
 <tt>InvokeInst</tt>s.</p>
 
-<p>This class is supposed to have "value semantics". So it should be passed by
-value, not by reference; it should not be dynamically allocated or deallocated
-using <tt>operator new</tt> or <tt>operator delete</tt>. It is efficiently
-copyable, assignable and constructable, with costs equivalents to that of a bare
-pointer. (You will notice, if you look at its definition, that it has only a
-single data member.)</p>
+<p>This class has "value semantics": it should be passed by value, not by
+reference and it should not be dynamically allocated or deallocated using
+<tt>operator new</tt> or <tt>operator delete</tt>. It is efficiently copyable,
+assignable and constructable, with costs equivalents to that of a bare pointer.
+If you look at its definition, it has only a single pointer member.</p>
 
 </div>
 
@@ -767,7 +759,7 @@
 
 <p><i>Instantiating Instructions</i></p>
 
-<p>Creation of <tt>Instruction</tt>s is straightforward: simply call the
+<p>Creation of <tt>Instruction</tt>s is straight-forward: simply call the
 constructor for the kind of instruction to instantiate and provide the necessary
 parameters. For example, an <tt>AllocaInst</tt> only <i>requires</i> a
 (const-ptr-to) <tt>Type</tt>. Thus:</p> 
@@ -851,7 +843,7 @@
 <div class="doc_text">
 
 <p>Deleting an instruction from an existing sequence of instructions that form a
-<a href="#BasicBlock"><tt>BasicBlock</tt></a> is very straightforward. First,
+<a href="#BasicBlock"><tt>BasicBlock</tt></a> is very straight-forward. First,
 you must have a pointer to the instruction that you wish to delete.  Second, you
 need to obtain the pointer to that instruction's basic block. You use the
 pointer to the basic block to get its list of instructions and then use the
@@ -1726,9 +1718,6 @@
 
 <ul>
 
-  <li><tt>PrimitiveID getPrimitiveID() const</tt>: Returns the base type of the
-  type.</li>
-
   <li><tt>bool isSigned() const</tt>: Returns whether an integral numeric type
   is signed. This is true for SByteTy, ShortTy, IntTy, LongTy. Note that this is
   not true for Float and Double. </li>
@@ -1747,29 +1736,11 @@
   <li><tt>bool isFloatingPoint()</tt>: Return true if this is one of the two
   floating point types.</li>
 
-  <li><tt>bool isRecursive() const</tt>: Returns rue if the type graph contains
-  a cycle.</li>
-
   <li><tt>isLosslesslyConvertableTo (const Type *Ty) const</tt>: Return true if
   this type can be converted to 'Ty' without any reinterpretation of bits. For
-  example, uint to int.</li>
+  example, uint to int or one pointer type to another.</li>
 
-  <li><tt>bool isPrimitiveType() const</tt>: Returns true if it is a primitive
-  type.</li>
-
-  <li><tt>bool isDerivedType() const</tt>: Returns true if it is a derived
-  type.</li>
-
-  <li><tt>const Type * getContainedType (unsigned i) const</tt>: This method is
-  used to implement the type iterator. For derived types, this returns the types
-  'contained' in the derived type, returning 0 when 'i' becomes invalid. This
-  allows the user to iterate over the types in a struct, for example, really
-  easily.</li>
-
-  <li><tt>unsigned getNumContainedTypes() const</tt>: Return the number of types
-  in the derived type.
-
-    <hr>
+<br>
     <p>Derived Types</p>
 
     <ul>
@@ -1830,7 +1801,7 @@
   <a href="mailto:dhurjati at cs.uiuc.edu">Dinakar Dhurjati</a> and
   <a href="mailto:sabre at nondot.org">Chris Lattner</a><br>
   <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/03/18 14:58:55 $
+  Last modified: $Date: 2004/05/23 21:06:58 $
 </address>
 
 </body>




From lattner at cs.uiuc.edu  Sun May 23 16:09:11 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:09:11 2004
Subject: [llvm-commits] CVS: llvm/docs/CFEBuildInstrs.html CommandLine.html ExtendingLLVM.html FAQ.html GettingStarted.html LLVMVsTheWorld.html ObjectFiles.html Projects.html ReleaseNotes.html SourceLevelDebugging.html Stacker.html TableGenFundamentals.html TestingGuide.html WritingAnLLVMPass.html
Message-ID: <200405232107.QAA03857@zion.cs.uiuc.edu>


Changes in directory llvm/docs:

CFEBuildInstrs.html updated: 1.15 -> 1.16
CommandLine.html updated: 1.21 -> 1.22
ExtendingLLVM.html updated: 1.5 -> 1.6
FAQ.html updated: 1.21 -> 1.22
GettingStarted.html updated: 1.55 -> 1.56
LLVMVsTheWorld.html updated: 1.7 -> 1.8
ObjectFiles.html updated: 1.2 -> 1.3
Projects.html updated: 1.10 -> 1.11
ReleaseNotes.html updated: 1.182 -> 1.183
SourceLevelDebugging.html updated: 1.4 -> 1.5
Stacker.html updated: 1.13 -> 1.14
TableGenFundamentals.html updated: 1.5 -> 1.6
TestingGuide.html updated: 1.7 -> 1.8
WritingAnLLVMPass.html updated: 1.26 -> 1.27

---
Log message:

Make use of the new doc_author class.  "Seperate content from presentation" they said.


---
Diffs of the changes:  (+63 -43)

Index: llvm/docs/CFEBuildInstrs.html
diff -u llvm/docs/CFEBuildInstrs.html:1.15 llvm/docs/CFEBuildInstrs.html:1.16
--- llvm/docs/CFEBuildInstrs.html:1.15	Wed May 12 13:49:47 2004
+++ llvm/docs/CFEBuildInstrs.html	Sun May 23 16:07:26 2004
@@ -18,8 +18,8 @@
   <li><a href="#license">License Information</a>
 </ol>
 
-<div class="doc_text">    
-  <p><b>Written by Brian R. Gaeke</b></p>
+<div class="doc_author">    
+  <p>Written by Brian R. Gaeke</p>
 </div>
 
 <!-- *********************************************************************** -->
@@ -274,7 +274,7 @@
 
   Brian Gaeke<br>
   <a href="http://llvm.cs.uiuc.edu">LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/05/12 18:49:47 $
+  Last modified: $Date: 2004/05/23 21:07:26 $
 </address>
 
 </body>


Index: llvm/docs/CommandLine.html
diff -u llvm/docs/CommandLine.html:1.21 llvm/docs/CommandLine.html:1.22
--- llvm/docs/CommandLine.html:1.21	Wed May 12 13:42:35 2004
+++ llvm/docs/CommandLine.html	Sun May 23 16:07:26 2004
@@ -84,8 +84,8 @@
     </ol></li>
 </ol>
 
-<div class="doc_text">
-  <p><b>Written by <a href="mailto:sabre at nondot.org">Chris Lattner</a></b></p>
+<div class="doc_author">
+  <p>Written by <a href="mailto:sabre at nondot.org">Chris Lattner</a></p>
 </div>
 
 <!-- *********************************************************************** -->
@@ -1707,7 +1707,7 @@
 
   <a href="mailto:sabre at nondot.org">Chris Lattner</a><br>
   <a href="http://llvm.cs.uiuc.edu">LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/05/12 18:42:35 $
+  Last modified: $Date: 2004/05/23 21:07:26 $
 </address>
 
 </body>


Index: llvm/docs/ExtendingLLVM.html
diff -u llvm/docs/ExtendingLLVM.html:1.5 llvm/docs/ExtendingLLVM.html:1.6
--- llvm/docs/ExtendingLLVM.html:1.5	Tue Apr 13 14:48:55 2004
+++ llvm/docs/ExtendingLLVM.html	Sun May 23 16:07:26 2004
@@ -23,8 +23,8 @@
   </ol></li>
 </ol>
 
-<div class="doc_text">    
-  <p><b>Written by <a href="http://misha.brukman.net">Misha Brukman</a></b></p>
+<div class="doc_author">    
+  <p>Written by <a href="http://misha.brukman.net">Misha Brukman</a></p>
 </div>
 
 <!-- *********************************************************************** -->
@@ -231,7 +231,7 @@
   <a href="http://misha.brukman.net">Misha Brukman</a><br>
   <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a>
   <br>
-  Last modified: $Date: 2004/04/13 19:48:55 $
+  Last modified: $Date: 2004/05/23 21:07:26 $
 </address>
 
 </body>


Index: llvm/docs/FAQ.html
diff -u llvm/docs/FAQ.html:1.21 llvm/docs/FAQ.html:1.22
--- llvm/docs/FAQ.html:1.21	Mon Mar 29 14:18:09 2004
+++ llvm/docs/FAQ.html	Sun May 23 16:07:26 2004
@@ -63,7 +63,7 @@
 
     <li>
     When I compile code using the LLVM GCC front end, it complains that it
-    cannot find crtend.o.
+    cannot find libcrtend.a.
     </li>
   </ol>
   </li>
@@ -80,6 +80,11 @@
   </li>
 </ol>
 
+<div class="doc_author">
+  <p>Written by <a href="http://llvm.cs.uiuc.edu">The LLVM Team</a></p>
+</div>
+
+
 <!-- *********************************************************************** -->
 <div class="doc_section">
   <a name="license">License</a>
@@ -511,10 +516,6 @@
 </p>
 </div>
 
-
-     <li>
-
-
 <!-- *********************************************************************** -->
 <!-- *********************************************************************** -->
 
@@ -522,7 +523,7 @@
 <div class="doc_footer">
   <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a>
   <br>
-  Last modified: $Date: 2004/03/29 20:18:09 $
+  Last modified: $Date: 2004/05/23 21:07:26 $
 </div>
 
 </body>


Index: llvm/docs/GettingStarted.html
diff -u llvm/docs/GettingStarted.html:1.55 llvm/docs/GettingStarted.html:1.56
--- llvm/docs/GettingStarted.html:1.55	Wed May 12 14:25:59 2004
+++ llvm/docs/GettingStarted.html	Sun May 23 16:07:26 2004
@@ -49,12 +49,15 @@
   <li><a href="#links">Links</a>
 </ul>
 
-<p>By: 
-  <a href="mailto:criswell at uiuc.edu">John Criswell</a>, 
-  <a href="mailto:sabre at nondot.org">Chris Lattner</a>,
-  <a href="http://misha.brukman.net">Misha Brukman</a>, 
-  <a href="http://www.cs.uiuc.edu/~vadve">Vikram Adve</a>, and
-  <a href="mailto:gshi1 at uiuc.edu">Guochun Shi</a>.</p>
+<div class="doc_author">
+  <p>Written by: 
+    <a href="mailto:criswell at uiuc.edu">John Criswell</a>, 
+    <a href="mailto:sabre at nondot.org">Chris Lattner</a>,
+    <a href="http://misha.brukman.net">Misha Brukman</a>, 
+    <a href="http://www.cs.uiuc.edu/~vadve">Vikram Adve</a>, and
+    <a href="mailto:gshi1 at uiuc.edu">Guochun Shi</a>.
+  </p>
+</div>
 
 
 <!-- *********************************************************************** -->
@@ -1204,7 +1207,7 @@
 
   <a href="mailto:sabre at nondot.org">Chris Lattner</a><br>
   <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/05/12 19:25:59 $
+  Last modified: $Date: 2004/05/23 21:07:26 $
 </address>
 
 </body>


Index: llvm/docs/LLVMVsTheWorld.html
diff -u llvm/docs/LLVMVsTheWorld.html:1.7 llvm/docs/LLVMVsTheWorld.html:1.8
--- llvm/docs/LLVMVsTheWorld.html:1.7	Tue Mar  9 01:20:26 2004
+++ llvm/docs/LLVMVsTheWorld.html	Sun May 23 16:07:26 2004
@@ -22,8 +22,8 @@
   <li><a href="#codeemission">Machine Code Emission</a></li>
 </ol>
 
-<div class="doc_text">    
-  <p><b>Written by Brian R. Gaeke</b></p>
+<div class="doc_author">
+  <p>Written by Brian R. Gaeke</p>
 </div>
 
 <!-- *********************************************************************** -->
@@ -173,7 +173,7 @@
   <address>Brian R. Gaeke</address>
   <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a>
   <br>
-  Last modified: $Date: 2004/03/09 07:20:26 $
+  Last modified: $Date: 2004/05/23 21:07:26 $
 </div>
 
 </body>


Index: llvm/docs/ObjectFiles.html
diff -u llvm/docs/ObjectFiles.html:1.2 llvm/docs/ObjectFiles.html:1.3
--- llvm/docs/ObjectFiles.html:1.2	Fri Feb 27 10:17:20 2004
+++ llvm/docs/ObjectFiles.html	Sun May 23 16:07:26 2004
@@ -18,11 +18,13 @@
 		</ol>
 	</li>
 </ol>
-<div class="doc_text">
-<p><b>Written by <a href="mailto:rspencer at x10sys.com">Reid Spencer</a> </b></p>
-<p> </p>
+
+<div class="doc_author">
+  <p>Written by <a href="mailto:rspencer at x10sys.com">Reid Spencer</a></p>
 </div>
+
 <hr>
+
 <!-- ======================================================================= -->
 <div class="doc_section"><a name="abstract">Abstract</a></div>
 <div class="doc_text">
@@ -291,7 +293,7 @@
 <div class="doc_footer">
 <address><a href="mailto:rspencer at x10sys.com">Reid Spencer</a></address>
 <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a> 
-<br>Last modified: $Date: 2004/02/27 16:17:20 $ </div>
+<br>Last modified: $Date: 2004/05/23 21:07:26 $ </div>
 </body>
 </html>
 <!-- vim: sw=2 ts=2 ai


Index: llvm/docs/Projects.html
diff -u llvm/docs/Projects.html:1.10 llvm/docs/Projects.html:1.11
--- llvm/docs/Projects.html:1.10	Wed May 12 15:57:43 2004
+++ llvm/docs/Projects.html	Sun May 23 16:07:26 2004
@@ -25,6 +25,10 @@
 <li><a href="#help">Further help</a></li>
 </ol>
 
+<div class="doc_author">
+  <p>Written by John Criswell</p>
+</div>
+
 <!-- *********************************************************************** -->
 <div class="doc_section"><a name="overview">Overview</a></div>
 <!-- *********************************************************************** -->
@@ -440,7 +444,7 @@
   <a href="mailto:criswell at uiuc.edu">John Criswell</a><br>
   <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a>
   <br>
-  Last modified: $Date: 2004/05/12 20:57:43 $
+  Last modified: $Date: 2004/05/23 21:07:26 $
 </address>
 
 </body>


Index: llvm/docs/ReleaseNotes.html
diff -u llvm/docs/ReleaseNotes.html:1.182 llvm/docs/ReleaseNotes.html:1.183
--- llvm/docs/ReleaseNotes.html:1.182	Sun May 23 15:42:08 2004
+++ llvm/docs/ReleaseNotes.html	Sun May 23 16:07:27 2004
@@ -28,8 +28,8 @@
   <li><a href="#additionalinfo">Additional Information</a></li>
 </ol>
 
-<div class="doc_text">
-  <p><b>Written by the <a href="http://llvm.cs.uiuc.edu">LLVM team</a></b><p>
+<div class="doc_author">
+  <p>Written by the <a href="http://llvm.cs.uiuc.edu">LLVM team</a><p>
 </div>
 
 <!-- *********************************************************************** -->
@@ -660,7 +660,7 @@
   src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
 
   <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/05/23 20:42:08 $
+  Last modified: $Date: 2004/05/23 21:07:27 $
 </address>
 
 </body>


Index: llvm/docs/SourceLevelDebugging.html
diff -u llvm/docs/SourceLevelDebugging.html:1.4 llvm/docs/SourceLevelDebugging.html:1.5
--- llvm/docs/SourceLevelDebugging.html:1.4	Wed May 12 16:26:16 2004
+++ llvm/docs/SourceLevelDebugging.html	Sun May 23 16:07:27 2004
@@ -68,6 +68,10 @@
 </tr>
 </table>
 
+<div class="doc_author">
+  <p>Written by <a href="mailto:sabre at nondot.org">Chris Lattner</a></p>
+</div>
+
 
 <!-- *********************************************************************** -->
 <div class="doc_section"><a name="introduction">Introduction</a></div> <!--
@@ -1107,7 +1111,7 @@
 
   <a href="mailto:sabre at nondot.org">Chris Lattner</a><br>
   <a href="http://llvm.cs.uiuc.edu">LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/05/12 21:26:16 $
+  Last modified: $Date: 2004/05/23 21:07:27 $
 </address>
 
 </body>


Index: llvm/docs/Stacker.html
diff -u llvm/docs/Stacker.html:1.13 llvm/docs/Stacker.html:1.14
--- llvm/docs/Stacker.html:1.13	Fri May 21 20:04:19 2004
+++ llvm/docs/Stacker.html	Sun May 23 16:07:27 2004
@@ -51,8 +51,8 @@
     </ol></li>
 </ol>
 
-<div class="doc_text">
-<p><b>Written by <a href="mailto:rspencer at x10sys.com">Reid Spencer</a></b></p>
+<div class="doc_author">
+  <p>Written by <a href="mailto:rspencer at x10sys.com">Reid Spencer</a></p>
 </div>
 
 <!-- ======================================================================= -->
@@ -1405,7 +1405,7 @@
 
   <a href="mailto:rspencer at x10sys.com">Reid Spencer</a><br>
   <a href="http://llvm.cs.uiuc.edu">LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/05/22 01:04:19 $
+  Last modified: $Date: 2004/05/23 21:07:27 $
 </address>
 
 </body>


Index: llvm/docs/TableGenFundamentals.html
diff -u llvm/docs/TableGenFundamentals.html:1.5 llvm/docs/TableGenFundamentals.html:1.6
--- llvm/docs/TableGenFundamentals.html:1.5	Wed May 12 13:31:21 2004
+++ llvm/docs/TableGenFundamentals.html	Sun May 23 16:07:27 2004
@@ -48,6 +48,10 @@
 </ul>
 </div>
 
+<div class="doc_author">
+  <p>Written by <a href="mailto:sabre at nondot.org">Chris Lattner</a></p>
+</div>
+
 <!-- *********************************************************************** -->
 <div class="doc_section"><a name="introduction">Introduction</a></div>
 <!-- *********************************************************************** -->
@@ -565,7 +569,7 @@
 
   <a href="mailto:sabre at nondot.org">Chris Lattner</a><br>
   <a href="http://llvm.cs.uiuc.edu">LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/05/12 18:31:21 $
+  Last modified: $Date: 2004/05/23 21:07:27 $
 </address>
 
 </body>


Index: llvm/docs/TestingGuide.html
diff -u llvm/docs/TestingGuide.html:1.7 llvm/docs/TestingGuide.html:1.8
--- llvm/docs/TestingGuide.html:1.7	Mon Mar  1 12:21:04 2004
+++ llvm/docs/TestingGuide.html	Sun May 23 16:07:27 2004
@@ -26,7 +26,9 @@
 <li><a href="#run">Running the LLVM Tests</a></li>
 </ol>
 
-<p><b>Written by John T. Criswell</b></p>
+<div class="doc_author">
+  <p>Written by John T. Criswell</p>
+</div>
 
 <!--===============================================================-->
 <div class="doc_section"><a name="overview">Overview</a></div>
@@ -377,7 +379,7 @@
 
   John T. Criswell<br>
   <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/03/01 18:21:04 $
+  Last modified: $Date: 2004/05/23 21:07:27 $
 </address>
 </body>
 </html>


Index: llvm/docs/WritingAnLLVMPass.html
diff -u llvm/docs/WritingAnLLVMPass.html:1.26 llvm/docs/WritingAnLLVMPass.html:1.27
--- llvm/docs/WritingAnLLVMPass.html:1.26	Wed Mar 17 15:33:32 2004
+++ llvm/docs/WritingAnLLVMPass.html	Sun May 23 16:07:27 2004
@@ -87,8 +87,8 @@
     </ul></li>
 </ol>
 
-<div class="doc_text">    
-  <p><b>Written by <a href="mailto:sabre at nondot.org">Chris Lattner</a></b><p>
+<div class="doc_author">
+  <p>Written by <a href="mailto:sabre at nondot.org">Chris Lattner</a></p>
 </div>
 
 <!-- *********************************************************************** -->
@@ -1471,7 +1471,7 @@
 
   <a href="mailto:sabre at nondot.org">Chris Lattner</a><br>
   <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/03/17 21:33:32 $
+  Last modified: $Date: 2004/05/23 21:07:27 $
 </address>
 
 </body>




From lattner at cs.uiuc.edu  Sun May 23 16:13:01 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:13:01 2004
Subject: [llvm-commits] CVS: llvm/lib/Analysis/AliasSetTracker.cpp
Message-ID: <200405232111.QAA03911@zion.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis:

AliasSetTracker.cpp updated: 1.15 -> 1.16

---
Log message:

Rename a method


---
Diffs of the changes:  (+2 -2)

Index: llvm/lib/Analysis/AliasSetTracker.cpp
diff -u llvm/lib/Analysis/AliasSetTracker.cpp:1.15 llvm/lib/Analysis/AliasSetTracker.cpp:1.16
--- llvm/lib/Analysis/AliasSetTracker.cpp:1.15	Tue Apr 27 10:12:45 2004
+++ llvm/lib/Analysis/AliasSetTracker.cpp	Sun May 23 16:10:58 2004
@@ -277,12 +277,12 @@
 }
 
 
-// remove method - This method is used to remove a pointer value from the
+// deleteValue method - This method is used to remove a pointer value from the
 // AliasSetTracker entirely.  It should be used when an instruction is deleted
 // from the program to update the AST.  If you don't use this, you would have
 // dangling pointers to deleted instructions.
 //
-void AliasSetTracker::remove(Value *PtrVal) {
+void AliasSetTracker::deleteValue(Value *PtrVal) {
   // First, look up the PointerRec for this pointer...
   hash_map<Value*, AliasSet::PointerRec>::iterator I = PointerMap.find(PtrVal);
   if (I == PointerMap.end()) return;  // Noop




From lattner at cs.uiuc.edu  Sun May 23 16:13:57 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:13:57 2004
Subject: [llvm-commits] CVS: llvm/include/llvm/Analysis/AliasSetTracker.h
Message-ID: <200405232110.QAA03897@zion.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Analysis:

AliasSetTracker.h updated: 1.12 -> 1.13

---
Log message:

Rename a method


---
Diffs of the changes:  (+3 -3)

Index: llvm/include/llvm/Analysis/AliasSetTracker.h
diff -u llvm/include/llvm/Analysis/AliasSetTracker.h:1.12 llvm/include/llvm/Analysis/AliasSetTracker.h:1.13
--- llvm/include/llvm/Analysis/AliasSetTracker.h:1.12	Sun Mar 14 22:08:18 2004
+++ llvm/include/llvm/Analysis/AliasSetTracker.h	Sun May 23 16:10:42 2004
@@ -264,12 +264,12 @@
   void add(BasicBlock &BB);       // Add all instructions in basic block
   void add(const AliasSetTracker &AST); // Add alias relations from another AST
 
-  /// remove method - This method is used to remove a pointer value from the
-  /// AliasSetTracker entirely.  It should be used when an instruction is
+  /// deleteValue method - This method is used to remove a pointer value from
+  /// the AliasSetTracker entirely.  It should be used when an instruction is
   /// deleted from the program to update the AST.  If you don't use this, you
   /// would have dangling pointers to deleted instructions.
   ///
-  void remove(Value *PtrVal);
+  void deleteValue(Value *PtrVal);
 
   /// getAliasSets - Return the alias sets that are active.
   const ilist<AliasSet> &getAliasSets() const { return AliasSets; }




From lattner at cs.uiuc.edu  Sun May 23 16:15:02 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:15:02 2004
Subject: [llvm-commits] CVS: llvm/lib/Analysis/LoadValueNumbering.cpp
Message-ID: <200405232113.QAA03955@zion.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis:

LoadValueNumbering.cpp updated: 1.18 -> 1.19

---
Log message:

Implement the interfaces to update value numbering information.  Add an
assert.


---
Diffs of the changes:  (+24 -2)

Index: llvm/lib/Analysis/LoadValueNumbering.cpp
diff -u llvm/lib/Analysis/LoadValueNumbering.cpp:1.18 llvm/lib/Analysis/LoadValueNumbering.cpp:1.19
--- llvm/lib/Analysis/LoadValueNumbering.cpp:1.18	Fri Apr  2 18:45:16 2004
+++ llvm/lib/Analysis/LoadValueNumbering.cpp	Sun May 23 16:13:24 2004
@@ -56,6 +56,24 @@
     virtual void getEqualNumberNodes(Value *V1,
                                      std::vector<Value*> &RetVals) const;
 
+    /// deleteValue - This method should be called whenever an LLVM Value is
+    /// deleted from the program, for example when an instruction is found to be
+    /// redundant and is eliminated.
+    ///
+    virtual void deleteValue(Value *V) {
+      getAnalysis<AliasAnalysis>().deleteValue(V);
+    }
+    
+    /// copyValue - This method should be used whenever a preexisting value in
+    /// the program is copied or cloned, introducing a new value.  Note that
+    /// analysis implementations should tolerate clients that use this method to
+    /// introduce the same value multiple times: if the analysis already knows
+    /// about a value, it should ignore the request.
+    ///
+    virtual void copyValue(Value *From, Value *To) {
+      getAnalysis<AliasAnalysis>().copyValue(From, To);
+    }
+
     /// getCallEqualNumberNodes - Given a call instruction, find other calls
     /// that have the same value number.
     void getCallEqualNumberNodes(CallInst *CI,
@@ -103,7 +121,7 @@
       TransparentBlocks.insert(TBI, std::make_pair(CurBlock, false));
       return false;
     }
-      TransparentBlocks.insert(TBI, std::make_pair(CurBlock, true));
+    TransparentBlocks.insert(TBI, std::make_pair(CurBlock, true));
   } else if (!TBI->second)
     // This block is known non-transparent, so that path can't be either.
     return false;
@@ -505,8 +523,12 @@
 
         BasicBlock::iterator BBI = I->first->end();
         while (1) {
+          assert(BBI != I->first->begin() &&
+                 "There is a store in this block of the pointer, but the store"
+                 " doesn't mod the address being stored to??  Must be a bug in"
+                 " the alias analysis implementation!");
           --BBI;
-          if (AA.getModRefInfo(BBI, LoadPtr, LoadSize)& AliasAnalysis::Mod){
+          if (AA.getModRefInfo(BBI, LoadPtr, LoadSize) & AliasAnalysis::Mod) {
             // If the invalidating instruction is one of the candidates,
             // then it provides the value the load loads.
             if (StoreInst *SI = dyn_cast<StoreInst>(BBI))




From lattner at cs.uiuc.edu  Sun May 23 16:15:11 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:15:11 2004
Subject: [llvm-commits] CVS: llvm/include/llvm/Analysis/AliasAnalysis.h
Message-ID: <200405232112.QAA03932@zion.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Analysis:

AliasAnalysis.h updated: 1.16 -> 1.17

---
Log message:

Several *major* changes to the AA interfaces:

1. Provide interfaces so that clients can update alias analyses to reflect 
   the changes made by the transformations.
2. Change how alias analysis implementations work overall.  In particular, 
   now clients will automatically forward to chained AA implementations: they 
   don't have to remember to do it themselves.


---
Diffs of the changes:  (+43 -11)

Index: llvm/include/llvm/Analysis/AliasAnalysis.h
diff -u llvm/include/llvm/Analysis/AliasAnalysis.h:1.16 llvm/include/llvm/Analysis/AliasAnalysis.h:1.17
--- llvm/include/llvm/Analysis/AliasAnalysis.h:1.16	Sun Apr 11 11:42:50 2004
+++ llvm/include/llvm/Analysis/AliasAnalysis.h	Sun May 23 16:12:38 2004
@@ -40,8 +40,10 @@
 class TargetData;
 
 class AliasAnalysis {
-  const TargetData *TD;
 protected:
+  const TargetData *TD;
+  AliasAnalysis *AA;       // Previous Alias Analysis to chain to.
+
   /// InitializeAliasAnalysis - Subclasses must call this method to initialize
   /// the AliasAnalysis interface before any other methods are called.  This is
   /// typically called by the run* methods of these subclasses.  This may be
@@ -55,7 +57,7 @@
   virtual void getAnalysisUsage(AnalysisUsage &AU) const;
 
 public:
-  AliasAnalysis() : TD(0) {}
+  AliasAnalysis() : TD(0), AA(0) {}
   virtual ~AliasAnalysis();  // We want to be subclassed
 
   /// getTargetData - Every alias analysis implementation depends on the size of
@@ -82,9 +84,7 @@
   /// analysis implementations.
   ///
   virtual AliasResult alias(const Value *V1, unsigned V1Size,
-                            const Value *V2, unsigned V2Size) {
-    return MayAlias;
-  }
+                            const Value *V2, unsigned V2Size);
 
   /// getMustAliases - If there are any pointers known that must alias this
   /// pointer, return them now.  This allows alias-set based alias analyses to
@@ -92,13 +92,13 @@
   /// alias analysis supports this, it should ADD any must aliased pointers to
   /// the specified vector.
   ///
-  virtual void getMustAliases(Value *P, std::vector<Value*> &RetVals) {}
+  virtual void getMustAliases(Value *P, std::vector<Value*> &RetVals);
 
   /// pointsToConstantMemory - If the specified pointer is known to point into
   /// constant global memory, return true.  This allows disambiguation of store
   /// instructions from constant pointers.
   ///
-  virtual bool pointsToConstantMemory(const Value *P) { return false; }
+  virtual bool pointsToConstantMemory(const Value *P);
 
   /// doesNotAccessMemory - If the specified function is known to never read or
   /// write memory, return true.  If the function only reads from known-constant
@@ -111,7 +111,7 @@
   ///
   /// This property corresponds to the GCC 'const' attribute.
   ///
-  virtual bool doesNotAccessMemory(Function *F) { return false; }
+  virtual bool doesNotAccessMemory(Function *F);
 
   /// onlyReadsMemory - If the specified function is known to only read from
   /// non-volatile memory (or not access memory at all), return true.  Functions
@@ -122,7 +122,7 @@
   ///
   /// This property corresponds to the GCC 'pure' attribute.
   ///
-  virtual bool onlyReadsMemory(Function *F) { return doesNotAccessMemory(F); }
+  virtual bool onlyReadsMemory(Function *F);
 
 
   //===--------------------------------------------------------------------===//
@@ -158,8 +158,7 @@
   /// Remember that if you override this and chain to another analysis, you must
   /// make sure that it doesn't have mod/ref info either.
   ///
-  virtual bool hasNoModRefInfoForCalls() const { return false; }
-
+  virtual bool hasNoModRefInfoForCalls() const;
 
   /// Convenience functions...
   ModRefResult getModRefInfo(LoadInst *L, Value *P, unsigned Size);
@@ -180,6 +179,10 @@
     }
   }
 
+  //===--------------------------------------------------------------------===//
+  /// Higher level methods for querying mod/ref information.
+  ///
+
   /// canBasicBlockModify - Return true if it is possible for execution of the
   /// specified basic block to modify the value pointed to by Ptr.
   ///
@@ -192,6 +195,35 @@
   ///
   bool canInstructionRangeModify(const Instruction &I1, const Instruction &I2,
                                  const Value *Ptr, unsigned Size);
+
+  //===--------------------------------------------------------------------===//
+  /// Methods that clients should call when they transform the program to allow
+  /// alias analyses to update their internal data structures.  Note that these
+  /// methods may be called on any instruction, regardless of whether or not
+  /// they have pointer-analysis implications.
+  ///
+
+  /// deleteValue - This method should be called whenever an LLVM Value is
+  /// deleted from the program, for example when an instruction is found to be
+  /// redundant and is eliminated.
+  ///
+  virtual void deleteValue(Value *V);
+
+  /// copyValue - This method should be used whenever a preexisting value in the
+  /// program is copied or cloned, introducing a new value.  Note that analysis
+  /// implementations should tolerate clients that use this method to introduce
+  /// the same value multiple times: if the analysis already knows about a
+  /// value, it should ignore the request.
+  ///
+  virtual void copyValue(Value *From, Value *To);
+
+  /// replaceWithNewValue - This method is the obvious combination of the two
+  /// above, and it provided as a helper to simplify client code.
+  ///
+  void replaceWithNewValue(Value *Old, Value *New) {
+    copyValue(Old, New);
+    deleteValue(Old);
+  }
 };
 
 // Because of the way .a files work, we must force the BasicAA implementation to




From lattner at cs.uiuc.edu  Sun May 23 16:15:20 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:15:20 2004
Subject: [llvm-commits] CVS: llvm/include/llvm/Analysis/ValueNumbering.h
Message-ID: <200405232112.QAA03942@zion.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Analysis:

ValueNumbering.h updated: 1.6 -> 1.7

---
Log message:

Add interfaces to update value numbering results


---
Diffs of the changes:  (+22 -5)

Index: llvm/include/llvm/Analysis/ValueNumbering.h
diff -u llvm/include/llvm/Analysis/ValueNumbering.h:1.6 llvm/include/llvm/Analysis/ValueNumbering.h:1.7
--- llvm/include/llvm/Analysis/ValueNumbering.h:1.6	Sat Apr 10 17:32:47 2004
+++ llvm/include/llvm/Analysis/ValueNumbering.h	Sun May 23 16:11:17 2004
@@ -39,13 +39,30 @@
 
   ///===-------------------------------------------------------------------===//
   /// Interfaces to update value numbering analysis information as the client
-  /// changes the program
+  /// changes the program.
   ///
 
-  /// deleteInstruction - Clients should invoke this method when they delete an
-  /// instruction from the program.  This allows the analysis implementations to
-  /// avoid having dangling pointers in their representation.
-  virtual void deleteInstruction(Instruction *I) {}
+  /// deleteValue - This method should be called whenever an LLVM Value is
+  /// deleted from the program, for example when an instruction is found to be
+  /// redundant and is eliminated.
+  ///
+  virtual void deleteValue(Value *V) {}
+
+  /// copyValue - This method should be used whenever a preexisting value in the
+  /// program is copied or cloned, introducing a new value.  Note that analysis
+  /// implementations should tolerate clients that use this method to introduce
+  /// the same value multiple times: if the analysis already knows about a
+  /// value, it should ignore the request.
+  ///
+  virtual void copyValue(Value *From, Value *To) {}
+
+  /// replaceWithNewValue - This method is the obvious combination of the two
+  /// above, and it provided as a helper to simplify client code.
+  ///
+  void replaceWithNewValue(Value *Old, Value *New) {
+    copyValue(Old, New);
+    deleteValue(Old);
+  }
 };
 
 extern void BasicValueNumberingStub();




From lattner at cs.uiuc.edu  Sun May 23 16:16:00 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:16:00 2004
Subject: [llvm-commits] CVS: llvm/lib/Analysis/DataStructure/DataStructureAA.cpp
Message-ID: <200405232114.QAA03991@zion.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis/DataStructure:

DataStructureAA.cpp updated: 1.17 -> 1.18

---
Log message:

Updates to work with the new auto-forwarding AA interface changes


---
Diffs of the changes:  (+3 -9)

Index: llvm/lib/Analysis/DataStructure/DataStructureAA.cpp
diff -u llvm/lib/Analysis/DataStructure/DataStructureAA.cpp:1.17 llvm/lib/Analysis/DataStructure/DataStructureAA.cpp:1.18
--- llvm/lib/Analysis/DataStructure/DataStructureAA.cpp:1.17	Mon Apr 26 09:44:08 2004
+++ llvm/lib/Analysis/DataStructure/DataStructureAA.cpp	Sun May 23 16:14:27 2004
@@ -44,7 +44,6 @@
       AU.setPreservesAll();                         // Does not transform code
       AU.addRequiredTransitive<TDDataStructures>(); // Uses TD Datastructures
       AU.addRequiredTransitive<BUDataStructures>(); // Uses BU Datastructures
-      AU.addRequired<AliasAnalysis>();              // Chains to another AA impl
     }
 
     //------------------------------------------------
@@ -56,12 +55,7 @@
 
     void getMustAliases(Value *P, std::vector<Value*> &RetVals);
 
-    bool pointsToConstantMemory(const Value *P) {
-      return getAnalysis<AliasAnalysis>().pointsToConstantMemory(P);
-    }
-    
-    AliasAnalysis::ModRefResult
-    getModRefInfo(CallSite CS, Value *P, unsigned Size);
+    ModRefResult getModRefInfo(CallSite CS, Value *P, unsigned Size);
 
   private:
     DSGraph *getGraphForValue(const Value *V);
@@ -155,7 +149,7 @@
 
   // FIXME: we could improve on this by checking the globals graph for aliased
   // global queries...
-  return getAnalysis<AliasAnalysis>().alias(V1, V1Size, V2, V2Size);
+  return AliasAnalysis::alias(V1, V1Size, V2, V2Size);
 }
 
 /// getModRefInfo - does a callsite modify or reference a value?
@@ -212,6 +206,6 @@
     }
   }
 #endif
-  return getAnalysis<AliasAnalysis>().getMustAliases(P, RetVals);
+  return AliasAnalysis::getMustAliases(P, RetVals);
 }
 




From lattner at cs.uiuc.edu  Sun May 23 16:16:09 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:16:09 2004
Subject: [llvm-commits] CVS: llvm/lib/Analysis/DataStructure/Steensgaard.cpp
Message-ID: <200405232114.QAA03982@zion.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis/DataStructure:

Steensgaard.cpp updated: 1.37 -> 1.38

---
Log message:

Update to match the autochaining interface that the AA interface uses


---
Diffs of the changes:  (+1 -6)

Index: llvm/lib/Analysis/DataStructure/Steensgaard.cpp
diff -u llvm/lib/Analysis/DataStructure/Steensgaard.cpp:1.37 llvm/lib/Analysis/DataStructure/Steensgaard.cpp:1.38
--- llvm/lib/Analysis/DataStructure/Steensgaard.cpp:1.37	Fri Jan 30 16:20:55 2004
+++ llvm/lib/Analysis/DataStructure/Steensgaard.cpp	Sun May 23 16:13:51 2004
@@ -47,7 +47,6 @@
       AliasAnalysis::getAnalysisUsage(AU);
       AU.setPreservesAll();                    // Does not transform code...
       AU.addRequired<LocalDataStructures>();   // Uses local dsgraph
-      AU.addRequired<AliasAnalysis>();         // Chains to another AA impl...
     }
 
     // print - Implement the Pass::print method...
@@ -63,10 +62,6 @@
     // alias - This is the only method here that does anything interesting...
     AliasResult alias(const Value *V1, unsigned V1Size,
                       const Value *V2, unsigned V2Size);
-
-    bool pointsToConstantMemory(const Value *P) {
-      return getAnalysis<AliasAnalysis>().pointsToConstantMemory(P);
-    }
     
   private:
     void ResolveFunctionCall(Function *F, const DSCallSite &Call,
@@ -238,5 +233,5 @@
   // If we cannot determine alias properties based on our graph, fall back on
   // some other AA implementation.
   //
-  return getAnalysis<AliasAnalysis>().alias(V1, V1Size, V2, V2Size);
+  return AliasAnalysis::alias(V1, V1Size, V2, V2Size);
 }




From lattner at cs.uiuc.edu  Sun May 23 16:17:00 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:17:00 2004
Subject: [llvm-commits] CVS: llvm/lib/Analysis/DataStructure/Local.cpp
Message-ID: <200405232114.QAA03998@zion.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis/DataStructure:

Local.cpp updated: 1.102 -> 1.103

---
Log message:

Fix a really nasty bug with the -disable-ds-field-sensitivity option


---
Diffs of the changes:  (+3 -0)

Index: llvm/lib/Analysis/DataStructure/Local.cpp
diff -u llvm/lib/Analysis/DataStructure/Local.cpp:1.102 llvm/lib/Analysis/DataStructure/Local.cpp:1.103
--- llvm/lib/Analysis/DataStructure/Local.cpp:1.102	Sun Apr  4 20:28:23 2004
+++ llvm/lib/Analysis/DataStructure/Local.cpp	Sun May 23 16:14:09 2004
@@ -130,6 +130,9 @@
     DSNode *createNode(const Type *Ty = 0) {
       DSNode *N = new DSNode(Ty, &G);   // Create the node
       if (DisableFieldSensitivity) {
+        // Create node handle referring to the old node so that it is
+        // immediately removed from the graph when the node handle is destroyed.
+        DSNodeHandle OldNNH = N;
         N->foldNodeCompletely();
         if (DSNode *FN = N->getForwardNode())
           N = FN;




From lattner at cs.uiuc.edu  Sun May 23 16:17:08 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:17:08 2004
Subject: [llvm-commits] CVS: llvm/lib/Analysis/BasicAliasAnalysis.cpp
Message-ID: <200405232115.QAA04010@zion.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis:

BasicAliasAnalysis.cpp updated: 1.40 -> 1.41

---
Log message:

Move the -no-aa AA implementation into this file since both of these 
alias analysis implementations are special: they do not autoforward to a 
chained implementation of alias analysis


---
Diffs of the changes:  (+46 -8)

Index: llvm/lib/Analysis/BasicAliasAnalysis.cpp
diff -u llvm/lib/Analysis/BasicAliasAnalysis.cpp:1.40 llvm/lib/Analysis/BasicAliasAnalysis.cpp:1.41
--- llvm/lib/Analysis/BasicAliasAnalysis.cpp:1.40	Mon Apr 12 00:16:42 2004
+++ llvm/lib/Analysis/BasicAliasAnalysis.cpp	Sun May 23 16:15:12 2004
@@ -36,13 +36,55 @@
 void llvm::BasicAAStub() {}
 
 namespace {
-  struct BasicAliasAnalysis : public ImmutablePass, public AliasAnalysis {
-    
+  /// NoAA - This class implements the -no-aa pass, which always returns "I
+  /// don't know" for alias queries.  NoAA is unlike other alias analysis
+  /// implementations, in that it does not chain to a previous analysis.  As
+  /// such it doesn't follow many of the rules that other alias analyses must.
+  ///
+  struct NoAA : public ImmutablePass, public AliasAnalysis {
+    virtual AliasResult alias(const Value *V1, unsigned V1Size,
+                              const Value *V2, unsigned V2Size) {
+      return MayAlias;
+    }
+
+    virtual void getMustAliases(Value *P, std::vector<Value*> &RetVals) { }
+    virtual bool pointsToConstantMemory(const Value *P) { return false; }
+    virtual bool doesNotAccessMemory(Function *F) { return false; }
+    virtual bool onlyReadsMemory(Function *F) { return false; }
+    virtual ModRefResult getModRefInfo(CallSite CS, Value *P, unsigned Size) {
+      return ModRef;
+    }
+    virtual ModRefResult getModRefInfo(CallSite CS1, CallSite CS2) {
+      return ModRef;
+    }
+    virtual bool hasNoModRefInfoForCalls() const { return true; }
+
+    virtual void deleteValue(Value *V) {}
+    virtual void copyValue(Value *From, Value *To) {}
+    virtual void getAnalysisUsage(AnalysisUsage &AU) const {}
+  };
+ 
+  // Register this pass...
+  RegisterOpt<NoAA>
+  U("no-aa", "No Alias Analysis (always returns 'may' alias)");
+
+  // Declare that we implement the AliasAnalysis interface
+  RegisterAnalysisGroup<AliasAnalysis, NoAA> V;
+}  // End of anonymous namespace
+
+
+namespace {
+  /// BasicAliasAnalysis - This is the default alias analysis implementation.
+  /// Because it doesn't chain to a previous alias analysis (like -no-aa), it
+  /// derives from the NoAA class.
+  struct BasicAliasAnalysis : public NoAA {
     virtual void getAnalysisUsage(AnalysisUsage &AU) const {
-      AliasAnalysis::getAnalysisUsage(AU);
+      AU.addRequired<TargetData>();
     }
     
-    virtual void initializePass();
+    virtual void initializePass() {
+      TD = &getAnalysis<TargetData>();
+    }
 
     AliasResult alias(const Value *V1, unsigned V1Size,
                       const Value *V2, unsigned V2Size);
@@ -78,10 +120,6 @@
   // Declare that we implement the AliasAnalysis interface
   RegisterAnalysisGroup<AliasAnalysis, BasicAliasAnalysis, true> Y;
 }  // End of anonymous namespace
-
-void BasicAliasAnalysis::initializePass() {
-  InitializeAliasAnalysis(this);
-}
 
 // hasUniqueAddress - Return true if the specified value points to something
 // with a unique, discernable, address.




From lattner at cs.uiuc.edu  Sun May 23 16:18:00 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:18:00 2004
Subject: [llvm-commits] CVS: llvm/include/llvm/Intrinsics.h
Message-ID: <200405232116.QAA04051@zion.cs.uiuc.edu>


Changes in directory llvm/include/llvm:

Intrinsics.h updated: 1.21 -> 1.22

---
Log message:

Add the new GC intrinsics


---
Diffs of the changes:  (+17 -12)

Index: llvm/include/llvm/Intrinsics.h
diff -u llvm/include/llvm/Intrinsics.h:1.21 llvm/include/llvm/Intrinsics.h:1.22
--- llvm/include/llvm/Intrinsics.h:1.21	Tue Apr 13 21:22:54 2004
+++ llvm/include/llvm/Intrinsics.h	Sun May 23 16:16:33 2004
@@ -26,27 +26,27 @@
   enum ID {
     not_intrinsic = 0,   // Must be zero
 
-    // Varargs handling intrinsics...
+    // Varargs handling intrinsics.
     vastart,        // Used to implement the va_start macro in C
     vaend,          // Used to implement the va_end macro in C
     vacopy,         // Used to implement the va_copy macro in C
 
-    // Code generator intrinsics...
+    // Code generator intrinsics.
     returnaddress,  // Yields the return address of a dynamic call frame
     frameaddress,   // Yields the frame address of a dynamic call frame
 
-    // Standard libc functions...
-    memcpy,         // Copy non-overlapping memory blocks
-    memmove,        // Copy potentially overlapping memory blocks
-    memset,         // Fill memory with a byte value
-
-    // Setjmp/Longjmp intrinsics...
+    // setjmp/longjmp intrinsics.
     setjmp,         // Used to represent a setjmp call in C
     longjmp,        // Used to represent a longjmp call in C
     sigsetjmp,      // Used to represent a sigsetjmp call in C
     siglongjmp,     // Used to represent a siglongjmp call in C
 
-    // Debugging intrinsics...
+    // Garbage Collection intrinsics.
+    gcroot,         // Defines a new GC root on the stack
+    gcread,         // Defines a read of a heap object  (for read barriers)
+    gcwrite,        // Defines a write to a heap object (for write barriers)
+
+    // Debugging intrinsics.
     dbg_stoppoint,    // Represents source lines and breakpointable places
     dbg_region_start, // Start of a region
     dbg_region_end,   // End of a region
@@ -54,10 +54,15 @@
     dbg_declare,      // Declare a local object
 
 
-    // Standard libm functions...
+    // Standard libc functions.
+    memcpy,         // Copy non-overlapping memory blocks
+    memmove,        // Copy potentially overlapping memory blocks
+    memset,         // Fill memory with a byte value
+
+    // Standard libm functions.
     
 
-    // Input/Output intrinsics
+    // Input/Output intrinsics.
     readport,
     writeport,
     readio,
@@ -65,7 +70,7 @@
 
     //===------------------------------------------------------------------===//
     // This section defines intrinsic functions used to represent Alpha
-    // instructions...
+    // instructions.
     //
     alpha_ctlz,     // CTLZ (count leading zero): counts the number of leading
                     // zeros in the given ulong value




From lattner at cs.uiuc.edu  Sun May 23 16:18:09 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:18:09 2004
Subject: [llvm-commits] CVS: llvm/lib/Analysis/AliasAnalysis.cpp
Message-ID: <200405232115.QAA04024@zion.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis:

AliasAnalysis.cpp updated: 1.17 -> 1.18

---
Log message:

Changes to work with the changes to the AliasAnalysis interface.  The -no-aa
class is now in the BasicAliasAnalysis.cpp file


---
Diffs of the changes:  (+67 -28)

Index: llvm/lib/Analysis/AliasAnalysis.cpp
diff -u llvm/lib/Analysis/AliasAnalysis.cpp:1.17 llvm/lib/Analysis/AliasAnalysis.cpp:1.18
--- llvm/lib/Analysis/AliasAnalysis.cpp:1.17	Sun Mar 14 22:07:29 2004
+++ llvm/lib/Analysis/AliasAnalysis.cpp	Sun May 23 16:15:48 2004
@@ -35,6 +35,64 @@
   RegisterAnalysisGroup<AliasAnalysis> Z("Alias Analysis");
 }
 
+//===----------------------------------------------------------------------===//
+// Default chaining methods
+//===----------------------------------------------------------------------===//
+
+AliasAnalysis::AliasResult
+AliasAnalysis::alias(const Value *V1, unsigned V1Size,
+                     const Value *V2, unsigned V2Size) {
+  assert(AA && "AA didn't call InitializeAliasAnalysis in its run method!");
+  return AA->alias(V1, V1Size, V2, V2Size);
+}
+
+void AliasAnalysis::getMustAliases(Value *P, std::vector<Value*> &RetVals) {
+  assert(AA && "AA didn't call InitializeAliasAnalysis in its run method!");
+  return AA->getMustAliases(P, RetVals);
+}
+
+bool AliasAnalysis::pointsToConstantMemory(const Value *P) {
+  assert(AA && "AA didn't call InitializeAliasAnalysis in its run method!");
+  return AA->pointsToConstantMemory(P);
+}
+
+bool AliasAnalysis::doesNotAccessMemory(Function *F) {
+  assert(AA && "AA didn't call InitializeAliasAnalysis in its run method!");
+  return AA->doesNotAccessMemory(F);
+}
+
+bool AliasAnalysis::onlyReadsMemory(Function *F) {
+  assert(AA && "AA didn't call InitializeAliasAnalysis in its run method!");
+  return doesNotAccessMemory(F) || AA->onlyReadsMemory(F);
+}
+
+bool AliasAnalysis::hasNoModRefInfoForCalls() const {
+  assert(AA && "AA didn't call InitializeAliasAnalysis in its run method!");
+  return AA->hasNoModRefInfoForCalls();
+}
+
+void AliasAnalysis::deleteValue(Value *V) {
+  assert(AA && "AA didn't call InitializeAliasAnalysis in its run method!");
+  AA->deleteValue(V);
+}
+
+void AliasAnalysis::copyValue(Value *From, Value *To) {
+  assert(AA && "AA didn't call InitializeAliasAnalysis in its run method!");
+  AA->copyValue(From, To);
+}
+
+AliasAnalysis::ModRefResult
+AliasAnalysis::getModRefInfo(CallSite CS1, CallSite CS2) {
+  // FIXME: we can do better.
+  assert(AA && "AA didn't call InitializeAliasAnalysis in its run method!");
+  return AA->getModRefInfo(CS1, CS2);
+}
+
+
+//===----------------------------------------------------------------------===//
+// AliasAnalysis non-virtual helper method implementation
+//===----------------------------------------------------------------------===//
+
 AliasAnalysis::ModRefResult
 AliasAnalysis::getModRefInfo(LoadInst *L, Value *P, unsigned Size) {
   return alias(L->getOperand(0), TD->getTypeSize(L->getType()),
@@ -56,24 +114,23 @@
 
 AliasAnalysis::ModRefResult
 AliasAnalysis::getModRefInfo(CallSite CS, Value *P, unsigned Size) {
+  ModRefResult Mask = ModRef;
   if (Function *F = CS.getCalledFunction())
     if (onlyReadsMemory(F)) {
       if (doesNotAccessMemory(F)) return NoModRef;
-      return Ref;
+      Mask = Ref;
     }
 
+  if (!AA) return Mask;
+
   // If P points to a constant memory location, the call definitely could not
   // modify the memory location.
-  return pointsToConstantMemory(P) ? Ref : ModRef;
-}
+  if ((Mask & Mod) && AA->pointsToConstantMemory(P))
+    Mask = Ref;
 
-AliasAnalysis::ModRefResult
-AliasAnalysis::getModRefInfo(CallSite CS1, CallSite CS2) {
-  // FIXME: could probably do better.
-  return ModRef;
+  return ModRefResult(Mask & AA->getModRefInfo(CS, P, Size));
 }
 
-
 // AliasAnalysis destructor: DO NOT move this to the header file for
 // AliasAnalysis or else clients of the AliasAnalysis class may not depend on
 // the AliasAnalysis.o file in the current .a file, causing alias analysis
@@ -86,6 +143,7 @@
 ///
 void AliasAnalysis::InitializeAliasAnalysis(Pass *P) {
   TD = &P->getAnalysis<TargetData>();
+  AA = &P->getAnalysis<AliasAnalysis>();
 }
 
 // getAnalysisUsage - All alias analysis implementations should invoke this
@@ -93,6 +151,7 @@
 // TargetData is required by the pass.
 void AliasAnalysis::getAnalysisUsage(AnalysisUsage &AU) const {
   AU.addRequired<TargetData>();            // All AA's need TargetData.
+  AU.addRequired<AliasAnalysis>();         // All AA's chain
 }
 
 /// canBasicBlockModify - Return true if it is possible for execution of the
@@ -130,23 +189,3 @@
 //
 extern void llvm::BasicAAStub();
 static IncludeFile INCLUDE_BASICAA_CPP((void*)&BasicAAStub);
-
-
-namespace {
-  struct NoAA : public ImmutablePass, public AliasAnalysis {
-    virtual void getAnalysisUsage(AnalysisUsage &AU) const {
-      AliasAnalysis::getAnalysisUsage(AU);
-    }
-    
-    virtual void initializePass() {
-      InitializeAliasAnalysis(this);
-    }
-  };
- 
-  // Register this pass...
-  RegisterOpt<NoAA>
-  X("no-aa", "No Alias Analysis (always returns 'may' alias)");
-
-  // Declare that we implement the AliasAnalysis interface
-  RegisterAnalysisGroup<AliasAnalysis, NoAA> Y;
-}  // End of anonymous namespace




From lattner at cs.uiuc.edu  Sun May 23 16:18:17 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:18:17 2004
Subject: [llvm-commits] CVS: llvm/include/llvm/Transforms/Scalar.h
Message-ID: <200405232116.QAA04038@zion.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Transforms:

Scalar.h updated: 1.39 -> 1.40

---
Log message:

Add a new prototype


---
Diffs of the changes:  (+11 -2)

Index: llvm/include/llvm/Transforms/Scalar.h
diff -u llvm/include/llvm/Transforms/Scalar.h:1.39 llvm/include/llvm/Transforms/Scalar.h:1.40
--- llvm/include/llvm/Transforms/Scalar.h:1.39	Mon Apr 19 01:28:37 2004
+++ llvm/include/llvm/Transforms/Scalar.h	Sun May 23 16:16:13 2004
@@ -279,11 +279,20 @@
 // handling mechanisms.  Note that after this pass runs the CFG is not entirely
 // accurate (exceptional control flow edges are not correct anymore) so only
 // very simple things should be done after the lowerinvoke pass has run (like
-// generation of native code).  This should not be used as a general purpose "my
-// LLVM-to-LLVM pass doesn't support the invoke instruction yet" lowering pass.
+// generation of native code).  This should *NOT* be used as a general purpose
+// "my LLVM-to-LLVM pass doesn't support the invoke instruction yet" lowering
+// pass.
 //
 FunctionPass *createLowerInvokePass();
 extern const PassInfo *LowerInvokePassID;
+
+  
+//===----------------------------------------------------------------------===//
+/// createLowerGCPass - This function returns an instance of the "lowergc"
+/// pass, which lowers garbage collection intrinsics to normal LLVM code.
+///
+FunctionPass *createLowerGCPass();
+  
 
 //===----------------------------------------------------------------------===//
 //




From lattner at cs.uiuc.edu  Sun May 23 16:19:01 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:19:01 2004
Subject: [llvm-commits] CVS: llvm/lib/VMCore/Function.cpp Verifier.cpp
Message-ID: <200405232117.QAA04065@zion.cs.uiuc.edu>


Changes in directory llvm/lib/VMCore:

Function.cpp updated: 1.68 -> 1.69
Verifier.cpp updated: 1.99 -> 1.100

---
Log message:

Recognize and verify the new GC intrinsics.


---
Diffs of the changes:  (+21 -6)

Index: llvm/lib/VMCore/Function.cpp
diff -u llvm/lib/VMCore/Function.cpp:1.68 llvm/lib/VMCore/Function.cpp:1.69
--- llvm/lib/VMCore/Function.cpp:1.68	Wed Apr 14 08:46:52 2004
+++ llvm/lib/VMCore/Function.cpp	Sun May 23 16:16:51 2004
@@ -217,6 +217,11 @@
   case 'f':
     if (getName() == "llvm.frameaddress")  return Intrinsic::frameaddress;
     break;
+  case 'g':
+    if (getName() == "llvm.gcwrite") return Intrinsic::gcwrite;
+    if (getName() == "llvm.gcread")  return Intrinsic::gcread;
+    if (getName() == "llvm.gcroot")  return Intrinsic::gcroot;
+    break;
   case 'l':
     if (getName() == "llvm.longjmp")  return Intrinsic::longjmp;
     break;


Index: llvm/lib/VMCore/Verifier.cpp
diff -u llvm/lib/VMCore/Verifier.cpp:1.99 llvm/lib/VMCore/Verifier.cpp:1.100
--- llvm/lib/VMCore/Verifier.cpp:1.99	Fri May 21 11:47:21 2004
+++ llvm/lib/VMCore/Verifier.cpp	Sun May 23 16:16:51 2004
@@ -653,15 +653,14 @@
     NumArgs = 1;
     break;
 
-  case Intrinsic:: readio: {
-    const Type * ParamType  = FT->getParamType(0);
-    const Type * ReturnType = FT->getReturnType();
+  case Intrinsic::readio: {
+    const PointerType *ParamType = dyn_cast<PointerType>(FT->getParamType(0));
+    const Type *ReturnType = FT->getReturnType();
 
     Assert1(FT->getNumParams() == 1,
             "Illegal # arguments for intrinsic function!", IF);
-    Assert1(isa<PointerType>(ParamType),
-            "First argument not a pointer!", IF);
-    Assert1(((cast<PointerType>(ParamType)->getElementType()) == ReturnType),
+    Assert1(ParamType, "First argument not a pointer!", IF);
+    Assert1(ParamType->getElementType() == ReturnType,
             "Pointer type doesn't match return type!", IF);
     NumArgs = 1;
     break;
@@ -671,6 +670,17 @@
   case Intrinsic::longjmp:         NumArgs = 2; break;
   case Intrinsic::sigsetjmp:       NumArgs = 2; break;
   case Intrinsic::siglongjmp:      NumArgs = 2; break;
+
+  case Intrinsic::gcroot:
+    Assert1(FT->getNumParams() == 2,
+            "Illegal # arguments for intrinsic function!", IF);
+    Assert1(isa<Constant>(CI.getOperand(2)) ||
+            isa<GlobalValue>(CI.getOperand(2)),
+            "Second argument to llvm.gcroot must be a constant!", &CI);
+    NumArgs = 2;
+    break;
+  case Intrinsic::gcread:          NumArgs = 1; break;
+  case Intrinsic::gcwrite:         NumArgs = 2; break;
 
   case Intrinsic::dbg_stoppoint:   NumArgs = 4; break;
   case Intrinsic::dbg_region_start:NumArgs = 1; break;




From lattner at cs.uiuc.edu  Sun May 23 16:21:01 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:21:01 2004
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Scalar/LowerGC.cpp
Message-ID: <200405232119.QAA04087@zion.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Scalar:

LowerGC.cpp added (r1.1)

---
Log message:

Implement the -lowergc pass which is used by code generators (like the CBE) 
that do not have builtin support for garbage collection.


---
Diffs of the changes:  (+326 -0)

Index: llvm/lib/Transforms/Scalar/LowerGC.cpp
diff -c /dev/null llvm/lib/Transforms/Scalar/LowerGC.cpp:1.1
*** /dev/null	Sun May 23 16:19:32 2004
--- llvm/lib/Transforms/Scalar/LowerGC.cpp	Sun May 23 16:19:22 2004
***************
*** 0 ****
--- 1,326 ----
+ //===-- LowerGC.cpp - Provide GC support for targets that don't -----------===//
+ //
+ //                     The LLVM Compiler Infrastructure
+ //
+ // This file was developed by the LLVM research group and is distributed under
+ // the University of Illinois Open Source License. See LICENSE.TXT for details.
+ //
+ //===----------------------------------------------------------------------===//
+ //
+ // This file implements lowering for the llvm.gc* intrinsics for targets that do
+ // not natively support them (which includes the C backend).  Note that the code
+ // generated is not as efficient as it would be for targets that natively
+ // support the GC intrinsics, but it is useful for getting new targets
+ // up-and-running quickly.
+ //
+ // This pass implements the code transformation described in this paper:
+ //   "Accurate Garbage Collection in an Uncooperative Environment"
+ //   Fergus Hendersen, ISMM, 2002
+ //
+ //===----------------------------------------------------------------------===//
+ 
+ #define DEBUG_TYPE "lowergc"
+ #include "llvm/Transforms/Scalar.h"
+ #include "llvm/Constants.h"
+ #include "llvm/DerivedTypes.h"
+ #include "llvm/Instructions.h"
+ #include "llvm/Module.h"
+ #include "llvm/Pass.h"
+ #include "llvm/Transforms/Utils/Cloning.h"
+ using namespace llvm;
+ 
+ namespace {
+   class LowerGC : public FunctionPass {
+     /// GCRootInt, GCReadInt, GCWriteInt - The function prototypes for the
+     /// llvm.gcread/llvm.gcwrite/llvm.gcroot intrinsics.
+     Function *GCRootInt, *GCReadInt, *GCWriteInt;
+ 
+     /// GCRead/GCWrite - These are the functions provided by the garbage
+     /// collector for read/write barriers.
+     Function *GCRead, *GCWrite;
+ 
+     /// RootChain - This is the global linked-list that contains the chain of GC
+     /// roots.
+     GlobalVariable *RootChain;
+ 
+     /// MainRootRecordType - This is the type for a function root entry if it
+     /// had zero roots.
+     const Type *MainRootRecordType;
+   public:
+     LowerGC() : GCRootInt(0), GCReadInt(0), GCWriteInt(0), 
+                 GCRead(0), GCWrite(0), RootChain(0), MainRootRecordType(0) {}
+     virtual bool doInitialization(Module &M);
+     virtual bool runOnFunction(Function &F);
+ 
+   private:
+     const StructType *getRootRecordType(unsigned NumRoots);
+   };
+ 
+   RegisterOpt<LowerGC>
+   X("lowergc", "Lower GC intrinsics, for GCless code generators");
+ }
+ 
+ /// createLowerGCPass - This function returns an instance of the "lowergc"
+ /// pass, which lowers garbage collection intrinsics to normal LLVM code.
+ FunctionPass *llvm::createLowerGCPass() {
+   return new LowerGC();
+ }
+ 
+ /// getRootRecordType - This function creates and returns the type for a root
+ /// record containing 'NumRoots' roots.
+ const StructType *LowerGC::getRootRecordType(unsigned NumRoots) {
+   // Build a struct that is a type used for meta-data/root pairs.
+   std::vector<const Type *> ST;
+   ST.push_back(GCRootInt->getFunctionType()->getParamType(0));
+   ST.push_back(GCRootInt->getFunctionType()->getParamType(1));
+   StructType *PairTy = StructType::get(ST);
+ 
+   // Build the array of pairs.
+   ArrayType *PairArrTy = ArrayType::get(PairTy, NumRoots);
+ 
+   // Now build the recursive list type.
+   PATypeHolder RootListH =
+     MainRootRecordType ? (Type*)MainRootRecordType : (Type*)OpaqueType::get();
+   ST.clear();
+   ST.push_back(PointerType::get(RootListH));         // Prev pointer
+   ST.push_back(Type::UIntTy);                        // NumElements in array
+   ST.push_back(PairArrTy);                           // The pairs
+   StructType *RootList = StructType::get(ST);
+   if (MainRootRecordType)
+     return RootList;
+ 
+   assert(NumRoots == 0 && "The main struct type should have zero entries!");
+   cast<OpaqueType>((Type*)RootListH.get())->refineAbstractTypeTo(RootList);
+   MainRootRecordType = RootListH;
+   return cast<StructType>(RootListH.get());
+ }
+ 
+ /// doInitialization - If this module uses the GC intrinsics, find them now.  If
+ /// not, this pass does not do anything.
+ bool LowerGC::doInitialization(Module &M) {
+   GCRootInt  = M.getNamedFunction("llvm.gcroot");
+   GCReadInt  = M.getNamedFunction("llvm.gcread");
+   GCWriteInt = M.getNamedFunction("llvm.gcwrite");
+   if (!GCRootInt && !GCReadInt && !GCWriteInt) return false;
+ 
+   PointerType *VoidPtr = PointerType::get(Type::SByteTy);
+   PointerType *VoidPtrPtr = PointerType::get(VoidPtr);
+ 
+   // If the program is using read/write barriers, find the implementations of
+   // them from the GC runtime library.
+   if (GCReadInt)        // Make:  sbyte* %llvm_gc_read(sbyte**)
+     GCRead = M.getOrInsertFunction("llvm_gc_read", VoidPtr, VoidPtrPtr, 0);
+   if (GCWriteInt)       // Make:  void %llvm_gc_write(sbyte*, sbyte**)
+     GCWrite = M.getOrInsertFunction("llvm_gc_write", Type::VoidTy,
+                                     VoidPtr, VoidPtrPtr, 0);
+ 
+   // If the program has GC roots, get or create the global root list.
+   if (GCRootInt) {
+     const StructType *RootListTy = getRootRecordType(0);
+     const Type *PRLTy = PointerType::get(RootListTy);
+     M.addTypeName("llvm_gc_root_ty", RootListTy);
+ 
+     // Get the root chain if it already exists.
+     RootChain = M.getGlobalVariable("llvm_gc_root_chain", PRLTy);
+     if (RootChain == 0) {
+       // If the root chain does not exist, insert a new one with linkonce
+       // linkage!
+       RootChain = new GlobalVariable(PRLTy, false, GlobalValue::LinkOnceLinkage,
+                                      Constant::getNullValue(RootListTy),
+                                      "llvm_gc_root_chain", &M);
+     } else if (RootChain->hasExternalLinkage() && RootChain->isExternal()) {
+       RootChain->setInitializer(Constant::getNullValue(PRLTy));
+       RootChain->setLinkage(GlobalValue::LinkOnceLinkage);
+     }
+   }
+   return true;
+ }
+ 
+ /// Coerce - If the specified operand number of the specified instruction does
+ /// not have the specified type, insert a cast.
+ static void Coerce(Instruction *I, unsigned OpNum, Type *Ty) {
+   if (I->getOperand(OpNum)->getType() != Ty) {
+     if (Constant *C = dyn_cast<Constant>(I->getOperand(OpNum)))
+       I->setOperand(OpNum, ConstantExpr::getCast(C, Ty));
+     else {
+       CastInst *C = new CastInst(I->getOperand(OpNum), Ty, "", I);
+       I->setOperand(OpNum, C);
+     }
+   }
+ }
+ 
+ /// runOnFunction - If the program is using GC intrinsics, replace any
+ /// read/write intrinsics with the appropriate read/write barrier calls, then
+ /// inline them.  Finally, build the data structures for 
+ bool LowerGC::runOnFunction(Function &F) {
+   // Quick exit for programs that are not using GC mechanisms.
+   if (!GCRootInt && !GCReadInt && !GCWriteInt) return false;
+ 
+   PointerType *VoidPtr    = PointerType::get(Type::SByteTy);
+   PointerType *VoidPtrPtr = PointerType::get(VoidPtr);
+ 
+   // If there are read/write barriers in the program, perform a quick pass over
+   // the function eliminating them.  While we are at it, remember where we see
+   // calls to llvm.gcroot.
+   std::vector<CallInst*> GCRoots;
+   std::vector<CallInst*> NormalCalls;
+ 
+   bool MadeChange = false;
+   for (Function::iterator BB = F.begin(), E = F.end(); BB != E; ++BB)
+     for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E;)
+       if (CallInst *CI = dyn_cast<CallInst>(II++)) {
+         if (!CI->getCalledFunction() ||
+             !CI->getCalledFunction()->getIntrinsicID())
+           NormalCalls.push_back(CI);   // Remember all normal function calls.
+ 
+         if (Function *F = CI->getCalledFunction())
+           if (F == GCRootInt)
+             GCRoots.push_back(CI);
+           else if (F == GCReadInt || F == GCWriteInt) {
+             if (F == GCWriteInt) {
+               // Change a llvm.gcwrite call to call llvm_gc_write instead.
+               CI->setOperand(0, GCWrite);
+               // Insert casts of the operands as needed.
+               Coerce(CI, 1, VoidPtr);
+               Coerce(CI, 2, VoidPtrPtr);
+             } else {
+               Coerce(CI, 1, VoidPtrPtr);
+               if (CI->getType() == VoidPtr) {
+                 CI->setOperand(0, GCRead);
+               } else {
+                 // Create a whole new call to replace the old one.
+                 CallInst *NC = new CallInst(GCRead, CI->getOperand(1),
+                                             CI->getName(), CI);
+                 Value *NV = new CastInst(NC, CI->getType(), "", CI);
+                 CI->replaceAllUsesWith(NV);
+                 BB->getInstList().erase(CI);
+                 CI = NC;
+               }
+             }
+ 
+             // Now that we made the replacement, inline expand the call if
+             // possible, otherwise things will be too horribly expensive.
+             InlineFunction(CI);
+             MadeChange = true;
+           }
+       }
+   
+   // If there are no GC roots in this function, then there is no need to create
+   // a GC list record for it.
+   if (GCRoots.empty()) return MadeChange;
+ 
+   // Okay, there are GC roots in this function.  On entry to the function, add a
+   // record to the llvm_gc_root_chain, and remove it on exit.
+ 
+   // Create the alloca, and zero it out.
+   const StructType *RootListTy = getRootRecordType(GCRoots.size());
+   AllocaInst *AI = new AllocaInst(RootListTy, 0, "gcroots", F.begin()->begin());
+ 
+   // Insert the memset call after all of the allocas in the function.
+   BasicBlock::iterator IP = AI;
+   while (isa<AllocaInst>(IP)) ++IP;
+ 
+   Constant *Zero = ConstantUInt::get(Type::UIntTy, 0);
+   Constant *One  = ConstantUInt::get(Type::UIntTy, 1);
+ 
+   // Get a pointer to the prev pointer.
+   std::vector<Value*> Par;
+   Par.push_back(Zero);
+   Par.push_back(Zero);
+   Value *PrevPtrPtr = new GetElementPtrInst(AI, Par, "prevptrptr", IP);
+ 
+   // Load the previous pointer.
+   Value *PrevPtr = new LoadInst(RootChain, "prevptr", IP);
+   // Store the previous pointer into the prevptrptr
+   new StoreInst(PrevPtr, PrevPtrPtr, IP);
+ 
+   // Set the number of elements in this record.
+   Par[1] = ConstantUInt::get(Type::UIntTy, 1);
+   Value *NumEltsPtr = new GetElementPtrInst(AI, Par, "numeltsptr", IP);
+   new StoreInst(ConstantUInt::get(Type::UIntTy, GCRoots.size()), NumEltsPtr,IP);
+ 
+   Par[1] = ConstantUInt::get(Type::UIntTy, 2);
+   Par.resize(4);
+ 
+   const PointerType *PtrLocTy =
+     cast<PointerType>(GCRootInt->getFunctionType()->getParamType(0));
+   Constant *Null = ConstantPointerNull::get(PtrLocTy);
+ 
+   // Initialize all of the gcroot records now, and eliminate them as we go.
+   for (unsigned i = 0, e = GCRoots.size(); i != e; ++i) {
+     // Initialize the meta-data pointer.
+     Par[2] = ConstantUInt::get(Type::UIntTy, i);
+     Par[3] = One;
+     Value *MetaDataPtr = new GetElementPtrInst(AI, Par, "MetaDataPtr", IP);
+     assert(isa<Constant>(GCRoots[i]->getOperand(2)) || 
+            isa<GlobalValue>(GCRoots[i]->getOperand(2)));
+     new StoreInst(GCRoots[i]->getOperand(2), MetaDataPtr, IP);
+ 
+     // Initialize the root pointer to null on entry to the function.
+     Par[3] = Zero;
+     Value *RootPtrPtr = new GetElementPtrInst(AI, Par, "RootEntPtr", IP);
+     new StoreInst(Null, RootPtrPtr, IP);
+     
+     // Each occurrance of the llvm.gcroot intrinsic now turns into an
+     // initialization of the slot with the address and a zeroing out of the
+     // address specified.
+     new StoreInst(Constant::getNullValue(PtrLocTy->getElementType()),
+                   GCRoots[i]->getOperand(1), GCRoots[i]);
+     new StoreInst(GCRoots[i]->getOperand(1), RootPtrPtr, GCRoots[i]);
+     GCRoots[i]->getParent()->getInstList().erase(GCRoots[i]);
+   }
+ 
+   // Now that the record is all initialized, store the pointer into the global
+   // pointer.
+   Value *C = new CastInst(AI, PointerType::get(MainRootRecordType), "", IP);
+   new StoreInst(C, RootChain, IP);
+ 
+   // On exit from the function we have to remove the entry from the GC root
+   // chain.  Doing this is straight-forward for return and unwind instructions:
+   // just insert the appropriate copy.
+   for (Function::iterator BB = F.begin(), E = F.end(); BB != E; ++BB)
+     if (isa<UnwindInst>(BB->getTerminator()) ||
+         isa<ReturnInst>(BB->getTerminator())) {
+       // We could reuse the PrevPtr loaded on entry to the function, but this
+       // would make the value live for the whole function, which is probably a
+       // bad idea.  Just reload the value out of our stack entry.
+       PrevPtr = new LoadInst(PrevPtrPtr, "prevptr", BB->getTerminator());
+       new StoreInst(PrevPtr, RootChain, BB->getTerminator());
+     }
+ 
+   // If an exception is thrown from a callee we have to make sure to
+   // unconditionally take the record off the stack.  For this reason, we turn
+   // all call instructions into invoke whose cleanup pops the entry off the
+   // stack.  We only insert one cleanup block, which is shared by all invokes.
+   if (!NormalCalls.empty()) {
+     // Create the shared cleanup block.
+     BasicBlock *Cleanup = new BasicBlock("gc_cleanup", &F);
+     UnwindInst *UI = new UnwindInst(Cleanup);
+     PrevPtr = new LoadInst(PrevPtrPtr, "prevptr", UI);
+     new StoreInst(PrevPtr, RootChain, UI);
+   
+     // Loop over all of the function calls, turning them into invokes.
+     while (!NormalCalls.empty()) {
+       CallInst *CI = NormalCalls.back();
+       BasicBlock *CBB = CI->getParent();
+       NormalCalls.pop_back();
+ 
+       // Split the basic block containing the function call.
+       BasicBlock *NewBB = CBB->splitBasicBlock(CI, CBB->getName()+".cont");
+ 
+       // Remove the unconditional branch inserted at the end of the CBB.
+       CBB->getInstList().pop_back();
+       NewBB->getInstList().remove(CI);
+       
+       // Create a new invoke instruction.
+       Value *II = new InvokeInst(CI->getCalledValue(), NewBB, Cleanup,
+                                  std::vector<Value*>(CI->op_begin()+1,
+                                                      CI->op_end()),
+                                  CI->getName(), CBB);
+       CI->replaceAllUsesWith(II);
+       delete CI;
+     }
+   }
+ 
+   return true;
+ }




From lattner at cs.uiuc.edu  Sun May 23 16:22:01 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:22:01 2004
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Scalar/LICM.cpp
Message-ID: <200405232120.QAA04134@zion.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Scalar:

LICM.cpp updated: 1.59 -> 1.60

---
Log message:

Adjust to the changes in the AliasSetTracker interface


---
Diffs of the changes:  (+14 -16)

Index: llvm/lib/Transforms/Scalar/LICM.cpp
diff -u llvm/lib/Transforms/Scalar/LICM.cpp:1.59 llvm/lib/Transforms/Scalar/LICM.cpp:1.60
--- llvm/lib/Transforms/Scalar/LICM.cpp:1.59	Sun Apr 18 17:46:08 2004
+++ llvm/lib/Transforms/Scalar/LICM.cpp	Sun May 23 16:20:19 2004
@@ -15,9 +15,9 @@
 //
 // This pass uses alias analysis for two purposes:
 //
-//  1. Moving loop invariant loads out of loops.  If we can determine that a
-//     load inside of a loop never aliases anything stored to, we can hoist it
-//     or sink it like any other instruction.
+//  1. Moving loop invariant loads and calls out of loops.  If we can determine
+//     that a load or call inside of a loop never aliases anything stored to,
+//     we can hoist it or sink it like any other instruction.
 //  2. Scalar Promotion of Memory - If there is a store instruction inside of
 //     the loop, we try to move the store to happen AFTER the loop instead of
 //     inside of the loop.  This can only happen if a few conditions are true:
@@ -32,20 +32,19 @@
 //===----------------------------------------------------------------------===//
 
 #include "llvm/Transforms/Scalar.h"
-#include "llvm/Transforms/Utils/PromoteMemToReg.h"
-#include "llvm/Transforms/Utils/Local.h"
+#include "llvm/DerivedTypes.h"
+#include "llvm/Instructions.h"
+#include "llvm/Target/TargetData.h"
 #include "llvm/Analysis/LoopInfo.h"
 #include "llvm/Analysis/AliasAnalysis.h"
 #include "llvm/Analysis/AliasSetTracker.h"
 #include "llvm/Analysis/Dominators.h"
-#include "llvm/Instructions.h"
-#include "llvm/DerivedTypes.h"
-#include "llvm/Target/TargetData.h"
 #include "llvm/Support/CFG.h"
+#include "llvm/Transforms/Utils/PromoteMemToReg.h"
+#include "llvm/Transforms/Utils/Local.h"
 #include "Support/CommandLine.h"
 #include "Support/Debug.h"
 #include "Support/Statistic.h"
-#include "llvm/Assembly/Writer.h"
 #include <algorithm>
 using namespace llvm;
 
@@ -136,7 +135,7 @@
       DominatorTree::Node *IDom            = DT->getNode(ExitBlock);
     
       // Because the exit block is not in the loop, we know we have to get _at
-      // least_ it's immediate dominator.
+      // least_ its immediate dominator.
       do {
         // Get next Immediate Dominator.
         IDom = IDom->getIDom();
@@ -442,7 +441,7 @@
   if (ExitBlocks.size() == 1) {
     if (!isExitBlockDominatedByBlockInLoop(ExitBlocks[0], I.getParent())) {
       // Instruction is not used, just delete it.
-      CurAST->remove(&I);
+      CurAST->deleteValue(&I);
       I.getParent()->getInstList().erase(&I);
     } else {
       // Move the instruction to the start of the exit block, after any PHI
@@ -455,7 +454,7 @@
     }
   } else if (ExitBlocks.size() == 0) {
     // The instruction is actually dead if there ARE NO exit blocks.
-    CurAST->remove(&I);
+    CurAST->deleteValue(&I);
     I.getParent()->getInstList().erase(&I);
   } else {
     // Otherwise, if we have multiple exits, use the PromoteMem2Reg function to
@@ -535,7 +534,7 @@
 
     // If the instruction doesn't dominate any exit blocks, it must be dead.
     if (InsertedBlocks.empty()) {
-      CurAST->remove(&I);
+      CurAST->deleteValue(&I);
       I.getParent()->getInstList().erase(&I);
     }
       
@@ -550,9 +549,8 @@
 /// that is safe to hoist, this instruction is called to do the dirty work.
 ///
 void LICM::hoist(Instruction &I) {
-  DEBUG(std::cerr << "LICM hoisting to";
-        WriteAsOperand(std::cerr, Preheader, false);
-        std::cerr << ": " << I);
+  DEBUG(std::cerr << "LICM hoisting to" << Preheader->getName() 
+                  << ": " << I);
 
   // Remove the instruction from its current basic block... but don't delete the
   // instruction.




From lattner at cs.uiuc.edu  Sun May 23 16:22:14 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:22:14 2004
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Scalar/GCSE.cpp
Message-ID: <200405232120.QAA04121@zion.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Scalar:

GCSE.cpp updated: 1.39 -> 1.40

---
Log message:

Add support for replacement of formal arguments with simpler expressions.


---
Diffs of the changes:  (+23 -2)

Index: llvm/lib/Transforms/Scalar/GCSE.cpp
diff -u llvm/lib/Transforms/Scalar/GCSE.cpp:1.39 llvm/lib/Transforms/Scalar/GCSE.cpp:1.40
--- llvm/lib/Transforms/Scalar/GCSE.cpp:1.39	Mon Apr 12 00:15:13 2004
+++ llvm/lib/Transforms/Scalar/GCSE.cpp	Sun May 23 16:19:55 2004
@@ -1,4 +1,4 @@
-//===-- GCSE.cpp - SSA based Global Common Subexpr Elimination ------------===//
+//===-- GCSE.cpp - SSA-based Global Common Subexpression Elimination ------===//
 // 
 //                     The LLVM Compiler Infrastructure
 //
@@ -33,6 +33,8 @@
   Statistic<> NumCallRemoved("gcse", "Number of calls removed");
   Statistic<> NumNonInsts   ("gcse", "Number of instructions removed due "
                              "to non-instruction values");
+  Statistic<> NumArgsRepl   ("gcse", "Number of function arguments replaced "
+                             "with constant values");
 
   struct GCSE : public FunctionPass {
     virtual bool runOnFunction(Function &F);
@@ -68,6 +70,25 @@
 
   std::vector<Value*> EqualValues;
 
+  // Check for value numbers of arguments.  If the value numbering
+  // implementation can prove that an incoming argument is a constant or global
+  // value address, substitute it, making the argument dead.
+  for (Function::aiterator AI = F.abegin(), E = F.aend(); AI != E; ++AI)
+    if (!AI->use_empty()) {
+      VN.getEqualNumberNodes(AI, EqualValues);
+      if (!EqualValues.empty()) {
+        for (unsigned i = 0, e = EqualValues.size(); i != e; ++i)
+          if (isa<Constant>(EqualValues[i]) ||
+              isa<GlobalValue>(EqualValues[i])) {
+            AI->replaceAllUsesWith(EqualValues[i]);
+            ++NumArgsRepl;
+            Changed = true;
+            break;
+          }
+        EqualValues.clear();
+      }
+    }
+
   // Traverse the CFG of the function in dominator order, so that we see each
   // instruction after we see its operands.
   for (df_iterator<DominatorTree::Node*> DI = df_begin(DT.getRootNode()),
@@ -161,7 +182,7 @@
   ++NumInstRemoved;   // Keep track of number of insts eliminated
 
   // Update value numbering
-  getAnalysis<ValueNumbering>().deleteInstruction(I);
+  getAnalysis<ValueNumbering>().deleteValue(I);
 
   // If we are not replacing the instruction with a constant, we cannot do
   // anything special.




From lattner at cs.uiuc.edu  Sun May 23 16:24:01 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:24:01 2004
Subject: [llvm-commits] CVS: llvm/lib/Transforms/IPO/Inliner.h
Message-ID: <200405232122.QAA04180@zion.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/IPO:

Inliner.h updated: 1.6 -> 1.7

---
Log message:

Simplify the interface and remove an unneeded #include


---
Diffs of the changes:  (+1 -9)

Index: llvm/lib/Transforms/IPO/Inliner.h
diff -u llvm/lib/Transforms/IPO/Inliner.h:1.6 llvm/lib/Transforms/IPO/Inliner.h:1.7
--- llvm/lib/Transforms/IPO/Inliner.h:1.6	Tue Apr 20 17:06:53 2004
+++ llvm/lib/Transforms/IPO/Inliner.h	Sun May 23 16:21:35 2004
@@ -19,11 +19,9 @@
 
 #define DEBUG_TYPE "inline"
 #include "llvm/CallGraphSCCPass.h"
-#include <set>
 
 namespace llvm {
-
-class CallSite;
+  class CallSite;
 
 /// Inliner - This class contains all of the helper code which is used to
 /// perform the inlining operations that does not depend on the policy.
@@ -51,12 +49,6 @@
   /// not inlined.
   ///
   virtual int getInlineCost(CallSite CS) = 0;
-  
-  /// getRecursiveInlineCost - This method can be implemented by subclasses if
-  /// it wants to treat calls to functions within the current SCC specially.  If
-  /// this method is not overloaded, it just chains to getInlineCost().
-  ///
-  virtual int getRecursiveInlineCost(CallSite CS);
 
 private:
   // InlineThreshold - Cache the value here for easy access.




From lattner at cs.uiuc.edu  Sun May 23 16:24:10 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:24:10 2004
Subject: [llvm-commits] CVS: llvm/lib/Transforms/IPO/ArgumentPromotion.cpp
Message-ID: <200405232122.QAA04172@zion.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/IPO:

ArgumentPromotion.cpp updated: 1.5 -> 1.6

---
Log message:

Fairly substantial changes to update the alias analysis we are querying as
we make the transformation.  This allows us to use interprocedural alias
analyses successfully.


---
Diffs of the changes:  (+92 -39)

Index: llvm/lib/Transforms/IPO/ArgumentPromotion.cpp
diff -u llvm/lib/Transforms/IPO/ArgumentPromotion.cpp:1.5 llvm/lib/Transforms/IPO/ArgumentPromotion.cpp:1.6
--- llvm/lib/Transforms/IPO/ArgumentPromotion.cpp:1.5	Sun Mar  7 19:04:36 2004
+++ llvm/lib/Transforms/IPO/ArgumentPromotion.cpp	Sun May 23 16:21:17 2004
@@ -1,4 +1,4 @@
-//===-- ArgumentPromotion.cpp - Promote 'by reference' arguments ----------===//
+//===-- ArgumentPromotion.cpp - Promote by-reference arguments ------------===//
 // 
 //                     The LLVM Compiler Infrastructure
 //
@@ -12,14 +12,14 @@
 // arguments.  If we can prove, through the use of alias analysis, that that an
 // argument is *only* loaded, then we can pass the value into the function
 // instead of the address of the value.  This can cause recursive simplification
-// of code, and lead to the elimination of allocas, especially in C++ template
-// code like the STL.
+// of code and lead to the elimination of allocas (especially in C++ template
+// code like the STL).
 //
 // This pass also handles aggregate arguments that are passed into a function,
 // scalarizing them if the elements of the aggregate are only loaded.  Note that
 // we refuse to scalarize aggregates which would require passing in more than
 // three operands to the function, because we don't want to pass thousands of
-// operands for a large array or something!
+// operands for a large array or structure!
 //
 // Note that this transformation could also be done for arguments that are only
 // stored to (returning the value instead), but we do not currently handle that
@@ -28,6 +28,7 @@
 //
 //===----------------------------------------------------------------------===//
 
+#define DEBUG_TYPE "argpromotion"
 #include "llvm/Transforms/IPO.h"
 #include "llvm/Constants.h"
 #include "llvm/DerivedTypes.h"
@@ -58,7 +59,7 @@
   class ArgPromotion : public Pass {
     // WorkList - The set of internal functions that we have yet to process.  As
     // we eliminate arguments from a function, we push all callers into this set
-    // so that the by reference argument can be bubbled out as far as possible.
+    // so that the by-reference argument can be bubbled out as far as possible.
     // This set contains only internal functions.
     std::set<Function*> WorkList;
   public:
@@ -128,7 +129,11 @@
   return Changed;
 }
 
-
+/// PromoteArguments - This method checks the specified function to see if there
+/// are any promotable arguments and if it is safe to promote the function (for
+/// example, all callers are direct).  If safe to promote some arguments, it
+/// calls the DoPromotion method.
+///
 bool ArgPromotion::PromoteArguments(Function *F) {
   assert(F->hasInternalLinkage() && "We can only process internal functions!");
 
@@ -144,15 +149,14 @@
   for (Value::use_iterator UI = F->use_begin(), E = F->use_end();
        UI != E; ++UI) {
     CallSite CS = CallSite::get(*UI);
-    if (Instruction *I = CS.getInstruction()) {
-      // Ensure that this call site is CALLING the function, not passing it as
-      // an argument.
-      for (CallSite::arg_iterator AI = CS.arg_begin(), E = CS.arg_end();
-           AI != E; ++AI)
-        if (*AI == F) return false;   // Passing the function address in!
-    } else {
-      return false;  // Cannot promote an indirect call!
-    }
+    if (!CS.getInstruction())       // "Taking the address" of the function
+      return false;
+
+    // Ensure that this call site is CALLING the function, not passing it as
+    // an argument.
+    for (CallSite::arg_iterator AI = CS.arg_begin(), E = CS.arg_end();
+         AI != E; ++AI)
+      if (*AI == F) return false;   // Passing the function address in!
   }
 
   // Check to see which arguments are promotable.  If an argument is not
@@ -171,6 +175,12 @@
   return true;
 }
 
+
+/// isSafeToPromoteArgument - As you might guess from the name of this method,
+/// it checks to see if it is both safe and useful to promote the argument.
+/// This method limits promotion of aggregates to only promote up to three
+/// elements of the aggregate in order to avoid exploding the number of
+/// arguments passed in.
 bool ArgPromotion::isSafeToPromoteArgument(Argument *Arg) const {
   // We can only promote this argument if all of the uses are loads, or are GEP
   // instructions (with constant indices) that are subsequently loaded.
@@ -185,6 +195,7 @@
       if (GEP->use_empty()) {
         // Dead GEP's cause trouble later.  Just remove them if we run into
         // them.
+        getAnalysis<AliasAnalysis>().deleteValue(GEP);
         GEP->getParent()->getInstList().erase(GEP);
         return isSafeToPromoteArgument(Arg);
       }
@@ -206,12 +217,15 @@
           return false;
         }
 
-      // See if there is already a GEP with these indices.  If so, check to make
-      // sure that we aren't promoting too many elements.  If not, nothing to
-      // do.
+      // See if there is already a GEP with these indices.  If not, check to
+      // make sure that we aren't promoting too many elements.  If so, nothing
+      // to do.
       if (std::find(GEPIndices.begin(), GEPIndices.end(), Operands) ==
           GEPIndices.end()) {
         if (GEPIndices.size() == 3) {
+          DEBUG(std::cerr << "argpromotion disable promoting argument '"
+                << Arg->getName() << "' because it would require adding more "
+                << "than 3 arguments to the function.\n");
           // We limit aggregate promotion to only promoting up to three elements
           // of the aggregate.
           return false;
@@ -222,11 +236,11 @@
       return false;  // Not a load or a GEP.
     }
 
-  if (Loads.empty()) return true;  // No users, dead argument.
+  if (Loads.empty()) return true;  // No users, this is a dead argument.
 
-  // Okay, now we know that the argument is only used by load instructions.
-  // Check to see if the pointer is guaranteed to not be modified from entry of
-  // the function to each of the load instructions.
+  // Okay, now we know that the argument is only used by load instructions.  Use
+  // alias analysis to check to see if the pointer is guaranteed to not be
+  // modified from entry of the function to each of the load instructions.
   Function &F = *Arg->getParent();
 
   // Because there could be several/many load instructions, remember which
@@ -265,7 +279,8 @@
   return true;
 }
 
-
+/// DoPromotion - This method actually performs the promotion of the specified
+/// arguments.  At this point, we know that it's safe to do so.
 void ArgPromotion::DoPromotion(Function *F, std::vector<Argument*> &Args2Prom) {
   std::set<Argument*> ArgsToPromote(Args2Prom.begin(), Args2Prom.end());
   
@@ -283,10 +298,17 @@
   //
   std::map<Argument*, std::set<std::vector<Value*> > > ScalarizedElements;
 
+  // OriginalLoads - Keep track of a representative load instruction from the
+  // original function so that we can tell the alias analysis implementation
+  // what the new GEP/Load instructions we are inserting look like.
+  std::map<std::vector<Value*>, LoadInst*> OriginalLoads;
+
   for (Function::aiterator I = F->abegin(), E = F->aend(); I != E; ++I)
     if (!ArgsToPromote.count(I)) {
       Params.push_back(I->getType());
-    } else if (!I->use_empty()) {
+    } else if (I->use_empty()) {
+      ++NumArgumentsDead;
+    } else {
       // Okay, this is being promoted.  Check to see if there are any GEP uses
       // of the argument.
       std::set<std::vector<Value*> > &ArgIndices = ScalarizedElements[I];
@@ -294,8 +316,14 @@
            ++UI) {
         Instruction *User = cast<Instruction>(*UI);
         assert(isa<LoadInst>(User) || isa<GetElementPtrInst>(User));
-        ArgIndices.insert(std::vector<Value*>(User->op_begin()+1,
-                                              User->op_end()));
+        std::vector<Value*> Indices(User->op_begin()+1, User->op_end());
+        ArgIndices.insert(Indices);
+        LoadInst *OrigLoad;
+        if (LoadInst *L = dyn_cast<LoadInst>(User))
+          OrigLoad = L;
+        else
+          OrigLoad = cast<LoadInst>(User->use_back());
+        OriginalLoads[Indices] = OrigLoad;
       }
 
       // Add a parameter to the function for each element passed in.
@@ -307,8 +335,6 @@
         ++NumArgumentsPromoted;
       else
         ++NumAggregatesPromoted;
-    } else {
-      ++NumArgumentsDead;
     }
 
   const Type *RetTy = FTy->getReturnType();
@@ -325,7 +351,11 @@
    // Create the new function body and insert it into the module...
   Function *NF = new Function(NFTy, F->getLinkage(), F->getName());
   F->getParent()->getFunctionList().insert(F, NF);
-  
+
+  // Get the alias analysis information that we need to update to reflect our
+  // changes.
+  AliasAnalysis &AA = getAnalysis<AliasAnalysis>();
+
   // Loop over all of the callers of the function, transforming the call sites
   // to pass in the loaded pointers.
   //
@@ -334,25 +364,30 @@
     CallSite CS = CallSite::get(F->use_back());
     Instruction *Call = CS.getInstruction();
 
-    // Make sure the caller of this function is revisited.
+    // Make sure the caller of this function is revisited now that we promoted
+    // arguments in a callee of it.
     if (Call->getParent()->getParent()->hasInternalLinkage())
       WorkList.insert(Call->getParent()->getParent());
     
-    // Loop over the operands, deleting dead ones...
+    // Loop over the operands, inserting GEP and loads in the caller as
+    // appropriate.
     CallSite::arg_iterator AI = CS.arg_begin();
     for (Function::aiterator I = F->abegin(), E = F->aend(); I != E; ++I, ++AI)
       if (!ArgsToPromote.count(I))
         Args.push_back(*AI);          // Unmodified argument
       else if (!I->use_empty()) {
-        // Non-dead argument.
+        // Non-dead argument: insert GEPs and loads as appropriate.
         std::set<std::vector<Value*> > &ArgIndices = ScalarizedElements[I];
         for (std::set<std::vector<Value*> >::iterator SI = ArgIndices.begin(),
                E = ArgIndices.end(); SI != E; ++SI) {
           Value *V = *AI;
-          if (!SI->empty())
+          LoadInst *OrigLoad = OriginalLoads[*SI];
+          if (!SI->empty()) {
             V = new GetElementPtrInst(V, *SI, V->getName()+".idx", Call);
-
+            AA.copyValue(OrigLoad->getOperand(0), V);
+          }
           Args.push_back(new LoadInst(V, V->getName()+".val", Call));
+          AA.copyValue(OrigLoad, Args.back());
         }
       }
 
@@ -372,6 +407,10 @@
     }
     Args.clear();
 
+    // Update the alias analysis implementation to know that we are replacing
+    // the old call with a new one.
+    AA.replaceWithNewValue(Call, New);
+
     if (!Call->use_empty()) {
       Call->replaceAllUsesWith(New);
       std::string Name = Call->getName();
@@ -399,8 +438,11 @@
       // new version.
       I->replaceAllUsesWith(I2);
       I2->setName(I->getName());
+      AA.replaceWithNewValue(I, I2);
       ++I2;
-    } else if (!I->use_empty()) {
+    } else if (I->use_empty()) {
+      AA.deleteValue(I);
+    } else {
       // Otherwise, if we promoted this argument, then all users are load
       // instructions, and all loads should be using the new argument that we
       // added.
@@ -412,9 +454,10 @@
                  "Load element should sort to front!");
           I2->setName(I->getName()+".val");
           LI->replaceAllUsesWith(I2);
+          AA.replaceWithNewValue(LI, I2);
           LI->getParent()->getInstList().erase(LI);
-          DEBUG(std::cerr << "*** Promoted argument '" << I->getName()
-                          << "' of function '" << F->getName() << "'\n");
+          DEBUG(std::cerr << "*** Promoted load of argument '" << I->getName()
+                          << "' in function '" << F->getName() << "'\n");
         } else {
           GetElementPtrInst *GEP = cast<GetElementPtrInst>(I->use_back());
           std::vector<Value*> Operands(GEP->op_begin()+1, GEP->op_end());
@@ -442,19 +485,29 @@
           while (!GEP->use_empty()) {
             LoadInst *L = cast<LoadInst>(GEP->use_back());
             L->replaceAllUsesWith(TheArg);
+            AA.replaceWithNewValue(L, TheArg);
             L->getParent()->getInstList().erase(L);
           }
+          AA.deleteValue(GEP);
           GEP->getParent()->getInstList().erase(GEP);
         }
       }
 
       // If we inserted a new pointer type, it's possible that IT could be
-      // promoted too.  Also, increment I2 past all of the arguments for this
-      // pointer.
+      // promoted too.  Also, increment I2 past all of the arguments added for
+      // this promoted pointer.
       for (unsigned i = 0, e = ArgIndices.size(); i != e; ++i, ++I2)
         if (isa<PointerType>(I2->getType()))
           WorkList.insert(NF);
     }
+
+  // Notify the alias analysis implementation that we inserted a new argument.
+  if (ExtraArgHack)
+    AA.copyValue(Constant::getNullValue(Type::IntTy), NF->abegin());
+
+
+  // Tell the alias analysis that the old function is about to disappear.
+  AA.replaceWithNewValue(F, NF);
 
   // Now that the old function is dead, delete it.
   F->getParent()->getFunctionList().erase(F);




From lattner at cs.uiuc.edu  Sun May 23 16:24:20 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:24:20 2004
Subject: [llvm-commits] CVS: llvm/lib/Transforms/IPO/Inliner.cpp
Message-ID: <200405232122.QAA04167@zion.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/IPO:

Inliner.cpp updated: 1.13 -> 1.14

---
Log message:

Fix cases where we missed inlining some more obvious candidates because the
caller was in an SCC.


---
Diffs of the changes:  (+108 -79)

Index: llvm/lib/Transforms/IPO/Inliner.cpp
diff -u llvm/lib/Transforms/IPO/Inliner.cpp:1.13 llvm/lib/Transforms/IPO/Inliner.cpp:1.14
--- llvm/lib/Transforms/IPO/Inliner.cpp:1.13	Wed Apr 21 15:44:33 2004
+++ llvm/lib/Transforms/IPO/Inliner.cpp	Sun May 23 16:22:17 2004
@@ -7,8 +7,9 @@
 // 
 //===----------------------------------------------------------------------===//
 //
-// This file implements the code shared between the LLVM inliners.  This
-// implements all of the boring mechanics of the bottom-up inlining.
+// This file implements the mechanics required to implement inlining without
+// missing any calls and updating the call graph.  The decisions of which calls
+// are profitable to inline are implemented elsewhere.
 //
 //===----------------------------------------------------------------------===//
 
@@ -23,6 +24,7 @@
 #include "Support/CommandLine.h"
 #include "Support/Debug.h"
 #include "Support/Statistic.h"
+#include <set>
 using namespace llvm;
 
 namespace {
@@ -35,8 +37,41 @@
 
 Inliner::Inliner() : InlineThreshold(InlineLimit) {}
 
-int Inliner::getRecursiveInlineCost(CallSite CS) {
-  return getInlineCost(CS);
+// InlineCallIfPossible - If it is possible to inline the specified call site,
+// do so and update the CallGraph for this operation.
+static bool InlineCallIfPossible(CallSite CS, CallGraph &CG,
+                                 const std::set<Function*> &SCCFunctions) {
+  Function *Caller = CS.getInstruction()->getParent()->getParent();
+  Function *Callee = CS.getCalledFunction();
+  if (!InlineFunction(CS)) return false;
+
+  // Update the call graph by deleting the edge from Callee to Caller
+  CallGraphNode *CalleeNode = CG[Callee];
+  CallGraphNode *CallerNode = CG[Caller];
+  CallerNode->removeCallEdgeTo(CalleeNode);
+
+  // Since we inlined all uninlined call sites in the callee into the caller,
+  // add edges from the caller to all of the callees of the callee.
+  for (CallGraphNode::iterator I = CalleeNode->begin(),
+         E = CalleeNode->end(); I != E; ++I)
+    CallerNode->addCalledFunction(*I);
+  
+  // If we inlined the last possible call site to the function,
+  // delete the function body now.
+  if (Callee->use_empty() && Callee->hasInternalLinkage() &&
+      !SCCFunctions.count(Callee)) {
+    DEBUG(std::cerr << "    -> Deleting dead function: "
+                    << Callee->getName() << "\n");
+    
+    // Remove any call graph edges from the callee to its callees.
+    while (CalleeNode->begin() != CalleeNode->end())
+      CalleeNode->removeCallEdgeTo(*(CalleeNode->end()-1));
+              
+    // Removing the node for callee from the call graph and delete it.
+    delete CG.removeFunctionFromModule(CalleeNode);
+    ++NumDeleted;
+  }
+  return true;
 }
 
 bool Inliner::runOnSCC(const std::vector<CallGraphNode*> &SCC) {
@@ -45,88 +80,82 @@
   std::set<Function*> SCCFunctions;
   DEBUG(std::cerr << "Inliner visiting SCC:");
   for (unsigned i = 0, e = SCC.size(); i != e; ++i) {
-    SCCFunctions.insert(SCC[i]->getFunction());
-    DEBUG(std::cerr << " " << (SCC[i]->getFunction() ?
-              SCC[i]->getFunction()->getName() : "INDIRECTNODE"));
+    Function *F = SCC[i]->getFunction();
+    if (F) SCCFunctions.insert(F);
+    DEBUG(std::cerr << " " << (F ? F->getName() : "INDIRECTNODE"));
   }
-  DEBUG(std::cerr << "\n");
 
-  bool Changed = false;
+  // Scan through and identify all call sites ahead of time so that we only
+  // inline call sites in the original functions, not call sites that result
+  // from inlining other functions.
+  std::vector<CallSite> CallSites;
+
   for (std::set<Function*>::iterator SCCI = SCCFunctions.begin(),
-         E = SCCFunctions.end(); SCCI != E; ++SCCI) {
-    Function *F = *SCCI;
-    if (F == 0 || F->isExternal()) continue;
-    DEBUG(std::cerr << "  Inspecting function: " << F->getName() << "\n");
-
-    // Scan through and identify all call sites ahead of time so that we only
-    // inline call sites in the original functions, not call sites that result
-    // in inlining other functions.
-    std::vector<CallSite> CallSites;
-    for (Function::iterator BB = F->begin(), E = F->end(); BB != E; ++BB)
-      for (BasicBlock::iterator I = BB->begin(); I != BB->end(); ++I) {
-        CallSite CS = CallSite::get(I);
-        if (CS.getInstruction() && CS.getCalledFunction() &&
-            !CS.getCalledFunction()->isExternal())
-          CallSites.push_back(CS);
-      }
+         E = SCCFunctions.end(); SCCI != E; ++SCCI)
+    if (Function *F = *SCCI)
+      for (Function::iterator BB = F->begin(), E = F->end(); BB != E; ++BB)
+        for (BasicBlock::iterator I = BB->begin(); I != BB->end(); ++I) {
+          CallSite CS = CallSite::get(I);
+          if (CS.getInstruction() && (!CS.getCalledFunction() ||
+                                      !CS.getCalledFunction()->isExternal()))
+            CallSites.push_back(CS);
+        }
 
-    // Now that we have all of the call sites, loop over them and inline them if
-    // it looks profitable to do so.
-    for (unsigned i = 0, e = CallSites.size(); i != e; ++i) {
-      CallSite CS = CallSites[i];
-      Function *Callee = CS.getCalledFunction();
-      // Determine whether this is a function IN the SCC...
-      bool inSCC = SCCFunctions.count(Callee);
-    
-      // If the policy determines that we should inline this function,
-      // try to do so...
-      int InlineCost = inSCC ? getRecursiveInlineCost(CS) : getInlineCost(CS);
-      if (InlineCost >= (int)InlineThreshold) {
-        DEBUG(std::cerr << "    NOT Inlining: cost=" << InlineCost
-              << ", Call: " << *CS.getInstruction());
-      } else {
-        DEBUG(std::cerr << "    Inlining: cost=" << InlineCost
-              << ", Call: " << *CS.getInstruction());
-      
-        Function *Caller = CS.getInstruction()->getParent()->getParent();
-
-        // Attempt to inline the function...
-        if (InlineFunction(CS)) {
-          ++NumInlined;
-
-          // Update the call graph by deleting the edge from Callee to Caller
-          CallGraphNode *CalleeNode = CG[Callee];
-          CallGraphNode *CallerNode = CG[Caller];
-          CallerNode->removeCallEdgeTo(CalleeNode);
-
-          // Since we inlined all uninlinable call sites in the callee into the
-          // caller, add edges from the caller to all of the callees of the
-          // callee.
-          for (CallGraphNode::iterator I = CalleeNode->begin(),
-                 E = CalleeNode->end(); I != E; ++I)
-            CallerNode->addCalledFunction(*I);
-        
-          // If we inlined the last possible call site to the function,
-          // delete the function body now.
-          if (Callee->use_empty() && Callee != Caller &&
-              Callee->hasInternalLinkage()) {
-            DEBUG(std::cerr << "    -> Deleting dead function: "
-                            << Callee->getName() << "\n");
-            SCCFunctions.erase(Callee);    // Remove function from this SCC.
-
-            // Remove any call graph edges from the callee to its callees.
-            while (CalleeNode->begin() != CalleeNode->end())
-              CalleeNode->removeCallEdgeTo(*(CalleeNode->end()-1));
-
-            // Removing the node for callee from the call graph and delete it.
-            delete CG.removeFunctionFromModule(CalleeNode);
-            ++NumDeleted;
+  DEBUG(std::cerr << ": " << CallSites.size() << " call sites.\n");
+  
+  // Now that we have all of the call sites, move the ones to functions in the
+  // current SCC to the end of the list.
+  unsigned FirstCallInSCC = CallSites.size();
+  for (unsigned i = 0; i < FirstCallInSCC; ++i)
+    if (Function *F = CallSites[i].getCalledFunction())
+      if (SCCFunctions.count(F))
+        std::swap(CallSites[i--], CallSites[--FirstCallInSCC]);
+  
+  // Now that we have all of the call sites, loop over them and inline them if
+  // it looks profitable to do so.
+  bool Changed = false;
+  bool LocalChange;
+  do {
+    LocalChange = false;
+    // Iterate over the outer loop because inlining functions can cause indirect
+    // calls to become direct calls.
+    for (unsigned CSi = 0; CSi != CallSites.size(); ++CSi)
+      if (Function *Callee = CallSites[CSi].getCalledFunction()) {
+        // Calls to external functions are never inlinable.
+        if (Callee->isExternal() ||
+            CallSites[CSi].getInstruction()->getParent()->getParent() ==Callee){
+          std::swap(CallSites[CSi], CallSites.back());
+          --CSi;
+          continue;
+        }
+
+        // If the policy determines that we should inline this function,
+        // try to do so.
+        CallSite CS = CallSites[CSi];
+        int InlineCost = getInlineCost(CS);
+        if (InlineCost >= (int)InlineThreshold) {
+          DEBUG(std::cerr << "    NOT Inlining: cost=" << InlineCost
+                << ", Call: " << *CS.getInstruction());
+        } else {
+          DEBUG(std::cerr << "    Inlining: cost=" << InlineCost
+                << ", Call: " << *CS.getInstruction());
+          
+          Function *Caller = CS.getInstruction()->getParent()->getParent();
+
+          // Attempt to inline the function...
+          if (InlineCallIfPossible(CS, CG, SCCFunctions)) {
+            // Remove this call site from the list.
+            std::swap(CallSites[CSi], CallSites.back());
+            CallSites.pop_back();
+            --CSi;
+
+            ++NumInlined;
+            Changed = true;
+            LocalChange = true;
           }
-          Changed = true;
         }
       }
-    }
-  }
+  } while (LocalChange);
 
   return Changed;
 }




From lattner at cs.uiuc.edu  Sun May 23 16:25:01 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:25:01 2004
Subject: [llvm-commits] CVS: llvm/lib/Target/X86/X86TargetMachine.cpp
Message-ID: <200405232123.QAA04234@zion.cs.uiuc.edu>


Changes in directory llvm/lib/Target/X86:

X86TargetMachine.cpp updated: 1.54 -> 1.55

---
Log message:

Add support for accurate garbage collection to the LLVM code generators


---
Diffs of the changes:  (+5 -0)

Index: llvm/lib/Target/X86/X86TargetMachine.cpp
diff -u llvm/lib/Target/X86/X86TargetMachine.cpp:1.54 llvm/lib/Target/X86/X86TargetMachine.cpp:1.55
--- llvm/lib/Target/X86/X86TargetMachine.cpp:1.54	Tue Apr  6 14:34:00 2004
+++ llvm/lib/Target/X86/X86TargetMachine.cpp	Sun May 23 16:23:35 2004
@@ -59,6 +59,9 @@
 // does to emit statically compiled machine code.
 bool X86TargetMachine::addPassesToEmitAssembly(PassManager &PM,
 					       std::ostream &Out) {
+  // FIXME: Implement efficient support for garbage collection intrinsics.
+  PM.add(createLowerGCPass());
+
   // FIXME: Implement the invoke/unwind instructions!
   PM.add(createLowerInvokePass());
 
@@ -117,6 +120,8 @@
 /// not supported for this target.
 ///
 void X86JITInfo::addPassesToJITCompile(FunctionPassManager &PM) {
+  // FIXME: Implement efficient support for garbage collection intrinsics.
+  PM.add(createLowerGCPass());
 
   // FIXME: Implement the invoke/unwind instructions!
   PM.add(createLowerInvokePass());




From lattner at cs.uiuc.edu  Sun May 23 16:25:11 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:25:11 2004
Subject: [llvm-commits] CVS: llvm/lib/Target/SparcV9/SparcV9TargetMachine.cpp
Message-ID: <200405232123.QAA04227@zion.cs.uiuc.edu>


Changes in directory llvm/lib/Target/SparcV9:

SparcV9TargetMachine.cpp updated: 1.110 -> 1.111

---
Log message:

Add support for accurate garbage collection to the LLVM code generators



---
Diffs of the changes:  (+6 -0)

Index: llvm/lib/Target/SparcV9/SparcV9TargetMachine.cpp
diff -u llvm/lib/Target/SparcV9/SparcV9TargetMachine.cpp:1.110 llvm/lib/Target/SparcV9/SparcV9TargetMachine.cpp:1.111
--- llvm/lib/Target/SparcV9/SparcV9TargetMachine.cpp:1.110	Sun Apr 25 02:04:49 2004
+++ llvm/lib/Target/SparcV9/SparcV9TargetMachine.cpp	Sun May 23 16:23:33 2004
@@ -125,6 +125,9 @@
 bool
 SparcV9TargetMachine::addPassesToEmitAssembly(PassManager &PM, std::ostream &Out)
 {
+  // FIXME: Implement efficient support for garbage collection intrinsics.
+  PM.add(createLowerGCPass());
+
   // Replace malloc and free instructions with library calls.
   PM.add(createLowerAllocationsPass());
   
@@ -199,6 +202,9 @@
 /// generation for the UltraSparcV9.
 ///
 void SparcV9JITInfo::addPassesToJITCompile(FunctionPassManager &PM) {
+  // FIXME: Implement efficient support for garbage collection intrinsics.
+  PM.add(createLowerGCPass());
+
   // Replace malloc and free instructions with library calls.
   PM.add(createLowerAllocationsPass());
   




From lattner at cs.uiuc.edu  Sun May 23 16:25:20 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:25:20 2004
Subject: [llvm-commits] CVS: llvm/lib/Target/CBackend/Writer.cpp
Message-ID: <200405232123.QAA04218@zion.cs.uiuc.edu>


Changes in directory llvm/lib/Target/CBackend:

Writer.cpp updated: 1.175 -> 1.176

---
Log message:

Add support for accurate garbage collection to the LLVM code generators


---
Diffs of the changes:  (+1 -0)

Index: llvm/lib/Target/CBackend/Writer.cpp
diff -u llvm/lib/Target/CBackend/Writer.cpp:1.175 llvm/lib/Target/CBackend/Writer.cpp:1.176
--- llvm/lib/Target/CBackend/Writer.cpp:1.175	Thu May 20 15:25:50 2004
+++ llvm/lib/Target/CBackend/Writer.cpp	Sun May 23 16:23:30 2004
@@ -1498,6 +1498,7 @@
 //===----------------------------------------------------------------------===//
 
 bool CTargetMachine::addPassesToEmitAssembly(PassManager &PM, std::ostream &o) {
+  PM.add(createLowerGCPass());
   PM.add(createLowerAllocationsPass());
   PM.add(createLowerInvokePass());
   PM.add(new CBackendNameAllUsedStructs());




From lattner at cs.uiuc.edu  Sun May 23 16:26:01 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:26:01 2004
Subject: [llvm-commits] CVS: llvm/lib/Target/X86/InstSelectSimple.cpp
Message-ID: <200405232123.QAA04240@zion.cs.uiuc.edu>


Changes in directory llvm/lib/Target/X86:

InstSelectSimple.cpp updated: 1.255 -> 1.256

---
Log message:

Add some notes to myself, no functional changes


---
Diffs of the changes:  (+19 -7)

Index: llvm/lib/Target/X86/InstSelectSimple.cpp
diff -u llvm/lib/Target/X86/InstSelectSimple.cpp:1.255 llvm/lib/Target/X86/InstSelectSimple.cpp:1.256
--- llvm/lib/Target/X86/InstSelectSimple.cpp:1.255	Fri May 14 01:54:56 2004
+++ llvm/lib/Target/X86/InstSelectSimple.cpp	Sun May 23 16:23:12 2004
@@ -3567,12 +3567,24 @@
     }
   }
 
-  // GEPTypes is empty, which means we have a single operand left.  See if we
-  // can set it as the base register.
+  // GEPTypes is empty, which means we have a single operand left.  Set it as
+  // the base register.
   //
-  // FIXME: When addressing modes are more powerful/correct, we could load
-  // global addresses directly as 32-bit immediates.
   assert(BaseReg == 0);
+
+#if 0   // FIXME: TODO!
+  if (AllocaInst *AI = dyn_castFixedAlloca(V)) {
+    // FIXME: When we can add FrameIndex values as the first operand, we can
+    // make GEP's of allocas MUCH more efficient!
+    unsigned FI = getFixedSizedAllocaFI(AI);
+    GEPOps.pop_back();
+    return;
+  } else if (GlobalValue *GV = dyn_cast<GlobalValue>(V)) {
+    // FIXME: When addressing modes are more powerful/correct, we could load
+    // global addresses directly as 32-bit immediates.
+  }
+#endif
+
   BaseReg = MBB ? getReg(GEPOps[0], MBB, IP) : 1;
   GEPOps.pop_back();        // Consume the last GEP operand
 }
@@ -3592,9 +3604,9 @@
   GEPOps[0] = Src;
   std::copy(IdxBegin, IdxEnd, GEPOps.begin()+1);
   
-  std::vector<const Type*> GEPTypes;
-  GEPTypes.assign(gep_type_begin(Src->getType(), IdxBegin, IdxEnd),
-                  gep_type_end(Src->getType(), IdxBegin, IdxEnd));
+  std::vector<const Type*>
+    GEPTypes(gep_type_begin(Src->getType(), IdxBegin, IdxEnd),
+             gep_type_end(Src->getType(), IdxBegin, IdxEnd));
 
   MachineBasicBlock::iterator IP;
   if (MBB) IP = MBB->end();




From lattner at cs.uiuc.edu  Sun May 23 16:26:11 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:26:11 2004
Subject: [llvm-commits] CVS: llvm/test/Regression/CodeGen/Generic/GC/
Message-ID: <200405232124.QAA04268@zion.cs.uiuc.edu>


Changes in directory llvm/test/Regression/CodeGen/Generic/GC:


---
Log message:

Directory /home/vadve/shared/PublicCVS/llvm/test/Regression/CodeGen/Generic/GC added to the repository


---
Diffs of the changes:  (+0 -0)



From lattner at cs.uiuc.edu  Sun May 23 16:26:20 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:26:20 2004
Subject: [llvm-commits] CVS: llvm/lib/Target/X86/README.txt
Message-ID: <200405232124.QAA04256@zion.cs.uiuc.edu>


Changes in directory llvm/lib/Target/X86:

README.txt updated: 1.12 -> 1.13

---
Log message:

minor wording change


---
Diffs of the changes:  (+1 -1)

Index: llvm/lib/Target/X86/README.txt
diff -u llvm/lib/Target/X86/README.txt:1.12 llvm/lib/Target/X86/README.txt:1.13
--- llvm/lib/Target/X86/README.txt:1.12	Sun Feb 29 12:44:03 2004
+++ llvm/lib/Target/X86/README.txt	Sun May 23 16:22:55 2004
@@ -132,7 +132,7 @@
 ------------------------------------
 
 The x86 has a very, uhm, flexible, way of accessing memory.  It is capable of
-addressing memory addresses of the following form directly in integer
+forming memory addresses of the following expression directly in integer
 instructions (which use ModR/M addressing):
 
    Base+[1,2,4,8]*IndexReg+Disp32




From lattner at cs.uiuc.edu  Sun May 23 16:27:04 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:27:04 2004
Subject: [llvm-commits] CVS: llvm/test/Regression/CodeGen/Generic/GC/alloc_loop.ll
Message-ID: <200405232125.QAA04286@zion.cs.uiuc.edu>


Changes in directory llvm/test/Regression/CodeGen/Generic/GC:

alloc_loop.ll added (r1.1)

---
Log message:

Add a simple testcase for garbage collection support


---
Diffs of the changes:  (+52 -0)

Index: llvm/test/Regression/CodeGen/Generic/GC/alloc_loop.ll
diff -c /dev/null llvm/test/Regression/CodeGen/Generic/GC/alloc_loop.ll:1.1
*** /dev/null	Sun May 23 16:25:00 2004
--- llvm/test/Regression/CodeGen/Generic/GC/alloc_loop.ll	Sun May 23 16:24:50 2004
***************
*** 0 ****
--- 1,52 ----
+ implementation
+ 
+ declare sbyte* %llvm_gc_allocate(uint)
+ declare void %llvm_gc_initialize()
+ 
+ declare void %llvm.gcroot(sbyte**, sbyte*)
+ declare void %llvm.gcwrite(sbyte*, sbyte**)
+ 
+ int %main() {
+ entry:
+ 	%A = alloca sbyte*
+ 	%B = alloca sbyte**
+ 
+ 	call void %llvm_gc_initialize()
+ 
+         ;; void *A;
+ 	call void %llvm.gcroot(sbyte** %A, sbyte* null)
+ 
+         ;; A = gcalloc(10);
+ 	%Aptr = call sbyte* %llvm_gc_allocate(uint 10)
+ 	store sbyte* %Aptr, sbyte** %A
+ 
+         ;; void **B;
+ 	%tmp.1 = cast sbyte*** %B to sbyte **
+ 	call void %llvm.gcroot(sbyte** %tmp.1, sbyte* null)
+ 
+ 	;; B = gcalloc(4);
+ 	%B = call sbyte* %llvm_gc_allocate(uint 8)
+ 	%tmp.2 = cast sbyte* %B to sbyte**
+ 	store sbyte** %tmp.2, sbyte*** %B
+ 
+ 	;; *B = A;
+ 	%B.1 = load sbyte*** %B
+ 	%A.1 = load sbyte** %A
+ 	call void %llvm.gcwrite(sbyte* %A.1, sbyte** %B.1)
+ 	
+ 	br label %AllocLoop
+ 
+ AllocLoop:
+ 	%i = phi uint [ 0, %entry ], [ %indvar.next, %AllocLoop ]
+         ;; Allocated mem: allocated memory is immediately dead.
+ 	call sbyte* %llvm_gc_allocate(uint 100)
+ 	
+ 	%indvar.next = add uint %i, 1
+ 	%exitcond = seteq uint %indvar.next, 10000000
+ 	br bool %exitcond, label %Exit, label %AllocLoop
+ 
+ Exit:
+ 	ret int 0
+ }
+ 
+ declare void %__main()




From lattner at cs.uiuc.edu  Sun May 23 16:27:14 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:27:14 2004
Subject: [llvm-commits] CVS: llvm/runtime/GC/SemiSpace/
Message-ID: <200405232125.QAA04309@zion.cs.uiuc.edu>


Changes in directory llvm/runtime/GC/SemiSpace:


---
Log message:

Directory /home/vadve/shared/PublicCVS/llvm/runtime/GC/SemiSpace added to the repository


---
Diffs of the changes:  (+0 -0)



From lattner at cs.uiuc.edu  Sun May 23 16:27:23 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:27:23 2004
Subject: [llvm-commits] CVS: llvm/runtime/GC/
Message-ID: <200405232125.QAA04299@zion.cs.uiuc.edu>


Changes in directory llvm/runtime/GC:


---
Log message:

Directory /home/vadve/shared/PublicCVS/llvm/runtime/GC added to the repository


---
Diffs of the changes:  (+0 -0)



From lattner at cs.uiuc.edu  Sun May 23 16:28:01 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:28:01 2004
Subject: [llvm-commits] CVS: llvm/runtime/GC/GCInterface.h Makefile gc_exported_symbols.lst
Message-ID: <200405232126.QAA04357@zion.cs.uiuc.edu>


Changes in directory llvm/runtime/GC:

GCInterface.h added (r1.1)
Makefile added (r1.1)
gc_exported_symbols.lst added (r1.1)

---
Log message:

Initial checkin of GC implementation support files


---
Diffs of the changes:  (+69 -0)

Index: llvm/runtime/GC/GCInterface.h
diff -c /dev/null llvm/runtime/GC/GCInterface.h:1.1
*** /dev/null	Sun May 23 16:26:09 2004
--- llvm/runtime/GC/GCInterface.h	Sun May 23 16:25:59 2004
***************
*** 0 ****
--- 1,42 ----
+ /*===-- GCInterface.h - Public interface exposed by garbage collectors ----===*\
+ |*
+ |*                     The LLVM Compiler Infrastructure
+ |*
+ |* This file was developed by the LLVM research group and is distributed under
+ |* the University of Illinois Open Source License. See LICENSE.TXT for details.
+ |* 
+ |*===----------------------------------------------------------------------===*|
+ |* 
+ |* This file defines the common public interface that must be exposed by all
+ |* LLVM garbage collectors.
+ |*
+ \*===----------------------------------------------------------------------===*/
+ 
+ #ifndef GCINTERFACE_H
+ #define GCINTERFACE_H
+ 
+ /* llvm_gc_initialize - This function is called to initalize the garbage
+  * collector.
+  */
+ void llvm_gc_initialize();
+ 
+ /* llvm_gc_allocate - This function allocates Size bytes from the heap and
+  * returns a pointer to it.
+  */
+ void *llvm_gc_allocate(unsigned Size);
+ 
+ /* llvm_gc_collect - This function forces a garbage collection cycle.
+  */
+ void llvm_gc_collect();
+ 
+ /* llvm_gc_read - This function should be implemented to include any read
+  * barrier code that is needed by the garbage collector.
+  */
+ void *llvm_gc_read(void **P);
+ 
+ /* llvm_gc_write - This function should be implemented to include any write
+  * barrier code that is needed by the garbage collector.
+  */
+ void llvm_gc_write(void *V, void **P);
+ 
+ #endif


Index: llvm/runtime/GC/Makefile
diff -c /dev/null llvm/runtime/GC/Makefile:1.1
*** /dev/null	Sun May 23 16:26:09 2004
--- llvm/runtime/GC/Makefile	Sun May 23 16:25:59 2004
***************
*** 0 ****
--- 1,20 ----
+ ##===- runtime/GC/Makefile ---------------------------------*- Makefile -*-===##
+ # 
+ #                     The LLVM Compiler Infrastructure
+ #
+ # This file was developed by the LLVM research group and is distributed under
+ # the University of Illinois Open Source License. See LICENSE.TXT for details.
+ # 
+ ##===----------------------------------------------------------------------===##
+ 
+ LEVEL = ../..
+ PARALLEL_DIRS := SemiSpace
+ 
+ include $(LEVEL)/Makefile.common
+ 
+ # Install target for libraries: Copy into the gcc install directory in chris's
+ # tree...
+ #
+ install::
+ 
+ clean::


Index: llvm/runtime/GC/gc_exported_symbols.lst
diff -c /dev/null llvm/runtime/GC/gc_exported_symbols.lst:1.1
*** /dev/null	Sun May 23 16:26:09 2004
--- llvm/runtime/GC/gc_exported_symbols.lst	Sun May 23 16:25:59 2004
***************
*** 0 ****
--- 1,7 ----
+ llvm_gc_initialize
+ llvm_gc_allocate
+ llvm_gc_collect
+ llvm_gc_write
+ llvm_gc_read
+ 
+ llvm_gc_root_chain
\ No newline at end of file




From llvm at cs.uiuc.edu  Sun May 23 16:28:12 2004
From: llvm at cs.uiuc.edu (LLVM)
Date: Sun May 23 16:28:12 2004
Subject: [llvm-commits] CVS: llvm/docs/llvm.css
Message-ID: <200405232126.QAA04344@zion.cs.uiuc.edu>


Changes in directory llvm/docs:

llvm.css updated: 1.17 -> 1.18

---
Log message:

Create a new style for tables that have no width (.doc_table_nw) so the 
table doesn't stretch across the whole page. Change for BytecodeFormat.html


---
Diffs of the changes:  (+3 -0)

Index: llvm/docs/llvm.css
diff -u llvm/docs/llvm.css:1.17 llvm/docs/llvm.css:1.18
--- llvm/docs/llvm.css:1.17	Sun May 23 16:03:30 2004
+++ llvm/docs/llvm.css	Sun May 23 16:25:50 2004
@@ -48,6 +48,9 @@
 .doc_table      { text-align: center; width: 90%; 
                   padding: 1px 1px 1px 1px; border: 1px; }
 
+.doc_table_nw   { text-align: center; border: 1px; 
+    		  padding: 1px 1px 1px 1px; }
+
 .doc_warning    { color: red; font-weight: bold }
 
 .doc_code       { border: solid 1px gray; background: #eeeeee;




From lattner at cs.uiuc.edu  Sun May 23 16:28:22 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:28:22 2004
Subject: [llvm-commits] CVS: llvm/runtime/Makefile
Message-ID: <200405232126.QAA04379@zion.cs.uiuc.edu>


Changes in directory llvm/runtime:

Makefile updated: 1.17 -> 1.18

---
Log message:

Build the garbage collectors


---
Diffs of the changes:  (+1 -1)

Index: llvm/runtime/Makefile
diff -u llvm/runtime/Makefile:1.17 llvm/runtime/Makefile:1.18
--- llvm/runtime/Makefile:1.17	Fri Apr 16 09:12:36 2004
+++ llvm/runtime/Makefile	Sun May 23 16:26:29 2004
@@ -12,7 +12,7 @@
 include $(LEVEL)/Makefile.config
 
 ifneq ($(wildcard $(LLVMGCCDIR)),)
-PARALLEL_DIRS  := GCCLibraries  libdummy  libprofile  libtrace 
+PARALLEL_DIRS  := GCCLibraries  libdummy  libprofile  libtrace GC
 else
 PARALLEL_DIRS  := 
 install all ::




From lattner at cs.uiuc.edu  Sun May 23 16:28:32 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:28:32 2004
Subject: [llvm-commits] CVS: llvm/runtime/GC/SemiSpace/Makefile semispace.c
Message-ID: <200405232126.QAA04369@zion.cs.uiuc.edu>


Changes in directory llvm/runtime/GC/SemiSpace:

Makefile added (r1.1)
semispace.c added (r1.1)

---
Log message:

Check in an EARLY START on a simple copying collector


---
Diffs of the changes:  (+117 -0)

Index: llvm/runtime/GC/SemiSpace/Makefile
diff -c /dev/null llvm/runtime/GC/SemiSpace/Makefile:1.1
*** /dev/null	Sun May 23 16:26:26 2004
--- llvm/runtime/GC/SemiSpace/Makefile	Sun May 23 16:25:45 2004
***************
*** 0 ****
--- 1,19 ----
+ ##===- runtime/GC/SemiSpace/Makefile -----------------------*- Makefile -*-===##
+ # 
+ #                     The LLVM Compiler Infrastructure
+ #
+ # This file was developed by the LLVM research group and is distributed under
+ # the University of Illinois Open Source License. See LICENSE.TXT for details.
+ # 
+ ##===----------------------------------------------------------------------===##
+ 
+ LEVEL = ../../..
+ BYTECODE_LIBRARY=1
+ #SHARED_LIBRARY=1
+ #DONT_BUILD_RELINKED=1
+ LIBRARYNAME=gcsemispace
+ 
+ EXPORTED_SYMBOL_FILE = $(SourceDir)/../gc_exported_symbols.lst
+ 
+ include $(LEVEL)/Makefile.common
+ 


Index: llvm/runtime/GC/SemiSpace/semispace.c
diff -c /dev/null llvm/runtime/GC/SemiSpace/semispace.c:1.1
*** /dev/null	Sun May 23 16:26:26 2004
--- llvm/runtime/GC/SemiSpace/semispace.c	Sun May 23 16:25:45 2004
***************
*** 0 ****
--- 1,98 ----
+ /*===-- semispace.c - Simple semi-space copying garbage collector ---------===*\
+ |*
+ |*                     The LLVM Compiler Infrastructure
+ |*
+ |* This file was developed by the LLVM research group and is distributed under
+ |* the University of Illinois Open Source License. See LICENSE.TXT for details.
+ |* 
+ |*===----------------------------------------------------------------------===*|
+ |* 
+ |* This garbage collector is an extremely simple copying collector.  It splits
+ |* the managed region of memory into two pieces: the current space to allocate
+ |* from, and the copying space.  When the portion being allocated from fills up,
+ |* a garbage collection cycle happens, which copies all live blocks to the other
+ |* half of the managed space.
+ |*
+ \*===----------------------------------------------------------------------===*/
+ 
+ #include "../GCInterface.h"
+ #include <stdio.h>
+ #include <stdlib.h>
+ 
+ /* FIXME: This should be in a code-generator specific library!
+  */
+ typedef struct GCRoot {
+   void **RootPtr;
+   void *Meta;
+ } GCRoot;
+ 
+ typedef struct GCRoots {
+   struct GCRoots *Next;
+   unsigned NumRoots;
+   GCRoot RootRecords[];
+ } GCRoots;
+ GCRoots *llvm_gc_root_chain;
+ 
+ static void llvm_cg_walk_gcroots(void (*FP)(void **Root, void *Meta)) {
+   GCRoots *R = llvm_gc_root_chain;
+   for (; R; R = R->Next) {
+     unsigned i, e;
+     for (i = 0, e = R->NumRoots; i != e; ++i)
+       FP(R->RootRecords[i].RootPtr, R->RootRecords[i].Meta);
+   }
+ }
+ /* END FIXME! */
+ 
+ 
+ 
+ /* We use no read/write barriers */
+ void *llvm_gc_read(void **P) { return *P; }
+ void llvm_gc_write(void *V, void **P) { *P = V; }
+ 
+ 
+ /* AllocPtr - This points to the next byte that is available for allocation.
+  */
+ static char *AllocPtr;
+ 
+ /* AllocEnd - This points to the first byte not available for allocation.  When
+  * AllocPtr passes this, we have run out of space.
+  */
+ static char *AllocEnd;
+ 
+ void llvm_gc_initialize() {
+   AllocPtr = calloc(1, 1000);
+   AllocEnd = AllocPtr + 1000;
+ }
+ 
+ /* We always want to inline the fast path, but never want to inline the slow
+  * path.
+  */
+ void *llvm_gc_allocate(unsigned Size) __attribute__((always_inline));
+ static void* llvm_gc_alloc_slow(unsigned Size) __attribute__((noinline));
+ 
+ void *llvm_gc_allocate(unsigned Size) {
+   char *OldAP = AllocPtr;
+   char *NewEnd = OldAP+Size;
+   if (NewEnd > AllocEnd)
+     return llvm_gc_alloc_slow(Size);
+   AllocPtr = NewEnd;
+   return OldAP;
+ }
+ 
+ static void* llvm_gc_alloc_slow(unsigned Size) {
+   llvm_gc_collect();
+   return llvm_gc_allocate(Size);
+ }
+ 
+ 
+ static void process_root(void **Root, void *Meta) {
+   printf("process_root[0x%X] = 0x%X\n", Root, *Root);
+ }
+ 
+ void llvm_gc_collect() {
+   printf("Garbage collecting!!\n");
+   llvm_cg_walk_gcroots(process_root);
+   abort();
+ 
+   /* TODO: Iterate through roots. */
+ }




From lattner at cs.uiuc.edu  Sun May 23 16:29:00 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:29:00 2004
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Scalar/LowerGC.cpp
Message-ID: <200405232127.QAA04396@zion.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Scalar:

LowerGC.cpp updated: 1.1 -> 1.2

---
Log message:

Spelling people's names right is kinda important


---
Diffs of the changes:  (+1 -1)

Index: llvm/lib/Transforms/Scalar/LowerGC.cpp
diff -u llvm/lib/Transforms/Scalar/LowerGC.cpp:1.1 llvm/lib/Transforms/Scalar/LowerGC.cpp:1.2
--- llvm/lib/Transforms/Scalar/LowerGC.cpp:1.1	Sun May 23 16:19:22 2004
+++ llvm/lib/Transforms/Scalar/LowerGC.cpp	Sun May 23 16:27:29 2004
@@ -15,7 +15,7 @@
 //
 // This pass implements the code transformation described in this paper:
 //   "Accurate Garbage Collection in an Uncooperative Environment"
-//   Fergus Hendersen, ISMM, 2002
+//   Fergus Henderson, ISMM, 2002
 //
 //===----------------------------------------------------------------------===//
 




From lattner at cs.uiuc.edu  Sun May 23 16:32:00 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:32:00 2004
Subject: [llvm-commits] CVS: llvm/test/Regression/Analysis/Andersens/
Message-ID: <200405232130.QAA04417@zion.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Analysis/Andersens:


---
Log message:

Directory /home/vadve/shared/PublicCVS/llvm/test/Regression/Analysis/Andersens added to the repository


---
Diffs of the changes:  (+0 -0)



From lattner at cs.uiuc.edu  Sun May 23 16:33:01 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 16:33:01 2004
Subject: [llvm-commits] CVS: llvm/test/Regression/Analysis/Andersens/arg-must-alias.ll basictest.ll external.ll
Message-ID: <200405232131.QAA04436@zion.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Analysis/Andersens:

arg-must-alias.ll added (r1.1)
basictest.ll added (r1.1)
external.ll added (r1.1)

---
Log message:

Testcases for andersen's alias analysis.

I think this is all of the stuff I had pending in my tree..


---
Diffs of the changes:  (+68 -0)

Index: llvm/test/Regression/Analysis/Andersens/arg-must-alias.ll
diff -c /dev/null llvm/test/Regression/Analysis/Andersens/arg-must-alias.ll:1.1
*** /dev/null	Sun May 23 16:31:10 2004
--- llvm/test/Regression/Analysis/Andersens/arg-must-alias.ll	Sun May 23 16:31:00 2004
***************
*** 0 ****
--- 1,17 ----
+ ; RUN: llvm-as < %s | opt -anders-aa -load-vn -gcse -deadargelim | llvm-dis | not grep ARG
+ 
+ %G = internal constant int* null
+ 
+ implementation
+ 
+ internal int %internal(int* %ARG) {
+ 	;; The 'Arg' argument must-aliases the null pointer, so it can be subsituted
+ 	;; directly here, making it dead.
+ 	store int* %ARG, int** %G
+ 	ret int 0
+ }
+ 
+ int %foo() {
+ 	%V = call int %internal(int* null)
+ 	ret int %V
+ }


Index: llvm/test/Regression/Analysis/Andersens/basictest.ll
diff -c /dev/null llvm/test/Regression/Analysis/Andersens/basictest.ll:1.1
*** /dev/null	Sun May 23 16:31:10 2004
--- llvm/test/Regression/Analysis/Andersens/basictest.ll	Sun May 23 16:31:00 2004
***************
*** 0 ****
--- 1,29 ----
+ 
+ implementation
+ 
+ void %test1() {
+ 	%X = malloc int*
+ 	%Y = malloc int
+ 	%Z = cast int* %Y to int
+ 	%W = cast int %Z to int*
+ 	store int* %W, int** %X
+ 	ret void
+ }
+ 
+ void %test2(int* %P) {
+ 	%X = malloc int*
+ 	%Y = malloc int
+ 	store int* %P, int** %X
+ 	ret void
+ }
+ 
+ internal int *%test3(int* %P) {
+ 	ret int* %P
+ }
+ 
+ void %test4() {
+ 	%X = malloc int
+ 	%Y = call int* %test3(int* %X)
+ 	%ZZ = getelementptr int* null, int 17
+ 	ret void
+ }


Index: llvm/test/Regression/Analysis/Andersens/external.ll
diff -c /dev/null llvm/test/Regression/Analysis/Andersens/external.ll:1.1
*** /dev/null	Sun May 23 16:31:10 2004
--- llvm/test/Regression/Analysis/Andersens/external.ll	Sun May 23 16:31:00 2004
***************
*** 0 ****
--- 1,22 ----
+ ; RUN: llvm-as < %s | opt -anders-aa -load-vn -gcse -deadargelim | llvm-dis | grep store | not grep null
+ 
+ ; Because the 'internal' function is passed to an external function, we don't
+ ; know what the incoming values will alias.  As such, we cannot do the 
+ ; optimization checked by the 'arg-must-alias.llx' test.
+ 
+ declare void %external(int(int*)*)
+ %G = internal constant int* null
+ 
+ implementation
+ 
+ internal int %internal(int* %ARG) {
+ 	;;; We *DON'T* know that ARG always points to null!
+ 	store int* %ARG, int** %G
+ 	ret int 0
+ }
+ 
+ int %foo() {
+ 	call void %external(int(int*)* %internal)
+ 	%V = call int %internal(int* null)
+ 	ret int %V
+ }




From alkis at cs.uiuc.edu  Sun May 23 18:05:03 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sun May 23 18:05:03 2004
Subject: [llvm-commits] CVS: llvm/runtime/GC/SemiSpace/semispace.c
Message-ID: <200405232302.SAA04864@zion.cs.uiuc.edu>


Changes in directory llvm/runtime/GC/SemiSpace:

semispace.c updated: 1.1 -> 1.2

---
Log message:

Eliminate warnings


---
Diffs of the changes:  (+1 -1)

Index: llvm/runtime/GC/SemiSpace/semispace.c
diff -u llvm/runtime/GC/SemiSpace/semispace.c:1.1 llvm/runtime/GC/SemiSpace/semispace.c:1.2
--- llvm/runtime/GC/SemiSpace/semispace.c:1.1	Sun May 23 16:25:45 2004
+++ llvm/runtime/GC/SemiSpace/semispace.c	Sun May 23 18:02:35 2004
@@ -86,7 +86,7 @@
 
 
 static void process_root(void **Root, void *Meta) {
-  printf("process_root[0x%X] = 0x%X\n", Root, *Root);
+  printf("process_root[0x%X] = 0x%X\n", (unsigned) Root, (unsigned) *Root);
 }
 
 void llvm_gc_collect() {




From alkis at cs.uiuc.edu  Sun May 23 18:06:01 2004
From: alkis at cs.uiuc.edu (Alkis Evlogimenos)
Date: Sun May 23 18:06:01 2004
Subject: [llvm-commits] CVS: llvm-java/lib/Compiler/Compiler.cpp
Message-ID: <200405232303.SAA04900@zion.cs.uiuc.edu>


Changes in directory llvm-java/lib/Compiler:

Compiler.cpp updated: 1.12 -> 1.13

---
Log message:

Add support for load and store.
Insert asserts to not implemented yet bycodes.


---
Diffs of the changes:  (+82 -17)

Index: llvm-java/lib/Compiler/Compiler.cpp
diff -u llvm-java/lib/Compiler/Compiler.cpp:1.12 llvm-java/lib/Compiler/Compiler.cpp:1.13
--- llvm-java/lib/Compiler/Compiler.cpp:1.12	Sun May 23 06:23:24 2004
+++ llvm-java/lib/Compiler/Compiler.cpp	Sun May 23 18:03:35 2004
@@ -241,39 +241,56 @@
         case ALOAD: {
             // FIXME: use opcodes to perform type checking
             unsigned index = readUByte(code, i);
-            opStack_.push(locals_[index]);
+            Instruction* in = new LoadInst(locals_[index]);
+            opStack_.push(in);
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
             break;
         }
         case ILOAD_0:
         case ILOAD_1:
         case ILOAD_2:
-        case ILOAD_3:
-            opStack_.push(locals_[code[i]-ILOAD_0]);
+        case ILOAD_3: {
+            Instruction* in = new LoadInst(locals_[code[i]-ILOAD_0]);
+            opStack_.push(in);
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
             break;
+        }
         case LLOAD_0:
         case LLOAD_1:
         case LLOAD_2:
-        case LLOAD_3:
-            opStack_.push(locals_[code[i]-LLOAD_0]);
+        case LLOAD_3: {
+            Instruction* in = new LoadInst(locals_[code[i]-LLOAD_0]);
+            opStack_.push(in);
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
             break;
+        }
         case FLOAD_0:
         case FLOAD_1:
         case FLOAD_2:
-        case FLOAD_3:
-            opStack_.push(locals_[code[i]-FLOAD_0]);
+        case FLOAD_3: {
+            Instruction* in = new LoadInst(locals_[code[i]-FLOAD_0]);
+            opStack_.push(in);
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
             break;
+        }
         case DLOAD_0:
         case DLOAD_1:
         case DLOAD_2:
-        case DLOAD_3:
-            opStack_.push(locals_[code[i]-DLOAD_0]);
+        case DLOAD_3: {
+            Instruction* in = new LoadInst(locals_[code[i]-DLOAD_0]);
+            opStack_.push(in);
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
             break;
+        }
         case ALOAD_0:
         case ALOAD_1:
         case ALOAD_2:
-        case ALOAD_3:
-            opStack_.push(locals_[code[i]-ALOAD_0]);
+        case ALOAD_3: {
+            Instruction* in = new LoadInst(locals_[code[i]-ALOAD_0]);
+            opStack_.push(in);
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
             break;
+        }
         case IALOAD:
         case LALOAD:
         case FALOAD:
@@ -282,31 +299,75 @@
         case BALOAD:
         case CALOAD:
         case SALOAD:
+            assert(0 && "not implemented");
         case ISTORE:
         case LSTORE:
         case FSTORE:
         case DSTORE:
-        case ASTORE:
+        case ASTORE: {
+            // FIXME: use opcodes to perform type checking
+            unsigned index = readUByte(code, i);
+            Value* v1 = opStack_.top(); opStack_.pop();
+            Instruction* in = new StoreInst(v1, locals_[index]);
+            opStack_.push(in);
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
+            break;
+        }
         case ISTORE_0:
         case ISTORE_1:
         case ISTORE_2:
-        case ISTORE_3:
+        case ISTORE_3: {
+            Value* v1 = opStack_.top(); opStack_.pop();
+            Instruction* in =
+                new StoreInst(v1, locals_[code[i]-ISTORE_0]);
+            opStack_.push(in);
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
+            break;
+        }
         case LSTORE_0:
         case LSTORE_1:
         case LSTORE_2:
-        case LSTORE_3:
+        case LSTORE_3: {
+            Value* v1 = opStack_.top(); opStack_.pop();
+            Instruction* in =
+                new StoreInst(v1, locals_[code[i]-LSTORE_0]);
+            opStack_.push(in);
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
+            break;
+        }
         case FSTORE_0:
         case FSTORE_1:
         case FSTORE_2:
-        case FSTORE_3:
+        case FSTORE_3: {
+            Value* v1 = opStack_.top(); opStack_.pop();
+            Instruction* in =
+                new StoreInst(v1, locals_[code[i]-FSTORE_0]);
+            opStack_.push(in);
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
+            break;
+        }
         case DSTORE_0:
         case DSTORE_1:
         case DSTORE_2:
-        case DSTORE_3:
+        case DSTORE_3: {
+            Value* v1 = opStack_.top(); opStack_.pop();
+            Instruction* in =
+                new StoreInst(v1, locals_[code[i]-DSTORE_0]);
+            opStack_.push(in);
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
+            break;
+        }
         case ASTORE_0:
         case ASTORE_1:
         case ASTORE_2:
-        case ASTORE_3:
+        case ASTORE_3: {
+            Value* v1 = opStack_.top(); opStack_.pop();
+            Instruction* in =
+                new StoreInst(v1, locals_[code[i]-ASTORE_0]);
+            opStack_.push(in);
+            bc2bbMap_[bcStart]->getInstList().push_back(in);
+            break;
+        }
         case IASTORE:
         case LASTORE:
         case FASTORE:
@@ -557,6 +618,7 @@
         case FCMPG:
         case DCMPL:
         case DCMPG:
+            assert(0 && "not implemented");
         case IFEQ:
         case IFNE:
         case IFLT:
@@ -623,6 +685,7 @@
             break;
         case JSR:
         case RET:
+            assert(0 && "not implemented");
         case TABLESWITCH: {
             Value* v1 = opStack_.top(); opStack_.pop();
             skipPadBytes(code, i);
@@ -680,6 +743,7 @@
         case MONITOREXIT:
 //      case WIDE:
         case MULTIANEWARRAY:
+            assert(0 && "not implemented");
         case IFNULL:
         case IFNONNULL: {
             static Instruction::BinaryOps java2llvm[] = {
@@ -700,6 +764,7 @@
             new BranchInst(bc2bbMap_[bcStart + readSInt(code, i)]);
             break;
         case JSR_W:
+            assert(0 && "not implemented");
         case BREAKPOINT:
         case IMPDEP1:
         case IMPDEP2:




From tbrethou at cs.uiuc.edu  Sun May 23 22:16:01 2004
From: tbrethou at cs.uiuc.edu (Tanya Brethour)
Date: Sun May 23 22:16:01 2004
Subject: [llvm-commits] CVS: llvm/include/llvm/CodeGen/MachineInstr.h
Message-ID: <200405240314.WAA12596@seraph.cs.uiuc.edu>


Changes in directory llvm/include/llvm/CodeGen:

MachineInstr.h updated: 1.148 -> 1.149

---
Log message:

Changed clone to be const.
Changed copy constructor to set parent, prev, and next pointers to null.


---
Diffs of the changes:  (+1 -1)

Index: llvm/include/llvm/CodeGen/MachineInstr.h
diff -u llvm/include/llvm/CodeGen/MachineInstr.h:1.148 llvm/include/llvm/CodeGen/MachineInstr.h:1.149
--- llvm/include/llvm/CodeGen/MachineInstr.h:1.148	Sun May 23 15:58:02 2004
+++ llvm/include/llvm/CodeGen/MachineInstr.h	Sun May 23 22:14:18 2004
@@ -460,7 +460,7 @@
 
   /// clone - Create a copy of 'this' instruction that is identical in
   /// all ways except the the instruction has no parent, prev, or next.
-  MachineInstr* clone();
+  MachineInstr* clone() const;
 
   //
   // Debugging support




From tbrethou at cs.uiuc.edu  Sun May 23 22:16:12 2004
From: tbrethou at cs.uiuc.edu (Tanya Brethour)
Date: Sun May 23 22:16:12 2004
Subject: [llvm-commits] CVS: llvm/lib/CodeGen/MachineInstr.cpp
Message-ID: <200405240314.WAA12587@seraph.cs.uiuc.edu>


Changes in directory llvm/lib/CodeGen:

MachineInstr.cpp updated: 1.98 -> 1.99

---
Log message:

Changed clone to be const.
Changed copy constructor to set parent, prev, and next pointers to null.


---
Diffs of the changes:  (+7 -1)

Index: llvm/lib/CodeGen/MachineInstr.cpp
diff -u llvm/lib/CodeGen/MachineInstr.cpp:1.98 llvm/lib/CodeGen/MachineInstr.cpp:1.99
--- llvm/lib/CodeGen/MachineInstr.cpp:1.98	Sun May 23 15:57:57 2004
+++ llvm/lib/CodeGen/MachineInstr.cpp	Sun May 23 22:14:16 2004
@@ -78,6 +78,12 @@
   //Add operands
   for(unsigned i=0; i < MI.getNumOperands(); ++i)
     operands.push_back(MachineOperand(MI.getOperand(i)));
+
+  //Set parent, next, and prev to null
+  parent = 0;
+  prev = 0;
+  next = 0;
+  
 }
 
 
@@ -89,7 +95,7 @@
 ///clone - Create a copy of 'this' instruction that is identical in
 ///all ways except the following: The instruction has no parent The
 ///instruction has no name
-MachineInstr* MachineInstr::clone() {
+MachineInstr* MachineInstr::clone() const {
   return new MachineInstr(*this);
 }
 




From lattner at cs.uiuc.edu  Sun May 23 22:47:01 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 22:47:01 2004
Subject: [llvm-commits] CVS: llvm/lib/CodeGen/MachineBasicBlock.cpp
Message-ID: <200405240345.WAA07512@zion.cs.uiuc.edu>


Changes in directory llvm/lib/CodeGen:

MachineBasicBlock.cpp updated: 1.10 -> 1.11

---
Log message:

Eliminate an explicit use of the LLVM basic block, using getParent instead,
which simplifies the code



---
Diffs of the changes:  (+1 -1)

Index: llvm/lib/CodeGen/MachineBasicBlock.cpp
diff -u llvm/lib/CodeGen/MachineBasicBlock.cpp:1.10 llvm/lib/CodeGen/MachineBasicBlock.cpp:1.11
--- llvm/lib/CodeGen/MachineBasicBlock.cpp:1.10	Wed May 12 16:35:22 2004
+++ llvm/lib/CodeGen/MachineBasicBlock.cpp	Sun May 23 22:44:52 2004
@@ -97,6 +97,6 @@
     OS << "\n" << LBB->getName() << " (" << (const void*)LBB << "):\n";
     for (const_iterator I = begin(); I != end(); ++I) {
         OS << "\t";
-        I->print(OS, MachineFunction::get(LBB->getParent()).getTarget());
+        I->print(OS, getParent()->getTarget());
     }
 }




From lattner at cs.uiuc.edu  Sun May 23 23:46:01 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 23:46:01 2004
Subject: [llvm-commits] CVS: llvm/docs/index.html
Message-ID: <200405240444.XAA28868@zion.cs.uiuc.edu>


Changes in directory llvm/docs:

index.html updated: 1.15 -> 1.16

---
Log message:

Add a link to the GC doc


---
Diffs of the changes:  (+2 -1)

Index: llvm/docs/index.html
diff -u llvm/docs/index.html:1.15 llvm/docs/index.html:1.16
--- llvm/docs/index.html:1.15	Fri May 21 21:28:36 2004
+++ llvm/docs/index.html	Sun May 23 23:43:49 2004
@@ -105,6 +105,7 @@
 <li><a href="ProgrammersManual.html">LLVM Programmers Manual</a></li>
 <li><a href="WritingAnLLVMPass.html">Writing an LLVM Pass</a></li>
 <li><a href="AliasAnalysis.html">Alias Analysis in LLVM</a></li>
+<li><a href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a></li>
 <li><a href="SourceLevelDebugging.html">Source Level Debugging with
 LLVM</a></li>
 <li><a href="TableGenFundamentals.html">TableGen Fundamentals</a></li>
@@ -179,7 +180,7 @@
 
   <a href="mailto:criswell [at] cs.uiuc.edu">John Criswell</a><br>
   <a href="http://llvm.cs.uiuc.edu">LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/05/22 02:28:36 $
+  Last modified: $Date: 2004/05/24 04:43:49 $
 </address>
 
 </body>




From lattner at cs.uiuc.edu  Sun May 23 23:48:00 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 23:48:00 2004
Subject: [llvm-commits] CVS: llvm/docs/ReleaseNotes.html
Message-ID: <200405240446.XAA29484@zion.cs.uiuc.edu>


Changes in directory llvm/docs:

ReleaseNotes.html updated: 1.183 -> 1.184

---
Log message:

Add some notes


---
Diffs of the changes:  (+6 -1)

Index: llvm/docs/ReleaseNotes.html
diff -u llvm/docs/ReleaseNotes.html:1.183 llvm/docs/ReleaseNotes.html:1.184
--- llvm/docs/ReleaseNotes.html:1.183	Sun May 23 16:07:27 2004
+++ llvm/docs/ReleaseNotes.html	Sun May 23 23:45:52 2004
@@ -101,6 +101,11 @@
 <li>LLVM now has new loop unrolling and loop unswitching passes.</li>
 <li>The induction variable substitution pass performs linear function test
 replacement and exit value replacement optimizations.</li>
+<li>LLVM now has first-class support for <a href="GarbageCollection.html">Accurate 
+Garbage Collection</a>, enabling the use of aggressive copying and generational 
+collectors.</li>
+<li>LLVM now includes an implementation of Andersen's interprocedural alias 
+analysis algorithm.</li>
 </ol>
 
 
@@ -660,7 +665,7 @@
   src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
 
   <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/05/23 21:07:27 $
+  Last modified: $Date: 2004/05/24 04:45:52 $
 </address>
 
 </body>




From lattner at cs.uiuc.edu  Sun May 23 23:53:01 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 23:53:01 2004
Subject: [llvm-commits] CVS: llvm/docs/ReleaseNotes.html
Message-ID: <200405240450.XAA30101@zion.cs.uiuc.edu>


Changes in directory llvm/docs:

ReleaseNotes.html updated: 1.184 -> 1.185

---
Log message:

Fix a div problem that was causing some wierd indentation.  Get more div'y


---
Diffs of the changes:  (+27 -3)

Index: llvm/docs/ReleaseNotes.html
diff -u llvm/docs/ReleaseNotes.html:1.184 llvm/docs/ReleaseNotes.html:1.185
--- llvm/docs/ReleaseNotes.html:1.184	Sun May 23 23:45:52 2004
+++ llvm/docs/ReleaseNotes.html	Sun May 23 23:50:25 2004
@@ -81,11 +81,15 @@
 with a broad variety of C++ programs, though it has still received less testing
 than the C front-end.</p>
 
+</div>
+
 <!--=========================================================================-->
 <div class="doc_subsubsection">
 <a name="newfeatures">This release implements the following new features:</a>
 </div>
 
+<div class="doc_text">
+
 <ol>
 <li>The LLVM <a href="LangRef.html#i_select"><tt>select</tt></a> instruction is
 now fully implemented and supported by all transformations, native code
@@ -108,12 +112,16 @@
 analysis algorithm.</li>
 </ol>
 
+</div>
+
 
 <!--=========================================================================-->
 <div class="doc_subsubsection">
 In this release, the following missing features were implemented:
 </div>
 
+<div class="doc_text">
+
 <ol>
 <li><a href="http://llvm.cs.uiuc.edu/PR82">LLVM cannot handle structures with
 more than 256 elements</a></li>
@@ -125,6 +133,7 @@
 uses of external fns by immediate constant exprs</a></li>
 </ol>
 
+</div>
 
 <!--=========================================================================-->
 <div class="doc_subsubsection">
@@ -132,6 +141,8 @@
 issues were fixed:</a>
 </div>
 
+<div class="doc_text">
+
 <ol>
 <li><a href="http://llvm.cs.uiuc.edu/PR305">LLVM tools will happily spew
 bytecode onto your terminal</a></li>
@@ -139,17 +150,21 @@
 for structure typedefs</a></li>
 </ol>
 
+</div>
+
 <!--=========================================================================-->
 <div class="doc_subsubsection">
 In this release, the following build problems were fixed:
 </div>
 
+<div class="doc_text">
 <ol>
 <li><a href="http://llvm.cs.uiuc.edu/PR301">Minor configure bugs with
 -disable/enable-povray and -disable-spec</a></li>
 <li><a href="http://llvm.cs.uiuc.edu/PR289">shell scripts output by gccld don't
 work if you change PATH</a></li>
 </ol>
+</div>
 
 <!--=========================================================================-->
 <div class="doc_subsubsection">
@@ -157,6 +172,7 @@
 improvements:</a>
 </div>
 
+<div class="doc_text">
 <ol>
 <li>Fixed: <a href="http://llvm.cs.uiuc.edu/PR309">[vmcore] Code quality problem
 due to long operand of getelementptr</a></li>
@@ -193,7 +209,6 @@
 propagation of function calls.  It currently supports a few math library
 functions like sqrt/sin/cos/etc.</li>
 </ol>
-
 </div>
 
 <!--=========================================================================-->
@@ -202,6 +217,8 @@
 were fixed:</a>
 </div>
 
+<div class="doc_text">
+
 <p>Bugs in the LLVM Core:</p>
 
 <ol>
@@ -230,7 +247,7 @@
 <li><a href="http://llvm.cs.uiuc.edu/PR298">[llvmgcc] Variable length array
 indexing miscompiled</a></li>
 </ol>
-
+</div>
 
 <!--=========================================================================-->
 <div class="doc_subsubsection">
@@ -238,12 +255,15 @@
 improvements:</a>
 </div>
 
+<div class="doc_text">
+
 <p>Documentation Improvements</p>
 <ol>
 <li><a href="BytecodeFormat.html">Documentation of the LLVM Bytecode file format
 was added.</a></li>
 <li>All documentation was brought up to HTML 4.01 (Strict) level.</li>
 </ol>
+</div>
 
 
 <!-- *********************************************************************** -->
@@ -273,6 +293,8 @@
 In this release, the following portability problems were fixed:
 </div>
 
+<div class="doc_text">
+
 <ol>
 <li><a href="http://llvm.cs.uiuc.edu/PR304">warnings compiling Stacker compiler
 on Mac OS X</a></li>
@@ -280,6 +302,8 @@
 understand abbreviated names in headers</a></li>
 </ol>
 
+</div>
+
 <!-- *********************************************************************** -->
 <div class="doc_section">
   <a name="knownproblems">Known Problems</a>
@@ -665,7 +689,7 @@
   src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
 
   <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/05/24 04:45:52 $
+  Last modified: $Date: 2004/05/24 04:50:25 $
 </address>
 
 </body>




From lattner at cs.uiuc.edu  Sun May 23 23:56:02 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 23:56:02 2004
Subject: [llvm-commits] CVS: llvm/docs/ReleaseNotes.html
Message-ID: <200405240453.XAA30725@zion.cs.uiuc.edu>


Changes in directory llvm/docs:

ReleaseNotes.html updated: 1.185 -> 1.186

---
Log message:

Personally, I think that documentation is as important as code.  Increase
the visibility of dox changes


---
Diffs of the changes:  (+4 -18)

Index: llvm/docs/ReleaseNotes.html
diff -u llvm/docs/ReleaseNotes.html:1.185 llvm/docs/ReleaseNotes.html:1.186
--- llvm/docs/ReleaseNotes.html:1.185	Sun May 23 23:50:25 2004
+++ llvm/docs/ReleaseNotes.html	Sun May 23 23:53:32 2004
@@ -105,6 +105,8 @@
 <li>LLVM now has new loop unrolling and loop unswitching passes.</li>
 <li>The induction variable substitution pass performs linear function test
 replacement and exit value replacement optimizations.</li>
+<li>The <a href="BytecodeFormat.html">LLVM Bytecode file format</a> is now 
+    documented.</a></li>
 <li>LLVM now has first-class support for <a href="GarbageCollection.html">Accurate 
 Garbage Collection</a>, enabling the use of aggressive copying and generational 
 collectors.</li>
@@ -148,6 +150,7 @@
 bytecode onto your terminal</a></li>
 <li><a href="http://llvm.cs.uiuc.edu/PR329">[llvmgcc] type names are not emitted
 for structure typedefs</a></li>
+<li>All documentation is now conformant to the HTML 4.01 (Strict) level.</li>
 </ol>
 
 </div>
@@ -249,23 +252,6 @@
 </ol>
 </div>
 
-<!--=========================================================================-->
-<div class="doc_subsubsection">
-<a name="misc">This release includes the following miscellaneous
-improvements:</a>
-</div>
-
-<div class="doc_text">
-
-<p>Documentation Improvements</p>
-<ol>
-<li><a href="BytecodeFormat.html">Documentation of the LLVM Bytecode file format
-was added.</a></li>
-<li>All documentation was brought up to HTML 4.01 (Strict) level.</li>
-</ol>
-</div>
-
-
 <!-- *********************************************************************** -->
 <div class="doc_section">
   <a name="portability">Portability and Supported Platforms</a>
@@ -689,7 +675,7 @@
   src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
 
   <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/05/24 04:50:25 $
+  Last modified: $Date: 2004/05/24 04:53:32 $
 </address>
 
 </body>




From lattner at cs.uiuc.edu  Sun May 23 23:57:01 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 23:57:01 2004
Subject: [llvm-commits] CVS: llvm/docs/GarbageCollection.html
Message-ID: <200405240454.XAA31335@zion.cs.uiuc.edu>


Changes in directory llvm/docs:

GarbageCollection.html updated: 1.1 -> 1.2

---
Log message:

Add a todo to make it clear that the section is not done


---
Diffs of the changes:  (+2 -2)

Index: llvm/docs/GarbageCollection.html
diff -u llvm/docs/GarbageCollection.html:1.1 llvm/docs/GarbageCollection.html:1.2
--- llvm/docs/GarbageCollection.html:1.1	Sun May 23 16:02:20 2004
+++ llvm/docs/GarbageCollection.html	Sun May 23 23:54:31 2004
@@ -214,7 +214,7 @@
 <div class="doc_text">
 
 <p>
-Either from root meta data, or from object headers.  Front-end can provide a
+TODO: Either from root meta data, or from object headers.  Front-end can provide a
 call-back to get descriptor from object without meta-data.
 </p>
 
@@ -411,7 +411,7 @@
 
   <a href="mailto:sabre at nondot.org">Chris Lattner</a><br>
   <a href="http://llvm.cs.uiuc.edu">LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/05/23 21:02:20 $
+  Last modified: $Date: 2004/05/24 04:54:31 $
 </address>
 
 </body>




From lattner at cs.uiuc.edu  Sun May 23 23:58:04 2004
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun May 23 23:58:04 2004
Subject: [llvm-commits] CVS: llvm/docs/BytecodeFormat.html
Message-ID: <200405240455.XAA31956@zion.cs.uiuc.edu>


Changes in directory llvm/docs:

BytecodeFormat.html updated: 1.4 -> 1.5

---
Log message:

Fix a couple of busted li's


---
Diffs of the changes:  (+3 -3)

Index: llvm/docs/BytecodeFormat.html
diff -u llvm/docs/BytecodeFormat.html:1.4 llvm/docs/BytecodeFormat.html:1.5
--- llvm/docs/BytecodeFormat.html:1.4	Sun May 23 12:12:45 2004
+++ llvm/docs/BytecodeFormat.html	Sun May 23 23:55:32 2004
@@ -32,8 +32,8 @@
       <li><a href="#typeool">Global Type Pool</a></li>
       <li><a href="#modinfo">Module Info Block</a></li>
       <li><a href="#constants">Global Constant Pool</a></li>
-      <li><a href="#functions">Function Blocks</a><li>
-      <li><a href="#symtab">Module Symbol Table</a><li>
+      <li><a href="#functions">Function Blocks</a></li>
+      <li><a href="#symtab">Module Symbol Table</a></li>
     </ol>
   </li>
 </ol>
@@ -447,7 +447,7 @@
   <a href="mailto:rspencer at x10sys.com">Reid Spencer</a> and 
   <a href="mailto:sabre at nondot.org">Chris Lattner</a><br>
   <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
-  Last modified: $Date: 2004/05/23 17:12:45 $
+  Last modified: $Date: 2004/05/24 04:55:32 $
 </address>
 </body>
 </html>