From bugzilla-daemon at cs.uiuc.edu  Thu Nov  1 05:53:06 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Thu, 1 Nov 2007 05:53:06 -0500
Subject: [LLVMbugs] [Bug 1568] segfault during domtree construction on
	invalid bytecode
In-Reply-To: <bug-1568-206@http.llvm.org/bugs/>
Message-ID: <200711011053.lA1Ar6cr009632@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1568


Duncan Sands <baldrick at free.fr> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #5 from Duncan Sands <baldrick at free.fr>  2007-11-01 05:53:05 ---
With the following additional changes it seems to be satisfactory:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071029/055061.html


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Thu Nov  1 21:19:42 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Thu, 1 Nov 2007 21:19:42 -0500
Subject: [LLVMbugs] [Bug 1760] test/Codegen/array.c fails with 'bus error'
	on Sparc
In-Reply-To: <bug-1760-206@http.llvm.org/bugs/>
Message-ID: <200711020219.lA22Jg2t021635@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1760


Chris Lattner <sabre at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #5 from Chris Lattner <sabre at nondot.org>  2007-11-01 21:19:42 ---
I believe this is fixed, please verify.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sat Nov  3 15:45:53 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sat, 3 Nov 2007 15:45:53 -0500
Subject: [LLVMbugs] [Bug 1761] New: llc generates invalid assembly: junk at
 end of line, first unrecognized character is `('
Message-ID: <bug-1761-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1761

           Summary: llc generates invalid assembly: junk at end of line,
                    first unrecognized character is `('
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: edwintorok at gmail.com
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=1181)
 --> (http://llvm.org/bugs/attachment.cgi?id=1181)
reduced C testcase

When I compile the original testcase from
http://llvm.org/bugs/show_bug.cgi?id=1748, I get:
$ llvm-gcc t.i
/tmp/cc9CxtGh.s: Assembler messages:
/tmp/cc9CxtGh.s:15: Error: junk at end of line, first unrecognized character is
`('

if I compile to bitcode and then run llc *and* assemble the resulting assembly
I get the same message.
Attached reduced testcases (with delta and bugpoint)


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sat Nov  3 16:56:41 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sat, 3 Nov 2007 16:56:41 -0500
Subject: [LLVMbugs] [Bug 1762] New: llvm-gcc sees a 128 bit integer,
	but there is none
Message-ID: <bug-1762-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1762

           Summary: llvm-gcc sees a 128 bit integer, but there is none
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: edwintorok at gmail.com
                CC: llvmbugs at cs.uiuc.edu


$ llvm-gcc -emit-llvm -c t2.i -o t2.bc
WARNING: 128-bit integers not supported!

Reduced test case follows, I don't see any 128-bit integers there:
[original testcase from linux kernel mm/slub.c, reduced using delta]

$ cat t2.i
     struct kmem_cache {
     int objects;
    };
     static void process_slab( struct kmem_cache *s   ) {
     unsigned long map[(((s->objects) + (64) - 1) / (64))];
    }

$ uname -a
Linux lightspeed2 2.6.23-rc8-hrt1-cfs-v22-g1bef7dc0-dirty #17 Sun Sep 30
18:49:34 EEST 2007 x86_64 GNU/Linux

$ llvm-gcc -v
Using built-in specs.
Target: x86_64-unknown-linux-gnu
Configured with: ../llvm-gcc4.0-2.1.source/configure
--prefix=/home/edwin/obj/../install --program-prefix=llvm-
--enable-llvm=/home/edwin/llvm-svn/llvm/ --enable-languages=c,c++
--disable-shared --disable-multilib : (reconfigured)
../llvm-gcc4.0-2.1.source/configure --prefix=/home/edwin/obj/../install
--program-prefix=llvm- --enable-llvm=/home/edwin/llvm-svn/llvm/
--disable-shared --disable-multilib
--with-gcc-version-trigger=/home/edwin/llvm-gcc4.0-2.1.source/gcc/version.c
--enable-languages=c,c++ --no-create --no-recursion
Thread model: posix
gcc version 4.0.1 (Apple Computer, Inc. build 5449)(LLVM build )


llvm-gcc is r43660, and llvm is r43663


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sat Nov  3 17:18:32 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sat, 3 Nov 2007 17:18:32 -0500
Subject: [LLVMbugs] [Bug 1763] New: llc: Couldn't allocate output reg for
	contraint 'q'!
Message-ID: <bug-1763-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1763

           Summary: llc: Couldn't allocate output reg for contraint 'q'!
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: normal
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: edwintorok at gmail.com
                CC: llvmbugs at cs.uiuc.edu


kernel/sched.c gives an error in code generator:

$ llvm-gcc testcase-min.i
Couldn't allocate output reg for contraint 'q'!
$ llc bugpoint-reduced-simplified.bc
Couldn't allocate output reg for contraint 'q'!

testcase-min.i:
static inline __attribute__((always_inline)) struct task_struct
*get_current(void) {
     }
       static inline __attribute__((always_inline)) unsigned long
__xchg(unsigned long x, volatile void * ptr, int size) {
      switch (size) {
      case 1:    __asm__ __volatile__("xchgb %b0,%1"     :"=q" (x)     :"m"
(*((volatile long *)(ptr))), "0" (x)     :"memory");
     }
     }
       struct task_struct {
      volatile long state;
     };
       void __attribute__((__section__(".sched.text"))) yield(void) {
      do {
    (void) ((__typeof__(*(&get_current()->state)))__xchg((unsigned
long)((0)),(&get_current()->state),sizeof(*(&get_current()->state))));
    }
     while (0);
     }

bugpoint_reduced_simplified.ll:
; ModuleID = 'bugpoint-reduced-simplified.bc'
target datalayout =
"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128"
target triple = "x86_64-unknown-linux-gnu"

define void @yield() {
entry:
        br i1 false, label %bb, label %return

bb:             ; preds = %entry
        %tmp9 = call i64 asm sideeffect "xchgb ${0:b},$1",
"=q,*m,0,~{dirflag},~{fpsr},~{flags},~{memory}"( i64* null, i64 0 )   ; <i64>
[#uses=0]
        ret void

return:         ; preds = %entry
        ret void
}


$ llc --version
Low Level Virtual Machine (http://llvm.org/):
  llvm version 2.2svn
  DEBUG build with assertions.
$ uname -a
Linux lightspeed2 2.6.23-rc8-hrt1-cfs-v22-g1bef7dc0-dirty #17 Sun Sep 30
18:49:34 EEST 2007 x86_64 GNU/Linux


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sat Nov  3 17:32:01 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sat, 3 Nov 2007 17:32:01 -0500
Subject: [LLVMbugs] [Bug 1764] New: using global symbols in asm with -O1/-Os
 leads to ' symbol already defined' error
Message-ID: <bug-1764-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1764

           Summary: using global symbols in asm with -O1/-Os leads to
                    'symbol already defined' error
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: edwintorok at gmail.com
                CC: llvmbugs at cs.uiuc.edu


Testcase is from kernel/sched.c, when I compile I get an error that
thread_return is already defined.

I am using llvm-gcc and llvm from latest svn version r43663 on x86-64.

$ llvm-gcc -O1  -c testcase-min.i -o foo.bc
/tmp/cck288u7.s: Assembler messages:
/tmp/cck288u7.s:56: Error: symbol `thread_return' is already defined

$ llvm-gcc -Os  -c testcase-min.i -o foo.bc
/tmp/cck288u7.s: Assembler messages:
/tmp/cck288u7.s:56: Error: symbol `thread_return' is already defined

$ llvm-gcc -O0  -c testcase-min.i -o foo.bc

Error is not reproducible if I compile to bitcode, and the optimize/codegen, or
just codegen.

cat testcase-min.i:

 typedef unsigned int __u32;
     struct x8664_pda {
     struct task_struct *pcurrent;
    }
     ;
     struct thread_info {
     __u32 flags;
    };
     struct thread_struct {
     unsigned long rsp;
    }
     __attribute__((aligned(16)));
     struct list_head {
    };
     struct task_struct {
     void *stack;
     struct mm_struct *mm, *active_mm;
     struct thread_struct thread;
    };
     struct rq {
     struct task_struct *curr, *idle;
    };
     static __typeof__(struct rq) per_cpu__runqueues;
     static inline __attribute__((always_inline)) void context_switch(struct rq
*rq, struct task_struct *prev,         struct task_struct *next) {
     if (__builtin_expect(!!(!prev->mm), 0)) {
     prev->active_mm = ((void *)0);
    }
     asm volatile("pushf ; pushq %%rbp ; movq %%rsi,%%rbp\n\t" "movq
%%rsp,%P[threadrsp](%[prev])\n\t" "movq %P[threadrsp](%[next]),%%rsp\n\t" "call
__switch_to\n\t" ".globl thread_return\n" "thread_return:\n\t" "movq
%%gs:%P[pda_pcurrent],%%rsi\n\t" "movq %P[thread_info](%%rsi),%%r8\n\t" "" "btr
 %[tif_fork],%P[ti_flags](%%r8)\n\t" "movq %%rax,%%rdi\n\t" "jc  
ret_from_fork\n\t" "movq %%rbp,%%rsi ; popq %%rbp ; popf\t" : "=a" (prev) :
[next] "S" (next), [prev] "D" (prev), [threadrsp] "i"
(__builtin_offsetof(struct task_struct,thread.rsp)), [ti_flags] "i"
(__builtin_offsetof(struct thread_info,flags)), [tif_fork] "i" (18),
[thread_info] "i" (__builtin_offsetof(struct task_struct,stack)),
[pda_pcurrent] "i" (__builtin_offsetof(struct x8664_pda,pcurrent)) : "memory",
"cc" ,"rcx","rbx","rdx","r8","r9","r10","r11","r12","r13","r14","r15");
    }
      void __attribute__((__section__(".sched.text"))) schedule(void) {
     struct task_struct *prev, *next;
     struct rq *rq;
     int cpu;
     rq = (&(*((void)((cpu)), &per_cpu__runqueues)));
     prev = rq->curr;
     if (__builtin_expect(!!(prev != next), 1)) {
     context_switch(rq, prev, next);
    }
    }


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sat Nov  3 18:49:19 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sat, 3 Nov 2007 18:49:19 -0500
Subject: [LLVMbugs] [Bug 1762] llvm-gcc sees a 128 bit integer,
	but there is none
In-Reply-To: <bug-1762-206@http.llvm.org/bugs/>
Message-ID: <200711032349.lA3NnJDn025352@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1762


Duncan Sands <baldrick at free.fr> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |INVALID




--- Comment #3 from Duncan Sands <baldrick at free.fr>  2007-11-03 18:49:18 ---
PS: On x86-32 I see that 64 bit arithmetic is being used.  Consulting
the gcc tree dumps it seems that this is because it is doing arithmetic
using "bitsizetype", a gcc internal type, which has twice as many bits
as the system word size (so 128 bits on your machine).  I checked that
mainline gcc-4.2 is also using bitsizetype for your testcase.  So while
I understand your surprise at this testcase doing 128 bit arithmetic,
this is really a gcc issue: mainline generates 128 bit arithmetic for it
too, only you never noticed :)  I'm closing this as invalid.  However it
could also be marked a duplicate of the bug asking for 128 bit support
(PR1462) if you like.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov  4 01:17:42 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 4 Nov 2007 01:17:42 -0500
Subject: [LLVMbugs] [Bug 1765] New: condition register bits and registers
	confused
Message-ID: <bug-1765-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1765

           Summary: condition register bits and registers confused
           Product: libraries
           Version: trunk
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: normal
          Priority: P2
         Component: Backend: PowerPC
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: nicholas at mxc.ca
                CC: llvmbugs at cs.uiuc.edu


The PPC backend currently emits code that doesn't distinguish between CR6 and
bit 6 (being a bit in CR1). This breaks PPC/Linux where "creqv 6, 6, 6" is
thought by the backend to be clobbering CR6 instead of CR1EQ.

Often the problem manifests in code like:
  creqv 2, 2, 2
  mr 6, 2
  bl somevararg1
  mr 6, 2
  bl somevararg2

and this crops up in 2005-03-07-VarArgs.c.

I'm not sure what the correct fix looks like. I tried modelling each CR-bit as
an i1 subreg of the CR and made a total mess of things. (I can attach the patch
if anyone wants it.)


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov  4 01:25:35 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 4 Nov 2007 01:25:35 -0500
Subject: [LLVMbugs] [Bug 1764] using global symbols in asm with -O1/-Os
 leads to ' symbol already defined' error
In-Reply-To: <bug-1764-206@http.llvm.org/bugs/>
Message-ID: <200711040625.lA46PZ60010423@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1764


Chris Lattner <sabre at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |INVALID




--- Comment #3 from Chris Lattner <sabre at nondot.org>  2007-11-04 01:25:34 ---
Duncan is right, the code is invalid.  It should use "local labels" with
forward/backward so that each copy of the asm block gets its own unique label.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov  4 01:52:04 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 4 Nov 2007 01:52:04 -0500
Subject: [LLVMbugs] [Bug 1763] 'q' inline asm constraint doesn't work on
	x86-64
In-Reply-To: <bug-1763-206@http.llvm.org/bugs/>
Message-ID: <200711040652.lA46q4GA011737@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1763


Chris Lattner <sabre at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
          Component|new bugs                    |Backend: X86
            Product|new-bugs                    |libraries
         Resolution|                            |FIXED
            Summary|llc: Couldn't allocate      |'q' inline asm constraint
                   |output reg for contraint    |doesn't work on x86-64
                   |'q'!                        |
   Target Milestone|---                         |2.2
            Version|unspecified                 |1.0




--- Comment #1 from Chris Lattner <sabre at nondot.org>  2007-11-04 01:52:03 ---
Fixed, patch here:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071029/055109.html

Thanks!

-Chris


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov  4 02:53:48 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 4 Nov 2007 02:53:48 -0600
Subject: [LLVMbugs] [Bug 1766] New: Assertion `Reg && "this is not a
	register!"' failed
Message-ID: <bug-1766-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1766

           Summary: Assertion `Reg && "this is not a register!"' failed
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: edwintorok at gmail.com
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=1187)
 --> (http://llvm.org/bugs/attachment.cgi?id=1187)
llc bugpoint-reduced-simplified.bc -f gives assertion failure

Compiling the attached reduced C testcase with llvm-gcc -O1 I get that
assertion failure.
If I compile without -O1 I don't get the error.
If I compile to bitcode, and then generate code I don't get the error.
If I also optimize the bitcode, I get the error at codegen time.

$ llc bugpoint-reduced-simplified.bc -f
llc: /home/edwin/llvm-svn/llvm/include/llvm/Target/MRegisterInfo.h:318: static
bool llvm::MRegisterInfo::isVirtualRegister(unsigned int): Assertion `Reg &&
"this is not a register!"' failed.
llc[0xce97ea]
llc[0xce9a56]
/lib/libc.so.6[0x32d1c31710]
/lib/libc.so.6(gsignal+0x35)[0x32d1c316a5]
/lib/libc.so.6(abort+0x110)[0x32d1c33100]
/lib/libc.so.6(__assert_fail+0xef)[0x32d1c2ac2f]
llc(llvm::MRegisterInfo::isVirtualRegister(unsigned int)+0x2a)[0x800990]
llc(llvm::MRegisterInfo::regsOverlap(unsigned int, unsigned int)
const+0x2b)[0xb5b397]
llc(llvm::LiveIntervals::conflictsWithPhysRegDef(llvm::LiveInterval const&,
llvm::VirtRegMap&, unsigned int)+0x13f)[0xb7811b]
llc[0xb40721]
llc[0xb40d47]
llc[0xb41068]
llc(llvm::MachineFunctionPass::runOnFunction(llvm::Function&)+0x33)[0x773699]
llc(llvm::FPPassManager::runOnFunction(llvm::Function&)+0x10e)[0xc7bf22]
llc(llvm::FunctionPassManagerImpl::run(llvm::Function&)+0x74)[0xc7c148]
llc(llvm::FunctionPassManager::run(llvm::Function&)+0x7f)[0xc7c2a5]
llc(main+0x8ac)[0x7593be]
/lib/libc.so.6(__libc_start_main+0xf4)[0x32d1c1db44]
llc(atan+0x139)[0x757c29]
Aborted

Attaching testcases.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov  4 03:15:21 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 4 Nov 2007 03:15:21 -0600
Subject: [LLVMbugs] [Bug 1767] New: llc segfaults when compiling optimized
	sse2 assembly
Message-ID: <bug-1767-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1767

           Summary: llc segfaults when compiling optimized sse2 assembly
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: normal
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: edwintorok at gmail.com
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=1190)
 --> (http://llvm.org/bugs/attachment.cgi?id=1190)
llvm-gcc -O1 -c testcase-min.i

kernel's crypto/xor.c:

$ llvm-gcc -O1 -c xor.i
crypto/xor.c:155: internal compiler error: Segmentation fault
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://llvm.org/bugs> for instructions.


$ llvm-gcc -emit-llvm -c xor.i -o xor.bc
$ opt -std-compile-opts xor.bc -o xor_o.bc
$ llc xor_o.bc
llc[0xce97ea]
llc[0xce9a56]
/lib/libc.so.6[0x32d1c31710]
llc(llvm::SimpleRegisterCoalescing::JoinIntervals(llvm::LiveInterval&,
llvm::LiveInterval&, bool&)+0x88c)[0xb539ba]
llc(llvm::SimpleRegisterCoalescing::JoinCopy(llvm::MachineInstr*, unsigned int,
unsigned int, bool&)+0xfe7)[0xb55853]
llc(llvm::SimpleRegisterCoalescing::CopyCoalesceInMBB(llvm::MachineBasicBlock*,
std::vector<llvm::SimpleRegisterCoalescing::CopyRec,
std::allocator<llvm::SimpleRegisterCoalescing::CopyRec> >&)+0x2e8)[0xb5620e]
llc(llvm::SimpleRegisterCoalescing::joinIntervals()+0x12b)[0xb5640d]
llc(llvm::SimpleRegisterCoalescing::runOnMachineFunction(llvm::MachineFunction&)+0x2cf)[0xb56bc9]
llc(llvm::MachineFunctionPass::runOnFunction(llvm::Function&)+0x33)[0x773699]
llc(llvm::FPPassManager::runOnFunction(llvm::Function&)+0x10e)[0xc7bf22]
llc(llvm::FunctionPassManagerImpl::run(llvm::Function&)+0x74)[0xc7c148]
llc(llvm::FunctionPassManager::run(llvm::Function&)+0x7f)[0xc7c2a5]
llc(main+0x8ac)[0x7593be]
/lib/libc.so.6(__libc_start_main+0xf4)[0x32d1c1db44]
llc(atan+0x139)[0x757c29]
Segmentation fault

Attaching reduced testcases. Let me know if you need something more.
I am not sure if the optimized bitcode is valid.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov  4 04:29:36 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 4 Nov 2007 04:29:36 -0600
Subject: [LLVMbugs] [Bug 1768] New: correlatedexpression optimization needs
 2 passes, and misses optimization due to function parameter order
Message-ID: <bug-1768-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1768

           Summary: correlatedexpression optimization needs 2 passes, and
                    misses optimization due to function parameter order
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: minor
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: edwintorok at gmail.com
                CC: llvmbugs at cs.uiuc.edu


Consider these 2 functions opt_fail and opt_ok, where only the order of the
parameters is different.
In opt_fail's case llvm doesn't recognize the optimization opportunity in
cond_true, while it recognizes it in opt_ok.
Also it takes an additional run of opt to actually optimize out the sub
instruction in opt_ok.

I think opt should optimize out the sub in one pass for opt_ok.
For opt_fail: if i==j, it should recognize that j==i too (unless this would
cause infinite loops).

; ModuleID = 'foo.bc'
target datalayout =
"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128"
target triple = "x86_64-unknown-linux-gnu"

define i32 @opt_fail(i32 %i, i32 %j) {
entry:
        %tmp7 = icmp eq i32 %i, %j              ; <i1> [#uses=1]
        br i1 %tmp7, label %cond_true, label %UnifiedReturnBlock1

cond_true:              ; preds = %entry
        %tmp11 = sub i32 %i, %j         ; <i32> [#uses=1]
        ret i32 %tmp11

UnifiedReturnBlock1:            ; preds = %entry
        ret i32 5
}

define i32 @opt_ok(i32 %j, i32 %i) {
entry:
        %tmp7 = icmp eq i32 %i, %j              ; <i1> [#uses=1]
        br i1 %tmp7, label %cond_true, label %UnifiedReturnBlock1

cond_true:              ; preds = %entry
        %tmp11 = sub i32 %i, %j         ; <i32> [#uses=1]
        ret i32 %tmp11

UnifiedReturnBlock1:            ; preds = %entry
        ret i32 5
}




$ opt -cee -std-compile-opts -debug-only=cee foo.bc -o foo2.bc -f
 RegionInfo for basic block: cond_true
  ValueInfo for: i32 %i
    Bounds = [-1,-1 )
    is == i32 %j
  ValueInfo for: i32 %j
    Bounds = [-1,-1 )
    is == i32 %j
  ValueInfo for: i1 %tmp7
    Bounds = [-1,0 )
    Replacement = i1 true
    is == i1 true

 RegionInfo for basic block: UnifiedReturnBlock
  ValueInfo for: i32 %i
    Bounds = [-1,-1 )
    is != i32 %j
  ValueInfo for: i32 %j
    Bounds = [-1,-1 )
    is != i32 %i
  ValueInfo for: i1 %tmp7
    Bounds = [0,-1 )
    Replacement = i1 false
    is == i1 false

 RegionInfo for basic block: cond_true
  ValueInfo for: i1 %tmp7
    Bounds = [-1,0 )
    Replacement = i1 true
    is == i1 true
  ValueInfo for: i32 %j
    Bounds = [-1,-1 )
    is == i32 %j
  ValueInfo for: i32 %i
    Bounds = [-1,-1 )
    Replacement = i32 %j
    is == i32 %j

In Inst:        %tmp11 = sub i32 %i, %j         ; <i32> [#uses=1]
  Replacing operand #0 with i32 %j
 RegionInfo for basic block: UnifiedReturnBlock
  ValueInfo for: i1 %tmp7
    Bounds = [0,-1 )
    Replacement = i1 false
    is == i1 false
  ValueInfo for: i32 %j
    Bounds = [-1,-1 )
    is != i32 %i
  ValueInfo for: i32 %i
    Bounds = [-1,-1 )
    is != i32 %j


; ModuleID = 'foo2.bc'
target datalayout =
"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128"
target triple = "x86_64-unknown-linux-gnu"

define i32 @opt_fail(i32 %i, i32 %j) {
entry:
        %tmp7 = icmp eq i32 %i, %j              ; <i1> [#uses=1]
        br i1 %tmp7, label %cond_true, label %UnifiedReturnBlock

cond_true:              ; preds = %entry
        %tmp11 = sub i32 %i, %j         ; <i32> [#uses=1]
        ret i32 %tmp11

UnifiedReturnBlock:             ; preds = %entry
        ret i32 5
}

define i32 @opt_ok(i32 %j, i32 %i) {
entry:
        %tmp7 = icmp eq i32 %i, %j              ; <i1> [#uses=1]
        br i1 %tmp7, label %cond_true, label %UnifiedReturnBlock

cond_true:              ; preds = %entry
        %tmp11 = sub i32 %j, %j         ; <i32> [#uses=1]
        ret i32 %tmp11

UnifiedReturnBlock:             ; preds = %entry
        ret i32 5
}


$ opt -std-compile-opts foo2.bc -o foo3.bc -f
 ModuleID = 'foo3.bc'
target datalayout =
"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128"
target triple = "x86_64-unknown-linux-gnu"

define i32 @opt_fail(i32 %i, i32 %j) {
entry:
        %tmp7 = icmp eq i32 %i, %j              ; <i1> [#uses=1]
        br i1 %tmp7, label %cond_true, label %UnifiedReturnBlock1

cond_true:              ; preds = %entry
        %tmp11 = sub i32 %i, %j         ; <i32> [#uses=1]
        ret i32 %tmp11

UnifiedReturnBlock1:            ; preds = %entry
        ret i32 5
}

define i32 @opt_ok(i32 %j, i32 %i) {
entry:
        %tmp7 = icmp eq i32 %i, %j              ; <i1> [#uses=1]
        %retval = select i1 %tmp7, i32 0, i32 5         ; <i32> [#uses=1]
        ret i32 %retval
}


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov  4 05:04:08 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 4 Nov 2007 05:04:08 -0600
Subject: [LLVMbugs] [Bug 1769] New: opt -cee crashes with " Instruction does
 not dominate all uses!"
Message-ID: <bug-1769-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1769

           Summary: opt -cee crashes with "Instruction does not dominate all
                    uses!"
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: normal
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: edwintorok at gmail.com
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=1194)
 --> (http://llvm.org/bugs/attachment.cgi?id=1194)
bugpoint simplified testcase, crashes on opt -cee

Standard optimizations, and code generation works, I assume the testcase
is valid llvm bitcode:
$ opt -verify -verify-each -std-compile-opts bugpoint-reduced-simplified.bc  -o
x.bc -f
$ llc bugpoint-reduced-simplified.bc -f

However opt -cee crashes:
$ opt -cee bugpoint-reduced-simplified.bc  -o x.bc -f
Instruction does not dominate all uses!
        %tmp604683 = icmp sgt i64 %tmp460, 0            ; <i1> [#uses=4]
        %tmp604683.fw_frontier = phi i1 [ %tmp604683.fw_merge1,
%bb482.bb484_crit_edge ], [ %tmp604683, %cond_next473 ]          ; <i1>
[#uses=0]
Broken module found, compilation aborted!
opt[0x9c16ba]
opt[0x9c1926]
/lib/libc.so.6[0x32d1c31710]
/lib/libc.so.6(gsignal+0x35)[0x32d1c316a5]
/lib/libc.so.6(abort+0x110)[0x32d1c33100]
opt[0x978940]
opt[0x982aa4]
opt(llvm::FPPassManager::runOnFunction(llvm::Function&)+0x10e)[0x953f92]
opt(llvm::FPPassManager::runOnModule(llvm::Module&)+0x52)[0x954100]
opt(llvm::MPPassManager::runOnModule(llvm::Module&)+0xec)[0x953c8c]
opt(llvm::PassManagerImpl::run(llvm::Module&)+0x74)[0x953e1c]
opt(llvm::PassManager::run(llvm::Module&)+0x20)[0x953e7e]
opt(main+0x986)[0x6bfe78]
/lib/libc.so.6(__libc_start_main+0xf4)[0x32d1c1db44]
opt(cos+0x119)[0x6b1609]
Aborted


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov  4 06:13:50 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 4 Nov 2007 06:13:50 -0600
Subject: [LLVMbugs] [Bug 1770] New: opt -loopunroll crash
Message-ID: <bug-1770-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1770

           Summary: opt -loopunroll crash
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: normal
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: edwintorok at gmail.com
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=1195)
 --> (http://llvm.org/bugs/attachment.cgi?id=1195)
simplified testcase

$ opt bugpoint-reduced-simplified.bc -loop-unroll -o f.bc -f
opt[0x9c16ba]
opt[0x9c1926]
/lib/libc.so.6[0x32d1c31710]
opt(llvm::PATypeHolder::get() const+0x10)[0x961b5c]
opt(llvm::PATypeHolder::operator llvm::Type*() const+0x15)[0x6d7037]
opt(llvm::Value::getType() const+0x19)[0x6d7053]
opt(llvm::PHINode::addIncoming(llvm::Value*, llvm::BasicBlock*)+0x2a)[0x6fc1f2]
opt[0x7c7eae]
opt[0x7c873c]
opt(llvm::LPPassManager::runOnFunction(llvm::Function&)+0x3b8)[0x8979ce]
opt(llvm::FPPassManager::runOnFunction(llvm::Function&)+0x10e)[0x953f92]
opt(llvm::FPPassManager::runOnModule(llvm::Module&)+0x52)[0x954100]
opt(llvm::MPPassManager::runOnModule(llvm::Module&)+0xec)[0x953c8c]
opt(llvm::PassManagerImpl::run(llvm::Module&)+0x74)[0x953e1c]
opt(llvm::PassManager::run(llvm::Module&)+0x20)[0x953e7e]
opt(main+0x986)[0x6bfe78]
/lib/libc.so.6(__libc_start_main+0xf4)[0x32d1c1db44]
opt(cos+0x119)[0x6b1609]
Segmentation fault

Attached bugpoint-reduced-simplified.ll


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov  4 10:55:18 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 4 Nov 2007 10:55:18 -0600
Subject: [LLVMbugs] [Bug 1771] New: symbols not in .bss section on ppc/linux
Message-ID: <bug-1771-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1771

           Summary: symbols not in .bss section on ppc/linux
           Product: libraries
           Version: trunk
          Platform: Macintosh
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: Backend: PowerPC
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: nicholas at mxc.ca
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=1197)
 --> (http://llvm.org/bugs/attachment.cgi?id=1197)
Put external linkage symbols in .bss

Linux expects to see symbols with initializers in the .bss (or .sbss) sections.
I haven't been able to find an ABI document describing the correct behaviour,
so I'm experimenting on GCC.

The attached patch makes LLVM more closely match GCC's behaviour on PPC/Linux.
I don't think it breaks anything, and fixes programs which #include <iostream>


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov  4 11:32:29 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 4 Nov 2007 11:32:29 -0600
Subject: [LLVMbugs] [Bug 1771] symbols not in .bss section on ppc/linux
In-Reply-To: <bug-1771-206@http.llvm.org/bugs/>
Message-ID: <200711041732.lA4HWT7L026776@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1771


Nick Lewycky <nicholas at mxc.ca> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #2 from Nick Lewycky <nicholas at mxc.ca>  2007-11-04 11:32:28 ---
Done. Thanks Chris!


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov  4 11:50:02 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 4 Nov 2007 11:50:02 -0600
Subject: [LLVMbugs] [Bug 1772] New: [meta] Missed optimizations with
 correlated/ related expressions
Message-ID: <bug-1772-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1772

           Summary: [meta] Missed optimizations with correlated/related
                    expressions
           Product: libraries
           Version: 2.1
          Platform: PC
        OS/Version: All
            Status: NEW
          Keywords: code-quality
          Severity: enhancement
          Priority: P2
         Component: Scalar Optimizations
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: sabre at nondot.org
                CC: llvmbugs at cs.uiuc.edu


This bug is intended to track missed optimizations related to expressions that
are related to each other, that the LLVM optimizer does not remove.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov  4 13:24:14 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 4 Nov 2007 13:24:14 -0600
Subject: [LLVMbugs] [Bug 1761] llc generates invalid assembly: junk at end
 of line, first unrecognized character is `('
In-Reply-To: <bug-1761-206@http.llvm.org/bugs/>
Message-ID: <200711041924.lA4JOEKI032170@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1761


Chris Lattner <sabre at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
          Component|new bugs                    |Backend: X86
            Product|new-bugs                    |libraries
         Resolution|                            |FIXED
   Target Milestone|---                         |2.2
            Version|unspecified                 |1.0




--- Comment #11 from Chris Lattner <sabre at nondot.org>  2007-11-04 13:24:13 ---
Ahh, I see.  This should work then, please verify:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071029/055123.html

-Chris


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov  4 14:15:22 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 4 Nov 2007 14:15:22 -0600
Subject: [LLVMbugs] [Bug 1773] New: llc generates movl access to %rsi
	instead of movq
Message-ID: <bug-1773-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1773

           Summary: llc generates movl access to %rsi instead of movq
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: normal
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: edwintorok at gmail.com
                CC: llvmbugs at cs.uiuc.edu


$ llvm-gcc testcase-min.i
/tmp/cczzO3SA.s: Assembler messages:
/tmp/cczzO3SA.s:14: Error: Incorrect register `%rsi' used with `l' suffix

I had no luck reproducing this by compiling to bitcode and then generating
code,
so here it is the assembly output.

Notice "movl %gs:0,%rsi", which should be a movq...

$ llvm-gcc testcase-min.i -S -o -
.ident  "GCC: (GNU) 4.0.1 (Apple Computer, Inc. build 5449)(LLVM build )"


        .text
        .align  16
        .globl  get_current
        .type   get_current, at function
get_current:
        pushq   %rbp
        movq    %rsp, %rbp
        subq    $40, %rsp
        jmp     .LBB1_2 # bb4
.LBB1_1:        # bb
        #APP
        movl %gs:0,%rsi
        #NO_APP
        movq    %rsi, -40(%rbp)
.LBB1_2:        # bb4
        movq    -40(%rbp), %rsi
        movq    %rsi, -16(%rbp)
        movq    %rsi, -32(%rbp)
        movq    %rsi, -24(%rbp)
        movq    %rsi, -8(%rbp)
.LBB1_3:        # return
        movq    -8(%rbp), %rsi
        movq    %rsi, %rax
        addq    $40, %rsp
        popq    %rbp
        ret
        .size   get_current, .-get_current


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov  4 14:37:22 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 4 Nov 2007 14:37:22 -0600
Subject: [LLVMbugs] [Bug 1774] New: Assertion `Offset1<Offset2 && " There is
 at least one different constant here!"' failed.
Message-ID: <bug-1774-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1774

           Summary: Assertion `Offset1<Offset2 && "There is at least one
                    different constant here!"' failed.
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: edwintorok at gmail.com
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=1199)
 --> (http://llvm.org/bugs/attachment.cgi?id=1199)
testcase

$ opt -gvn bugpoint-reduced-simplified.bc
WARNING: You're attempting to print out a bitcode file.
This is inadvisable as it may cause display problems. If
you REALLY want to taste LLVM bitcode first-hand, you
can force output with the `-f' option.

opt: BasicAliasAnalysis.cpp:798:
llvm::AliasAnalysis::AliasResult<unnamed>::BasicAliasAnalysis::CheckGEPInstructions(const
llvm::Type*, llvm::Value**, unsigned int, unsigned int, const llvm::Type*,
llvm::Value**, unsigned int, unsigned int): Assertion `Offset1<Offset2 &&
"There is at least one different constant here!"' failed.
opt[0x9c16ba]
opt[0x9c1926]
/lib/libc.so.6[0x32d1c31710]
/lib/libc.so.6(gsignal+0x35)[0x32d1c316a5]
/lib/libc.so.6(abort+0x110)[0x32d1c33100]
/lib/libc.so.6(__assert_fail+0xef)[0x32d1c2ac2f]
opt[0x8819dc]
opt[0x88253d]
opt[0x881df8]
opt[0x881d8b]
opt(llvm::MemoryDependenceAnalysis::getDependency(llvm::Instruction*,
llvm::Instruction*, llvm::BasicBlock*)+0x963)[0x89c92f]
opt(llvm::MemoryDependenceAnalysis::nonLocalHelper(llvm::Instruction*,
llvm::BasicBlock*, llvm::DenseMap<llvm::BasicBlock*, llvm::Value*,
llvm::DenseMapInfo<llvm::BasicBlock*>, llvm::DenseMapInfo<llvm::Value*>
>&)+0x199)[0x89cd13]
opt(llvm::MemoryDependenceAnalysis::getNonLocalDependency(llvm::Instruction*,
llvm::DenseMap<llvm::BasicBlock*, llvm::Value*,
llvm::DenseMapInfo<llvm::BasicBlock*>, llvm::DenseMapInfo<llvm::Value*>
>&)+0x21a)[0x89d23c]
opt[0x74ffa1]
opt[0x7504e7]
opt[0x7506e2]
opt[0x750a5c]
opt[0x750b94]
opt(llvm::FPPassManager::runOnFunction(llvm::Function&)+0x10e)[0x953f92]
opt(llvm::FPPassManager::runOnModule(llvm::Module&)+0x52)[0x954100]
opt(llvm::MPPassManager::runOnModule(llvm::Module&)+0xec)[0x953c8c]
opt(llvm::PassManagerImpl::run(llvm::Module&)+0x74)[0x953e1c]
opt(llvm::PassManager::run(llvm::Module&)+0x20)[0x953e7e]
opt(main+0x986)[0x6bfe78]
/lib/libc.so.6(__libc_start_main+0xf4)[0x32d1c1db44]
opt(cos+0x119)[0x6b1609]
Aborted

$ llvm-gcc -O1 -c testcase-min.i
testcase-min.i:43: warning: ???struct urb??? declared inside parameter list
testcase-min.i:43: warning: its scope is only this definition or declaration,
which is probably not what you want
testcase-min.i:44: warning: ???struct urb??? declared inside parameter list
testcase-min.i:45: warning: ???struct usb_host_endpoint??? declared inside
parameter list
testcase-min.i:69: warning: ???struct urb??? declared inside parameter list
testcase-min.i:71: warning: ???struct urb??? declared inside parameter list
testcase-min.i:99: warning: ???struct usb_host_endpoint??? declared inside
parameter list
testcase-min.i:105: warning: initialization from incompatible pointer type
testcase-min.i:105: warning: initialization from incompatible pointer type
testcase-min.i:105: warning: initialization from incompatible pointer type
cc1: BasicAliasAnalysis.cpp:798:
llvm::AliasAnalysis::AliasResult<unnamed>::BasicAliasAnalysis::CheckGEPInstructions(const
llvm::Type*, llvm::Value**, unsigned int, unsigned int, const llvm::Type*,
llvm::Value**, unsigned int, unsigned int): Assertion `Offset1<Offset2 &&
"There is at least one different constant here!"' failed.
testcase-min.i:110: internal compiler error: Aborted
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://llvm.org/bugs> for instructions.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov  4 19:00:53 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 4 Nov 2007 19:00:53 -0600
Subject: [LLVMbugs] [Bug 1766] Assertion `Reg && "this is not a register!"'
	failed
In-Reply-To: <bug-1766-206@http.llvm.org/bugs/>
Message-ID: <200711050100.lA510rYC015575@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1766


Evan Cheng <evan.cheng at apple.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #4 from Evan Cheng <evan.cheng at apple.com>  2007-11-04 19:00:53 ---
Fixed.
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071029/055137.html


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov  4 19:38:26 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 4 Nov 2007 19:38:26 -0600
Subject: [LLVMbugs] [Bug 1775] New: ExtractMappedBlocksFromModule discards
 the " blocks to not extract" list
Message-ID: <bug-1775-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1775

           Summary: ExtractMappedBlocksFromModule discards the "blocks to
                    not extract" list
           Product: tools
           Version: trunk
          Platform: All
        OS/Version: All
            Status: NEW
          Severity: normal
          Priority: P2
         Component: bugpoint
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: gordonhenriksen at mac.com
                CC: llvmbugs at cs.uiuc.edu


In tools/bugpoint/ExtractFunction.cpp:ExtractMappedBlocksFromModule, bugpoint
is discarding the "blocks to not extract" list since it's doing a fork+exec to
run the block extractor pass. bugpoint used to just fork, so it used to work.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov  4 20:09:40 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 4 Nov 2007 20:09:40 -0600
Subject: [LLVMbugs] [Bug 1773] llc generates movl access to %rsi instead of
	movq
In-Reply-To: <bug-1773-206@http.llvm.org/bugs/>
Message-ID: <200711050209.lA529eZd018387@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1773


Evan Cheng <evan.cheng at apple.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |evan.cheng at apple.com
             Status|NEW                         |RESOLVED
         Resolution|                            |INVALID




--- Comment #2 from Evan Cheng <evan.cheng at apple.com>  2007-11-04 20:09:40 ---
I think this is pilot error. Here is the reduced test case:

void f() {
  void *ret__;
  asm volatile("mov" "l %%gs:0,%0": "=r" (ret__));
}

x86-64 pointer type is i64 so a 64-bit register is assigned.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov  4 21:14:49 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 4 Nov 2007 21:14:49 -0600
Subject: [LLVMbugs] [Bug 1767] llc segfaults when compiling optimized sse2
	assembly
In-Reply-To: <bug-1767-206@http.llvm.org/bugs/>
Message-ID: <200711050314.lA53EnQl021034@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1767


Evan Cheng <evan.cheng at apple.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED




--- Comment #6 from Evan Cheng <evan.cheng at apple.com>  2007-11-04 21:14:48 ---
Fixed.
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071029/055144.html


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov  4 21:15:21 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 4 Nov 2007 21:15:21 -0600
Subject: [LLVMbugs] [Bug 770] Permit coallescing of registers in subset
	register classes
In-Reply-To: <bug-770-206@http.llvm.org/bugs/>
Message-ID: <200711050315.lA53FLte021073@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=770


Evan Cheng <evan.cheng at apple.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|REOPENED                    |RESOLVED
         Resolution|                            |FIXED




--- Comment #3 from Evan Cheng <evan.cheng at apple.com>  2007-11-04 21:15:20 ---
This is now done.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Mon Nov  5 01:16:08 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Mon, 5 Nov 2007 01:16:08 -0600
Subject: [LLVMbugs] [Bug 1776] New: movaps being generated despite alignment
 1 being specified
Message-ID: <bug-1776-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1776

           Summary: movaps being generated despite alignment 1 being
                    specified
           Product: libraries
           Version: trunk
          Platform: PC
        OS/Version: All
            Status: NEW
          Severity: major
          Priority: P2
         Component: Backend: X86
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: evan.cheng at apple.com
                CC: llvmbugs at cs.uiuc.edu


Reported by Chuck Rose. See
http://lists.cs.uiuc.edu/pipermail/llvmdev/2007-October/011167.html

JITTed code contains a movaps which is bad news for targets where stack is not
being aligned on 16-byte boundary"

15D10012  sub         esp,4Ch
15D10015  mov         eax,dword ptr [esp+60h]
15D10019  movups      xmm0,xmmword ptr [eax]
15D1001C  movaps      xmmword ptr [esp+8],xmm0    ?? why did this become a
movaps?
15D10021  movups      xmmword ptr [esp+28h],xmm0
15D10026  mov         esi,dword ptr [esp+58h]
15D1002A  mov         edi,dword ptr [esp+5Ch]
15D1002E  mov         dword ptr [esp],40000000h
15D10035  call        X86CompilationCallback (1335030h)

The movaps is generated for spill /restore.

Windows align stack on 8-byte boundary, so until the x86 target can align stack
for functions that use SSE on 16-byte on 16-byte, it should use movups for SSE
register spill / restore.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Mon Nov  5 01:32:16 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Mon, 5 Nov 2007 01:32:16 -0600
Subject: [LLVMbugs] [Bug 1776] movaps being generated despite alignment 1
 being specified
In-Reply-To: <bug-1776-206@http.llvm.org/bugs/>
Message-ID: <200711050732.lA57WGoj001199@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1776


Evan Cheng <evan.cheng at apple.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #1 from Evan Cheng <evan.cheng at apple.com>  2007-11-05 01:32:15 ---
Fixed.
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071105/055148.html

We are using movups for SSE spill / restore on targets where stack alignment is
< 16. This should be reverted once dynamic stack alignment is implemented (i.e.
PR1636).


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Mon Nov  5 01:49:07 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Mon, 5 Nov 2007 01:49:07 -0600
Subject: [LLVMbugs] [Bug 1777] New: Optimizer not handling phi nodes for
 specific control flow
Message-ID: <bug-1777-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1777

           Summary: Optimizer not handling phi nodes for specific control
                    flow
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Windows XP
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: sharparrow1 at yahoo.com
                CC: llvmbugs at cs.uiuc.edu


target datalayout = "e-p:32:32-f64:32:64-i64:32:64-f80:32:32"

declare i1 @rrr()

define i1 @zxcv() {
entry:
%a = alloca i32
%i = ptrtoint i32* %a to i32
%b = call i1 @rrr()
br i1 %b, label %one, label %two

one:
%x = phi i32 [%i, %entry], [%y, %two]
%c = call i1 @rrr()
br i1 %c, label %two, label %end

two:
%y = phi i32 [%i, %entry], [%x, %one]
%d = call i1 @rrr()
br i1 %d, label %one, label %end

end:
%f = phi i32 [ %x, %one], [%y, %two]
; Change the %f to %i, and the optimizer suddenly becomes a lot smarter
; even though %f must equal %i at this point
%g = inttoptr i32 %f to i32*
store i32 10, i32* %g
%z = call i1 @rrr()
ret i1 %z
}

opt -std-compile-opts does nothing to this code, whereas if the %f is changed
to %i, the alloca+inttoptr+ptrtoint+store+phi all get eliminated.

I'm not sure how the optimizer should be dealing with this.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Mon Nov  5 11:06:32 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Mon, 5 Nov 2007 11:06:32 -0600
Subject: [LLVMbugs] [Bug 1778] New: XPASS
Message-ID: <bug-1778-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1778

           Summary: XPASS
           Product: Test Suite
           Version: trunk
          Platform: PC
        OS/Version: All
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: DejaGNU
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: dpatel at apple.com
                CC: llvmbugs at cs.uiuc.edu


This is an enhancement request to implement XPASS support to enable test on
selected platform.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Mon Nov  5 13:33:00 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Mon, 5 Nov 2007 13:33:00 -0600
Subject: [LLVMbugs] [Bug 1770] opt -loopunroll crash
In-Reply-To: <bug-1770-206@http.llvm.org/bugs/>
Message-ID: <200711051933.lA5JX084011835@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1770


Devang Patel <dpatel at apple.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #2 from Devang Patel <dpatel at apple.com>  2007-11-05 13:33:00 ---
Fixed.
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071105/055169.html


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Mon Nov  5 23:59:55 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Mon, 5 Nov 2007 23:59:55 -0600
Subject: [LLVMbugs] [Bug 1774] basicaa crash on array index with value so
 large that its scaled value becomes negative
In-Reply-To: <bug-1774-206@http.llvm.org/bugs/>
Message-ID: <200711060559.lA65xt9S012700@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1774


Chris Lattner <sabre at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
          Component|new bugs                    |Global Analyses
           Keywords|                            |compile-fail
            Product|new-bugs                    |libraries
         Resolution|                            |FIXED
            Summary|Assertion `Offset1<Offset2  |basicaa crash on array index
                   |&& "There is at least one   |with value so large that its
                   |different constant here!"'  |scaled value becomes
                   |failed.                     |negative
   Target Milestone|---                         |2.2
            Version|unspecified                 |trunk




--- Comment #2 from Chris Lattner <sabre at nondot.org>  2007-11-05 23:59:54 ---
Fixed, patch here:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071105/055195.html

Thanks!

-Chris


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Tue Nov  6 12:37:31 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Tue, 6 Nov 2007 12:37:31 -0600
Subject: [LLVMbugs] [Bug 1654] Broken arm_struct_value_rtx confuses
 objc_act.c in LLVM-GCC 4.0
In-Reply-To: <bug-1654-206@http.llvm.org/bugs/>
Message-ID: <200711061837.lA6IbVjH022636@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1654


Dale Johannesen <dalej at apple.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #2 from Dale Johannesen <dalej at apple.com>  2007-11-06 12:37:30 ---
I think the patch I checked in as 43768 fixes the problem and is cleaner. 
Patrick, could you verify this works for you?  Thanks for the report.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Tue Nov  6 15:17:40 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Tue, 6 Nov 2007 15:17:40 -0600
Subject: [LLVMbugs] [Bug 1779] New: Assertion
 `!OpInfo.AssignedRegs.Regs.empty() && "Couldn' t allocate input reg!"'
 failed
Message-ID: <bug-1779-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1779

           Summary: Assertion `!OpInfo.AssignedRegs.Regs.empty() &&
                    "Couldn't allocate input reg!"' failed
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: normal
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: edwintorok at gmail.com
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=1201)
 --> (http://llvm.org/bugs/attachment.cgi?id=1201)
testcase

On a 32-bit machine (!important, the inline assembly is for 32-bit):
$ llvm-gcc testcase-min.i
cc1: SelectionDAGISel.cpp:3732: void
llvm::SelectionDAGLowering::visitInlineAsm(llvm::CallInst&): Assertion
`!OpInfo.AssignedRegs.Regs.empty() && "Couldn't allocate input reg!"' failed.
testcase-min.i:47: internal compiler error: Aborted
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://llvm.org/bugs> for instructions.

$ llc bugpoint-reduced-simplified.bc
llc: SelectionDAGISel.cpp:3726: void
llvm::SelectionDAGLowering::visitInlineAsm(llvm::CallInst&): Assertion
`!OpInfo.AssignedRegs.Regs.empty() && "Couldn't allocate input reg!"' failed.
llc[0x86e7f7c]
/lib/i686/cmov/libc.so.6(abort+0x101)[0xb7ce2181]
/lib/i686/cmov/libc.so.6(__assert_fail+0xee)[0xb7cd9b6e]
llc(llvm::SelectionDAGLowering::visitInlineAsm(llvm::CallInst&)+0x2217)[0x84566e7]

Note: the testcase contains the __builtin_choose_expr suggested by Chris
Lattner (thanks!). This bug has nothing to do with that, if you remove it, the
bug is still there.
I left it there though, because it illustrates the operand sizes.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Tue Nov  6 16:33:51 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Tue, 6 Nov 2007 16:33:51 -0600
Subject: [LLVMbugs] [Bug 1778] XPASS
In-Reply-To: <bug-1778-206@http.llvm.org/bugs/>
Message-ID: <200711062233.lA6MXpMb006158@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1778


Tanya Lattner <tonic at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #6 from Tanya Lattner <tonic at nondot.org>  2007-11-06 16:33:51 ---
Implemented:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071105/055224.html


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Wed Nov  7 04:28:43 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Wed, 7 Nov 2007 04:28:43 -0600
Subject: [LLVMbugs] [Bug 1777] Optimizer not handling phi nodes for specific
 control flow
In-Reply-To: <bug-1777-206@http.llvm.org/bugs/>
Message-ID: <200711071028.lA7AShaF021809@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1777


Wojciech Matyjewicz <wmatyjewicz at fastmail.fm> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|wmatyjewicz at fastmail.fm     |
             Status|NEW                         |RESOLVED
         OS/Version|Windows XP                  |All
         Resolution|                            |FIXED




--- Comment #9 from Wojciech Matyjewicz <wmatyjewicz at fastmail.fm>  2007-11-07 04:28:38 ---
That fixes it with much simpler idea than mine;)


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Wed Nov  7 14:48:45 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Wed, 7 Nov 2007 14:48:45 -0600
Subject: [LLVMbugs] [Bug 1780] New: instcombine crash
Message-ID: <bug-1780-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1780

           Summary: instcombine crash
           Product: libraries
           Version: trunk
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: Scalar Optimizations
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: alenhar2 at uiuc.edu
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=1204)
 --> (http://llvm.org/bugs/attachment.cgi?id=1204)
reduced file

Instcombine crash in file from the linux kernel (x86, llvm head).


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Wed Nov  7 15:02:38 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Wed, 7 Nov 2007 15:02:38 -0600
Subject: [LLVMbugs] [Bug 1781] New: memcpy Defaulted to Alignment 1
Message-ID: <bug-1781-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1781

           Summary: memcpy Defaulted to Alignment 1
           Product: new-bugs
           Version: unspecified
          Platform: Macintosh
        OS/Version: All
            Status: NEW
          Severity: normal
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: isanbard at gmail.com
                CC: llvmbugs at cs.uiuc.edu


To get around a major problem with the FE (4.0) and the alignment of src & dest
pointers to a memcpy, we default the alignment to 1 in all cases. This is a
hack and we should find a real fix for this. Basically, both the src & dest
pointers' alignments need to be taken into account when generating the
alignment for the memcpy; picking the smallest alignment.

See CFrontend/2007-11-07-AlignedMemcpy.c


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Wed Nov  7 15:44:09 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Wed, 7 Nov 2007 15:44:09 -0600
Subject: [LLVMbugs] [Bug 1782] New: Assertion `Offset1 != Offset2 && " There
 is at least one different constant here!"' failed
Message-ID: <bug-1782-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1782

           Summary: Assertion `Offset1 != Offset2 && "There is at least one
                    different constant here!"' failed
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: normal
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: edwintorok at gmail.com
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=1205)
 --> (http://llvm.org/bugs/attachment.cgi?id=1205)
testcase

The testcase from http://llvm.org/bugs/show_bug.cgi?id=1774 compiles now, but
the original file doesn't.
New testcase attached.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Wed Nov  7 15:49:07 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Wed, 7 Nov 2007 15:49:07 -0600
Subject: [LLVMbugs] [Bug 1783] New: detect text files as text/plain even
 with LF line ending
Message-ID: <bug-1783-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1783

           Summary: detect text files as text/plain even with LF line ending
           Product: Bugzilla Admin
           Version: unspecified
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: UI
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: edwintorok at gmail.com
                CC: llvmbugs at cs.uiuc.edu


Any text file (.i, .ll, ...) I attach is detected as application/octet-stream,
instead of text/plain.
I am using Linux, and line endings are LF-only. 
Can bugzilla be configured to correctly detect content-type?


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Thu Nov  8 11:40:58 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Thu, 8 Nov 2007 11:40:58 -0600
Subject: [LLVMbugs] [Bug 1780] instcombine crash
In-Reply-To: <bug-1780-206@http.llvm.org/bugs/>
Message-ID: <200711081740.lA8Hew0g027613@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1780


Andrew Lenharth <alenhar2 at uiuc.edu> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #3 from Andrew Lenharth <alenhar2 at uiuc.edu>  2007-11-08 11:40:57 ---
Thanks for the further reduction.

I think your patch is overkill.

--- lib/Transforms/Scalar/InstructionCombining.cpp      (revision 43891)
+++ lib/Transforms/Scalar/InstructionCombining.cpp      (working copy)
@@ -7582,7 +7582,7 @@
                                            unsigned PrefAlign = 0) {
   if (GlobalVariable *GV = dyn_cast<GlobalVariable>(V)) {
     unsigned Align = GV->getAlignment();
-    if (Align == 0 && TD)
+    if (Align == 0 && TD && !isa<OpaqueType>(GV->getType()->getElementType()))
       Align = TD->getPrefTypeAlignment(GV->getType()->getElementType());

     // If there is a large requested alignment and we can, bump up the
alignment

Should do it.

Since align is already 0 and there is no initializer, it will return zero with
just that change.

closed.

http://llvm.org/viewvc/llvm-project?rev=43893&view=rev


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Thu Nov  8 15:42:45 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Thu, 8 Nov 2007 15:42:45 -0600
Subject: [LLVMbugs] [Bug 1783] detect text files as text/plain even with LF
	line ending
In-Reply-To: <bug-1783-206@http.llvm.org/bugs/>
Message-ID: <200711082142.lA8LgjS6014574@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1783


Tanya Lattner <tonic at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |INVALID




-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Fri Nov  9 02:29:10 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Fri, 9 Nov 2007 02:29:10 -0600
Subject: [LLVMbugs] [Bug 1784] New: vectors of x86 long double will not work
	properly
Message-ID: <bug-1784-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1784

           Summary: vectors of x86 long double will not work properly
           Product: new-bugs
           Version: unspecified
          Platform: Other
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: baldrick at free.fr
                CC: llvmbugs at cs.uiuc.edu


I noticed that a lot of the vector code assumes that there is
no padding between vector elements.  For example, the size is
assumed to be the primitive size of the element times the number
of elements.  However for x86 long double the primitive size is
80 bits while elements will be spaced by 96/128 bits (depending
on the os) in order to maintain alignment.  There's a similar
problem for vectors of apints.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Fri Nov  9 02:54:08 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Fri, 9 Nov 2007 02:54:08 -0600
Subject: [LLVMbugs] [Bug 1785] New: llvm-gcc-4.2 fails to build after latest
	updates
Message-ID: <bug-1785-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1785

           Summary: llvm-gcc-4.2 fails to build after latest updates
           Product: new-bugs
           Version: unspecified
          Platform: Other
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: baldrick at free.fr
                CC: llvmbugs at cs.uiuc.edu


On linux x86 the build fails here:

../../gcc-4.2.llvm/gcc/c-common.c: In function ???iasm_stmt???:
../../gcc-4.2.llvm/gcc/c-common.c:7577: error: ???MACHO_DYNAMIC_NO_PIC_P???
undeclared (first use in this function)
../../gcc-4.2.llvm/gcc/c-common.c:7577: error: (Each undeclared identifier is
reported only once
../../gcc-4.2.llvm/gcc/c-common.c:7577: error: for each function it appears
in.)


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Fri Nov  9 06:05:36 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Fri, 9 Nov 2007 06:05:36 -0600
Subject: [LLVMbugs] [Bug 1785] llvm-gcc-4.2 fails to build after latest
	updates
In-Reply-To: <bug-1785-206@http.llvm.org/bugs/>
Message-ID: <200711091205.lA9C5arZ010166@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1785


Anton Korobeynikov <asl at math.spbu.ru> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED




--- Comment #2 from Anton Korobeynikov <asl at math.spbu.ru>  2007-11-09 04:20:59 ---
First part fixed in:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071105/055466.html

But now compiler cannot create any executable. The weird thing is:

gcc -v
Using built-in specs.
Target: i686-pc-linux-gnu
Configured with: ../src/configure linux gnu
Thread model: single
gcc version 4.2.1(llvm) (Based on Apple Inc. build 5546)


--- Comment #3 from Anton Korobeynikov <asl at math.spbu.ru>  2007-11-09 06:05:35 ---
Finally fixed with:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071105/055468.html

Please reopen, if needed.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Fri Nov  9 08:12:32 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Fri, 9 Nov 2007 08:12:32 -0600
Subject: [LLVMbugs] [Bug 1786] New: SimplifyCFG leaves unreachable basic
	blocks
Message-ID: <bug-1786-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1786

           Summary: SimplifyCFG leaves unreachable basic blocks
           Product: libraries
           Version: 2.0
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: Scalar Optimizations
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: jay.foad at antixlabs.com
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=1209)
 --> (http://llvm.org/bugs/attachment.cgi?id=1209)
bitcode for test case

I run this on the attached bitcode file pre.bc:

$ opt -f -o post.bc pre.bc -simplifycfg

Then in post.bc, basic block bb17 is only reachable from cond_next27, and
cond_next27 is only reachable from bb17. Neither is reachable from the entry
block so they should have been removed.

Running "opt ... -simplifycfg -simplifycfg" still doesn't remove them, but I
guess that's because the pass manager still only runs the pass once, or
something.

If I run simplifycfg again on post.bc, they do get removed.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Fri Nov  9 08:17:59 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Fri, 9 Nov 2007 08:17:59 -0600
Subject: [LLVMbugs] [Bug 1787] New: x86 long double not passed according to
	the ABI
Message-ID: <bug-1787-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1787

           Summary: x86 long double not passed according to the ABI
           Product: new-bugs
           Version: unspecified
          Platform: Other
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: baldrick at free.fr
                CC: llvmbugs at cs.uiuc.edu


If I pass two long doubles from a routine compiled with llvm-gcc to a routine
compiled with mainline gcc, then the second value is not passed right.

$ gcc -c ld_gcc.c
$ llvm-gcc -c ld_llvm.c
$ gcc -o x ld_llvm.o ld_gcc.o && ./x
1       0

If I recompile using only gcc, then all is well:
$ gcc -c ld_llvm.c
$ gcc -o x ld_llvm.o ld_gcc.o && ./x
1       1

First observed when calling the long double math library functions.

ld_gcc.c:

#include <stdio.h>
void p(long double d, long double e) {
  printf("%g    %g\n", (double)d, (double)e);
}

ld_llvm.c:

extern void p(long double, long double);
int main(void) {
  p(1.0, 1.0);
  return 0;
}


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Fri Nov  9 09:36:00 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Fri, 9 Nov 2007 09:36:00 -0600
Subject: [LLVMbugs] [Bug 1733] llc crash in
	SimpleRegisterCoalescing::JoinCopy
In-Reply-To: <bug-1733-206@http.llvm.org/bugs/>
Message-ID: <200711091536.lA9Fa0Kx027419@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1733


Duncan Sands <baldrick at free.fr> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #2 from Duncan Sands <baldrick at free.fr>  2007-11-09 09:35:59 ---
Yup, it seems to be fixed.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Fri Nov  9 16:42:30 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Fri, 9 Nov 2007 16:42:30 -0600
Subject: [LLVMbugs] [Bug 1788] New: assertion failure in lib/transforms/IPO/
 DeadArgumentElimination
Message-ID: <bug-1788-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1788

           Summary: assertion failure in
                    lib/transforms/IPO/DeadArgumentElimination
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Windows XP
            Status: NEW
          Severity: normal
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: richard.smith at antixlabs.com
                CC: llvmbugs at cs.uiuc.edu


llvm-ld gives an assertion failure "CS.instruction() && Unknown instruction for
the I2I list" in lib/transforms/IPO/DeadArgumentElimination. The app which
provoked it is large and under NDA so cannot be attached, but the reason is
thus:

runOnModule in lib/transforms/IPO/DeadArgumentElimination.cpp iterates through
every function and -

(a) checks whether it has varargs but no va_start. If that is so the varargs
are removed and every call to the function is replaced with one which omits the
params;

(b) calls SurveyFunction, and if the funtion has "maybe alive" args, all calls
within the function are pushed to the vector InstructionsToInspect.

In this case, a function processed provokes (b) and pushes the addresses of a
number of call intructions to InstructionsToInspect, one of which is to a
function with unused varargs. Later in the iteration loop that function is
processed provoking (a) - it is modified and all calls to it are replaced. At
this point, InstructionsToInspect still contains a pointer to the original
call, which is now invalid.

When InstructionsToInspect is processed it reaches the invalid pointer and the
random data there happens to provoke the asserion failure reported.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sat Nov 10 16:10:06 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sat, 10 Nov 2007 16:10:06 -0600
Subject: [LLVMbugs] [Bug 1787] x86 long double not passed according to the
	ABI
In-Reply-To: <bug-1787-206@http.llvm.org/bugs/>
Message-ID: <200711102210.lAAMA6wp011101@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1787


Dale Johannesen <dalej at apple.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED




-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov 11 02:48:28 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 11 Nov 2007 02:48:28 -0600
Subject: [LLVMbugs] [Bug 1786] SimplifyCFG leaves unreachable basic blocks
In-Reply-To: <bug-1786-206@http.llvm.org/bugs/>
Message-ID: <200711110848.lAB8mShd020092@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1786


Chris Lattner <sabre at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |INVALID




--- Comment #3 from Chris Lattner <sabre at nondot.org>  2007-11-11 02:48:28 ---
simplifycfg only handles trivially dead blocks. For more aggressive dce please
use the adce pass.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov 11 15:03:49 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 11 Nov 2007 15:03:49 -0600
Subject: [LLVMbugs] [Bug 1789] New: Cannot have spaces in LLVM build path
Message-ID: <bug-1789-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1789

           Summary: Cannot have spaces in LLVM build path
           Product: Build scripts
           Version: trunk
          Platform: PC
        OS/Version: Windows XP
            Status: NEW
          Severity: normal
          Priority: P2
         Component: Makefiles
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: grrussel at googlemail.com
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=1212)
 --> (http://llvm.org/bugs/attachment.cgi?id=1212)
Patch for the VS getting started web page

Cannot build LLVM on Windows with a directory path containing spaces when
building from the MSVC solution file - looks like copy commands in the project
do not enclose arguments in quotes, and therefore fail


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov 11 15:13:19 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 11 Nov 2007 15:13:19 -0600
Subject: [LLVMbugs] [Bug 1790] New: Loop ICM: Missed optimization on ctype.h
	functions
Message-ID: <bug-1790-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1790

           Summary: Loop ICM: Missed optimization on ctype.h functions
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: edwintorok at gmail.com
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=1213)
 --> (http://llvm.org/bugs/attachment.cgi?id=1213)
C example program

Functions like isspace from <ctype.h> access a constant array returned by
__ctype_b_loc.

Declaration of __ctype_b_loc is:
extern __const unsigned short int **__ctype_b_loc (void)
     __attribute__ ((__const));

Yet, in the generated llvm bytecode and assembly __ctype_b_loc is called each
time
(if I call isspace in a loop).
Gcc generates similar code (__ctype_b_loc inside of the loop), but I think llvm
could do better, and optimize this by moving function calls out of loops *iff*
they are "const" functions.

The problem is that in the generated bitcode the constness of __ctype_b_loc is
lost, it is simply an: declare i16** @__ctype_b_loc()

Attached small C example program, preprocessed C, and bitcode.

I didn't do any benchmarks, but I think avoiding an extra call is a win, and
doesn't need a benchmark to prove it ;)


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Mon Nov 12 07:25:51 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Mon, 12 Nov 2007 07:25:51 -0600
Subject: [LLVMbugs] [Bug 1791] New: exception handling broken
Message-ID: <bug-1791-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1791

           Summary: exception handling broken
           Product: new-bugs
           Version: unspecified
          Platform: Other
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: baldrick at free.fr
                CC: llvmbugs at cs.uiuc.edu


Some recent change has broken exception handling.  Running the
following testcase results in a segmentation fault:

int main()
{
  throw 1;
}

Program received signal SIGSEGV, Segmentation fault.
0x08057ffc in _GLOBAL_OFFSET_TABLE_ ()


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Mon Nov 12 11:11:51 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Mon, 12 Nov 2007 11:11:51 -0600
Subject: [LLVMbugs] [Bug 1786] SimplifyCFG leaves unreachable basic blocks
In-Reply-To: <bug-1786-206@http.llvm.org/bugs/>
Message-ID: <200711121711.lACHBpTC010359@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1786


Chris Lattner <sabre at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |REOPENED
         Resolution|INVALID                     |




--- Comment #5 from Chris Lattner <sabre at nondot.org>  2007-11-12 11:11:50 ---
Ah, I didn't get it that "opt -simplifycfg -simplifycfg" didn't work but "opt
-simplifycfg | opt -simplifycfg" did.  Reopening.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Mon Nov 12 17:09:56 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Mon, 12 Nov 2007 17:09:56 -0600
Subject: [LLVMbugs] [Bug 1792] New: LLVM unable to unroll trivial code
Message-ID: <bug-1792-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1792

           Summary: LLVM unable to unroll trivial code
           Product: libraries
           Version: 2.1
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: major
          Priority: P2
         Component: Loop Optimizer
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: nadav256 at gmail.com
                CC: llvmbugs at cs.uiuc.edu


The trivial code below is not unrolled.

nlewycky: okay, so -scalar-evolutions correctly detects that the loop runs
exactly 31 iterations. that's good.
 hm, but it doesn't unroll anyways. That's not right.

Code:

unsigned int bitCount(unsigned int input) {
        unsigned int count = 0;
        unsigned int i;
        for (i=0; i<sizeof(input)*8; i++) {
                count += (input>>i) & 1;
        }
        return count;
}


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Mon Nov 12 17:12:38 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Mon, 12 Nov 2007 17:12:38 -0600
Subject: [LLVMbugs] [Bug 1792] LLVM unable to unroll trivial code
In-Reply-To: <bug-1792-206@http.llvm.org/bugs/>
Message-ID: <200711122312.lACNCcKV028301@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1792


Nick Lewycky <nlewycky at google.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |DUPLICATE




--- Comment #1 from Nick Lewycky <nlewycky at google.com>  2007-11-12 17:12:37 ---
Here's what -debug -loop-unroll has to say:

  Loop Unroll: F[f] Loop %bb
    Trip Count = 32
    Loop Size = 5
    TOO LARGE TO UNROLL: 160>100

You can crank up the threshold with -unroll-threshold=500 but the result is a
huge mess. This code is a pop-counter, marking it a dupe of PR1488.


*** This bug has been marked as a duplicate of bug 1488 ***


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Tue Nov 13 01:33:55 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Tue, 13 Nov 2007 01:33:55 -0600
Subject: [LLVMbugs] [Bug 1786] two passes of simplifycfg needed to simplify
	code
In-Reply-To: <bug-1786-206@http.llvm.org/bugs/>
Message-ID: <200711130733.lAD7Xtbi029283@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1786


Chris Lattner <sabre at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Depends on|217                         |
             Status|NEW                         |RESOLVED
           Keywords|                            |code-quality
         Resolution|                            |FIXED
            Summary|SimplifyCFG leaves          |two passes of simplifycfg
                   |unreachable basic blocks    |needed to simplify code
   Target Milestone|---                         |2.2




--- Comment #6 from Chris Lattner <sabre at nondot.org>  2007-11-13 01:33:54 ---
Fixed, patch here:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071112/055567.html

Thanks for reporting this!

-Chris


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Tue Nov 13 02:45:10 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Tue, 13 Nov 2007 02:45:10 -0600
Subject: [LLVMbugs] [Bug 1790] Loop ICM: Missed optimization on ctype.h
	functions
In-Reply-To: <bug-1790-206@http.llvm.org/bugs/>
Message-ID: <200711130845.lAD8jAlx008000@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1790


Anton Korobeynikov <asl at math.spbu.ru> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |DUPLICATE




--- Comment #10 from Anton Korobeynikov <asl at math.spbu.ru>  2007-11-13 02:45:10 ---
Yes it indeed is.

*** This bug has been marked as a duplicate of bug 1383 ***


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Tue Nov 13 08:33:12 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Tue, 13 Nov 2007 08:33:12 -0600
Subject: [LLVMbugs] [Bug 1793] New: [: 1: ==: unexpected operator
Message-ID: <bug-1793-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1793

           Summary: [: 1: ==: unexpected operator
           Product: new-bugs
           Version: unspecified
          Platform: Other
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: baldrick at free.fr
                CC: llvmbugs at cs.uiuc.edu


This occurs several times during the gcc-4.2 build.  I'm pretty sure it
only started happening recently.

make[3]: Leaving directory 'libdecnumber'
[: 1: ==: unexpected operator


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Tue Nov 13 10:18:06 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Tue, 13 Nov 2007 10:18:06 -0600
Subject: [LLVMbugs] [Bug 1794] New: Add option to print all functions
Message-ID: <bug-1794-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1794

           Summary: Add option to print all functions
           Product: tools
           Version: trunk
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: llvm2cpp
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: zack at tungstengraphics.com
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=1217)
 --> (http://llvm.org/bugs/attachment.cgi?id=1217)
add print functions

It would be nice to have an option in llvm2cpp which made it generate all
defined functions from a file. The attached patch does that.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Tue Nov 13 12:13:23 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Tue, 13 Nov 2007 12:13:23 -0600
Subject: [LLVMbugs] [Bug 1795] New: Missed optimization: eliminating loads
 around ptrtoint/ inttoptr
Message-ID: <bug-1795-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1795

           Summary: Missed optimization: eliminating loads around
                    ptrtoint/inttoptr
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Windows XP
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: sharparrow1 at yahoo.com
                CC: llvmbugs at cs.uiuc.edu


define void @myfunc(i32 %.val24) {
EntryBlock:
        add i32 %.val24, -12
        inttoptr i32 %0 to i32*
        store i32 1, i32* %1
        add i32 %.val24, -16
        inttoptr i32 %2 to i32*
        getelementptr i32* %3, i32 1
        load i32* %4
        tail call i32 @ZZZ( i32 %5 )
        ret void
}

The "load i32* %4" always results in a "1", so it should be possible to
optimize it out (on x86).

At one point I had a hack in my tree to push addition forwards across inttoptr
instances in VisitIntToPtr, but I'm not sure if that's the right approach.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Tue Nov 13 12:22:54 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Tue, 13 Nov 2007 12:22:54 -0600
Subject: [LLVMbugs] [Bug 1794] Add option to print all functions
In-Reply-To: <bug-1794-206@http.llvm.org/bugs/>
Message-ID: <200711131822.lADIMsol005432@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1794


Chris Lattner <sabre at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #1 from Chris Lattner <sabre at nondot.org>  2007-11-13 12:22:53 ---
applied, thanks!
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071112/055573.html


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Tue Nov 13 14:48:32 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Tue, 13 Nov 2007 14:48:32 -0600
Subject: [LLVMbugs] [Bug 1796] New: Optimizer doesn't insert unreachable
 after noreturn call
Message-ID: <bug-1796-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1796

           Summary: Optimizer doesn't insert unreachable after noreturn call
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Windows XP
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: sharparrow1 at yahoo.com
                CC: llvmbugs at cs.uiuc.edu


declare void @Finisher(i32) noreturn

define void @YYY(i32) {
  tail call void @Finisher(i32 %0) noreturn
  tail call void @Finisher(i32 %0) noreturn
  ret void
}

is left untouched by opt -std-compile-opts.  YYY should be optimized to:

tail call void @Finisher(i32 %0) noreturn
unreachable


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Tue Nov 13 17:17:07 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Tue, 13 Nov 2007 17:17:07 -0600
Subject: [LLVMbugs] [Bug 1793] [: 1: ==: unexpected operator
In-Reply-To: <bug-1793-206@http.llvm.org/bugs/>
Message-ID: <200711132317.lADNH78R019692@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1793


Duncan Sands <baldrick at free.fr> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #1 from Duncan Sands <baldrick at free.fr>  2007-11-13 17:17:06 ---
Fixed here:

http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071112/055592.html


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Wed Nov 14 00:20:14 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Wed, 14 Nov 2007 00:20:14 -0600
Subject: [LLVMbugs] [Bug 1796] Optimizer doesn't insert unreachable after
	noreturn call
In-Reply-To: <bug-1796-206@http.llvm.org/bugs/>
Message-ID: <200711140620.lAE6KEUo005735@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1796


Chris Lattner <sabre at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
          Component|new bugs                    |Scalar Optimizations
           Keywords|                            |code-quality
            Product|new-bugs                    |libraries
         Resolution|                            |FIXED
   Target Milestone|---                         |2.2
            Version|unspecified                 |1.0




--- Comment #1 from Chris Lattner <sabre at nondot.org>  2007-11-14 00:20:14 ---
Implemented, patch here:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071112/055600.html


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Wed Nov 14 00:49:29 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Wed, 14 Nov 2007 00:49:29 -0600
Subject: [LLVMbugs] [Bug 1797] New: BlockExtractor's "blocks to not extract"
 list doesn' t allow for anon blocks
Message-ID: <bug-1797-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1797

           Summary: BlockExtractor's "blocks to not extract" list doesn't
                    allow for anon blocks
           Product: libraries
           Version: trunk
          Platform: All
        OS/Version: All
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: Interprocedural Optimizations
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: nicholas at mxc.ca
                CC: llvmbugs at cs.uiuc.edu


This is fallout from fixing PR1775. The current file contains function/basic
block like as:
  file << F->getName() << " " << BB->getName() << "\n";
which might not handle anonymous blocks or functions, or either one containing
spaces or newlines. This needs to be changed, preferrably to use the
asmprinter/parser code for quoting.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Wed Nov 14 00:49:51 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Wed, 14 Nov 2007 00:49:51 -0600
Subject: [LLVMbugs] [Bug 1775] ExtractMappedBlocksFromModule discards the "
 blocks to not extract" list
In-Reply-To: <bug-1775-206@http.llvm.org/bugs/>
Message-ID: <200711140649.lAE6npL8008638@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1775


Nick Lewycky <nicholas at mxc.ca> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED




--- Comment #3 from Nick Lewycky <nicholas at mxc.ca>  2007-11-14 00:49:50 ---
Applied. The bug for the possible things it won't handle is PR1797.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Wed Nov 14 01:05:11 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Wed, 14 Nov 2007 01:05:11 -0600
Subject: [LLVMbugs] [Bug 1789] Cannot have spaces in LLVM build path
In-Reply-To: <bug-1789-206@http.llvm.org/bugs/>
Message-ID: <200711140705.lAE75BvJ009633@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1789


Chris Lattner <sabre at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #1 from Chris Lattner <sabre at nondot.org>  2007-11-14 01:05:11 ---
Applied, thanks!
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071112/055602.html


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Wed Nov 14 01:19:16 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Wed, 14 Nov 2007 01:19:16 -0600
Subject: [LLVMbugs] [Bug 1798] New: wrong integer overflow calculation on
 signed/ unsigned int when using optimizer
Message-ID: <bug-1798-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1798

           Summary: wrong integer overflow calculation on signed/unsigned
                    int when using optimizer
           Product: new-bugs
           Version: unspecified
          Platform: Macintosh
        OS/Version: MacOS X
            Status: NEW
          Severity: normal
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: joseph.bebel+llvm at gmail.com
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=1219)
 --> (http://llvm.org/bugs/attachment.cgi?id=1219)
small test program showing the overflow

While experimenting with LLVM as a gcc-drop-in I noticed some
interesting behavior related to integer overflow. Consider the
attached test program which sums up large numbers in both signed and unsigned
ints.

Compiled using the latest llvm-g++ (2.1, mac os x universal tarball)
with no optimization, and under all levels of optimization under Apple
GCC 4.0.1 the program correctly outputs "1206807378" for both signed and
unsigned int.

However, when using llvm-g++ under any level of optimization
(-O1/2/3), it outputs instead "-940676270" (which is the above answer
minus 2^31) or "3354291026" for the unsigned sum. Always reproducible.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Wed Nov 14 03:20:05 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Wed, 14 Nov 2007 03:20:05 -0600
Subject: [LLVMbugs] [Bug 1791] exception handling broken
In-Reply-To: <bug-1791-206@http.llvm.org/bugs/>
Message-ID: <200711140920.lAE9K5hv023025@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1791


Anton Korobeynikov <asl at math.spbu.ru> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #6 from Anton Korobeynikov <asl at math.spbu.ru>  2007-11-14 03:20:04 ---
Fixed in:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071112/055610.html


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Wed Nov 14 05:20:19 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Wed, 14 Nov 2007 05:20:19 -0600
Subject: [LLVMbugs] [Bug 1799] New: "Node emitted out of order - early" llc
	failure
Message-ID: <bug-1799-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1799

           Summary: "Node emitted out of order - early" llc failure
           Product: new-bugs
           Version: unspecified
          Platform: Other
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: baldrick at free.fr
                CC: llvmbugs at cs.uiuc.edu


$ llc node.bc
llc: lib/CodeGen/SelectionDAG/ScheduleDAG.cpp:398: void
llvm::ScheduleDAG::CreateVirtualRegisters(llvm::SDNode*, llvm::MachineInstr*,
const llvm::TargetInstrDescriptor&, llvm::DenseMap<llvm::SDOperand, unsigned
int, llvm::DenseMapInfo<llvm::SDOperand>, llvm::DenseMapInfo<unsigned int> >&):
Assertion `isNew && "Node emitted out of order - early"' failed.
llc[0x88d4026]
llc[0x88d42ec]
[0xffffe420]
/lib/libc.so.6(abort+0x101)[0xb7cbd221]
/lib/libc.so.6(__assert_fail+0xf0)[0xb7cb4db0]
llc(llvm::ScheduleDAG::CreateVirtualRegisters(llvm::SDNode*,
llvm::MachineInstr*, llvm::TargetInstrDescriptor const&,
llvm::DenseMap<llvm::SDOperand, unsigned int,
llvm::DenseMapInfo<llvm::SDOperand>, llvm::DenseMapInfo<unsigned int>
>&)+0x24f)[0x86aec39]
llc(llvm::ScheduleDAG::EmitNode(llvm::SDNode*, unsigned int,
llvm::DenseMap<llvm::SDOperand, unsigned int,
llvm::DenseMapInfo<llvm::SDOperand>, llvm::DenseMapInfo<unsigned int>
>&)+0x16b)[0x86afd0b]
llc(llvm::ScheduleDAG::EmitSchedule()+0x266)[0x86b07cc]
llc[0x86170a2]
llc(llvm::ScheduleDAG::Run()+0x94)[0x86ad2d4]
llc(llvm::SelectionDAGISel::ScheduleAndEmitDAG(llvm::SelectionDAG&)+0x73)[0x863cc05]
llc[0x8566db0]
llc(llvm::SelectionDAGISel::CodeGenAndEmitDAG(llvm::SelectionDAG&)+0x1e7)[0x863ce47]
llc(llvm::SelectionDAGISel::SelectBasicBlock(llvm::BasicBlock*,
llvm::MachineFunction&, llvm::FunctionLoweringInfo&)+0xb5)[0x8651c45]
llc(llvm::SelectionDAGISel::runOnFunction(llvm::Function&)+0x1e4)[0x86538c4]
llc[0x84bf765]
llc(llvm::FPPassManager::runOnFunction(llvm::Function&)+0x11a)[0x88665e8]
llc(llvm::FunctionPassManagerImpl::run(llvm::Function&)+0x6e)[0x8866846]
llc(llvm::FunctionPassManager::run(llvm::Function&)+0x88)[0x886699c]
llc(main+0x95b)[0x8367ea9]
/lib/libc.so.6(__libc_start_main+0xe0)[0xb7ca8050]
llc[0x8366421]
Aborted (core dumped)


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Wed Nov 14 14:31:38 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Wed, 14 Nov 2007 14:31:38 -0600
Subject: [LLVMbugs] [Bug 1800] New: after instcombine,
	loop is only executed once
Message-ID: <bug-1800-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1800

           Summary: after instcombine, loop is only executed once
           Product: new-bugs
           Version: unspecified
          Platform: Other
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: baldrick at free.fr
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=1221)
 --> (http://llvm.org/bugs/attachment.cgi?id=1221)
testcase .ll

$ lli ic.bc
hello
$ opt -instcombine ic.bc | lli
h???

After instcombine the copy loop is only executed once, rather
than five times, meaning only the first character (h) is copied.
I think this is because instcombine is turning this:

       %max13512 = select i1 %tmp13411, i32 %tmp13310, i32 0
       %tmp13713 = icmp eq i32 %max13512, 1
       br i1 %tmp13713, label %bb143.loopexit, label %bb.nph

into

       br i1 %tmp13411, label %bb143.loopexit, label %bb.nph

which is wrong if %tmp13310 is not equal to 1.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Wed Nov 14 15:02:20 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Wed, 14 Nov 2007 15:02:20 -0600
Subject: [LLVMbugs] [Bug 1781] memcpy Defaulted to Alignment 1
In-Reply-To: <bug-1781-206@http.llvm.org/bugs/>
Message-ID: <200711142102.lAEL2KKO028536@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1781


Bill Wendling <isanbard at gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #2 from Bill Wendling <isanbard at gmail.com>  2007-11-14 15:02:19 ---
Fixed:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071112/055619.html


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Thu Nov 15 00:11:34 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Thu, 15 Nov 2007 00:11:34 -0600
Subject: [LLVMbugs] [Bug 1788] assertion failure in
 lib/transforms/IPO/DeadArgumentElimination
In-Reply-To: <bug-1788-206@http.llvm.org/bugs/>
Message-ID: <200711150611.lAF6BY46024728@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1788


Chris Lattner <sabre at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #3 from Chris Lattner <sabre at nondot.org>  2007-11-15 00:11:31 ---
Thank you for your analysis of the problem.  You are exactly right in all
counts.  I applied this patch, please verify that it fixes the problem:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071112/055643.html

-Chris


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Thu Nov 15 01:12:33 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Thu, 15 Nov 2007 01:12:33 -0600
Subject: [LLVMbugs] [Bug 1798] wrong integer overflow calculation on signed/
 unsigned int when using optimizer
In-Reply-To: <bug-1798-206@http.llvm.org/bugs/>
Message-ID: <200711150712.lAF7CXWI027823@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1798


Nick Lewycky <nicholas at mxc.ca> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED




--- Comment #4 from Nick Lewycky <nicholas at mxc.ca>  2007-11-15 01:12:33 ---
Patch checked in.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Thu Nov 15 05:42:52 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Thu, 15 Nov 2007 05:42:52 -0600
Subject: [LLVMbugs] [Bug 1801] New: Crash with functions with empty names
Message-ID: <bug-1801-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1801

           Summary: Crash with functions with empty names
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Windows XP
            Status: NEW
          Severity: minor
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: sharparrow1 at yahoo.com
                CC: llvmbugs at cs.uiuc.edu


declare void @""(i32, i32)
declare void @""(i32, i32)
define void @EWEWEW(i32) {
  call void @0(i32 1, i32 2)
  call void @1(i32 1, i32 2)
  ret void
}

Crashes when run through opt -std-compile-opts.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Thu Nov 15 12:10:05 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Thu, 15 Nov 2007 12:10:05 -0600
Subject: [LLVMbugs] [Bug 1798] wrong integer overflow calculation on signed/
 unsigned int when using optimizer
In-Reply-To: <bug-1798-206@http.llvm.org/bugs/>
Message-ID: <200711151810.lAFIA5h9004351@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1798


Nick Lewycky <nlewycky at google.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |REOPENED
         Resolution|FIXED                       |




--- Comment #5 from Nick Lewycky <nlewycky at google.com>  2007-11-15 12:10:04 ---
Patch caused PR726 to regress. Further analysis needed. Reopening.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Thu Nov 15 12:58:17 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Thu, 15 Nov 2007 12:58:17 -0600
Subject: [LLVMbugs] [Bug 1802] New: CodeGen/X86/2007-11-14-Coalescer-Bug.ll
 fails on X86-64
Message-ID: <bug-1802-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1802

           Summary: CodeGen/X86/2007-11-14-Coalescer-Bug.ll fails on X86-64
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: normal
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: edwintorok at gmail.com
                CC: llvmbugs at cs.uiuc.edu


FAIL: /home/edwin/llvm-svn/llvm/test/CodeGen/X86/2007-11-14-Coalescer-Bug.ll
Failed with exit(1) at line 1
while running: llvm-as <
/home/edwin/llvm-svn/llvm/test/CodeGen/X86/2007-11-14-Coalescer-Bug.ll | llc
-march=x86 -x86-asm-syntax=att | grep movl | grep 4
child process exited abnormally


LLVM is SVN r44177, debug build, using gcc (GCC) 4.2.3 20071014 (prerelease)
(Debian 4.2.2-3).

$ uname -a
Linux lightspeed2 2.6.23-rc8-hrt1-cfs-v22-g1bef7dc0-dirty #17 Sun Sep 30
18:49:34 EEST 2007 x86_64 GNU/Linux

$ llvm-as <
/home/edwin/llvm-svn/llvm/test/CodeGen/X86/2007-11-14-Coalescer-Bug.ll | llc
-march=x86 -x86-asm-syntax=att|grep movl
        movl    8(%esp), %eax
        movl    %ecx, (%esp)
        movl    (%esp), %edx

$ llvm-as <
/home/edwin/llvm-svn/llvm/test/CodeGen/X86/2007-11-14-Coalescer-Bug.ll | llc
-march=x86 -x86-asm-syntax=att|grep movl|grep 4
$


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Thu Nov 15 16:09:29 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Thu, 15 Nov 2007 16:09:29 -0600
Subject: [LLVMbugs] [Bug 1803] New: Debug info doesn't work on x86-64
Message-ID: <bug-1803-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1803

           Summary: Debug info doesn't work on x86-64
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: normal
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: edwintorok at gmail.com
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=1227)
 --> (http://llvm.org/bugs/attachment.cgi?id=1227)
assembly generated by llvm-gcc with incorrect debug info

On x86-64 the generated debug info is unusable by gdb.
It works on x86-32.

$ cat p.c
int main()
{
}

$ llvm-gcc -g p.c

 gdb ./a.out
GNU gdb 6.6.90.20070912-debian
Copyright (C) 2007 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.  Type "show copying"
and "show warranty" for details.
This GDB was configured as "x86_64-linux-gnu"...
Using host libthread_db library "/lib/libthread_db.so.1".
(gdb) b main
Breakpoint 1 at 0x400434
(gdb) r
Starting program: /home/edwin/llvm-svn/llvm/a.out
warning: no loadable sections found in added symbol-file system-supplied DSO at
0x7ffff93fd000

Breakpoint 1, 0x0000000000400434 in main ()
Single stepping until exit from function main,
which has no line number information.
0x00000032d1c1db44 in __libc_start_main () from /lib/libc.so.6


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Fri Nov 16 00:05:06 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Fri, 16 Nov 2007 00:05:06 -0600
Subject: [LLVMbugs] [Bug 1800] after instcombine, loop is only executed once
In-Reply-To: <bug-1800-206@http.llvm.org/bugs/>
Message-ID: <200711160605.lAG656vU021475@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1800


Chris Lattner <sabre at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|enhancement                 |normal
             Status|NEW                         |RESOLVED
          Component|new bugs                    |Scalar Optimizations
           Keywords|                            |miscompilation
            Product|new-bugs                    |libraries
         Resolution|                            |FIXED
   Target Milestone|---                         |2.2
            Version|unspecified                 |1.0




--- Comment #4 from Chris Lattner <sabre at nondot.org>  2007-11-16 00:05:06 ---
Wow, very nice catch, patch here:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071112/055664.html

Please verify, thanks!

-Chris


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Fri Nov 16 05:57:43 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Fri, 16 Nov 2007 05:57:43 -0600
Subject: [LLVMbugs] [Bug 1804] New: APFloat wrong handling of smallest
 normalized positive 32 bit float
Message-ID: <bug-1804-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1804

           Summary: APFloat wrong handling of smallest normalized positive
                    32 bit float
           Product: new-bugs
           Version: unspecified
          Platform: Other
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: baldrick at free.fr
                CC: llvmbugs at cs.uiuc.edu


Here's a C testcase showing the problem (originally spotted in the
Ada testsuite):

float F(void) {
  const int i = 8388608;
  const float f = *(float *)&i;
  return f;
}

$ llvm-gcc -S -O4 af.c
->
define float @F() pure  {
entry:
        ret float 0.000000e+00
}

The float has become zero, but 8388608 = 2^23 corresponds to
2^(-126), the smallest positive normalized float, and is non-zero.
It is also output as zero in the final assembler:
        .long   0       # float 0


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Fri Nov 16 11:18:28 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Fri, 16 Nov 2007 11:18:28 -0600
Subject: [LLVMbugs] [Bug 1805] New: llvm-g++ emits trees for computed labels
	too late
Message-ID: <bug-1805-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1805

           Summary: llvm-g++ emits trees for computed labels too late
           Product: tools
           Version: trunk
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: normal
          Priority: P2
         Component: llvm-g++
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: asl at math.spbu.ru
                CC: llvmbugs at cs.uiuc.edu


Consider the following easy testcase:

int baz(void*);

int foo() {
 static void* bar[] = { &&bb1, &&bb2, &&bb3 };
 switch (baz(bar)) {
 case 1:
   goto bb1;
 case 2:
   goto bb2;
 default:
   goto bb3;
 }
 bb1: return 1;
 bb2: return 2;
 bb3: return 3;
}

It works with both llvm-g++ and llvm-gcc for 4.0, works for llvm-gcc 4.2, but
not for llvm-g++ 4.2 :( Looks like some sort of "regression".


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Fri Nov 16 16:46:14 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Fri, 16 Nov 2007 16:46:14 -0600
Subject: [LLVMbugs] [Bug 1801] basicAA crash on functions with empty names
In-Reply-To: <bug-1801-206@http.llvm.org/bugs/>
Message-ID: <200711162246.lAGMkE0B032569@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1801


Tanya Lattner <tonic at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #4 from Tanya Lattner <tonic at nondot.org>  2007-11-16 16:46:14 ---
Fixed:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071112/055675.html


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Fri Nov 16 19:03:16 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Fri, 16 Nov 2007 19:03:16 -0600
Subject: [LLVMbugs] [Bug 1804] APFloat wrong handling of smallest normalized
 positive 32 bit float
In-Reply-To: <bug-1804-206@http.llvm.org/bugs/>
Message-ID: <200711170103.lAH13GnB006351@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1804


Dale Johannesen <dalej at apple.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #1 from Dale Johannesen <dalej at apple.com>  2007-11-16 19:03:15 ---
Fixed.
http://llvm.org/viewvc/llvm-project?view=rev&revision=44201


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sat Nov 17 01:57:11 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sat, 17 Nov 2007 01:57:11 -0600
Subject: [LLVMbugs] [Bug 1802] CodeGen/X86/2007-11-14-Coalescer-Bug.ll fails
	on X86-64
In-Reply-To: <bug-1802-206@http.llvm.org/bugs/>
Message-ID: <200711170757.lAH7vBik024911@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1802


T??r??k Edwin <edwintorok at gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #4 from T??r??k Edwin <edwintorok at gmail.com>  2007-11-17 01:57:10 ---
(In reply to comment #3)
> There was a typo in the run line, now fixed:
> -; RUN: llvm-as < %s | llc -march=x86 -x86-asm-syntax=att | grep movl | grep 4
> +; RUN: llvm-as < %s | llc -march=x86 -x86-asm-syntax=att | grep movl | count 4
> 
> Can you update your tree and let me know if it still fails?
> 

The test succeeds now. Closing bug.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sat Nov 17 02:28:45 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sat, 17 Nov 2007 02:28:45 -0600
Subject: [LLVMbugs] [Bug 1806] New: -anders-aa crash with asm rorw
Message-ID: <bug-1806-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1806

           Summary: -anders-aa crash with asm rorw
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: edwintorok at gmail.com
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=1243)
 --> (http://llvm.org/bugs/attachment.cgi?id=1243)
bugpoint reduced testcase

$ Debug/bin/opt -anders-aa bugpoint-reduced-simplified.bc
WARNING: You're attempting to print out a bitcode file.
This is inadvisable as it may cause display problems. If
you REALLY want to taste LLVM bitcode first-hand, you
can force output with the `-f' option.

i16 (i16)* asm "rorw $$8, ${0:w}", "=r,0,~{dirflag},~{fpsr},~{flags},~{cc}"
opt: Andersens.cpp:408: unsigned
int<unnamed>::Andersens::getNode(llvm::Value*): Assertion `0 &&
"Value does not have a node in the points-to graph!"' failed.
/home/edwin/llvm-svn/llvm/Debug/bin/opt[0x9cb54a]
/home/edwin/llvm-svn/llvm/Debug/bin/opt[0x9cb7b6]
/lib/libc.so.6[0x32d1c31710]
/lib/libc.so.6(gsignal+0x35)[0x32d1c316a5]
/lib/libc.so.6(abort+0x110)[0x32d1c33100]
/lib/libc.so.6(__assert_fail+0xef)[0x32d1c2ac2f]
/home/edwin/llvm-svn/llvm/Debug/bin/opt[0x8648cb]
/home/edwin/llvm-svn/llvm/Debug/bin/opt[0x86807d]
/home/edwin/llvm-svn/llvm/Debug/bin/opt[0x868238]
/home/edwin/llvm-svn/llvm/Debug/bin/opt[0x8682b0]
/home/edwin/llvm-svn/llvm/Debug/bin/opt[0x8682cf]
/home/edwin/llvm-svn/llvm/Debug/bin/opt[0x868618]
/home/edwin/llvm-svn/llvm/Debug/bin/opt[0x8686c1]
/home/edwin/llvm-svn/llvm/Debug/bin/opt[0x868716]
/home/edwin/llvm-svn/llvm/Debug/bin/opt[0x86875b]
/home/edwin/llvm-svn/llvm/Debug/bin/opt[0x8687b0]
/home/edwin/llvm-svn/llvm/Debug/bin/opt[0x8687d5]
/home/edwin/llvm-svn/llvm/Debug/bin/opt[0x868c60]
/home/edwin/llvm-svn/llvm/Debug/bin/opt[0x868f02]
/home/edwin/llvm-svn/llvm/Debug/bin/opt(llvm::MPPassManager::runOnModule(llvm::Module&)+0xec)[0x9
                                      5d5d4]
/home/edwin/llvm-svn/llvm/Debug/bin/opt(llvm::PassManagerImpl::run(llvm::Module&)+0x74)[0x95d764]
/home/edwin/llvm-svn/llvm/Debug/bin/opt(llvm::PassManager::run(llvm::Module&)+0x20)[0x95d7c6]
/home/edwin/llvm-svn/llvm/Debug/bin/opt(main+0x986)[0x6c5a32]
/lib/libc.so.6(__libc_start_main+0xf4)[0x32d1c1db44]
/home/edwin/llvm-svn/llvm/Debug/bin/opt(cos+0x121)[0x6b7139]
Aborted

The call in the attached bugpoint-reduced-simplified.bc is dead code, but just
because bugpoint simplified it, in the original it is not dead code.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sat Nov 17 02:44:13 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sat, 17 Nov 2007 02:44:13 -0600
Subject: [LLVMbugs] [Bug 1807] New: llvm test-suite make error
Message-ID: <bug-1807-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1807

           Summary: llvm test-suite make error
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: edwintorok at gmail.com
                CC: llvmbugs at cs.uiuc.edu


I tried to compile the llvm-testsuite, however I can't:

make[1]: Entering directory
`/home/edwin/llvm-svn/llvm/projects/llvm-test/SingleSource'
make[2]: Entering directory
`/home/edwin/llvm-svn/llvm/projects/llvm-test/SingleSource/UnitTests'
make[3]: Entering directory
`/home/edwin/llvm-svn/llvm/projects/llvm-test/SingleSource/UnitTests/Vector'
make[3]: *** No rule to make target `Output/build2.linked.rbc', needed by
`Output/build2.linked.bc'.  Stop.
make[3]: Leaving directory
`/home/edwin/llvm-svn/llvm/projects/llvm-test/SingleSource/UnitTests/Vector'
make[2]: *** [all] Error 1
make[2]: Leaving directory
`/home/edwin/llvm-svn/llvm/projects/llvm-test/SingleSource/UnitTests'
make[1]: *** [UnitTests/.makeall] Error 2
make[1]: Leaving directory
`/home/edwin/llvm-svn/llvm/projects/llvm-test/SingleSource'
make: *** [SingleSource/.makeall] Error 2

$ make -v
GNU Make 3.81
Copyright (C) 2006  Free Software Foundation, Inc.
This is free software; see the source for copying conditions.
There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A
PARTICULAR PURPOSE.

This program built for x86_64-pc-linux-gnu

$ uname -a
Linux lightspeed2 2.6.23-rc8-hrt1-cfs-v22-g1bef7dc0-dirty #17 Sun Sep 30
18:49:34 EEST 2007 x86_64 GNU/Linux

The problem seems to be because of:
Output/%.linked.rbc: Output/%.bc

I am no Makefile expert, but IIRC that rule is a rule on how to create a .rbc
file out of a .linked file.
However I think what is meant here is a rule how to make .linked.rbc files out
of .bc files.

Are there any flags I can pass to make to have it compiled?


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sat Nov 17 09:17:26 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sat, 17 Nov 2007 09:17:26 -0600
Subject: [LLVMbugs] [Bug 1808] New: CBE generates code that will SIGILL:
 "ote: if this code is reached, the program will abort"
Message-ID: <bug-1808-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1808

           Summary: CBE generates code that will SIGILL: "ote: if this code
                    is reached, the program will abort"
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: edwintorok at gmail.com
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=1244)
 --> (http://llvm.org/bugs/attachment.cgi?id=1244)
delta reduced testcase

If I compile the attached testcase with CBE, I get a warning from gcc
"<stdin>:320: warning: function called through a non-compatible type
<stdin>:320: note: if this code is reached, the program will abort".

When I run the program I do indeed get a SIGILL, Illegal Instruction.
Looking at the generated assembly file (by gcc), there is a 0x0b0f (which is
ud2a)   which causes this.

Generating assembly directly with llvm-gcc works, the problem is only with the
C code generated by llc.

$ llvm-gcc -Wfatal-errors testcase-min.i -O0 -emit-llvm -c -o - |
/home/edwin/llvm-svn/llvm/Release/bin/llc -march=c | gcc -O0 -x c - -S -o - |
grep 0x0b0f
testcase-min.i:16: warning: ???struct cl_limits??? declared inside parameter
list
testcase-min.i:16: warning: its scope is only this definition or declaration,
which is probably not what you want
testcase-min.i:16: warning: ???struct cl_engine??? declared inside parameter
list
testcase-min.i:23: warning: ???struct optstruct??? declared inside parameter
list
testcase-min.i: In function ???clamav_unpack???:
testcase-min.i:31: warning: incompatible implicit declaration of built-in
function ???fprintf???
<stdin>:148: warning: conflicting types for built-in function ???malloc???
<stdin>: In function ???clamav_unpack???:
<stdin>:320: warning: function called through a non-compatible type
<stdin>:320: note: if this code is reached, the program will abort
                .value  0x0b0f


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sat Nov 17 13:12:53 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sat, 17 Nov 2007 13:12:53 -0600
Subject: [LLVMbugs] [Bug 1809] New: CBE generates code that makes wrong use
 of __builtin_stack_restore/save
Message-ID: <bug-1809-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1809

           Summary: CBE generates code that makes wrong use of
                    __builtin_stack_restore/save
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: major
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: edwintorok at gmail.com
                CC: llvmbugs at cs.uiuc.edu


When I compile oggenc.c from the llvm testsuite on x86-64 I get an ICE in gcc
when trying to compile the generated C code.
GCC developers say llvm got the pointer sizes wrong:

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34134#c3

$ llvm-gcc oggenc.c -c -emit-llvm -o -|opt -std-compile-opts|llc -march=c|gcc
-x c - -O1 -S
....
<stdin>: In function ???_ve_envelope_search???:
<stdin>:48980: internal compiler error: in lhd_set_decl_assembler_name, at
langhooks.c:165
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://gcc.gnu.org/bugs.html> for instructions.
For Debian GNU/Linux specific bug reporting instructions,
see <URL:file:///usr/share/doc/gcc-4.2/README.Bugs>.


I am trying to get a reduced testcase with delta/bugpoint, but it is slow. I'll
attach it when done.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov 18 07:50:10 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 18 Nov 2007 07:50:10 -0600
Subject: [LLVMbugs] [Bug 1810] New: Scalar evolution incorrectly handles Or
	instruction
Message-ID: <bug-1810-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1810

           Summary: Scalar evolution incorrectly handles Or instruction
           Product: new-bugs
           Version: unspecified
          Platform: All
        OS/Version: All
            Status: NEW
          Keywords: miscompilation
          Severity: normal
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: wmatyjewicz at fastmail.fm
                CC: llvmbugs at cs.uiuc.edu


Given the following code:

define void @fun() {
entry:
        br label %header
header:
        %i = phi i32 [ 1, %entry ], [ %i.next, %body ]
        %cond = icmp eq i32 %i, 10
        br i1 %cond, label %exit, label %body
body:
        %a = mul i32 %i, 5
        %b = or i32 %a, 1
        %i.next = add i32 %i, 1
        br label %header
exit:        
        ret void
}

scalar evolution pass (opt -analyze -scalar-evolution) produces incorrect chrec
for %b:
  %a = mul i32 %i, 5   
--> { 5,+, 5}<header>
  %b = or i32 %a, 1
--> { 6,+, 5}<header>
It is incorrect, because at the first iteration %b should be assigned 5, not 6.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Mon Nov 19 11:01:51 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Mon, 19 Nov 2007 11:01:51 -0600
Subject: [LLVMbugs] [Bug 1420] llvm-gcc asserts on address of field in bit
 field structure
In-Reply-To: <bug-1420-206@http.llvm.org/bugs/>
Message-ID: <200711191701.lAJH1p8a024640@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1420


Duncan Sands <baldrick at free.fr> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #8 from Duncan Sands <baldrick at free.fr>  2007-11-19 11:01:51 ---
This compiles to something sensible looking with llvm-gcc-4.2.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Mon Nov 19 12:22:28 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Mon, 19 Nov 2007 12:22:28 -0600
Subject: [LLVMbugs] [Bug 1811] New: Assertion hit on ppc
Message-ID: <bug-1811-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1811

           Summary: Assertion hit on ppc
           Product: libraries
           Version: trunk
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: major
          Priority: P2
         Component: Backend: PowerPC
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: zack at tungstengraphics.com
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=1254)
 --> (http://llvm.org/bugs/attachment.cgi?id=1254)
testcase which hits the assertion

The attached testcase (llvm-as mrcooper.ll && lli mrcooper.bc) hits the
following assertion:
0x10340a30: v4f32 = undef vp-tris: LegalizeDAG.cpp:6209:
void<unnamed>::SelectionDAGLegalize::SplitVectorOp(llvm::SDOperand,
llvm::SDOperand&, llvm::SDOperand&): Assertion `0 && "Unhandled operation in
SplitVectorOp!"' failed.
i86 works fine.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Mon Nov 19 15:45:40 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Mon, 19 Nov 2007 15:45:40 -0600
Subject: [LLVMbugs] [Bug 1812] New: ExpandUnalignedLoad doesn't handle
	vector loads
Message-ID: <bug-1812-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1812

           Summary: ExpandUnalignedLoad doesn't handle vector loads
           Product: libraries
           Version: 1.0
          Platform: PC
        OS/Version: All
            Status: NEW
          Keywords: compile-fail
          Severity: normal
          Priority: P2
         Component: Common Code Generator Code
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: sabre at nondot.org
                CC: natebegeman at mac.com, dalej at apple.com,
                    llvmbugs at cs.uiuc.edu


ExpandUnalignedLoad doesn't handle unaligned vectors right, which causes a
codegen crash when compiling this code to the G5 (-march=ppc32 -mcpu=g5):

define void @execute_shader(<4 x float>* %OUT, <4 x float>* %IN, <4 x float>*
%CONST) {
entry:
        %input2 = load <4 x float>* null, align 8               ; <<4 x float>>
        %shuffle7 = shufflevector <4 x float> %input2, <4 x float> < float
0.000000e+00, float 1.000000e+00, float 0.000000e+00, float 1.000000e+00 >, <4
x i32> < i32 2, i32 2, i32 2, i32 2 >; <<4 x float>> [#uses=1]

        %mul1 = mul <4 x float> %shuffle7, zeroinitializer              ; <<4 x
        %add2 = add <4 x float> %mul1, %input2          ; <<4 x float>>
        store <4 x float> %add2, <4 x float>* null, align 8
        ret void
}


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Mon Nov 19 15:46:58 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Mon, 19 Nov 2007 15:46:58 -0600
Subject: [LLVMbugs] [Bug 1811] Assertion hit on ppc
In-Reply-To: <bug-1811-206@http.llvm.org/bugs/>
Message-ID: <200711192146.lAJLkw3q007514@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1811


Chris Lattner <sabre at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
          Component|Backend: PowerPC            |Common Code Generator Code
           Keywords|                            |compile-fail
         Resolution|                            |FIXED
   Target Milestone|---                         |2.2




--- Comment #4 from Chris Lattner <sabre at nondot.org>  2007-11-19 15:46:58 ---
In the spirit of one bugzilla per problem, I filed the remaining issue as Bug
1812.  It can be worked around by giving your vectors 16-byte alignment instead
of 8-byte alignment.

Testcase for this bug here:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071119/055724.html

-Chris


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Tue Nov 20 11:58:34 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Tue, 20 Nov 2007 11:58:34 -0600
Subject: [LLVMbugs] [Bug 1813] New: llvm-gcc thinks that A*abs(B) == abs(A*B)
Message-ID: <bug-1813-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1813

           Summary: llvm-gcc thinks that A*abs(B) == abs(A*B)
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: major
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: edwintorok at gmail.com
                CC: llvmbugs at cs.uiuc.edu


#include <stdio.h>
#include <stdlib.h>
int main( void )
{
  int i=2;
  if( -10*abs (i-1) == 10*abs(i-1) ) {
    printf ("-10==10!\n");
    abort();
  }
}

I just noticed this gcc bug (which affects all 3.x and 4.x except trunk):
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34130

llvm-gcc is affected by it too, even if I pass it -O0, and don't optimize the
bitcode at all. I think the bug is in the gcc frontend.

I think it would be useful to backport the gcc change to llvm-gcc 4.0/4.2.

I am a bit puzzled though, doesn't llvm use its own constant folding
implementation? [the gcc bug was in const-fold.c according to the commit on
gcc]


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Tue Nov 20 23:10:28 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Tue, 20 Nov 2007 23:10:28 -0600
Subject: [LLVMbugs] [Bug 1814] New: indvars crash on vector subtract
Message-ID: <bug-1814-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1814

           Summary: indvars crash on vector subtract
           Product: libraries
           Version: trunk
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: normal
          Priority: P2
         Component: Loop Optimizer
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: nicholas at mxc.ca
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=1257)
 --> (http://llvm.org/bugs/attachment.cgi?id=1257)
testcase

This bitcode file crashes -indvars. Here's the stack:

opt: /home/nicholas/llvm-commit/include/llvm/Support/Casting.h:199: typename
llvm::cast_retty<To, From>::ret_type llvm::cast(const Y&) [with X =
llvm::IntegerType, Y = const llvm::Type*]: Assertion `isa<X>(Val) &&
"cast<Ty>() argument of incompatible type!"' failed.

Program received signal SIGABRT, Aborted.
[Switching to Thread 0xf7c9d6c0 (LWP 16580)]
0xffffe405 in __kernel_vsyscall ()
(gdb) bt
#0  0xffffe405 in __kernel_vsyscall ()
#1  0xf7cc97f5 in raise () from /lib/i686/cmov/libc.so.6
#2  0xf7ccb181 in abort () from /lib/i686/cmov/libc.so.6
#3  0xf7cc2b6e in __assert_fail () from /lib/i686/cmov/libc.so.6
#4  0x0833c6b6 in llvm::cast<llvm::IntegerType, llvm::Type const*> (
    Val=@0xffaad940)
    at /home/nicholas/llvm-commit/include/llvm/Support/Casting.h:199
#5  0x085221a6 in llvm::ConstantInt::get (Ty=0x86ca390, 
    V=18446744073709551615, isSigned=false) at Constants.cpp:223
#6  0x084bf97b in llvm::ScalarEvolution::getIntegerSCEV (this=0x86ca028, 
    Val=-1, Ty=0x86ca390) at ScalarEvolution.cpp:513
#7  0x084c2cce in llvm::ScalarEvolution::getNegativeSCEV (this=0x86ca028, 
    V=@0xffaada50) at ScalarEvolution.cpp:538
#8  0x084c2f3e in llvm::ScalarEvolution::getMinusSCEV (this=0x86ca028, 
    LHS=@0xffaada54, RHS=@0xffaada50) at ScalarEvolution.cpp:546
#9  0x084c390d in createSCEV (this=0x86cdaf0, V=0x86cb1a8)
    at ScalarEvolution.cpp:1531
#10 0x084c4051 in getSCEV (this=0x86cdaf0, V=0x86cb1a8)
    at ScalarEvolution.cpp:1325
#11 0x084c7287 in getSCEVAtScope (this=0x86cdaf0, V=0x86ce020, L=0x0)
    at ScalarEvolution.cpp:2129
#12 0x084c7d01 in llvm::ScalarEvolution::getSCEVAtScope (this=0x86ca028, 
    V=0x86cb1f8, L=0x0) at ScalarEvolution.cpp:2672
#13 0x0837d0dd in RewriteLoopExitValues (this=0x86cb288, L=0x86cd8e8)
    at IndVarSimplify.cpp:376
#14 0x0837d4a4 in runOnLoop (this=0x86cb288, L=0x86cd8e8, LPM=@0x86c9f58)
    at IndVarSimplify.cpp:461
#15 0x084ac1b3 in llvm::LPPassManager::runOnFunction (this=0x86c9f58, 
    F=@0x86ca4c8) at LoopPass.cpp:225
#16 0x0856a14a in llvm::FPPassManager::runOnFunction (this=0x86c9ad0, 
    F=@0x86ca4c8) at PassManager.cpp:1171
#17 0x0856a2ec in llvm::FPPassManager::runOnModule (this=0x86c9ad0, 
    M=@0x86c9870) at PassManager.cpp:1191
#18 0x08569e20 in llvm::MPPassManager::runOnModule (this=0x86cb2f8, 
    M=@0x86c9870) at PassManager.cpp:1240
#19 0x08569fd8 in llvm::PassManagerImpl::run (this=0x86caec0, M=@0x86c9870)
    at PassManager.cpp:1313
#20 0x0856a02a in llvm::PassManager::run (this=0xffaae1c8, M=@0x86c9870)
    at PassManager.cpp:1345
#21 0x082dfde3 in main (argc=3, argv=0xffaae3c4) at opt.cpp:427
(gdb) up 9
#9  0x084c390d in createSCEV (this=0x86cdaf0, V=0x86cb1a8)
    at ScalarEvolution.cpp:1531
1531                                 getSCEV(I->getOperand(1)));
(gdb) p V->dump()
        sub <4 x i16> zeroinitializer, zeroinitializer          ; <<4 x
i16>>:22 [#uses=1]

$1 = void


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Wed Nov 21 03:00:14 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Wed, 21 Nov 2007 03:00:14 -0600
Subject: [LLVMbugs] [Bug 1815] New: codegen crash without DAG combiner
Message-ID: <bug-1815-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1815

           Summary: codegen crash without DAG combiner
           Product: new-bugs
           Version: unspecified
          Platform: Other
        OS/Version: Linux
            Status: NEW
          Severity: minor
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: baldrick at free.fr
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=1259)
 --> (http://llvm.org/bugs/attachment.cgi?id=1259)
testcase .ll

If you turn off the pre-legalize DAG combiner thusly:

--- lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp       (revision 44262)
+++ lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp       (working copy)
@@ -4739,7 +4739,7 @@
   DEBUG(DAG.dump());

   // Run the DAG combiner in pre-legalize mode.
-  DAG.Combine(false, *AA);
+//  DAG.Combine(false, *AA);

then the attached testcase causes a crash (X86):

$ llc -f udiv.bc
Cannot yet select: 0x8a54950: i8 = udiv 0x8a54410, 0x8a54478

That doesn't seem right.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Wed Nov 21 09:26:44 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Wed, 21 Nov 2007 09:26:44 -0600
Subject: [LLVMbugs] [Bug 1561] g++-4.2 produces a bunch of warnings in
	lib/System/ltdl.c
In-Reply-To: <bug-1561-206@http.llvm.org/bugs/>
Message-ID: <200711211526.lALFQiKt006042@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1561


Duncan Sands <baldrick at free.fr> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED




--- Comment #2 from Duncan Sands <baldrick at free.fr>  2007-11-21 09:26:43 ---
Some kind person fixed this.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Wed Nov 21 13:38:00 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Wed, 21 Nov 2007 13:38:00 -0600
Subject: [LLVMbugs] [Bug 1816] New: APInt::countTrailingZeros() >
	APInt::getBitWidth()
Message-ID: <bug-1816-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1816

           Summary: APInt::countTrailingZeros() > APInt::getBitWidth()
           Product: libraries
           Version: trunk
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: Core LLVM classes
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: nlewycky at google.com
                CC: llvmbugs at cs.uiuc.edu


APInt's "countTailingZeros" function is either buggy, or at best misleading.
Consider the following standalone program (link with LLVM core):

  #include <iostream>
  #include "llvm/ADT/APInt.h"

  int main() {
    llvm::APInt myint = llvm::APInt::getNullValue(290);
    uint32_t zeros = myint.countTrailingZeros();

    std::cout << zeros << "\n";
  }

So, how many "trailing zeros" are there in a 290-bit integer?

320.

What's slightly worrying is that this behaviour is documented. The
documentation http://llvm.org/doxygen/classllvm_1_1APInt.html#z9_9 says
"Returns: getNumWords() * APINT_BITS_PER_WORD if the value is zero."

So the code matches the documentation, but I'm thinking that both are wrong
here. It doesn't make sense for it to return a value greater than its BitWidth.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Wed Nov 21 21:07:47 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Wed, 21 Nov 2007 21:07:47 -0600
Subject: [LLVMbugs] [Bug 1806] Andersen's Analysis crashes on inline asm
In-Reply-To: <bug-1806-206@http.llvm.org/bugs/>
Message-ID: <200711220307.lAM37ld4010667@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1806


Nick Lewycky <nicholas at mxc.ca> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED




--- Comment #7 from Nick Lewycky <nicholas at mxc.ca>  2007-11-21 21:07:47 ---
Patch checked in (got review from Dan Berlin on IRC).


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Thu Nov 22 02:00:45 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Thu, 22 Nov 2007 02:00:45 -0600
Subject: [LLVMbugs] [Bug 1810] Scalar evolution incorrectly handles Or
	instruction
In-Reply-To: <bug-1810-206@http.llvm.org/bugs/>
Message-ID: <200711220800.lAM80j7v027943@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1810


Nick Lewycky <nicholas at mxc.ca> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #8 from Nick Lewycky <nicholas at mxc.ca>  2007-11-22 02:00:45 ---
Patch checked in.
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071119/055744.html

I forked the countTrailingZeros issue to a new bug. Thanks for the patch!


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Thu Nov 22 15:54:52 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Thu, 22 Nov 2007 15:54:52 -0600
Subject: [LLVMbugs] [Bug 1817] New: Instcombine assertion
Message-ID: <bug-1817-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1817

           Summary: Instcombine assertion
           Product: libraries
           Version: trunk
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: Scalar Optimizations
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: asl at math.spbu.ru
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=1262)
 --> (http://llvm.org/bugs/attachment.cgi?id=1262)
Bytecode in question

This one was caught by pypy folks. It looks like some internal stuff was
corrupted. valgrind is clean. Steps to reproduce:

 opt foo.bc -inline -tailduplicate -scalarrepl -gvn -instcombine -simplifycfg

Bugpoint wasn't be able to reduce stuff: it was fooled by another assertion
caught after removing of global initializers.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Thu Nov 22 15:58:18 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Thu, 22 Nov 2007 15:58:18 -0600
Subject: [LLVMbugs] [Bug 1818] New: memdep should eliminate successive
 identical pure calls
Message-ID: <bug-1818-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1818

           Summary: memdep should eliminate successive identical pure calls
           Product: new-bugs
           Version: unspecified
          Platform: Other
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: baldrick at free.fr
                CC: llvmbugs at cs.uiuc.edu


The testcase test/Analysis/BasicAA/pure-const-dce.ll has two "pure" calls
in a row:
        %tmp5 = call i32 @TestPure( i32 6 ) readonly
        %tmp6 = call i32 @TestPure( i32 6 ) readonly
One of them could be eliminated, but isn't.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Thu Nov 22 16:06:51 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Thu, 22 Nov 2007 16:06:51 -0600
Subject: [LLVMbugs] [Bug 1819] New: Struct-returning const functions do not
 benefit from being const
Message-ID: <bug-1819-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1819

           Summary: Struct-returning const functions do not benefit from
                    being const
           Product: new-bugs
           Version: unspecified
          Platform: Other
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: baldrick at free.fr
                CC: llvmbugs at cs.uiuc.edu


Normally if a function is const/pure in gcc it gets turned into
a readnone/readonly function in LLVM.  The exception to this is
functions that return a struct: they get an extra pointer parameter
that is used to write out the return value, so can't be marked
readnone or readonly (since otherwise optimizers think that the
no return value is written, because readnone/readonly functions
can't write via pointers...).  There's a similar problems for
byval functions: they read from the pointer used to pass the
struct argument by value, so can't be marked readnone.  It would
be nice to do something about this so that such functions get
optimized better.

This problem was split out of PR1383.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Thu Nov 22 17:47:52 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Thu, 22 Nov 2007 17:47:52 -0600
Subject: [LLVMbugs] [Bug 1817] Instcombine assertion
In-Reply-To: <bug-1817-206@http.llvm.org/bugs/>
Message-ID: <200711222347.lAMNlqDn016627@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1817


Chris Lattner <sabre at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|enhancement                 |normal
             Status|NEW                         |RESOLVED
           Keywords|                            |compile-fail
         Resolution|                            |FIXED
   Target Milestone|---                         |2.2




--- Comment #5 from Chris Lattner <sabre at nondot.org>  2007-11-22 17:47:51 ---
Fixed, patch here:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071119/055760.html

-Chris


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Thu Nov 22 22:26:05 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Thu, 22 Nov 2007 22:26:05 -0600
Subject: [LLVMbugs] [Bug 1820] New: Macro expansion bug
Message-ID: <bug-1820-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1820

           Summary: Macro expansion bug
           Product: cfe
           Version: unspecified
          Platform: PC
        OS/Version: NetBSD
            Status: NEW
          Severity: normal
          Priority: P2
         Component: preprocessor
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: neil at daikokuya.co.uk
                CC: llvmbugs at cs.uiuc.edu


clang doesn't implement the rescan rule of C99 6.10.3.4p2 correctly.

#define f(x) h(x
#define h(x) x(void) 
extern int f(f));

should preprocess to

   extern int f(void);

clang outputs

   extern int h(void;

Compare to GCC, ucpp, sparse and Comeau online at
http://www.comeaucomputing.com/tryitout/ for example.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Fri Nov 23 00:51:37 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Fri, 23 Nov 2007 00:51:37 -0600
Subject: [LLVMbugs] [Bug 1820] Macro expansion bug
In-Reply-To: <bug-1820-206@http.llvm.org/bugs/>
Message-ID: <200711230651.lAN6pblu008954@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1820


Chris Lattner <sabre at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #1 from Chris Lattner <sabre at nondot.org>  2007-11-23 00:51:37 ---
Fixed, patch here:
http://lists.cs.uiuc.edu/pipermail/cfe-commits/Week-of-Mon-20071119/002954.html

Thanks for reporting this incredibly subtle bug, please review the fix and see
if it makes sense to you.

-Chris


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Fri Nov 23 02:02:37 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Fri, 23 Nov 2007 02:02:37 -0600
Subject: [LLVMbugs] [Bug 1821] New: llvm spills like crazy on fft code
Message-ID: <bug-1821-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1821

           Summary: llvm spills like crazy on fft code
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: All
            Status: NEW
          Keywords: code-quality
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: sabre at nondot.org
                CC: evan.cheng at apple.com, llvmbugs at cs.uiuc.edu


On Anton's new oourafft.c example, we run consistently ~20% slower than
GCC-compiled code.  Here's a sample cut-down testcase, we spill far more than
GCC on x86-32.  On x86-64, we generate an identical number of instructions,
presumably due to the extra regs available.

  void test(int n, double *a, double *w)
{
  int j, k1, k2;
  double wk1r, wk1i, wk2r, wk2i, wk3r, wk3i;
  double x0r, x0i, x1r, x1i, x2r, x2i, x3r, x3i;

  x0r = a[0] + a[2];
  x0i = a[1] + a[3];
  x1r = a[0] - a[2];
  x1i = a[1] - a[3];
  x2r = a[4] + a[6];
  x2i = a[5] + a[7];
  x3r = a[4] - a[6];
  x3i = a[5] - a[7];
  a[0] = x0r + x2r;
  a[1] = x0i + x2i;
  a[4] = x0r - x2r;
  a[5] = x0i - x2i;
  a[2] = x1r - x3i;
  a[3] = x1i + x3r;
  a[6] = x1r + x3i;
  a[7] = x1i - x3r;
}


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Fri Nov 23 02:47:02 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Fri, 23 Nov 2007 02:47:02 -0600
Subject: [LLVMbugs] [Bug 1814] indvars crash on vector subtract
In-Reply-To: <bug-1814-206@http.llvm.org/bugs/>
Message-ID: <200711230847.lAN8l2EV021950@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1814


Chris Lattner <sabre at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
          Component|Loop Optimizer              |Global Analyses
           Keywords|                            |compile-fail
         Resolution|                            |FIXED
   Target Milestone|---                         |2.2




--- Comment #1 from Chris Lattner <sabre at nondot.org>  2007-11-23 02:47:01 ---
Fixed, patch here:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071119/055762.html

Thanks!


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Fri Nov 23 14:04:45 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Fri, 23 Nov 2007 14:04:45 -0600
Subject: [LLVMbugs] [Bug 1383] new per-function attributes for pure/const
In-Reply-To: <bug-1383-206@http.llvm.org/bugs/>
Message-ID: <200711232004.lANK4jlm021226@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1383


Duncan Sands <baldrick at free.fr> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #36 from Duncan Sands <baldrick at free.fr>  2007-11-23 14:04:44 ---
Fixed, thanks to Anton, Edwin and yours truly:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071112/055613.html
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071112/055613.html
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071112/055622.html
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071112/055623.html
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071119/055711.html
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071119/055712.html
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071119/055745.html
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071119/055746.html
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071119/055747.html
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071119/055748.html
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071119/055749.html
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071119/055754.html
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071119/055755.html
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071119/055756.html
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071119/055763.html
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071119/055769.html


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Fri Nov 23 16:44:27 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Fri, 23 Nov 2007 16:44:27 -0600
Subject: [LLVMbugs] [Bug 1816] APInt::countTrailingZeros(0) defined wrong
In-Reply-To: <bug-1816-206@http.llvm.org/bugs/>
Message-ID: <200711232244.lANMiRj5028592@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1816


Chris Lattner <sabre at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|enhancement                 |normal
             Status|NEW                         |RESOLVED
           Keywords|                            |miscompilation
         Resolution|                            |FIXED
            Summary|APInt::countTrailingZeros() |APInt::countTrailingZeros(0)
                   |> APInt::getBitWidth()      |defined wrong
   Target Milestone|---                         |2.2




--- Comment #2 from Chris Lattner <sabre at nondot.org>  2007-11-23 16:44:27 ---
Fixed, patches here:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071119/055773.html
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071119/055775.html

This bug actually manifested it as a miscompilation when constant folding
llvm.cttz also.  I verified that ctlz was already doing the right thing, even
though it was documented not to.

Thanks!

-Chris


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sat Nov 24 18:09:36 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sat, 24 Nov 2007 18:09:36 -0600
Subject: [LLVMbugs] [Bug 1822] New: instcombine misses select i1 %X, %Y, %X
Message-ID: <bug-1822-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1822

           Summary: instcombine misses select i1 %X, %Y, %X
           Product: libraries
           Version: trunk
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: Scalar Optimizations
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: nicholas at mxc.ca
                CC: llvmbugs at cs.uiuc.edu


I found this line of code in a testcase from Pypy. Not sure if this is actually
there or if it's just a bugpoint artefact, but regardless:

  %v1650 = select i1 %v1652, i1 %v1653, i1 %v1652

should become %v1650 = and i1 %v1653, i1 %v1652. Similarly, the opposite
(switch the last two args to select) should become an Or. The codegen catches
these cases anyways.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov 25 02:36:30 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 25 Nov 2007 02:36:30 -0600
Subject: [LLVMbugs] [Bug 731] Enhanced Makefile System To Use SCONS
In-Reply-To: <bug-731-206@http.llvm.org/bugs/>
Message-ID: <200711250836.lAP8aU8W026403@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=731


Reid Spencer <rspencer at reidspencer.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |REOPENED
         Resolution|LATER                       |




--- Comment #10 from Reid Spencer <rspencer at reidspencer.com>  2007-11-25 02:36:29 ---
I think using scons at this point is just too much of a change for llvm. What I
would like to do is the original proposal of modularizing the makefiles so that
each directory builds one kind of thing, etc. I've already started that work in
the support module (currently unused). I'd like to see all the supportish type
things moved there and all the other modules start using it. Of course, it
needs significant upgrade since I last worked on it. I need a makefile system
for HLVM and I'm abandoning scons there because I don't want HLVM looking any
different from any of the other LLVM projects. So, I'll be working on this in
the support module so that HLVM and sample projects can use it. When those are
sufficiently working well, I'll retro-fit LLVM .. don't hold your breath :) 


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov 25 08:11:52 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 25 Nov 2007 08:11:52 -0600
Subject: [LLVMbugs] [Bug 1716] noalias attribute pessimization
In-Reply-To: <bug-1716-206@http.llvm.org/bugs/>
Message-ID: <200711251411.lAPEBqFZ009060@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1716


Duncan Sands <baldrick at free.fr> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #12 from Duncan Sands <baldrick at free.fr>  2007-11-25 08:11:51 ---
Fixed here:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071119/055792.html


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov 25 08:14:09 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 25 Nov 2007 08:14:09 -0600
Subject: [LLVMbugs] [Bug 1823] New: 'noalias' attribute lost when compiling
 with optimization
Message-ID: <bug-1823-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1823

           Summary: 'noalias' attribute lost when compiling with
                    optimization
           Product: new-bugs
           Version: unspecified
          Platform: Other
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: baldrick at free.fr
                CC: llvmbugs at cs.uiuc.edu


Testcase:

#include <stdio.h>
#include <stdlib.h>

int power(int X) {
  if (X == 0) return 1;
  return X*power(X-1);
}

int main(int argc, char **argv) {
  printf("%d\n", power(atoi(argv[0])));
}

When compiled at -O0 the declaration of printf is:

declare i32 @printf(i8* noalias , ...) nounwind

When compiled at -O1 it is:

declare i32 @printf(i8*, ...) nounwind

Somehow the noalias attribute gets lost at -O1.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov 25 11:24:58 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 25 Nov 2007 11:24:58 -0600
Subject: [LLVMbugs] [Bug 1824] New: sema segfaults on conditional operator
	without LHS
Message-ID: <bug-1824-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1824

           Summary: sema segfaults on conditional operator without LHS
           Product: cfe
           Version: unspecified
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: normal
          Priority: P2
         Component: Semantic Analyzer
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: nunoplopes at sapo.pt
                CC: llvmbugs at cs.uiuc.edu


this makes clang segfault:
int a = 1?:3;

This is because Sema::UsualUnaryConversions() dereferences LHS (NULL) several
times.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov 25 15:28:31 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 25 Nov 2007 15:28:31 -0600
Subject: [LLVMbugs] [Bug 1822] instcombine misses select i1 %X, %Y, %X
In-Reply-To: <bug-1822-206@http.llvm.org/bugs/>
Message-ID: <200711252128.lAPLSVxq003662@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1822


Chris Lattner <sabre at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #1 from Chris Lattner <sabre at nondot.org>  2007-11-25 15:28:30 ---
Implemented, patch here:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071119/055795.html


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov 25 16:42:09 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 25 Nov 2007 16:42:09 -0600
Subject: [LLVMbugs] [Bug 1614] SCEV needs min/max expressions for trip count
	analysis
In-Reply-To: <bug-1614-206@http.llvm.org/bugs/>
Message-ID: <200711252242.lAPMg9Ws008486@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1614


Nick Lewycky <nicholas at mxc.ca> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED




--- Comment #10 from Nick Lewycky <nicholas at mxc.ca>  2007-11-25 16:42:08 ---
Patch checked in!


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov 25 18:29:30 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 25 Nov 2007 18:29:30 -0600
Subject: [LLVMbugs] [Bug 1383] new per-function attributes for pure/const
In-Reply-To: <bug-1383-206@http.llvm.org/bugs/>
Message-ID: <200711260029.lAQ0TUUK014744@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1383


Owen Anderson <resistor at mac.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |REOPENED
         Resolution|FIXED                       |




--- Comment #39 from Owen Anderson <resistor at mac.com>  2007-11-25 18:29:28 ---
I don't believe the AA patch is correct.  In
test/Analysis/BasicAA/pure-const-dce.ll, I'm seeing a return of ModRef for the
function called by %tmp5.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov 25 19:41:34 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 25 Nov 2007 19:41:34 -0600
Subject: [LLVMbugs] [Bug 1824] sema segfaults on conditional operator
	without LHS
In-Reply-To: <bug-1824-206@http.llvm.org/bugs/>
Message-ID: <200711260141.lAQ1fYvN020005@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1824


Chris Lattner <sabre at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #1 from Chris Lattner <sabre at nondot.org>  2007-11-25 19:41:33 ---
Fixed, patch here:
http://lists.cs.uiuc.edu/pipermail/cfe-commits/Week-of-Mon-20071119/002974.html


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Sun Nov 25 23:29:50 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Sun, 25 Nov 2007 23:29:50 -0600
Subject: [LLVMbugs] [Bug 1825] New: GVN miscompiles siod
Message-ID: <bug-1825-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1825

           Summary: GVN miscompiles siod
           Product: libraries
           Version: trunk
          Platform: All
        OS/Version: All
            Status: NEW
          Severity: normal
          Priority: P2
         Component: Scalar Optimizations
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: nicholas at mxc.ca
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=1267)
 --> (http://llvm.org/bugs/attachment.cgi?id=1267)
bugpoint reduced testcase

GVN collapses the 2 calls to strlen into one. In between them, there's a call
to "@htqs_arg_bb27" with i8* %tmp25 as a parameter. That pointer could be used
to modify the contents of %tmpbuff, and make the result of strlen different
from the first call to the second.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Mon Nov 26 02:52:16 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Mon, 26 Nov 2007 02:52:16 -0600
Subject: [LLVMbugs] [Bug 1825] GVN miscompiles siod
In-Reply-To: <bug-1825-206@http.llvm.org/bugs/>
Message-ID: <200711260852.lAQ8qGaA022127@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1825


Nick Lewycky <nicholas at mxc.ca> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
           Keywords|                            |miscompilation
         Resolution|                            |FIXED




--- Comment #2 from Nick Lewycky <nicholas at mxc.ca>  2007-11-26 02:52:16 ---
He didn't close the bug, but Owen fixed it with this commit:

http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071126/055800.html


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Mon Nov 26 02:54:03 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Mon, 26 Nov 2007 02:54:03 -0600
Subject: [LLVMbugs] [Bug 1383] new per-function attributes for pure/const
In-Reply-To: <bug-1383-206@http.llvm.org/bugs/>
Message-ID: <200711260854.lAQ8s38Z022226@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1383


Owen Anderson <resistor at mac.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|REOPENED                    |RESOLVED
         Resolution|                            |FIXED




-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Mon Nov 26 17:57:40 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Mon, 26 Nov 2007 17:57:40 -0600
Subject: [LLVMbugs] [Bug 1815] codegen crash without DAG combiner
In-Reply-To: <bug-1815-206@http.llvm.org/bugs/>
Message-ID: <200711262357.lAQNvevB007937@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1815


Dan Gohman <djg at cray.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #4 from Dan Gohman <djg at cray.com>  2007-11-26 17:57:39 ---
The xform is producing a divide only because it expects the divide will be
optimized into something else, so checking for legality of the divide isn't
exactly what's needed.

It looks like LLVM's division-by-integer-constant logic only handles i32 and
i64 right now, so that's why this i8 udiv isn't being optimized.

I fixed the bug by disabling the xform in the case that the division isn't
optimized. When it is optimized, the expanded code is already be legal.

The commit is here:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071126/055807.html


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Mon Nov 26 19:57:59 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Mon, 26 Nov 2007 19:57:59 -0600
Subject: [LLVMbugs] [Bug 1826] New: llvm.exp mishandles & in RUN lines
Message-ID: <bug-1826-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1826

           Summary: llvm.exp mishandles & in RUN lines
           Product: Test Suite
           Version: trunk
          Platform: All
        OS/Version: All
            Status: NEW
          Severity: normal
          Priority: P2
         Component: DejaGNU
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: djg at cray.com
                CC: llvmbugs at cs.uiuc.edu


This one-line test:                                                             

; RUN: echo hi |& false                                                         

passes; it should fail.

The problem appears to be on this line in test/lib/llvm.exp:

      } elseif {[regexp {RUN: *([^&]+)(&&)?} $line match oneline suffix]} {

I just removed the last use of && from the llvm regression tests, so the
bug can probably be fixed by removing the special handling for &. However,
this uncovers many failures due to malformed RUN lines and failing tests.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Tue Nov 27 01:25:45 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Tue, 27 Nov 2007 01:25:45 -0600
Subject: [LLVMbugs] [Bug 1827] New: crash on simple loop example
Message-ID: <bug-1827-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1827

           Summary: crash on simple loop example
           Product: libraries
           Version: trunk
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: Loop Optimizer
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: nicholas at mxc.ca
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=1271)
 --> (http://llvm.org/bugs/attachment.cgi?id=1271)
hand-written crasher (not reduced)

I got this example from the comment in the top of GCC's
tree-scalar-evolution.c. Amazingly, it triggers an assert in LLVM:

$ llvm-commit/Debug/bin/opt -analyze -scalar-evolution gcc3.bc

Printing analysis 'Scalar Evolution Analysis' for function 'foo':
Classifying expressions for: foo
        %a = phi i32 [ 2, %entry ], [ %b, %loop_1 ]             ; <i32>
[#uses=2]
opt: APInt.cpp:1725: llvm::APInt llvm::APInt::udiv(const llvm::APInt&) const:
Assertion `RHS.VAL != 0 && "Divide by zero?"' failed.
  --> { 2,+, 1}<loop_1>         Exits: llvm-commit/Debug/bin/opt[0x85d259a]
llvm-commit/Debug/bin/opt[0x85d2860]
[0xffffe500]
Aborted
(gdb) bt
#0  0xffffe405 in __kernel_vsyscall ()
#1  0xf7c3f7f5 in raise () from /lib/i686/cmov/libc.so.6
#2  0xf7c41181 in abort () from /lib/i686/cmov/libc.so.6
#3  0xf7c38b6e in __assert_fail () from /lib/i686/cmov/libc.so.6
#4  0x085a6500 in llvm::APInt::udiv (this=0xffe45c9c, RHS=@0xffe45bf8)
    at APInt.cpp:1725
#5  0x084cfbd1 in llvm::APInt::sdiv (this=0xffe45c9c, RHS=@0xffe45bf8)
    at /home/nicholas/llvm-commit/include/llvm/ADT/APInt.h:608
#6  0x084b9217 in SolveQuadraticEquation (AddRec=0x86c7618, SE=@0x86c4448)
    at ScalarEvolution.cpp:2342
#7  0x084c0700 in llvm::SCEVAddRecExpr::getNumIterationsInRange (
    this=0x86c75c8, Range=@0xffe45f20, SE=@0x86c4448)
    at ScalarEvolution.cpp:2563
#8  0x084c0176 in llvm::SCEVAddRecExpr::getNumIterationsInRange (
    this=0x86c7558, Range=@0xffe460c4, SE=@0x86c4448)
    at ScalarEvolution.cpp:2503
#9  0x084c1870 in ComputeIterationCount (this=0x86c5000, L=0x86c4f70)
    at ScalarEvolution.cpp:1797
#10 0x084c1edf in getIterationCount (this=0x86c5000, L=0x86c4f70)
    at ScalarEvolution.cpp:1676
#11 0x084c2e04 in getSCEVAtScope (this=0x86c5000, V=0x86c72a8, L=0x0)
    at ScalarEvolution.cpp:2271
#12 0x084c3003 in llvm::ScalarEvolution::getSCEVAtScope (this=0x86c4448, 
    V=0x86c45b8, L=0x0) at ScalarEvolution.cpp:2673
#13 0x084c326c in llvm::ScalarEvolution::print (this=0x86c4448, OS=@0x86bd0a0)
    at ScalarEvolution.cpp:2723
#14 0x082dcafe in llvm::Pass::print (this=0x86c4448, O=0x86bd0a0, M=0x86c3870)
    at /home/nicholas/llvm-commit/include/llvm/Pass.h:119
#15 0x082dcaad in runOnFunction (this=0x86c31b0, F=@0x86c5b80) at opt.cpp:166
#16 0x085645f6 in llvm::FPPassManager::runOnFunction (this=0x86c5608, 
    F=@0x86c5b80) at PassManager.cpp:1171
#17 0x08564798 in llvm::FPPassManager::runOnModule (this=0x86c5608, 
    M=@0x86c3870) at PassManager.cpp:1191
#18 0x085642cc in llvm::MPPassManager::runOnModule (this=0x86c52f8, 
    M=@0x86c3870) at PassManager.cpp:1240
#19 0x08564484 in llvm::PassManagerImpl::run (this=0x86c2fe0, M=@0x86c3870)
    at PassManager.cpp:1313
#20 0x085644d6 in llvm::PassManager::run (this=0xffe46548, M=@0x86c3870)
    at PassManager.cpp:1345
#21 0x082dc603 in main (argc=4, argv=0xffe46744) at opt.cpp:426


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Tue Nov 27 03:31:15 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Tue, 27 Nov 2007 03:31:15 -0600
Subject: [LLVMbugs] [Bug 1828] New: loop-index-split produce bogus code
Message-ID: <bug-1828-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1828

           Summary: loop-index-split produce bogus code
           Product: libraries
           Version: trunk
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: normal
          Priority: P2
         Component: Loop Optimizer
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: asl at math.spbu.ru
                CC: dpatel at apple.com, llvmbugs at cs.uiuc.edu


Created an attachment (id=1272)
 --> (http://llvm.org/bugs/attachment.cgi?id=1272)
Bytecode in question

Consider the attached bytecode (reduced from 90M pypy-generated .ll).

Running opt -loop-index.split resulted:

Instruction does not dominate all uses!
        %v63822.clone = add i32 %j_167.reg2mem.0.clone, -1              ; <i32>
[#uses=4]
        %lsplit.ev.clone = icmp slt i32 0, %v63822.clone                ; <i1>
[#uses=1]
Instruction does not dominate all uses!
        %lsplit.ev.clone = icmp slt i32 0, %v63822.clone                ; <i1>
[#uses=1]
        %lsplit.ev1.clone = select i1 %lsplit.ev.clone, i32 0, i32
%v63822.clone                ; <i32> [#uses=0]
Instruction does not dominate all uses!
        %v63822 = add i32 %j_167.reg2mem.0, -1          ; <i32> [#uses=3]
        %lsplit.ev = icmp slt i32 0, %v63822            ; <i1> [#uses=1]
Instruction does not dominate all uses!
        %lsplit.ev = icmp slt i32 0, %v63822            ; <i1> [#uses=1]
        %lsplit.ev1 = select i1 %lsplit.ev, i32 0, i32 %v63822          ; <i32>
[#uses=1]
Instruction does not dominate all uses!
        %lsplit.ev1 = select i1 %lsplit.ev, i32 0, i32 %v63822          ; <i32>
[#uses=1]
        %v6377959 = icmp slt i32 %j_167.reg2mem.0, %lsplit.ev1          ; <i1>
[#uses=1]
Instruction does not dominate all uses!
        %v6377959 = icmp slt i32 %j_167.reg2mem.0, %lsplit.ev1          ; <i1>
[#uses=1]
        br i1 %v6377959, label %block46, label %block44.block42.clone_crit_edge
Broken module found, compilation aborted!


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Tue Nov 27 07:30:44 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Tue, 27 Nov 2007 07:30:44 -0600
Subject: [LLVMbugs] [Bug 1146] Move function attributes out of FunctionType
 into calls and functions
In-Reply-To: <bug-1146-206@http.llvm.org/bugs/>
Message-ID: <200711271330.lARDUiSw025958@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1146


Duncan Sands <baldrick at free.fr> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED




--- Comment #55 from Duncan Sands <baldrick at free.fr>  2007-11-27 07:30:35 ---
Fixed here (original patches by Reid Spencer, updated by yours truly):
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071126/055824.html
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071126/055825.html
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071126/055826.html


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Tue Nov 27 07:38:51 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Tue, 27 Nov 2007 07:38:51 -0600
Subject: [LLVMbugs] [Bug 1393] llvm-upgrade fails to upgrade these tests
In-Reply-To: <bug-1393-206@http.llvm.org/bugs/>
Message-ID: <200711271338.lARDcpFn026392@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1393


Duncan Sands <baldrick at free.fr> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |baldrick at free.fr
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #1 from Duncan Sands <baldrick at free.fr>  2007-11-27 07:38:28 ---
Resolved by PR1146 fix.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Tue Nov 27 22:56:58 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Tue, 27 Nov 2007 22:56:58 -0600
Subject: [LLVMbugs] [Bug 1829] New: CodeGen/ARM/2007-03-13-InstrSched.ll
	failing
Message-ID: <bug-1829-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1829

           Summary: CodeGen/ARM/2007-03-13-InstrSched.ll failing
           Product: new-bugs
           Version: unspecified
          Platform: Macintosh
        OS/Version: MacOS X
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: tonic at nondot.org
                CC: llvmbugs at cs.uiuc.edu


llvm-as < /Users/lattner/work/llvm/test/CodeGen/ARM/2007-03-13-InstrSched.ll |
llc -mtriple=arm-apple-darwin -relocation-model=pic  -mattr=+v6 -stats

outputs:
 51 asm-printer   - Number of machine instrs printed

The test is looking for 41.

I've XFAILed this for now.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Tue Nov 27 22:58:32 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Tue, 27 Nov 2007 22:58:32 -0600
Subject: [LLVMbugs] [Bug 1830] New: CodeGen/ARM/fp.ll failing
Message-ID: <bug-1830-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1830

           Summary: CodeGen/ARM/fp.ll failing
           Product: new-bugs
           Version: unspecified
          Platform: Macintosh
        OS/Version: Mac System 9.x
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: tonic at nondot.org
                CC: llvmbugs at cs.uiuc.edu


FAIL: test/CodeGen/ARM/fp.ll
Failed with exit(1) at line 6
while running: grep fmrrd fp.ll.tmp | count 5
count: expected 5 lines and got        3.

I get this output:
grep fmrrd fp.ll.tmp
        fmrrd r0, r1, d0
        fmrrd r0, r1, d0
        fmrrd r0, r1, d0


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Tue Nov 27 22:59:53 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Tue, 27 Nov 2007 22:59:53 -0600
Subject: [LLVMbugs] [Bug 1831] New: CodeGen/ARM/lsr-code-insertion.ll failing
Message-ID: <bug-1831-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1831

           Summary: CodeGen/ARM/lsr-code-insertion.ll failing
           Product: new-bugs
           Version: unspecified
          Platform: Macintosh
        OS/Version: Mac System 9.x
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: tonic at nondot.org
                CC: llvmbugs at cs.uiuc.edu


RUN: llvm-as < %s | llc -stats |& grep {40.*Number of machine instrs printed}

CodeGen/ARM/lsr-code-insertion.ll produces:
42 asm-printer    - Number of machine instrs printed

The test is looking for 40 though.

XFAILed for now.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Tue Nov 27 23:00:52 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Tue, 27 Nov 2007 23:00:52 -0600
Subject: [LLVMbugs] [Bug 1826] llvm.exp mishandles & in RUN lines
In-Reply-To: <bug-1826-206@http.llvm.org/bugs/>
Message-ID: <200711280500.lAS50qla015865@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1826


Tanya Lattner <tonic at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #4 from Tanya Lattner <tonic at nondot.org>  2007-11-27 23:00:51 ---
Fixed:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071126/055851.html


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Tue Nov 27 23:15:51 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Tue, 27 Nov 2007 23:15:51 -0600
Subject: [LLVMbugs] [Bug 1807] documentation: need to reconfigure llvm to
 compile llvm test-suite
In-Reply-To: <bug-1807-206@http.llvm.org/bugs/>
Message-ID: <200711280515.lAS5FpbR016472@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1807


Tanya Lattner <tonic at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #5 from Tanya Lattner <tonic at nondot.org>  2007-11-27 23:15:51 ---
Fixed:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071126/055852.html


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Tue Nov 27 23:27:36 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Tue, 27 Nov 2007 23:27:36 -0600
Subject: [LLVMbugs] [Bug 1830] CodeGen/ARM/fp.ll failing
In-Reply-To: <bug-1830-206@http.llvm.org/bugs/>
Message-ID: <200711280527.lAS5RaPv017132@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1830


Chris Lattner <sabre at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #1 from Chris Lattner <sabre at nondot.org>  2007-11-27 23:27:35 ---
Fixed, yay for better code:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071126/055854.html


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Wed Nov 28 13:38:00 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Wed, 28 Nov 2007 13:38:00 -0600
Subject: [LLVMbugs] [Bug 1832] New: dejagnu should fail a test with no RUN:
	lines
Message-ID: <bug-1832-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1832

           Summary: dejagnu should fail a test with no RUN: lines
           Product: Test Suite
           Version: trunk
          Platform: Macintosh
        OS/Version: Mac System 9.x
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: DejaGNU
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: sabre at nondot.org
                CC: llvmbugs at cs.uiuc.edu


Tests with no run lines pass, they should fail, ideally with a message
complaining about no  run lines :)

e.g.:
http://lists.cs.uiuc.edu/pipermail/cfe-commits/Week-of-Mon-20071126/003016.html

-Chris


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Wed Nov 28 15:16:56 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Wed, 28 Nov 2007 15:16:56 -0600
Subject: [LLVMbugs] [Bug 1832] dejagnu should fail a test with no RUN: lines
In-Reply-To: <bug-1832-206@http.llvm.org/bugs/>
Message-ID: <200711282116.lASLGubk005148@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1832


Tanya Lattner <tonic at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |tonic at nondot.org
             Status|NEW                         |RESOLVED
         Resolution|                            |INVALID




--- Comment #2 from Tanya Lattner <tonic at nondot.org>  2007-11-28 15:16:55 ---
Tests with no RUN lines fail in the tcl stuff.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Wed Nov 28 15:35:34 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Wed, 28 Nov 2007 15:35:34 -0600
Subject: [LLVMbugs] [Bug 1809] GCC ICEs on CBE output that uses
	__builtin_stack_save
In-Reply-To: <bug-1809-206@http.llvm.org/bugs/>
Message-ID: <200711282135.lASLZYYu006147@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1809


T??r??k Edwin <edwintorok at gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #12 from T??r??k Edwin <edwintorok at gmail.com>  2007-11-28 15:35:34 ---
(In reply to comment #11)
> Please try this patch, if it works, please close this bug as fixed:
> http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071126/055870.html
> 

It works. Thanks!


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Wed Nov 28 16:48:57 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Wed, 28 Nov 2007 16:48:57 -0600
Subject: [LLVMbugs] [Bug 1744] llvm-gcc crash on valid initializer
In-Reply-To: <bug-1744-206@http.llvm.org/bugs/>
Message-ID: <200711282248.lASMmvIJ010077@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1744


Chris Lattner <sabre at nondot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|enhancement                 |normal
             Status|NEW                         |RESOLVED
          Component|new bugs                    |llvm-gcc
           Keywords|                            |compile-fail
            Product|new-bugs                    |tools
         Resolution|                            |FIXED
   Target Milestone|---                         |2.2
            Version|unspecified                 |1.0




--- Comment #7 from Chris Lattner <sabre at nondot.org>  2007-11-28 16:48:56 ---
Fixed, patch here:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071126/055872.html
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071126/055873.html

Testcase here: test/CFrontend/2007-11-28-GlobalInitializer.c

-Chris


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Thu Nov 29 02:11:38 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Thu, 29 Nov 2007 02:11:38 -0600
Subject: [LLVMbugs] [Bug 1833] New: llc -enable-eh crash in LegalizeDAG
Message-ID: <bug-1833-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1833

           Summary: llc -enable-eh crash in LegalizeDAG
           Product: new-bugs
           Version: unspecified
          Platform: Other
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: baldrick at free.fr
                CC: llvmbugs at cs.uiuc.edu


Created an attachment (id=1273)
 --> (http://llvm.org/bugs/attachment.cgi?id=1273)
bugpoint reduced .ll

$ llc -f -enable-eh crash.bc
llc: lib/CodeGen/SelectionDAG/LegalizeDAG.cpp:340:
void<unnamed>::SelectionDAGLegalize::LegalizeDAG(): Assertion
`LegalizedNodes.count(OldRoot) && "Root didn't get legalized?"' failed.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From asl at math.spbu.ru  Thu Nov 29 10:07:52 2007
From: asl at math.spbu.ru (Anton Korobeynikov)
Date: Thu, 29 Nov 2007 19:07:52 +0300
Subject: [LLVMbugs] Build is broken
Message-ID: <1196352472.23227.30.camel@asl.dorms.spbu.ru>

Hello, Everyone.

The error message listed here:
http://lists.cs.uiuc.edu/pipermail/llvmdev/2007-November/011573.html

seems to be specific to all gcc 3.x compilers (At least it looks so,
verified with 3.4.6 and 4.2 on x86-32/linux).

The workaround is:

diff --git a/include/llvm/CodeGen/MachineLoopInfo.h
b/include/llvm/CodeGen/MachineLoopInfo.h
index 1f6b1a1..ca05e01 100644
--- a/include/llvm/CodeGen/MachineLoopInfo.h
+++ b/include/llvm/CodeGen/MachineLoopInfo.h
@@ -71,9 +71,6 @@ bool LoopBase<MachineBasicBlock>::isLCSSAForm() const
{
   return false;
 }
 
-EXTERN_TEMPLATE_INSTANTIATION(class LoopBase<MachineBasicBlock>);
-EXTERN_TEMPLATE_INSTANTIATION(class LoopInfoBase<MachineBasicBlock>);
-
 typedef LoopBase<MachineBasicBlock> MachineLoop;
 
 class MachineLoopInfo : public MachineFunctionPass {


However, I don't think it's correct.

-- 
With best regards, Anton Korobeynikov.

Faculty of Mathematics & Mechanics, Saint Petersburg State University.



From bugzilla-daemon at cs.uiuc.edu  Thu Nov 29 13:37:05 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Thu, 29 Nov 2007 13:37:05 -0600
Subject: [LLVMbugs] [Bug 1834] New: GCC PCH support is broken when -gdwarf-2
	is used
Message-ID: <bug-1834-206@http.llvm.org/bugs/>

http://llvm.org/bugs/show_bug.cgi?id=1834

           Summary: GCC PCH support is broken when -gdwarf-2 is used
           Product: tools
           Version: trunk
          Platform: PC
        OS/Version: All
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: llvm-gcc
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: dpatel at apple.com
                CC: llvmbugs at cs.uiuc.edu


Here is simple test case that causes ICE.

--- a.h ---

int foo() {
  return 42;
}
--- a.c ---
#include "a.h"

int main() {
  return foo();
  return 0;
}
---
$ llvm-gcc -gdwarf-2 a.h
$ llvm-gcc -gdwarf-2 a.c -o a.o


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Thu Nov 29 19:07:57 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Thu, 29 Nov 2007 19:07:57 -0600
Subject: [LLVMbugs] [Bug 1834] GCC PCH support is broken when -gdwarf-2 is
	used
In-Reply-To: <bug-1834-206@http.llvm.org/bugs/>
Message-ID: <200711300107.lAU17vHW026833@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1834


Devang Patel <dpatel at apple.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #1 from Devang Patel <dpatel at apple.com>  2007-11-29 19:07:57 ---
Fixed.
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071126/055892.html
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071126/055893.html


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

From bugzilla-daemon at cs.uiuc.edu  Fri Nov 30 02:51:57 2007
From: bugzilla-daemon at cs.uiuc.edu (bugzilla-daemon at cs.uiuc.edu)
Date: Fri, 30 Nov 2007 02:51:57 -0600
Subject: [LLVMbugs] [Bug 1818] gvn should eliminate successive identical
	pure calls
In-Reply-To: <bug-1818-206@http.llvm.org/bugs/>
Message-ID: <200711300851.lAU8pvgw027916@zion.cs.uiuc.edu>

http://llvm.org/bugs/show_bug.cgi?id=1818


Owen Anderson <resistor at mac.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED




--- Comment #3 from Owen Anderson <resistor at mac.com>  2007-11-30 02:51:57 ---
Fixed now.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.