[llvm-commits] [llvm] r144267 - in /llvm/trunk: lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp test/CodeGen/X86/2006-05-11-InstrSched.ll test/CodeGen/X86/2009-04-21-NoReloadImpDef.ll test/CodeGen/X86/change-compare-stride-1.ll test/CodeGen/X86/fold-pcmpeqd-0.ll test/CodeGen/X86/iv-users-in-other-loops.ll test/CodeGen/X86/lsr-loop-exit-cond.ll test/CodeGen/X86/lsr-reuse-trunc.ll test/CodeGen/X86/masked-iv-safe.ll test/CodeGen/X86/multiple-loop-post-inc.ll test/CodeGen/X86/sse2.ll test/CodeGen/X86/sse3.ll

[Rafael Ávila de Espíndola]

On 11/10/2011 10:59 PM, Duncan Sands wrote:
> Hi Evan,
> 
>> That's really bizarre. It's obviously exposing a unrelated bug. Thanks for looking into this.
> 
> I think this is a case of PR11200: the code generators make decisions based on
> floating point computations.  On i386, GCC uses the floating point stack, while
> LLVM uses xmm registers, when building LLVM.  This results in codegen sometimes
> making different decisions due to different rounding.

Can't we just use -msse when building stage1 or do a bootstrap4?

> I've disabled the part of the dragonegg bootstrap that compares object files
> produced by GCC compiled LLVM with those produced by LLVM compiled LLVM as a
> workaround.
> 
> Ciao, Duncan.

Cheers,
Rafael