[llvm-commits] [llvm] r164281 - in /llvm/trunk: lib/Target/X86/X86ISelLowering.cpp lib/Target/X86/X86ISelLowering.h lib/Target/X86/X86InstrCompiler.td lib/Target/X86/X86InstrInfo.cpp lib/Target/X86/X86InstrInfo.h test/CodeGen/X86/2010-01-08-Atomic64Bug.ll test/CodeGen/X86/atomic16.ll test/CodeGen/X86/atomic32.ll test/CodeGen/X86/atomic64.ll test/CodeGen/X86/atomic6432.ll test/CodeGen/X86/atomic8.ll test/CodeGen/X86/atomic_op.ll test/CodeGen/X86/pr13458.ll

[Cameron Zwarich]

I'm on vacation for 2 weeks. I attached a patch to the bug, but I didn't doublecheck every codegen difference or run execution tests. The register allocator creates a few more subregister copies than before; it might be possible to get rid of them by changing the patch a bit.

Cameron

On Mar 3, 2013, at 2:34 PM, Evan Cheng <evan.cheng at apple.com> wrote:

> Any update on this. Michael, are you working on fixing this? It looks like a potentially significant problem.
> 
> Thanks,
> 
> Evan
> 
> On Feb 25, 2013, at 12:56 PM, Cameron Zwarich <zwarich at apple.com> wrote:
> 
>> I filed a bug: http://llvm.org/bugs/show_bug.cgi?id=15355
>> 
>> Cameron
>> 
>> On Feb 25, 2013, at 10:57 AM, Michael Liao <michael.liao at intel.com> wrote:
>> 
>>> Could you elaborate more on this? Do you have bug reported? - michael
>>> 
>>> On Sun, 2013-02-24 at 15:24 -0800, Cameron Zwarich wrote:
>>>> I guess it is a bit late to say this now, but this commit has a lot of problems. It sticks store memory operands onto loads, copies kill flags from one use to multiple uses, and uses a physical register across basic blocks prior to register allocation. I have a patch ready for the first two, and I'll probably fix the last one and commit it.
>>>> 
>>>> Cameron
>> _______________________________________________
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>