[llvm-commits] [llvm] r57257 - /llvm/trunk/lib/Target/X86/X86InstrInfo.td
Evan Cheng
evan.cheng at apple.com
Wed Oct 8 02:49:16 PDT 2008
Yes, Owen is working on fixing (or rather has fixed) local RA.
Evan
On Oct 7, 2008, at 11:34 PM, Chris Lattner wrote:
> On Oct 7, 2008, at 2:29 PM, Evan Cheng wrote:
>>>>> We could fix it, but on the other hand it's arguably sloppy for
>>>>> instruction selectors to not precisely describe their physical
>>>>> register uses.
>>>>
>>>> I agree. In this case DVI8r should say it defines AL, AH and uses
>>>> AL,
>>>> AH which makes it clear it reads 2 values and outputs 2.
>>>
>>> No, it reads 1 value and outputs two. It is not the same as DIV16
>>> and
>>> DIV32.
>>> The way I have it matches Intel docs.
>>
>> Ok. It's good to match what Intel manual says. But that's additional
>> goodness. Previously DIV16r is marked as using AL, AH and the
>> implicit
>> use is copied to AX. While this is not ideal, it's not incorrect.
>> Local liveness cannot mark AX as dead.
>
> To put it another way, it looks like (as Dale said) both changes are
> useful. The change to make div more closely follow the docs is useful
> on its own merits, but the local RA should still be fixed.
>
> -Chris
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
More information about the llvm-commits
mailing list