[PATCH] X86: Allow pextr to use a memory destination

[Jim Grosbach]

Hi Adam,

This look good to me. Applied as r200957.

Thanks,
-Jim

On Feb 4, 2014, at 1:56 PM, Adam Nemet <anemet at apple.com> wrote:

> Hi,
> 
> This patch (my first to llvm :) is an attempt to resolve the following FIXME:
> 
> -// FIXME:
> -// There's an AssertZext in the way of writing the store pattern
> -// (store (i8 (trunc (X86pextrb (v16i8 VR128:$src1), imm:$src2))), addr:$dst)
> 
> My approach is to introduce the assertzext operation to .td and then include that in the pattern.  Conveniently, assertzext is already defined in AArch64.td so I just expose that to all targets.
> 
> The test widen_load-2.ll requires a slight change because with the stores gone, the remaining instructions are scheduled in a different order.
> 
> I’ve tested with llvm/test and llvm-test on x86_64.
> 
> Adam
> 
> <pextr-2.patch>_______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits