[llvm-commits] [PATCH] Fixes for ARM vst and vld instructions

[Jim Grosbach]

On Sep 3, 2012, at 8:54 AM, Tim Northover <t.p.northover at gmail.com> wrote:

> Hi all,
> 
>> Please review and commit these patches, as I don’t have commit access.
> 
> Just to say that I've run these patches through the MC Hammer
> test-suite we've got (Richard Barton reported the first of these
> issues as http://llvm.org/bugs/show_bug.cgi?id=13038) and I believe
> they fix all the encode/decode cycle problems found on lane-specifying
> NEON loads/stores: "vld1.32 {d[1]}, [r0]" etc.
> 
> I'm a little uncomfortable doing our own official review, but can
> certainly do any committing once someone gives the nod.

LGTM in general, and resolving the MC Hammer failures is a really good sign.

Small tweak I'd like is for the test cases to use register values other than zero (r0, d0). The encoding will default to zero even if we totally screw up mapping the register operand into the instruction encoding, so I prefer to use pretty much any other register. Many of the existing test cases are absolutely horrible examples in that regard. :(

Thanks!

-jim

> 
> Regards.
> 
> Tim.
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits