R600/SI: Prettier display of input modifiers and fold of fabs

[Vincent Lejeune]

I updated the patch set against newer llvm revision, no regression from piglit.




> Le Mardi 22 avril 2014 3h44, "Daenzer, Michel" <Michel.Daenzer at amd.com> a écrit :
> 
>>  +; SI-CHECK-LABEL: @fneg_fabs_fold
>>  +
>>  +define void @fneg_fabs_fold(float addrspace(1)* %out, float %in, float 
> %in2) {
> 
> This should also check that the modifiers are used as expected.
> 

Actually this was a left over test, I initially expected that folding abs and neg operands would happens automatically because the function that does the folding can be called several time but it turned out not being so easy.
As fneg(fabs) is not a common pattern I though it's ok if it's not folded at the moment.

Vincent.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-R600-SI-Use-pseudo-instruction-for-fabs-clamp-fneg.patch
Type: text/x-patch
Size: 4085 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20140505/2086b36f/attachment.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0002-R600-SI-Prettier-display-of-input-modifiers.patch
Type: text/x-patch
Size: 15076 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20140505/2086b36f/attachment-0001.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0003-R600-SI-Fold-fabs-fneg-into-src-input-modifier.patch
Type: text/x-patch
Size: 3793 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20140505/2086b36f/attachment-0002.bin>