projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 4808914) | patch
[X86][AVX512] Change mask ops on vpermi2var tests to not use zeroinitializer.
authorSimon Pilgrim <llvm-dev@redking.me.uk>
Fri, 2 Nov 2018 19:39:41 +0000 (19:39 +0000)
committerSimon Pilgrim <llvm-dev@redking.me.uk>
Fri, 2 Nov 2018 19:39:41 +0000 (19:39 +0000)
commit88e8763baef3d6af562ce1549245878adc6bc84b
tree2864e014fcb46f45f25d910c70cb7be0a3f5f039tree | snapshot
parent4808914a5829ebdb19d0eb5ab93281169db8ad2dcommit | diff
[X86][AVX512] Change mask ops on vpermi2var tests to not use zeroinitializer.

This is necessary as I'm wanting to remove the 'Constant Pool' shuffle decoding from getTargetShuffleMask - but using getTargetShuffleMaskIndices allows the shuffle combiner to realize that these calls are really broadcasts.....

As with a lot of the X86ISD::VPERMV3 code this causes some vperm2i/vperm2t shuffles to flip depending on optimal commutation.

llvm-svn: 346032
llvm/test/CodeGen/X86/avx512vbmi-intrinsics.ll diff | blob | history
llvm/test/CodeGen/X86/avx512vbmivl-intrinsics.ll diff | blob | history
Domain: System / Toolchain;