projects / oota-llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 3e00ff9) | patch
[X86][AVX] Fix lowering of X86ISD::VZEXT_MOVL for 128-bit -> 256-bit extension
authorSimon Pilgrim <llvm-dev@redking.me.uk>
Thu, 19 Nov 2015 12:18:37 +0000 (12:18 +0000)
committerSimon Pilgrim <llvm-dev@redking.me.uk>
Thu, 19 Nov 2015 12:18:37 +0000 (12:18 +0000)
commit175ea25f68a3eb84a917e004dd3b12352d3ae281
tree8c363951d6ce48fdbdeb70348756fcaf30987b52tree | snapshot
parent3e00ff9fdcaab9ecdfb36b84ab3ec25286ad2d52commit | diff
[X86][AVX] Fix lowering of X86ISD::VZEXT_MOVL for 128-bit -> 256-bit extension

The lowering patterns for X86ISD::VZEXT_MOVL for 128-bit to 256-bit vectors were just copying the lower xmm instead of actually masking off the first scalar using a blend.

Fix for PR25320.

Differential Revision: http://reviews.llvm.org/D14151

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@253561 91177308-0d34-0410-b5e6-96231b3b80d8
lib/Target/X86/X86InstrSSE.td diff | blob | history
test/CodeGen/X86/vec_extract-avx.ll diff | blob | history
C/C++ LLVM-based compilers that forbids OOTA behaviors