projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: e16b1e2) | patch
[X86][SSE] Add SSE41/AVX1 costs for vector shifts.
authorSimon Pilgrim <llvm-dev@redking.me.uk>
Sun, 23 Oct 2016 16:49:04 +0000 (16:49 +0000)
committerSimon Pilgrim <llvm-dev@redking.me.uk>
Sun, 23 Oct 2016 16:49:04 +0000 (16:49 +0000)
commit6ac1e98b09ab38adde003620cc77172196b2ef62
tree572fa3020b26d715e1af2581d58a9005d60d03cftree | snapshot
parente16b1e2271b955764ed44211160e1a213e45fa6ecommit | diff
[X86][SSE] Add SSE41/AVX1 costs for vector shifts.

We were defaulting to SSE2 costs which weren't taking into account the availability of PBLENDW/PBLENDVB to improve merging of per-element shift results.

llvm-svn: 284939
llvm/lib/Target/X86/X86TargetTransformInfo.cpp diff | blob | history
llvm/test/Analysis/CostModel/X86/vshift-ashr-cost.ll diff | blob | history
llvm/test/Analysis/CostModel/X86/vshift-lshr-cost.ll diff | blob | history
llvm/test/Analysis/CostModel/X86/vshift-shl-cost.ll diff | blob | history
Domain: System / Toolchain;