projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 9586656) | patch
[x86] add AVX2 cost model entries for maxnum of 256-bit vectors
authorSanjay Patel <spatel@rotateright.com>
Mon, 2 Nov 2020 17:01:49 +0000 (12:01 -0500)
committerSanjay Patel <spatel@rotateright.com>
Mon, 2 Nov 2020 17:20:17 +0000 (12:20 -0500)
commit35fa3c474f4e6685da0d883bb6707ac5aa4d37cc
tree4db524e2b25ac44b85d8bb047f98d2e9e4d969a6tree | snapshot
parent9586656f4e3cf53a2755ad965773909b52728332commit | diff
[x86] add AVX2 cost model entries for maxnum of 256-bit vectors

As noticed in D90554 ,
the AVX2 costs for 256-bit vectors did not include FMAXNUM entries,
so we fell back to AVX1 which assumes those ops will be split into
128-bit halves or something close to that.

Differential Revision: https://reviews.llvm.org/D90613
llvm/lib/Target/X86/X86TargetTransformInfo.cpp diff | blob | history
llvm/test/Analysis/CostModel/X86/fmaxnum.ll diff | blob | history
llvm/test/Analysis/CostModel/X86/fminnum.ll diff | blob | history
Domain: System / Toolchain;