projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 82090cb) | patch
[CostModel][X86] Add CostKinds handling for fmul ops
authorSimon Pilgrim <llvm-dev@redking.me.uk>
Sat, 3 Sep 2022 09:42:20 +0000 (10:42 +0100)
committerSimon Pilgrim <llvm-dev@redking.me.uk>
Sat, 3 Sep 2022 09:42:20 +0000 (10:42 +0100)
commit0735200e3f50de1cab4d2fff0ebff9aec52ff074
tree2996d625c986e27252a1a98073c0c733920c5792tree | snapshot
parent82090cb85eaa14ca2203aa9b234f132aa7a7a961commit | diff
[CostModel][X86] Add CostKinds handling for fmul ops

This was achieved with an updated version of the 'cost-tables vs llvm-mca' script D103695

As we're using 'typical' worst case values, not all cost entries come from a single CPU - e.g. the latency/throughput from haswell but the size-latency(uops) from zen1/alderlake-e due to 'double pumping'
llvm/lib/Target/X86/X86TargetTransformInfo.cpp diff | blob | history
llvm/test/Analysis/CostModel/X86/arith-fp-latency.ll diff | blob | history
llvm/test/Analysis/CostModel/X86/arith-fp-sizelatency.ll diff | blob | history
llvm/test/Analysis/CostModel/X86/intrinsic-cost-kinds.ll diff | blob | history
Domain: System / Toolchain;