On AMDGPU, alloca instructions have penalty that can
be avoided when SROA is applied after inlining.
This patch introduces the default implementation of
TargetTransformInfo::getCallerAllocaCost.
Differential D149740
[InlineCost][TargetTransformInfo][AMDGPU] Consider cost of alloca instructions in the caller (1/2) jmmartinez on May 3 2023, 5:22 AM. Authored by
Details On AMDGPU, alloca instructions have penalty that can This patch introduces the default implementation of
Diff Detail
Event TimelineComment Actions I'm missing the AMDGPU bit of this patch, or maybe I'm not understanding something? Is it possible to add a small unit test, maybe in unittests/Analysis/InlineCostTest.cpp? Comment Actions Sorry, I put it on the patch stack: https://reviews.llvm.org/D149741
Will do! Thanks! Comment Actions
Comment Actions
|
nit: can you do this in 2 steps, easier to read. same in onInitializeSROAArg. thanks