projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: be88ba0) | patch
[AMDGPU] Promote alloca to vector in opt
authorStanislav Mekhanoshin <Stanislav.Mekhanoshin@amd.com>
Wed, 20 May 2020 23:24:06 +0000 (16:24 -0700)
committerStanislav Mekhanoshin <Stanislav.Mekhanoshin@amd.com>
Thu, 21 May 2020 20:49:51 +0000 (13:49 -0700)
commit689e616ed0b1210517dda65dc3528f8fdb9a5959
tree90a6a648ded0c5a7a58a660e31290561d9ad5210tree | snapshot
parentbe88ba09d59112b2df25a35f906e55026fa93128commit | diff
[AMDGPU] Promote alloca to vector in opt

Promote alloca to vector before SROA and loop unroll. If we manage
to eliminate allocas before unroll we may choose to unroll less.

Differential Revision: https://reviews.llvm.org/D80386
llvm/lib/Target/AMDGPU/AMDGPU.h diff | blob | history
llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp diff | blob | history
llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp diff | blob | history
llvm/lib/Target/AMDGPU/sroa-before-unroll.ll [new file with mode: 0644] blob
llvm/test/CodeGen/AMDGPU/opt-pipeline.ll diff | blob | history
Domain: System / Toolchain;