projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: e47d292) | patch
RegAllocFast: Make self loop live-out heuristic more aggressive
authorMatt Arsenault <Matthew.Arsenault@amd.com>
Mon, 31 Aug 2020 19:09:50 +0000 (15:09 -0400)
committerMatt Arsenault <Matthew.Arsenault@amd.com>
Wed, 16 Sep 2020 17:12:38 +0000 (13:12 -0400)
commit738c73a454881ca78214816754c1b82941d0cd26
treee51fc8c2aa6a579240a16d0953dbff4e85019544tree | snapshot
parente47d2927de79767663f0a0ece0581522fbe40ac4commit | diff
RegAllocFast: Make self loop live-out heuristic more aggressive

This currently has no impact on code, but prevents sizeable code size
regressions after D52010. This prevents spilling and reloading all
values inside blocks that loop back. Add a baseline test which would
regress without this patch.
llvm/lib/CodeGen/RegAllocFast.cpp diff | blob | history
llvm/test/CodeGen/AMDGPU/fastregalloc-self-loop-heuristic.mir [new file with mode: 0644] blob
Domain: System / Toolchain;