projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 5c1b8de) | patch
[AMDGPU] Do not apply schedule metric for regions with spilling
authorStanislav Mekhanoshin <Stanislav.Mekhanoshin@amd.com>
Mon, 13 Feb 2023 19:36:12 +0000 (11:36 -0800)
committerStanislav Mekhanoshin <Stanislav.Mekhanoshin@amd.com>
Tue, 14 Feb 2023 20:16:46 +0000 (12:16 -0800)
commit12b4f9e2af957bd248ee3a39143fbef7a483fd35
treea62471d4a19f195b65188f0cfb6737f5e745bac9tree | snapshot
parent5c1b8de77d1c6abc384e376a664a898608b9b470commit | diff
[AMDGPU] Do not apply schedule metric for regions with spilling

D139710 has added a metric to increase schedule's ILP while
staying within the same occupancy. Do not bother to apply this
metric to a region which is known to have spilling, it may result
in spilling to reappear after the previous stage and will do no
good if we already spilling anyway. It may also reduce compile
time a bit for such regions.

Fixes: SWDEV-377300
Differential Revision: https://reviews.llvm.org/D143934
llvm/lib/Target/AMDGPU/GCNSchedStrategy.cpp diff | blob | history
llvm/test/CodeGen/AMDGPU/schedule-regpressure-ilp-metric-spills.mir [new file with mode: 0644] blob
Domain: System / Toolchain;