[PATCH] D69654: AMDGPU: Switch backend default max workgroup size to 1024
Stanislav Mekhanoshin via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Tue Nov 12 12:19:47 PST 2019
rampitec accepted this revision.
rampitec added a comment.
This revision is now accepted and ready to land.
LGTM
================
Comment at: llvm/test/CodeGen/AMDGPU/occupancy-levels.ll:265
; GCN-LABEL: {{^}}used_lds_6556:
-; GFX9: ; Occupancy: 9
-; GFX1010W64: ; Occupancy: 19
+; GFX9: ; Occupancy: 10
+; GFX1010W64: ; Occupancy: 20
----------------
rampitec wrote:
> This needs to be investigated first I believe. There must be some wrong logic somewhere with LDS accounting for occupancy.
I have recollected the logic behind LDS sharing and occupancy calculations. The increase seems to be right because LDS allocation is per WG and with a bigger WG you have less of them.
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D69654/new/
https://reviews.llvm.org/D69654
More information about the llvm-commits
mailing list