aco: consider that GFX10.3 allocates LDS in 1024 byte blocks
authorRhys Perry <pendingchaos02@gmail.com>
Fri, 11 Dec 2020 17:33:03 +0000 (17:33 +0000)
committerRhys Perry <pendingchaos02@gmail.com>
Mon, 15 Feb 2021 13:35:38 +0000 (13:35 +0000)
commitb759557caccd28ae811830ef22aab16391c19dba
tree639527b1db251c1edf2e922e2dc8aba9db553f56
parent7ff805a19d9df606f28a408364c208d79170d273
aco: consider that GFX10.3 allocates LDS in 1024 byte blocks

fossil-db (GFX10.3):
Totals from 3 (0.00% of 139391) affected shaders:
VMEM: 513 -> 511 (-0.39%)
SMEM: 94 -> 92 (-2.13%)
VClause: 31 -> 30 (-3.23%)

fossil-db (GFX10.3, wave32):
Totals from 4 (0.00% of 139391) affected shaders:
VClause: 82 -> 81 (-1.22%)

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8761>
src/amd/compiler/aco_instruction_selection_setup.cpp
src/amd/compiler/aco_ir.cpp
src/amd/compiler/aco_ir.h
src/amd/compiler/aco_live_var_analysis.cpp