Module: Mesa
Branch: main
Commit: b1e59646de7a572f4f7b363c3c3cefe840bdacbd
URL:    
http://cgit.freedesktop.org/mesa/mesa/commit/?id=b1e59646de7a572f4f7b363c3c3cefe840bdacbd

Author: Rhys Perry <[email protected]>
Date:   Fri Dec  9 15:42:44 2022 +0000

aco/gfx11: increase vgpr_limit to 256

fossil-db (gfx1100):
Totals from 280 (0.21% of 134574) affected shaders:
MaxWaves: 3124 -> 2846 (-8.90%); split: +3.46%, -12.36%
Instrs: 1139038 -> 1091407 (-4.18%); split: -4.18%, +0.00%
CodeSize: 5809332 -> 5486812 (-5.55%); split: -5.55%, +0.00%
VGPRs: 35004 -> 42864 (+22.45%); split: -1.85%, +24.31%
SpillSGPRs: 1896 -> 1865 (-1.64%); split: -2.37%, +0.74%
SpillVGPRs: 17807 -> 2382 (-86.62%)
Scratch: 2573312 -> 736256 (-71.39%)
Latency: 27470485 -> 17981296 (-34.54%); split: -34.54%, +0.00%
InvThroughput: 5606102 -> 6527051 (+16.43%); split: -4.19%, +20.61%
VClause: 32319 -> 19927 (-38.34%); split: -39.13%, +0.78%
SClause: 15014 -> 14897 (-0.78%); split: -0.95%, +0.17%
Copies: 102977 -> 93511 (-9.19%); split: -9.93%, +0.74%
Branches: 15164 -> 14969 (-1.29%)
PreSGPRs: 19132 -> 19014 (-0.62%)
PreVGPRs: 30494 -> 37460 (+22.84%)

Signed-off-by: Rhys Perry <[email protected]>
Reviewed-by: Georg Lehmann <[email protected]>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20251>

---

 src/amd/compiler/aco_ir.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/amd/compiler/aco_ir.cpp b/src/amd/compiler/aco_ir.cpp
index eb1daf4d3d1..34c9ea535c8 100644
--- a/src/amd/compiler/aco_ir.cpp
+++ b/src/amd/compiler/aco_ir.cpp
@@ -105,7 +105,7 @@ init_program(Program* program, Stage stage, const struct 
aco_shader_info* info,
    /* apparently gfx702 also has 16-bank LDS but I can't find a family for 
that */
    program->dev.has_16bank_lds = family == CHIP_KABINI || family == 
CHIP_STONEY;
 
-   program->dev.vgpr_limit = gfx_level >= GFX11 ? 128 : 256; //TODO: fix 
encoding for 16-bit v128+
+   program->dev.vgpr_limit = 256;
    program->dev.physical_vgprs = 256;
    program->dev.vgpr_alloc_granule = 4;
 

Reply via email to