KonstantinSeurer
/
mesa
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

radv: don't lower_pack() after load-store-vectorization 

Totals from 7 (0.01% of 139391) affected shaders:
CodeSize: 282900 -> 283324 (+0.15%); split: -0.01%, +0.16%
Instrs: 45287 -> 45338 (+0.11%); split: -0.01%, +0.12%
Cycles: 11496332 -> 11510396 (+0.12%); split: -0.00%, +0.12%
VMEM: 2355 -> 2335 (-0.85%)
Copies: 15506 -> 15561 (+0.35%)

A bit of noise in some parallel-rdp shaders.

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8425>
This commit is contained in:
Daniel Schürmann 2021-01-08 22:05:26 +01:00 committed by Marge Bot
parent 987a0e6a67
commit e92bd57008
1 changed files with 1 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
src/amd/vulkan/radv_pipeline.c
View File

@ -3291,7 +3291,7 @@ VkResult radv_create_shaders(struct radv_pipeline *pipeline,
NIR_PASS_V(nir[i], nir_lower_memory_model);
bool lower_to_scalar = false;
bool lower_pack = false;
nir_load_store_vectorize_options vectorize_opts = {
.modes = nir_var_mem_ssbo | nir_var_mem_ubo |
nir_var_mem_push_const | nir_var_mem_shared |
@ -3309,7 +3309,6 @@ VkResult radv_create_shaders(struct radv_pipeline *pipeline,
if (nir_opt_load_store_vectorize(nir[i], &vectorize_opts)) {
lower_to_scalar = true;
lower_pack = true;
}
/* do this again since information such as outputs_read can be out-of-date */
@ -3321,8 +3320,6 @@ VkResult radv_create_shaders(struct radv_pipeline *pipeline,
if (lower_to_scalar)
nir_lower_alu_to_scalar(nir[i], NULL, NULL);
if (lower_pack)
nir_lower_pack(nir[i]);
/* lower ALU operations */
/* TODO: Some 64-bit tests crash inside LLVM. */