nir/opt_load_store_vectorize: vectorize load_smem_amd

radeonsi+ACO with the new vectorization callback:

TOTALS FROM AFFECTED SHADERS (19508/58918)
  VGPRs: 708672 -> 708864 (0.03 %)
  Code Size: 31458688 -> 31217160 (-0.77 %) bytes
  Max Waves: 305960 -> 305952 (-0.00 %)

Reviewed-by: Qiang Yu <yuq825@gmail.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29398>
This commit is contained in:
Marek Olšák 2024-05-24 23:40:43 -04:00 committed by Marge Bot
parent a44e5cfccf
commit 0727634443

View file

@ -111,6 +111,7 @@ get_info(nir_intrinsic_op op)
STORE(nir_var_uniform, const_ir3, -1, -1, -1, 0, 4)
INFO(nir_var_mem_shared, shared_append_amd, true, -1, -1, -1, -1, 1)
INFO(nir_var_mem_shared, shared_consume_amd, true, -1, -1, -1, -1, 1)
LOAD(nir_var_mem_global, smem_amd, 0, 1, -1, 1)
default:
break;
#undef ATOMIC