fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-01-10 03:50:13 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 2

radeonsi: set SHARED_VGPR_CNT for compute for ACO

Browse source 
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32910>
This commit is contained in:
Marek Olšák 2025-01-01 14:14:20 -05:00 • committed by Marge Bot
parent a962979baa
commit cbac2e4c75
2 changed files with 6 additions and 7 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

9
src/gallium/drivers/radeonsi/si_compute.c
View file

@ -201,10 +201,13 @@ static void si_create_compute_state_async(void *job, void *gdata, int thread_ind
: sel->info.uses_thread_id[1] ? 1 : 0) |
S_00B84C_LDS_SIZE(shader->config.lds_size);
/* COMPUTE_PGM_RSRC3 is only present on GFX10+ and GFX940+. */
shader->config.rsrc3 = S_00B8A0_SHARED_VGPR_CNT(shader->config.num_shared_vgprs / 8);
if (sscreen->info.gfx_level >= GFX12)
shader->config.rsrc3 = S_00B8A0_INST_PREF_SIZE_GFX12(si_get_shader_prefetch_size(shader));
shader->config.rsrc3 |= S_00B8A0_INST_PREF_SIZE_GFX12(si_get_shader_prefetch_size(shader));
else if (sscreen->info.gfx_level >= GFX11)
shader->config.rsrc3 = S_00B8A0_INST_PREF_SIZE_GFX11(si_get_shader_prefetch_size(shader));
shader->config.rsrc3 |= S_00B8A0_INST_PREF_SIZE_GFX11(si_get_shader_prefetch_size(shader));
simple_mtx_lock(&sscreen->shader_cache_mutex);
si_shader_cache_insert_shader(sscreen, ir_sha1_cache_key, shader, true);
@ -580,7 +583,7 @@ static bool si_switch_compute_shader(struct si_context *sctx, struct si_compute
sctx->compute_scratch_buffer->gpu_address >> 40);
}
if (sctx->gfx_level >= GFX11) {
if (sctx->gfx_level >= GFX10) {
radeon_opt_set_sh_reg(R_00B8A0_COMPUTE_PGM_RSRC3,
SI_TRACKED_COMPUTE_PGM_RSRC3, config->rsrc3);
}

4
src/gallium/drivers/radeonsi/si_state.c
View file

@ -4927,12 +4927,8 @@ static void si_init_compute_preamble_state(struct si_context *sctx,
};
ac_init_compute_preamble_state(&preamble_state, &pm4->base);
if (sctx->gfx_level == GFX10 || sctx->gfx_level == GFX10_3)
ac_pm4_set_reg(&pm4->base, R_00B8A0_COMPUTE_PGM_RSRC3, 0);
}
static void si_init_graphics_preamble_state(struct si_context *sctx,
struct si_pm4_state *pm4)
{