summaryrefslogtreecommitdiffstats
path: root/src/gallium/drivers/radeonsi/si_state.c
diff options
context:
space:
mode:
authorAndres Rodriguez <[email protected]>2017-10-20 18:02:13 -0400
committerBas Nieuwenhuizen <[email protected]>2017-10-21 01:01:44 +0200
commit557de3b9ae2e0a1889442bb48f731d69c993a027 (patch)
tree56268752dfaf301a4a8447a81a9c1f0f604e023a /src/gallium/drivers/radeonsi/si_state.c
parent986c4b0bd4eb5115d656ccbc0351bd824ff7e413 (diff)
radeonsi: hardcode shader WAVE_LIMIT to the maximum value
This is part of a cooperative scheduling approach used by radv. All drivers in the stack must opt-in to resource arbitration, otherwise GL based apps will be able to ignore system priorities. We always hardcode the field to its maximum value, instead of attempting to calculate an approximate usage. In testing, there were no benefits to using anything other than the maximum. Signed-off-by: Andres Rodriguez <[email protected]> Reviewed-by: Bas Nieuwenhuizen <[email protected]>
Diffstat (limited to 'src/gallium/drivers/radeonsi/si_state.c')
-rw-r--r--src/gallium/drivers/radeonsi/si_state.c21
1 files changed, 14 insertions, 7 deletions
diff --git a/src/gallium/drivers/radeonsi/si_state.c b/src/gallium/drivers/radeonsi/si_state.c
index ae45e1a6b26..8e3717e97f4 100644
--- a/src/gallium/drivers/radeonsi/si_state.c
+++ b/src/gallium/drivers/radeonsi/si_state.c
@@ -4994,11 +4994,15 @@ static void si_init_config(struct si_context *sctx)
if (sctx->b.chip_class >= CIK) {
if (sctx->b.chip_class >= GFX9) {
- si_pm4_set_reg(pm4, R_00B41C_SPI_SHADER_PGM_RSRC3_HS, S_00B41C_CU_EN(0xffff));
+ si_pm4_set_reg(pm4, R_00B41C_SPI_SHADER_PGM_RSRC3_HS,
+ S_00B41C_CU_EN(0xffff) | S_00B41C_WAVE_LIMIT(0x3F));
} else {
- si_pm4_set_reg(pm4, R_00B51C_SPI_SHADER_PGM_RSRC3_LS, S_00B51C_CU_EN(0xffff));
- si_pm4_set_reg(pm4, R_00B41C_SPI_SHADER_PGM_RSRC3_HS, 0);
- si_pm4_set_reg(pm4, R_00B31C_SPI_SHADER_PGM_RSRC3_ES, S_00B31C_CU_EN(0xffff));
+ si_pm4_set_reg(pm4, R_00B51C_SPI_SHADER_PGM_RSRC3_LS,
+ S_00B51C_CU_EN(0xffff) | S_00B51C_WAVE_LIMIT(0x3F));
+ si_pm4_set_reg(pm4, R_00B41C_SPI_SHADER_PGM_RSRC3_HS,
+ S_00B41C_WAVE_LIMIT(0x3F));
+ si_pm4_set_reg(pm4, R_00B31C_SPI_SHADER_PGM_RSRC3_ES,
+ S_00B31C_CU_EN(0xffff) | S_00B31C_WAVE_LIMIT(0x3F));
/* If this is 0, Bonaire can hang even if GS isn't being used.
* Other chips are unaffected. These are suboptimal values,
@@ -5008,7 +5012,8 @@ static void si_init_config(struct si_context *sctx)
S_028A44_ES_VERTS_PER_SUBGRP(64) |
S_028A44_GS_PRIMS_PER_SUBGRP(4));
}
- si_pm4_set_reg(pm4, R_00B21C_SPI_SHADER_PGM_RSRC3_GS, S_00B21C_CU_EN(0xffff));
+ si_pm4_set_reg(pm4, R_00B21C_SPI_SHADER_PGM_RSRC3_GS,
+ S_00B21C_CU_EN(0xffff) | S_00B21C_WAVE_LIMIT(0x3F));
/* Compute LATE_ALLOC_VS.LIMIT. */
unsigned num_cu_per_sh = sscreen->b.info.num_good_compute_units /
@@ -5040,10 +5045,12 @@ static void si_init_config(struct si_context *sctx)
/* VS can't execute on one CU if the limit is > 2. */
si_pm4_set_reg(pm4, R_00B118_SPI_SHADER_PGM_RSRC3_VS,
- S_00B118_CU_EN(late_alloc_limit > 2 ? 0xfffe : 0xffff));
+ S_00B118_CU_EN(late_alloc_limit > 2 ? 0xfffe : 0xffff) |
+ S_00B118_WAVE_LIMIT(0x3F));
si_pm4_set_reg(pm4, R_00B11C_SPI_SHADER_LATE_ALLOC_VS,
S_00B11C_LIMIT(late_alloc_limit));
- si_pm4_set_reg(pm4, R_00B01C_SPI_SHADER_PGM_RSRC3_PS, S_00B01C_CU_EN(0xffff));
+ si_pm4_set_reg(pm4, R_00B01C_SPI_SHADER_PGM_RSRC3_PS,
+ S_00B01C_CU_EN(0xffff) | S_00B01C_WAVE_LIMIT(0x3F));
}
if (sctx->b.chip_class >= VI) {