fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2025-12-20 05:10:11 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 2

radeonsi: double pixel throughput in certain cases of PS without inputs

Browse source 
When no barycentric VGPRs are needed, we always enabled one of the pairs
(e.g. PERSP_SAMPLE_ENA) because it's a HW requirement. However,
the requirement says that LINE_STIPPLE_TEX_ENA can be enabled instead,
which occupies only 1 VGPR.

To get maximum pixel throughput, we can only have 2 initialized VGPRs
at most. By reducing initialized VGPRs from 2 (with PERSP_SAMPLE_ENA) to 1
(with LINE_STIPPLE_TEX_ENA), we can have 1 additional initialized VGPR
for free with maximum pixel throughput, such as POS_FIXED_PT for
frag_coord.xy without MSAA.

Only ACO gets this perf improvement because the change would be more
complicated with LLVM.

Reviewed-by: Qiang Yu <yuq825@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38915>
This commit is contained in:
Marek Olšák 2025-12-11 16:09:08 -05:00 • committed by Marge Bot
parent 6592a18cd7
commit 5acabdd1f8
3 changed files with 13 additions and 4 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
src/gallium/drivers/radeonsi/si_shader_args.c
View file

@ -671,7 +671,7 @@ void si_get_ps_prolog_args(struct si_shader_args *args,
ac_add_arg(&args->ac, AC_ARG_VGPR, 2, AC_ARG_VALUE, &args->ac.linear_sample); ac_add_arg(&args->ac, AC_ARG_VGPR, 2, AC_ARG_VALUE, &args->ac.linear_sample);
ac_add_arg(&args->ac, AC_ARG_VGPR, 2, AC_ARG_VALUE, &args->ac.linear_center); ac_add_arg(&args->ac, AC_ARG_VGPR, 2, AC_ARG_VALUE, &args->ac.linear_center);
ac_add_arg(&args->ac, AC_ARG_VGPR, 2, AC_ARG_VALUE, &args->ac.linear_centroid); ac_add_arg(&args->ac, AC_ARG_VGPR, 2, AC_ARG_VALUE, &args->ac.linear_centroid);
/* skip LINE_STIPPLE_TEX */ ac_add_arg(&args->ac, AC_ARG_VGPR, 1, AC_ARG_VALUE, NULL); /* LINE_STIPPLE_TEX */
/* POS_X|Y|Z|W_FLOAT */ /* POS_X|Y|Z|W_FLOAT */
u_foreach_bit(i, key->ps_prolog.fragcoord_usage_mask) u_foreach_bit(i, key->ps_prolog.fragcoord_usage_mask)

1
src/gallium/drivers/radeonsi/si_shader_internal.h
View file

@ -16,6 +16,7 @@
S_0286D0_LINEAR_SAMPLE_ENA(1) | \ S_0286D0_LINEAR_SAMPLE_ENA(1) | \
S_0286D0_LINEAR_CENTER_ENA(1) | \ S_0286D0_LINEAR_CENTER_ENA(1) | \
S_0286D0_LINEAR_CENTROID_ENA(1) | \ S_0286D0_LINEAR_CENTROID_ENA(1) | \
S_0286D0_LINE_STIPPLE_TEX_ENA(1) | \
S_0286D0_FRONT_FACE_ENA(1) | \ S_0286D0_FRONT_FACE_ENA(1) | \
S_0286D0_ANCILLARY_ENA(1) | \ S_0286D0_ANCILLARY_ENA(1) | \
S_0286D0_SAMPLE_COVERAGE_ENA(1) | \ S_0286D0_SAMPLE_COVERAGE_ENA(1) | \

14
src/gallium/drivers/radeonsi/si_shader_variant_info.c
View file

@ -423,7 +423,15 @@ void si_fixup_spi_ps_input_config(struct si_shader *shader)
shader->config.spi_ps_input_ena |= S_0286CC_PERSP_SAMPLE_ENA(1); shader->config.spi_ps_input_ena |= S_0286CC_PERSP_SAMPLE_ENA(1);
} }
/* At least one pair of interpolation weights must be enabled. */ /* At least one pair of barycentric coordinates or LINE_STIPPLE_TEX_ENA must be enabled.
if (!(shader->config.spi_ps_input_ena & 0x7f)) * Since LINE_STIPPLE_TEX_ENA is the only one that loads only 1 VGPR, use it.
shader->config.spi_ps_input_ena |= S_0286CC_PERSP_SAMPLE_ENA(1); */
if (!(shader->config.spi_ps_input_ena & 0x7f) &&
!G_0286CC_LINE_STIPPLE_TEX_ENA(shader->config.spi_ps_input_ena)) {
/* LLVM sets PERSP_SAMPLE_ENA in this case, so we have to do the same. */
if (shader->selector->info.base.use_aco_amd)
shader->config.spi_ps_input_ena |= S_0286CC_LINE_STIPPLE_TEX_ENA(1);
else
shader->config.spi_ps_input_ena |= S_0286CC_PERSP_SAMPLE_ENA(1);
}
} }