NeroReflex/mesa
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

radv: add barycentric_at_sample lowering when the number of samples is dynamic

Browse source 
Use two different paths (static vs dynamic) to avoid running more NIR
pass to remove dead CF code when static is used.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18677>
This commit is contained in:
Samuel Pitoiset 2022-09-19 15:08:52 +02:00 • committed by Marge Bot
parent 68bb58a46e
commit deb2dccc75
2 changed files with 41 additions and 14 deletions
Download patch file Download diff file Expand all files Collapse all files

7
src/amd/vulkan/radv_nir_lower_abi.c
View file

@ -263,8 +263,11 @@ lower_abi_instr(nir_builder *b, nir_instr *instr, void *state)
nir_ssa_def *offset = nir_ishl_imm(b, sample_id, 3); /* 2 floats containing samplepos.xy */
nir_const_value *const_num_samples = nir_src_as_const_value(intrin->src[1]);
assert(const_num_samples);
sample_pos_offset += (const_num_samples->u32 << 3);
if (const_num_samples) {
sample_pos_offset += (const_num_samples->u32 << 3);
} else {
offset = nir_iadd(b, offset, nir_ishl_imm(b, intrin->src[1].ssa, 3));
}
replacement = nir_load_global_amd(b, 2, 32, addr, offset,
.base = sample_pos_offset, .access = ACCESS_NON_WRITEABLE);

48
src/amd/vulkan/radv_shader.c
View file

@ -599,20 +599,44 @@ radv_lower_fs_intrinsics(nir_shader *nir, const struct radv_pipeline_stage *fs_s
nir_ssa_def *num_samples = nir_load_rasterization_samples_amd(&b);
nir_ssa_def *new_dest;
if (!key->ps.num_samples) {
new_dest =
nir_load_barycentric_pixel(&b, 32,
.interp_mode = nir_intrinsic_interp_mode(intrin));
if (key->dynamic_rasterization_samples) {
nir_ssa_def *res1, *res2;
nir_push_if(&b, nir_ieq_imm(&b, num_samples, 1));
{
res1 = nir_load_barycentric_pixel(&b, 32,
.interp_mode = nir_intrinsic_interp_mode(intrin));
}
nir_push_else(&b, NULL);
{
nir_ssa_def *sample_pos =
nir_load_sample_positions_amd(&b, 32, intrin->src[0].ssa, num_samples);
/* sample_pos -= 0.5 */
sample_pos = nir_fsub(&b, sample_pos, nir_imm_float(&b, 0.5f));
res2 = nir_load_barycentric_at_offset(&b, 32, sample_pos,
.interp_mode = nir_intrinsic_interp_mode(intrin));
}
nir_pop_if(&b, NULL);
new_dest = nir_if_phi(&b, res1, res2);
} else {
nir_ssa_def *sample_pos =
nir_load_sample_positions_amd(&b, 32, intrin->src[0].ssa, num_samples);
if (!key->ps.num_samples) {
new_dest =
nir_load_barycentric_pixel(&b, 32,
.interp_mode = nir_intrinsic_interp_mode(intrin));
} else {
nir_ssa_def *sample_pos =
nir_load_sample_positions_amd(&b, 32, intrin->src[0].ssa, num_samples);
/* sample_pos -= 0.5 */
sample_pos = nir_fsub(&b, sample_pos, nir_imm_float(&b, 0.5f));
/* sample_pos -= 0.5 */
sample_pos = nir_fsub(&b, sample_pos, nir_imm_float(&b, 0.5f));
new_dest =
nir_load_barycentric_at_offset(&b, 32, sample_pos,
.interp_mode = nir_intrinsic_interp_mode(intrin));
new_dest =
nir_load_barycentric_at_offset(&b, 32, sample_pos,
.interp_mode = nir_intrinsic_interp_mode(intrin));
}
}
nir_ssa_def_rewrite_uses(&intrin->dest.ssa, new_dest);
@ -628,7 +652,7 @@ radv_lower_fs_intrinsics(nir_shader *nir, const struct radv_pipeline_stage *fs_s
}
if (progress)
nir_metadata_preserve(impl, nir_metadata_block_index | nir_metadata_dominance);
nir_metadata_preserve(impl, 0);
else
nir_metadata_preserve(impl, nir_metadata_all);