fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-08 17:48:10 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

nak/nir: Make interpolate_at_sample more efficient

Browse source 
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30218>
This commit is contained in:
Faith Ekstrand 2024-07-16 18:01:10 -05:00 • committed by Marge Bot
parent 24d5acf052
commit 1f430b1111
1 changed files with 24 additions and 16 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

40
src/nouveau/compiler/nak_nir_lower_fs_inputs.c
View file

@ -104,8 +104,8 @@ interp_fs_input(nir_builder *b, unsigned num_components, uint32_t addr,
}
static nir_def *
load_sample_pos_at(nir_builder *b, nir_def *sample_id,
const struct nak_fs_key *fs_key)
load_sample_pos_u4_at(nir_builder *b, nir_def *sample_id,
const struct nak_fs_key *fs_key)
{
nir_def *loc = nir_ldc_nv(b, 1, 8,
nir_imm_int(b, fs_key->sample_locations_cb),
@ -117,7 +117,14 @@ load_sample_pos_at(nir_builder *b, nir_def *sample_id,
loc = nir_u2u32(b, loc);
nir_def *loc_x_u4 = nir_iand_imm(b, loc, 0xf);
nir_def *loc_y_u4 = nir_iand_imm(b, nir_ushr_imm(b, loc, 4), 0xf);
nir_def *loc_u4 = nir_vec2(b, loc_x_u4, loc_y_u4);
return nir_vec2(b, loc_x_u4, loc_y_u4);
}
static nir_def *
load_sample_pos_at(nir_builder *b, nir_def *sample_id,
const struct nak_fs_key *fs_key)
{
nir_def *loc_u4 = load_sample_pos_u4_at(b, sample_id, fs_key);
nir_def *result = nir_fmul_imm(b, nir_i2f32(b, loc_u4), 1.0 / 16.0);
return result;
@ -127,26 +134,27 @@ static nir_def *
load_barycentric_offset(nir_builder *b, nir_intrinsic_instr *bary,
const struct nak_fs_key *fs_key)
{
nir_def *offset_f;
nir_def *offset_s12;
if (bary->intrinsic == nir_intrinsic_load_barycentric_coord_at_sample ||
bary->intrinsic == nir_intrinsic_load_barycentric_at_sample) {
nir_def *sample_id = bary->src[0].ssa;
nir_def *sample_pos = load_sample_pos_at(b, sample_id, fs_key);
offset_f = nir_fadd_imm(b, sample_pos, -0.5);
nir_def *offset_u4 = load_sample_pos_u4_at(b, sample_id, fs_key);
/* The sample position we loaded is a u4 from the upper-left and the
* sample position wanted by ipa.offset is s12
*/
offset_s12 = nir_iadd_imm(b, nir_ishl_imm(b, offset_u4, 8), -2048);
} else {
offset_f = bary->src[0].ssa;
nir_def *offset_f = bary->src[0].ssa;
offset_f = nir_fclamp(b, offset_f, nir_imm_float(b, -0.5),
nir_imm_float(b, 0.437500));
offset_s12 = nir_f2i32(b, nir_fmul_imm(b, offset_f, 4096.0));
}
offset_f = nir_fclamp(b, offset_f, nir_imm_float(b, -0.5),
nir_imm_float(b, 0.437500));
nir_def *offset_fixed =
nir_f2i32(b, nir_fmul_imm(b, offset_f, 4096.0));
nir_def *offset = nir_prmt_nv(b, nir_imm_int(b, 0x5410),
nir_channel(b, offset_fixed, 0),
nir_channel(b, offset_fixed, 1));
return offset;
return nir_prmt_nv(b, nir_imm_int(b, 0x5410),
nir_channel(b, offset_s12, 0),
nir_channel(b, offset_s12, 1));
}
struct lower_fs_input_ctx {