fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-02-23 05:20:30 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 2

nak/nir: Use an indirect load for sample locations

Browse source 
A single ldc is probably more efficient than a 64-bit load and the pile
of math we were generating before.  The only reason for the old method
was that it let us avoid indirect cbuf loads because we didn't support
them for a while.  Now that we can support all cbuf loads, we can just
do an indirect 1B load and call it good.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30218>
This commit is contained in:
Faith Ekstrand 2024-07-16 15:53:34 -05:00 • committed by Marge Bot
parent f673e2bf68
commit cc33cafcac
1 changed files with 5 additions and 4 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

9
src/nouveau/compiler/nak_nir_lower_fs_inputs.c
View file

@ -107,13 +107,14 @@ static nir_def *
load_sample_pos_at(nir_builder *b, nir_def *sample_id,
const struct nak_fs_key *fs_key)
{
nir_def *loc = nir_ldc_nv(b, 1, 64,
nir_def *loc = nir_ldc_nv(b, 1, 8,
nir_imm_int(b, fs_key->sample_locations_cb),
nir_imm_int(b, fs_key->sample_locations_offset),
nir_iadd_imm(b, sample_id,
fs_key->sample_locations_offset),
.align_mul = 8, .align_offset = 0);
/* Yay little endian */
loc = nir_ushr(b, loc, nir_imul_imm(b, sample_id, 8));
/* The rest of these calculations are in 32-bit */
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);