fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-06 11:38:05 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

ac/nir/ngg: Fix scalarized mesh primitive indices
Some checks are pending
macOS-CI / macOS-CI (dri) (push) Waiting to run
Details
macOS-CI / macOS-CI (xlib) (push) Waiting to run
Details

Browse source 
Take the write_mask into account when storing primitive indices,
otherwise they will end up being stored in the wrong place.

Fixes: 8e24d3426d ("ac/nir/ngg: Refactor MS primitive indices for scalarized IO.")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37610>
This commit is contained in:
Timur Kristóf 2025-09-24 19:43:23 +02:00 • committed by Marge Bot
parent 3dc9c1a91e
commit d3579190d6
1 changed files with 8 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

9
src/amd/common/nir/ac_nir_lower_ngg_mesh.c
View file

@ -136,15 +136,20 @@ ms_store_prim_indices(nir_builder *b,
assert(nir_src_is_const(*nir_get_io_offset_src(intrin)));
assert(nir_src_as_uint(*nir_get_io_offset_src(intrin)) == 0);
const unsigned write_mask = nir_intrinsic_write_mask(intrin);
const unsigned component_offset = nir_intrinsic_component(intrin);
nir_def *store_val = intrin->src[0].ssa;
assert(store_val->num_components <= 3);
assert(write_mask && write_mask <= BITFIELD_MASK(s->vertices_per_prim));
if (store_val->num_components > s->vertices_per_prim)
store_val = nir_trim_vector(b, store_val, s->vertices_per_prim);
if (s->layout.var.prm_attr.mask & VARYING_BIT_PRIMITIVE_INDICES) {
for (unsigned c = 0; c < store_val->num_components; ++c) {
if (!(write_mask & BITFIELD_BIT(c)))
continue;
const unsigned i = VARYING_SLOT_PRIMITIVE_INDICES * 4 + c + component_offset;
nir_store_var(b, s->out_variables[i], nir_channel(b, store_val, c), 0x1);
}
@ -157,7 +162,9 @@ ms_store_prim_indices(nir_builder *b,
/* The max vertex count is 256, so these indices always fit 8 bits.
* To reduce LDS use, store these as a flat array of 8-bit values.
*/
nir_store_shared(b, nir_u2u8(b, store_val), offset, .base = s->layout.lds.indices_addr + component_offset);
nir_store_shared(b, nir_u2u8(b, store_val), offset,
.base = s->layout.lds.indices_addr + component_offset,
.write_mask = write_mask);
}
static void