fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-05 07:28:11 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 2

nir/opt_varyings: fix compaction with sparse indirect FS inputs

Browse source 
Without this, compaction can put inputs into vec4 slots already occupied
by indirectly-accessed inputs while ignoring their interpolation qualifier,
which is incorrect.

All input components sharing the same vec4 slot must use interpolation
qualifiers that are compatible with each other.

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32424>
This commit is contained in:
Marek Olšák 2024-11-30 04:32:05 -05:00 • committed by Marge Bot
parent b01f3cea7a
commit 1aa9fec542
1 changed files with 32 additions and 13 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

45
src/compiler/nir/nir_opt_varyings.c
View file

@ -4222,19 +4222,6 @@ fs_assign_slots(struct linkage_info *linkage,
continue;
}
/* Copy the FS vec4 type if indexed indirectly, and move to
* the next slot.
*/
if (BITSET_TEST32(linkage->indirect_mask, slot_index)) {
if (assigned_fs_vec4_type) {
assigned_fs_vec4_type[vec4_slot(slot_index)] =
linkage->fs_vec4_type[vec4_slot(slot_index)];
}
assert(slot_index % 2 == 0);
slot_index += 2; /* increment by 32 bits */
continue;
}
/* This slot is already assigned (assigned_mask is set). Move to
* the next one.
*/
@ -4559,6 +4546,38 @@ compact_varyings(struct linkage_info *linkage,
BITSET_DECLARE(assigned_mask, NUM_SCALAR_SLOTS);
BITSET_ZERO(assigned_mask);
/* Iterate over all indirectly accessed inputs and set the assigned vec4
* type of each occupied slot to the vec4 type of indirect inputs, so
* that compaction doesn't put inputs of a different vec4 type in
* the same vec4.
*
* We don't try to compact indirect input arrays, though we could.
*/
unsigned i;
BITSET_FOREACH_SET(i, linkage->indirect_mask, NUM_SCALAR_SLOTS) {
struct scalar_slot *slot = &linkage->slot[i];
/* The slot of the first array element contains all loads for all
* elements, including all direct accesses, while all other array
* elements are empty (on purpose).
*/
if (list_is_empty(&linkage->slot[i].consumer.loads))
continue;
assert(slot->num_slots >= 2);
for (unsigned array_index = 0; array_index < slot->num_slots;
array_index++) {
unsigned vec4_index = vec4_slot(i) + array_index;
unsigned scalar_index = i + array_index * 8;
assigned_fs_vec4_type[vec4_index] = linkage->fs_vec4_type[vec4_index];
/* Indirectly-indexed slots are marked to always occupy 32 bits
* (2 16-bit slots), though we waste the high 16 bits if they are unused.
*/
BITSET_SET_RANGE_INSIDE_WORD(assigned_mask, scalar_index, scalar_index + 1);
}
}
if (linkage->has_flexible_interp) {
/* This codepath packs convergent varyings with both interpolated and
* flat, whichever has free space.