fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-18 18:08:15 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 2
202482 commits 66 branches 935 tags 949 MiB
Commit graph

10345 commits

Author SHA1 Message Date
Job Noorman
739ca77e66 nir/lower_subgroups: use build_cluster_mask for quad mask 
build_subgroup_quad_mask can now be written in terms of
build_cluster_mask.

Signed-off-by: Job Noorman <jnoorman@igalia.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31732>
 2025-02-27 18:53:19 +00:00
Benjamin Lee
252c59602e panfrost: implement 16-bit ldexp 
Bifrost LDEXP.v2f16 takes a 16-bit exponent, which requires messy
lowering. The codegen for this is quite bad currently, but would be
improved by implementing unpack_32_2x16_split_*, and by fusing
comparisons with CSEL.

The main alternative is converting to F32, then LDEXP.f32, then
converting back to F16. This has better codegen for dynamic exponents
currently, but worse in the common case with a constant exponent where
all the saturating cast logic can be folded.

Fixes dEQP-VK.glsl.builtin.precision_fp16_storage16b.ldexp.compute.vec2
when shaderFloat16 is enabled in panvk.

Signed-off-by: Benjamin Lee <benjamin.lee@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Rebecca Mckeever <rebecca.mckeever@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33637>
 2025-02-27 16:49:11 +00:00
Alejandro Piñeiro
142311258d nir: aliasing checks should be also done with index != 0 
Right now the aliasing/overlapping checks are only done with index
0. I guess that was done because variables don't get a different
internal location even if you have a different index.

But doing that, the checks would not detect a case like this:
  layout(location = 0, index = 1) out vec4 color;
  layout(location = 0, index = 1) out vec4 factor;

That was used on the following piglit parser test:
spec/arb_explicit_attrib_location/1.10/compiler/layout-13.frag

And as the spec included on that test, is a link error case:

" * if more than one varying out variable is bound to the same
    number and index; or"

This commit executes the aliasing checks for index 1 too, and moves
the skip down, to only skip if the current variable and all previous
location-assigned variables has different index and location.

The bad news is that now such assigned variables need to be tracked on
OpenGL-ES. Before that commit that was avoided.

With this commit the mentioned parser test properly fails to link in
any driver.

Reviewed-by: Juan A. Suarez <jasuarez@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33093>
 2025-02-27 15:10:52 +00:00
Job Noorman
2619d576e7 nir/lower_phis_to_scalar: don't create moves for undef sources 
Creating moves out of undefs makes it more difficult for other passes to
detects undefs without having to chase moves. Instead, just create a new
1-component undef.

Signed-off-by: Job Noorman <jnoorman@igalia.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29889>
 2025-02-27 13:18:14 +00:00
Job Noorman
5ae12b6a5a nir/lower_phis_to_scalar: use nir_builder API where possible 
Signed-off-by: Job Noorman <jnoorman@igalia.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29889>
 2025-02-27 13:18:14 +00:00
Job Noorman
66407e3d24 nir/lower_phis_to_scalar: remove unused mem_ctx 
Signed-off-by: Job Noorman <jnoorman@igalia.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29889>
 2025-02-27 13:18:14 +00:00
Lionel Landwerlin
b72c772f83 spirv: fix racy build 
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reported-by: Ki'sak
Acked-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33769>
 2025-02-26 23:09:51 +00:00
Alyssa Rosenzweig
266638d10a nir: clean up progress 
semantic patch made a few bad choices.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33722>
 2025-02-26 15:19:53 +00:00
Alyssa Rosenzweig
593308a685 nir: eliminate nir_metadata_preserve 
Everybody uses the wrapper now.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33722>
 2025-02-26 15:19:53 +00:00
Alyssa Rosenzweig
24d088104f nir,nak: update comments referencing nir_metadata_preserve 
in prep for removing this method.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33722>
 2025-02-26 15:19:53 +00:00
Alyssa Rosenzweig
ff94b155ab treewide: port remaining nir_metadata_preserve users 
apply our semantic patch manually to the remaining users. Coccinelle bailed on
these files for whatever reason, I guess.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33722>
 2025-02-26 15:19:53 +00:00
Alyssa Rosenzweig
9a58a8257e treewide: Switch to nir_progress 
Via the Coccinelle patch at the end of the commit message, followed by

sed -ie 's/progress = progress | /progress |=/g' $(git grep -l 'progress = prog')
ninja -C ~/mesa/build clang-format
cd ~/mesa/src/compiler/nir && clang-format -i *.c
agxfmt

    @@
    identifier prog;
    expression impl, metadata;
    @@

    -if (prog) {
    -nir_metadata_preserve(impl, metadata);
    -} else {
    -nir_metadata_preserve(impl, nir_metadata_all);
    -}
    -return prog;
    +return nir_progress(prog, impl, metadata);

    @@
    expression prog_expr, impl, metadata;
    @@

    -if (prog_expr) {
    -nir_metadata_preserve(impl, metadata);
    -return true;
    -} else {
    -nir_metadata_preserve(impl, nir_metadata_all);
    -return false;
    -}
    +bool progress = prog_expr;
    +return nir_progress(progress, impl, metadata);

    @@
    identifier prog;
    expression impl, metadata;
    @@

    -nir_metadata_preserve(impl, prog ? (metadata) : nir_metadata_all);
    -return prog;
    +return nir_progress(prog, impl, metadata);

    @@
    identifier prog;
    expression impl, metadata;
    @@

    -nir_metadata_preserve(impl, prog ? (metadata) : nir_metadata_all);
    +nir_progress(prog, impl, metadata);

    @@
    expression impl, metadata;
    @@

    -nir_metadata_preserve(impl, metadata);
    -return true;
    +return nir_progress(true, impl, metadata);

    @@
    expression impl;
    @@

    -nir_metadata_preserve(impl, nir_metadata_all);
    -return false;
    +return nir_no_progress(impl);

    @@
    identifier other_prog, prog;
    expression impl, metadata;
    @@

    -if (prog) {
    -nir_metadata_preserve(impl, metadata);
    -} else {
    -nir_metadata_preserve(impl, nir_metadata_all);
    -}
    -other_prog |= prog;
    +other_prog = other_prog | nir_progress(prog, impl, metadata);

    @@
    identifier prog;
    expression impl, metadata;
    @@

    -if (prog) {
    -nir_metadata_preserve(impl, metadata);
    -} else {
    -nir_metadata_preserve(impl, nir_metadata_all);
    -}
    +nir_progress(prog, impl, metadata);

    @@
    identifier other_prog, prog;
    expression impl, metadata;
    @@

    -if (prog) {
    -nir_metadata_preserve(impl, metadata);
    -other_prog = true;
    -} else {
    -nir_metadata_preserve(impl, nir_metadata_all);
    -}
    +other_prog = other_prog | nir_progress(prog, impl, metadata);

    @@
    expression prog_expr, impl, metadata;
    identifier prog;
    @@

    -if (prog_expr) {
    -nir_metadata_preserve(impl, metadata);
    -prog = true;
    -} else {
    -nir_metadata_preserve(impl, nir_metadata_all);
    -}
    +bool impl_progress = prog_expr;
    +prog = prog | nir_progress(impl_progress, impl, metadata);

    @@
    identifier other_prog, prog;
    expression impl, metadata;
    @@

    -if (prog) {
    -other_prog = true;
    -nir_metadata_preserve(impl, metadata);
    -} else {
    -nir_metadata_preserve(impl, nir_metadata_all);
    -}
    +other_prog = other_prog | nir_progress(prog, impl, metadata);

    @@
    expression prog_expr, impl, metadata;
    identifier prog;
    @@

    -if (prog_expr) {
    -prog = true;
    -nir_metadata_preserve(impl, metadata);
    -} else {
    -nir_metadata_preserve(impl, nir_metadata_all);
    -}
    +bool impl_progress = prog_expr;
    +prog = prog | nir_progress(impl_progress, impl, metadata);

    @@
    expression prog_expr, impl, metadata;
    @@

    -if (prog_expr) {
    -nir_metadata_preserve(impl, metadata);
    -} else {
    -nir_metadata_preserve(impl, nir_metadata_all);
    -}
    +bool impl_progress = prog_expr;
    +nir_progress(impl_progress, impl, metadata);

    @@
    identifier prog;
    expression impl, metadata;
    @@

    -nir_metadata_preserve(impl, metadata);
    -prog = true;
    +prog = nir_progress(true, impl, metadata);

    @@
    identifier prog;
    expression impl, metadata;
    @@

    -if (prog) {
    -nir_metadata_preserve(impl, metadata);
    -}
    -return prog;
    +return nir_progress(prog, impl, metadata);

    @@
    identifier prog;
    expression impl, metadata;
    @@

    -if (prog) {
    -nir_metadata_preserve(impl, metadata);
    -}
    +nir_progress(prog, impl, metadata);

    @@
    expression impl;
    @@

    -nir_metadata_preserve(impl, nir_metadata_all);
    +nir_no_progress(impl);

    @@
    expression impl, metadata;
    @@

    -nir_metadata_preserve(impl, metadata);
    +nir_progress(true, impl, metadata);

squashme! sed -ie 's/progress = progress | /progress |=/g' $(git grep -l 'progress = prog')

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Acked-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33722>
 2025-02-26 15:19:53 +00:00
Alyssa Rosenzweig
91872c9c51 nir: clang-format 
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Acked-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33722>
 2025-02-26 15:19:53 +00:00
Alyssa Rosenzweig
469b8bbf3c nir: add nir_progress/nir_no_progress helpers 
These will replace nir_metadata_preserve as more ergonomic replacements that
convey a notion of impl progress instead of simply updating metadata.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33722>
 2025-02-26 15:19:52 +00:00
Georg Lehmann
a237a3def8 nir/opt_algebraic: optimize b2i(a) != -b2i(b) 
Foz-DB Navi21:
Totals from 4 (0.01% of 79377) affected shaders:
Instrs: 881 -> 861 (-2.27%)
CodeSize: 4968 -> 4836 (-2.66%)
Latency: 6127 -> 6006 (-1.97%)
InvThroughput: 1128 -> 1068 (-5.32%)
VALU: 564 -> 534 (-5.32%)
SALU: 111 -> 121 (+9.01%)

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33498>
 2025-02-25 20:38:09 +00:00
Georg Lehmann
4141043295 nir/opt_algebraic: optimize constant shift of DXBC booleans 
Can be combined with further iand.

Foz-DB Navi21:
Totals from 190 (0.24% of 79377) affected shaders:
Instrs: 100628 -> 100225 (-0.40%); split: -0.41%, +0.01%
CodeSize: 567828 -> 565884 (-0.34%); split: -0.35%, +0.00%
Latency: 968415 -> 968052 (-0.04%); split: -0.09%, +0.06%
InvThroughput: 285804 -> 285210 (-0.21%); split: -0.25%, +0.04%
VClause: 1959 -> 1958 (-0.05%)
Copies: 5696 -> 5711 (+0.26%)
PreSGPRs: 7567 -> 7569 (+0.03%)
VALU: 77161 -> 76751 (-0.53%); split: -0.54%, +0.01%
SALU: 7831 -> 7840 (+0.11%); split: -0.09%, +0.20%

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33498>
 2025-02-25 20:38:09 +00:00
Georg Lehmann
1e522e7d75 nir/opt_algebraic: optimize dxbc boolean not 
Foz-DB Navi21:
Totals from 237 (0.30% of 79377) affected shaders:
Instrs: 486690 -> 486146 (-0.11%); split: -0.11%, +0.00%
CodeSize: 2629516 -> 2626052 (-0.13%); split: -0.13%, +0.00%
VGPRs: 18744 -> 18736 (-0.04%)
Latency: 7404763 -> 7399806 (-0.07%); split: -0.07%, +0.01%
InvThroughput: 1800282 -> 1798388 (-0.11%); split: -0.11%, +0.00%
VClause: 12101 -> 12106 (+0.04%); split: -0.01%, +0.05%
Copies: 34225 -> 34170 (-0.16%); split: -0.21%, +0.05%
PreSGPRs: 14634 -> 14639 (+0.03%)
PreVGPRs: 16713 -> 16706 (-0.04%)
VALU: 317523 -> 316693 (-0.26%); split: -0.26%, +0.00%
SALU: 53814 -> 54097 (+0.53%); split: -0.38%, +0.90%

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33498>
 2025-02-25 20:38:09 +00:00
Georg Lehmann
f9722e35be nir/opt_algebraic: optimize more boolean bcsel with constants 
Foz-DB Navi21:
Totals from 667 (0.84% of 79377) affected shaders:
Instrs: 3890980 -> 3886878 (-0.11%); split: -0.11%, +0.00%
CodeSize: 21088576 -> 21065848 (-0.11%); split: -0.11%, +0.00%
SpillSGPRs: 458 -> 446 (-2.62%); split: -3.49%, +0.87%
Latency: 26160728 -> 26162856 (+0.01%); split: -0.02%, +0.02%
InvThroughput: 6999254 -> 7000593 (+0.02%); split: -0.01%, +0.03%
VClause: 103745 -> 103743 (-0.00%)
SClause: 93113 -> 93109 (-0.00%)
Copies: 344097 -> 344794 (+0.20%); split: -0.05%, +0.25%
Branches: 134546 -> 134764 (+0.16%); split: -0.01%, +0.17%
PreSGPRs: 40677 -> 40298 (-0.93%); split: -0.93%, +0.00%
PreVGPRs: 40185 -> 40190 (+0.01%)
VALU: 2584477 -> 2584468 (-0.00%); split: -0.00%, +0.00%
SALU: 573587 -> 569353 (-0.74%); split: -0.75%, +0.01%
SMEM: 124794 -> 124790 (-0.00%)

v2 (idr): Remove a pattern that is made redundant by this commit
combined with the previous commit.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33498>
 2025-02-25 20:38:09 +00:00
Georg Lehmann
9785fa460c nir/opt_algebraic: optimize DXBC boolean bcsel 
Foz-DB Navi21:
Totals from 1749 (2.20% of 79377) affected shaders:
Instrs: 1695408 -> 1685149 (-0.61%); split: -0.68%, +0.07%
CodeSize: 9241312 -> 9174180 (-0.73%); split: -0.79%, +0.06%
VGPRs: 90688 -> 90664 (-0.03%); split: -0.04%, +0.01%
SpillSGPRs: 278 -> 298 (+7.19%)
Latency: 9560167 -> 9540386 (-0.21%); split: -0.29%, +0.08%
InvThroughput: 2236022 -> 2220411 (-0.70%); split: -0.72%, +0.02%
VClause: 29910 -> 29917 (+0.02%)
Copies: 146365 -> 145230 (-0.78%); split: -1.03%, +0.25%
Branches: 59545 -> 59560 (+0.03%)
PreSGPRs: 78858 -> 79242 (+0.49%); split: -0.10%, +0.59%
PreVGPRs: 78643 -> 78560 (-0.11%); split: -0.11%, +0.00%
VALU: 1127861 -> 1113990 (-1.23%); split: -1.24%, +0.01%
SALU: 249535 -> 253237 (+1.48%); split: -0.15%, +1.63%

v2 (idr): Remove a pattern that is now redundant.

v3 (idr): Don't undistribute ineg from bcsel. On platforms where ineg
is a free source modifier, this can be harmful.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33498>
 2025-02-25 20:38:09 +00:00
Georg Lehmann
674d970861 nir/opt_algebraic: 0 >= a -> 0 == a 
Foz-DB Navi21:
Totals from 2179 (2.75% of 79377) affected shaders:
MaxWaves: 40987 -> 40917 (-0.17%); split: +0.00%, -0.18%
Instrs: 5950981 -> 5949310 (-0.03%); split: -0.04%, +0.01%
CodeSize: 32120808 -> 32110328 (-0.03%); split: -0.04%, +0.00%
VGPRs: 141704 -> 141768 (+0.05%); split: -0.01%, +0.05%
SpillSGPRs: 1750 -> 1746 (-0.23%)
Latency: 56667295 -> 56562916 (-0.18%); split: -0.19%, +0.00%
InvThroughput: 13292128 -> 13288691 (-0.03%); split: -0.03%, +0.00%
VClause: 151845 -> 151755 (-0.06%); split: -0.06%, +0.00%
SClause: 172316 -> 172443 (+0.07%); split: -0.02%, +0.09%
Copies: 458724 -> 458951 (+0.05%); split: -0.08%, +0.13%
Branches: 195239 -> 195351 (+0.06%); split: -0.00%, +0.06%
PreSGPRs: 135304 -> 135317 (+0.01%); split: -0.01%, +0.02%
PreVGPRs: 122430 -> 122428 (-0.00%); split: -0.01%, +0.01%
VALU: 3924585 -> 3924062 (-0.01%); split: -0.02%, +0.01%
SALU: 820666 -> 819414 (-0.15%); split: -0.17%, +0.02%
SMEM: 247036 -> 247142 (+0.04%); split: -0.00%, +0.04%

v2 (idr): Remove a pattern that is now redundant. This was originaly
removed in a commit later in the MR.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33498>
 2025-02-25 20:38:09 +00:00
Georg Lehmann
000f14f7fd nir/opt_algebraic: optimize ineg(a) == #b 
No Foz-DB changes.

v2 (idr): Remove some patterns that are now redundant. These were
originally removed in a commit later in the MR.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33498>
 2025-02-25 20:38:08 +00:00
Georg Lehmann
3e4ac92298 nir/opt_algebraic: optimize ineg(a) == ineg(b) 
DXBC boolean cleanup.

Foz-DB Navi21:
Totals from 19 (0.02% of 79188) affected shaders:
Instrs: 9720 -> 9652 (-0.70%)
CodeSize: 54056 -> 53640 (-0.77%)
Latency: 95357 -> 94377 (-1.03%); split: -1.03%, +0.00%
InvThroughput: 17331 -> 16939 (-2.26%)
Copies: 604 -> 605 (+0.17%)
PreSGPRs: 832 -> 838 (+0.72%)
PreVGPRs: 701 -> 699 (-0.29%)
VALU: 6551 -> 6485 (-1.01%)
SALU: 893 -> 891 (-0.22%); split: -1.68%, +1.46%

v2 (idr): Remove a pattern that is now redundant. The version without
ineg already exists much earlier in the file. Search for b2iN.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33498>
 2025-02-25 20:38:08 +00:00
Georg Lehmann
dd1a7f0e8c nir: add a pass to optimize phis to 1bit 
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33498>
 2025-02-25 20:38:08 +00:00
Alyssa Rosenzweig
2504e7951b treewide: use nir_shader_tex_pass 
Adapted the Coccinelle rules from the nir_shader_intrinsics_pass commit a while
ago, which was buggy then and buggy now, so then I fixed stuff up manually
(including formatting).

Via Coccinelle patch:

    @def@
    typedef bool;
    typedef nir_builder;
    typedef nir_instr;
    typedef nir_def;
    identifier fn, instr, intr, x, builder, data;
    @@

    static fn(nir_builder* builder,
    -nir_instr *instr,
    +nir_tex_instr *intr,
    ...)
    {
    (
    -   if (instr->type != nir_instr_type_tex)
    -      return false;
    -   nir_tex_instr *intr = nir_instr_as_tex(instr);
    |
    -   nir_tex_instr *intr = nir_instr_as_tex(instr);
    -   if (instr->type != nir_instr_type_tex)
    -      return false;
    )

    <...
    (
    -instr->x
    +intr->instr.x
    |
    -instr
    +&intr->instr
    )
    ...>

    }

    @pass depends on def@
    identifier def.fn;
    expression shader, progress;
    @@

    (
    -nir_shader_instructions_pass(shader, fn,
    +nir_shader_tex_pass(shader, fn,
    ...)
    |
    -NIR_PASS_V(shader, nir_shader_instructions_pass, fn,
    +NIR_PASS_V(shader, nir_shader_tex_pass, fn,
    ...)
    |
    -NIR_PASS(progress, shader, nir_shader_instructions_pass, fn,
    +NIR_PASS(progress, shader, nir_shader_tex_pass, fn,
    ...)
    )

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Acked-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com> [v3d]
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33672>
 2025-02-24 19:33:26 +00:00
Georg Lehmann
940e87f225 nir/opt_remove_phis: use nir_shader_phi_pass 
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33672>
 2025-02-24 19:33:26 +00:00
Georg Lehmann
e4f0de89a5 nir/opt_phi_precision: use nir_shader_phi_pass 
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33672>
 2025-02-24 19:33:26 +00:00
Georg Lehmann
5a0702f351 nir/builder: add nir_shader_phi_pass 
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33672>
 2025-02-24 19:33:26 +00:00
Alyssa Rosenzweig
dda2dadb98 nir/builder: add nir_shader_tex_pass 
after the intrinsic and ALU passes. why not?

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Acked-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33672>
 2025-02-24 19:33:26 +00:00
Georg Lehmann
5da76df4cd nir/search_helpers: check tex source type in is_only_used_as_float 
Foz-DB Navi21:
Totals from 164 (0.21% of 79377) affected shaders:
Instrs: 197477 -> 197035 (-0.22%); split: -0.23%, +0.01%
CodeSize: 1052944 -> 1051140 (-0.17%); split: -0.18%, +0.01%
VGPRs: 8104 -> 8080 (-0.30%)
Latency: 1115663 -> 1115567 (-0.01%); split: -0.06%, +0.05%
InvThroughput: 265822 -> 265158 (-0.25%); split: -0.26%, +0.01%
VClause: 3792 -> 3789 (-0.08%); split: -0.11%, +0.03%
SClause: 5738 -> 5744 (+0.10%); split: -0.02%, +0.12%
Copies: 12223 -> 12200 (-0.19%); split: -0.53%, +0.34%
PreVGPRs: 6807 -> 6801 (-0.09%); split: -0.15%, +0.06%
VALU: 139206 -> 138785 (-0.30%); split: -0.31%, +0.01%
SALU: 27852 -> 27853 (+0.00%)

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33674>
 2025-02-24 16:34:53 +00:00
Georg Lehmann
3d8585e4fc nir/search_helpers: look through vecs in is_only_used_as_float 
Will be useful with the next commit, or for backends that don't lower
alu to scalar.

No changes on Navi21.

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33674>
 2025-02-24 16:34:53 +00:00
Georg Lehmann
e0cebac14f nir/opt_algebraic: optimize b2f(a != 0) * a 
Just D3D9 things.

Foz-DB Navi21:
Totals from 137 (0.17% of 79377) affected shaders:
MaxWaves: 3366 -> 3370 (+0.12%); split: +0.24%, -0.12%
Instrs: 76462 -> 72091 (-5.72%)
CodeSize: 411584 -> 380792 (-7.48%)
Latency: 279472 -> 275505 (-1.42%); split: -2.01%, +0.59%
InvThroughput: 71311 -> 65369 (-8.33%)
VClause: 1587 -> 1612 (+1.58%); split: -1.01%, +2.58%
SClause: 1111 -> 1105 (-0.54%); split: -1.08%, +0.54%
Copies: 5621 -> 5602 (-0.34%); split: -1.39%, +1.05%
PreSGPRs: 5266 -> 5241 (-0.47%); split: -0.51%, +0.04%
PreVGPRs: 4249 -> 4236 (-0.31%); split: -0.35%, +0.05%
VALU: 50049 -> 45901 (-8.29%)
SALU: 8948 -> 8818 (-1.45%)

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33674>
 2025-02-24 16:34:53 +00:00
Lionel Landwerlin
23e14ac41e nir: track lowered image intrinsics to globals 
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32676>
 2025-02-23 15:16:50 +00:00
Eric Engestrom
dd7bac4e31 vtn_bindgen2: keep the printf blob local to avoid LTO issues 
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33442>
 2025-02-23 01:25:23 +00:00
Timur Kristóf
d74d316f60 glsl: Don't use deprecated NIR_PASS_V macro anymore. 
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33609>
 2025-02-22 08:54:39 +01:00
Timur Kristóf
093e68b518 compiler/clc: Stop using deprecated NIR_PASS_V macro. 
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33609>
 2025-02-22 08:54:23 +01:00
Timur Kristóf
9e7609b0ff ac/nir: Add bool return value to ac_nir_lower_task_outputs_to_mem. 
And fixup its NIR counterparts too.

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33609>
 2025-02-22 08:54:17 +01:00
Timur Kristóf
65139305e2 nir: Don't use deprecated NIR_PASS_V macro anymore. 
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33609>
 2025-02-22 08:54:16 +01:00
Timur Kristóf
7222bb397d nir: Add bool return value to nir_fixup_deref_types. 
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33609>
 2025-02-22 08:54:16 +01:00
Timur Kristóf
2fa7711a2b nir: Add bool return value to nir_fixup_deref_modes. 
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33609>
 2025-02-22 08:54:16 +01:00
Timur Kristóf
4f744998ef nir: Add comment to indicate that NIR_PASS_V is deprecated. 
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33609>
 2025-02-22 08:54:16 +01:00
Danylo Piliaiev
763ddd0fd3 nir/nir_lower_multiview: Don't assert if load_deref doesn't have var 
If deref chain has nir_deref_type_cast nir_intrinsic_get_var will
return null, which is valid for e.g. shader inputs, since the pass
only care about outputs.

NIR excerpt that caused issues:

```
    32x3    %6 = deref_cast (block *)%5 (ubo block)  (ptr_stride=0, align_mul=0, align_offset=0)
    32x3    %7 = deref_struct &%6->field0 (ubo vec4[4])  // &((block *)%5)->field0
    32      %8 = load_const (0x00000001)
    32x3    %9 = deref_array &(*%7)[1] (ubo vec4)  // &((block *)%5)->field0[1]
    32x4   %10 = @load_deref (%9) (access=none)
```

Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33647>
 2025-02-21 11:09:22 +00:00
Lorenzo Rossi
26079c1a93 nir: support shared atomics in nir_lower_atomics 
Add support to rewrite shared atomics into compare-and-swap loops,
previously the nir_lower_atomics pass only supported global and ssbo
atomics.

Only freedreno irc3 reuses nir_lower_atomics, this change does not
impact their usage since they do not support shared atomics.

Signed-off-by: Lorenzo Rossi <snowycoder@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33572>
 2025-02-21 00:33:16 +00:00
Ian Romanick
15544ed858 nir/algebraic: Undistribute b2i from logic-ops 
shader-db:
All Intel platforms had similar results. (Lunar Lake shown)
total instructions in shared programs: 16973309 -> 16973173 (<.01%)
instructions in affected programs: 13780 -> 13644 (-0.99%)
helped: 31 / HURT: 0

total cycles in shared programs: 915620550 -> 915618604 (<.01%)
cycles in affected programs: 185962 -> 184016 (-1.05%)
helped: 30 / HURT: 1

fossil-db:

All Intel platforms had similar results. (Lunar Lake shown)
Totals:
Instrs: 209748003 -> 209745278 (-0.00%)
Cycle count: 30514920400 -> 30514716506 (-0.00%); split: -0.00%, +0.00%
Max live registers: 65477183 -> 65477584 (+0.00%)
Non SSA regs after NIR: 237334710 -> 237333632 (-0.00%)

Totals from 1257 (0.18% of 706651) affected shaders:
Instrs: 693039 -> 690314 (-0.39%)
Cycle count: 39792504 -> 39588610 (-0.51%); split: -0.97%, +0.46%
Max live registers: 194170 -> 194571 (+0.21%)
Non SSA regs after NIR: 821978 -> 820900 (-0.13%)

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33648>
 2025-02-21 00:01:11 +00:00
Ian Romanick
a48a044cf6 nir/algebraic: Simplify equality comparisons of b2T with 1 or 0 
Adding the b2i(a) == 1 and b2i(a) != 1 patterns also helps prevent
regressions when spurious negations are removed from integer equality
comparisons, as is done in !33498.

v2: Make all variables part of the iteration instead of calculating some
of them. Suggested by Alyssa.

shader-db:

All Intel platforms had similar results. (Lunar Lake shown)
total instructions in shared programs: 16973331 -> 16973309 (<.01%)
instructions in affected programs: 266 -> 244 (-8.27%)
helped: 2 / HURT: 0

total cycles in shared programs: 915620774 -> 915620550 (<.01%)
cycles in affected programs: 4360 -> 4136 (-5.14%)
helped: 2 / HURT: 0

fossil-db:

All Intel platforms had similar results. (Lunar Lake shown)
Totals:
Instrs: 209748011 -> 209748003 (-0.00%)
Cycle count: 30514920286 -> 30514920400 (+0.00%); split: -0.00%, +0.00%
Non SSA regs after NIR: 237334726 -> 237334710 (-0.00%)

Totals from 8 (0.00% of 706651) affected shaders:
Instrs: 16956 -> 16948 (-0.05%)
Cycle count: 261052 -> 261166 (+0.04%); split: -0.92%, +0.96%
Non SSA regs after NIR: 20000 -> 19984 (-0.08%)

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33648>
 2025-02-21 00:01:11 +00:00
Ian Romanick
3f39d8f4ff nir/algebraic: Optimize zero comparisons of umax or umin 
I observered some of the existing patterns stopped being applied after
some of the ult-to-ieq optimizations in !33498. It turns out that these
patterns occur even without those changes.

shader-db:

All Intel platforms had similar results. (Lunar Lake shown)
total instructions in shared programs: 16973339 -> 16973331 (<.01%)
instructions in affected programs: 7977 -> 7969 (-0.10%)
helped: 2 / HURT: 0

total cycles in shared programs: 915620938 -> 915620774 (<.01%)
cycles in affected programs: 136022 -> 135858 (-0.12%)
helped: 2 / HURT: 0

fossil-db:

Lunar Lake
Totals:
Instrs: 209748173 -> 209748011 (-0.00%); split: -0.00%, +0.00%
Cycle count: 30514361348 -> 30514920286 (+0.00%); split: -0.00%, +0.00%
Spill count: 511813 -> 511808 (-0.00%)
Fill count: 622537 -> 622533 (-0.00%)
Max live registers: 65477033 -> 65477183 (+0.00%); split: -0.00%, +0.00%
Non SSA regs after NIR: 237334728 -> 237334726 (-0.00%); split: -0.00%, +0.00%

Totals from 26 (0.00% of 706651) affected shaders:
Instrs: 332073 -> 331911 (-0.05%); split: -0.05%, +0.00%
Cycle count: 959758560 -> 960317498 (+0.06%); split: -0.03%, +0.09%
Spill count: 10293 -> 10288 (-0.05%)
Fill count: 23784 -> 23780 (-0.02%)
Max live registers: 9682 -> 9832 (+1.55%); split: -0.08%, +1.63%
Non SSA regs after NIR: 232135 -> 232133 (-0.00%); split: -0.03%, +0.03%

Meteor Lake and DG2 had similar results. (Meteor Lake shown)
Totals:
Instrs: 233538532 -> 233536113 (-0.00%); split: -0.00%, +0.00%
Cycle count: 24428142259 -> 24426705655 (-0.01%); split: -0.01%, +0.00%
Spill count: 513128 -> 512923 (-0.04%)
Fill count: 557329 -> 557108 (-0.04%)
Max live registers: 42129806 -> 42129881 (+0.00%); split: -0.00%, +0.00%
Non SSA regs after NIR: 256711720 -> 256711718 (-0.00%); split: -0.00%, +0.00%

Totals from 26 (0.00% of 805759) affected shaders:
Instrs: 325629 -> 323210 (-0.74%); split: -0.74%, +0.00%
Cycle count: 893896782 -> 892460178 (-0.16%); split: -0.21%, +0.05%
Spill count: 10467 -> 10262 (-1.96%)
Fill count: 24291 -> 24070 (-0.91%)
Max live registers: 4946 -> 5021 (+1.52%); split: -0.08%, +1.60%
Non SSA regs after NIR: 232980 -> 232978 (-0.00%); split: -0.03%, +0.03%

Tiger Lake, Ice Lake, and Skylake had similar results. (Tiger Lake shown)
Totals:
Instrs: 237289818 -> 237289714 (-0.00%); split: -0.00%, +0.00%
Cycle count: 22959586058 -> 22960049302 (+0.00%); split: -0.00%, +0.00%
Max live registers: 42182257 -> 42182337 (+0.00%)
Non SSA regs after NIR: 255579974 -> 255579970 (-0.00%); split: -0.00%, +0.00%

Totals from 23 (0.00% of 802019) affected shaders:
Instrs: 27051 -> 26947 (-0.38%); split: -0.39%, +0.01%
Cycle count: 10545917 -> 11009161 (+4.39%); split: -0.09%, +4.49%
Max live registers: 2198 -> 2278 (+3.64%)
Non SSA regs after NIR: 31741 -> 31737 (-0.01%); split: -0.20%, +0.19%

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33648>
 2025-02-21 00:01:11 +00:00
Ian Romanick
4311121e73 nir/algebraic: More (a == 0 || a == 1 || ...) patterns 
At least some Total War: Warhammer3 vertex shaders associate the
comparisons differntly, so the existing patterns were not triggered.

No shader-db changes on any Intel platform.

fossil-db:

All Intel platforms had similar results. (Lunar Lake shown)
Totals:
Instrs: 209748654 -> 209748173 (-0.00%)
Cycle count: 30514333964 -> 30514361348 (+0.00%); split: -0.00%, +0.00%
Fill count: 622688 -> 622537 (-0.02%)
Max live registers: 65477039 -> 65477033 (-0.00%)
Non SSA regs after NIR: 237334768 -> 237334728 (-0.00%)

Totals from 512 (0.07% of 706651) affected shaders:
Instrs: 1000693 -> 1000212 (-0.05%)
Cycle count: 42174312 -> 42201696 (+0.06%); split: -0.15%, +0.21%
Fill count: 11456 -> 11305 (-1.32%)
Max live registers: 121599 -> 121593 (-0.00%)
Non SSA regs after NIR: 1253445 -> 1253405 (-0.00%)

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33648>
 2025-02-21 00:01:11 +00:00
Georg Lehmann
f26069fdd9 nir: replace nir_opt_conditional_discard with nir_opt_peephole_select 
Foz-DB Navi21:
Totals from 118 (0.15% of 79377) affected shaders:
Instrs: 208001 -> 207355 (-0.31%); split: -0.33%, +0.01%
CodeSize: 1080428 -> 1078432 (-0.18%); split: -0.20%, +0.02%
SpillSGPRs: 202 -> 211 (+4.46%)
Latency: 1923508 -> 1919093 (-0.23%); split: -0.62%, +0.39%
InvThroughput: 407475 -> 407081 (-0.10%); split: -0.12%, +0.02%
SClause: 7050 -> 7033 (-0.24%); split: -0.31%, +0.07%
Copies: 12156 -> 11821 (-2.76%); split: -3.04%, +0.28%
PreSGPRs: 8198 -> 8331 (+1.62%); split: -0.02%, +1.65%
PreVGPRs: 7628 -> 7528 (-1.31%)
VALU: 155747 -> 155657 (-0.06%); split: -0.06%, +0.00%
SALU: 18295 -> 17782 (-2.80%); split: -2.98%, +0.18%
SMEM: 10521 -> 10519 (-0.02%)

Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33590>
 2025-02-20 21:59:17 +00:00
Georg Lehmann
8251a5b846 nir/peephole_select: don't completely ignore ifs with dont_flatten 
Apps are misusing this for cases where the if-else are empty (except for phis)
or for conditional discard which will become relevant in the next commit.

Foz-DB Navi21:
Totals from 173 (0.22% of 79188) affected shaders:
Instrs: 1465214 -> 1464987 (-0.02%); split: -0.04%, +0.03%
CodeSize: 7960472 -> 7965188 (+0.06%); split: -0.01%, +0.07%
Latency: 10001176 -> 10012782 (+0.12%); split: -0.01%, +0.12%
InvThroughput: 2336017 -> 2338979 (+0.13%); split: -0.00%, +0.13%
Copies: 140105 -> 138225 (-1.34%)
Branches: 49746 -> 49732 (-0.03%)
VALU: 975632 -> 976322 (+0.07%); split: -0.01%, +0.08%
SALU: 201369 -> 200688 (-0.34%)

Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33590>
 2025-02-20 21:59:16 +00:00
Georg Lehmann
cfee9e1d9f nir/peephole_select: add option to allow discard without ~0 limit 
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33590>
 2025-02-20 21:59:16 +00:00
Georg Lehmann
ca8147edbe nir/peephole_select: add options struct 
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33590>
 2025-02-20 21:59:16 +00:00