fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-19 13:38:19 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 2
217632 commits 66 branches 935 tags 966 MiB
Commit graph

11583 commits

Author SHA1 Message Date
Emma Anholt
6b6cf7b89f nir/opt_algebraic_tests: Fix annotating uint values. 
In debugging the CI flakes, it was disconcerting to see:

1     %1 = fisnormal %0 (0.000000) // 0x55e7c4147c00

instead of:

1     %1 = fisnormal %0 (0.000000) // 0x0

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39369>
 2026-01-26 05:39:42 +00:00
Emma Anholt
77cfb20d0a nir/opt_algebraic_tests: Fix leak of the variable conds ht. 
We'd end the test with 500MB leaked.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39369>
 2026-01-26 05:39:42 +00:00
Emma Anholt
cdec063d37 nir/opt_algebraic: Fix a bit of imad24_ir3's optimization. 
The mul is 24-bit sign-extended, so in simplifying we should retain that.
If nothing else, this keeps us on the happy path of mul24s.

I didn't fix the other broken pattern, since it's not really part of this
MR.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39369>
 2026-01-26 05:39:41 +00:00
Emma Anholt
e5a9eae2b5 nir/opt_algebraic_tests: Fix fuzzing levels for multi-component inputs. 
We were enumerating enough for a single component, but not all the
combinations.  This helps show that our fdots fail pretty consistently.
And triggers more skipping from the fany_equal16s thanks to varied inputs.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39369>
 2026-01-26 05:39:41 +00:00
Emma Anholt
7fd0287a89 nir/opt_algebraic_tests: Test !nir_fp_preserve_signed_zero behavior. 
Iterate over a set of sign-flips for 0.0s to see if we can find a set that
makes the search and replace sides match.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39369>
 2026-01-26 05:39:41 +00:00
Emma Anholt
d9eb9a3eef nir/opt_algebraic_tests: Make sure we test the same inputs on BE as LE. 
With the inexact signed zero support, we ended up with diverging inputs
producing diverging "all skipped" results for the matrix mul tests.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39369>
 2026-01-26 05:39:40 +00:00
Emma Anholt
68f5bc4f12 nir/opt_algebraic_tests: Rename and use the enum result type more. 
As I introduced another layer of iteration for signed zero testing, the
former logic got unwieldy.  In fact, it was already unwieldy enough that I
forgot to clear all_skipped when the assert failed, allowing a failing
test to be marked UNSUPPORTED instead of XFAIL.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39369>
 2026-01-26 05:39:40 +00:00
Emma Anholt
7f1a64e7f5 nir/opt_algebraic_tests: Move more of the base class code to be methods. 
Less passing the *test around separately.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39369>
 2026-01-26 05:39:40 +00:00
Emma Anholt
845e2b3954 nir/opt_algebraic_tests: Remove unnecessary input_count. 
Originally this was for sizing the array, but now we're doing std::vector.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39369>
 2026-01-26 05:39:40 +00:00
Emma Anholt
a90163a15a nir/opt_algebraic_tests: Add support for expression swizzles. 
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39369>
 2026-01-26 05:39:40 +00:00
Emma Anholt
c30c383d4d nir/opt_algebraic_tests: Allow testing of fdot*_replicated opcodes. 
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39369>
 2026-01-26 05:39:40 +00:00
Emma Anholt
173295adf4 nir/opt_algebraic_tests: Allow testing udiv_aligned_4. 
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39369>
 2026-01-26 05:39:40 +00:00
Emma Anholt
231a477402 nir/opcodes: Define udiv_aligned_4 to return poison for not-aligned-4. 
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39369>
 2026-01-26 05:39:40 +00:00
Emma Anholt
94237c3ea3 nir/opt_algebraic_tests: Allow testing mul/mad_relaxed opcodes. 
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39369>
 2026-01-26 05:39:40 +00:00
Emma Anholt
f82621bfdc nir/opcodes: Define the mul/mad_relaxed opcodes to return poison for OOB. 
For the "defined results for >24b", you want the non-relaxed opcodes.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39369>
 2026-01-26 05:39:39 +00:00
Emma Anholt
fd7754fba1 nir/opt_algebraic_tests: Allow testing imad24_ir3. 
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39369>
 2026-01-26 05:39:39 +00:00
Emma Anholt
ce7ad2639a nir: Fix C UB in imad24_ir3 evaluation. 
Same fix as imul24, technically you can't shift into the top bit of the
int32, but the util helper does it right.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39369>
 2026-01-26 05:39:39 +00:00
Caio Oliveira
231747b7fa spirv: Check Capability for identifying SPV_NV_mesh_shader
Some checks are pending
macOS-CI / macOS-CI (dri) (push) Waiting to run
Details
macOS-CI / macOS-CI (xlib) (push) Waiting to run
Details 
Take advantage of the new enabled_capabilities.  Note MeshShadingNV
is not an alias to MeshShadingEXT.

Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39505>
 2026-01-25 05:11:00 +00:00
Christian Gmeiner
4fedf6bedd nir/lower_blend: Add advanced blending support 
Add support for advanced blending (VK_EXT_blend_operation_advanced and
GL_KHR_blend_equation_advanced), enabling around 40 advanced blend modes
including multiply, screen, overlay, HSL modes (hue, saturation, color,
luminosity), Porter-Duff modes, and extended modes like lineardodge
and vividlight.

Advanced blending slots into the existing blending logic alongside logic
operations and standard blending. The implementation supports both
premultiplied and non-premultiplied alpha for source and destination, and
provides three overlap modes (uncorrelated, conjoint, disjoint).

Signed-off-by: Christian Gmeiner <cgmeiner@igalia.com>
Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38929>
 2026-01-24 21:44:01 +00:00
Christian Gmeiner
9c7088f41c nir: Extract blend equation helpers to shared header 
Move the blend equation helper functions (blend_multiply, blend_screen,
blend_overlay, etc.) from gl_nir_lower_blend_equation_advanced.c to a
new shared header file nir_blend_equation_advanced_helper.h.

These helpers implement the mathematical blend operations defined by
KHR_blend_equation_advanced and will be reused by the new NIR lowering
pass for VK_EXT_blend_operation_advanced.

Signed-off-by: Christian Gmeiner <cgmeiner@igalia.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38929>
 2026-01-24 21:44:01 +00:00
Christian Gmeiner
4294e0bbd4 util/blend: Add advanced blend modes 
VK_EXT_blend_operation_advanced and GL_NV_blend_equation_advanced
defines additional blend operations beyond what OpenGL KHR_blend_equation_advanced
provides. Add these modes to pipe_advanced_blend_mode.

Also add a default case to gl_nir_lower_blend_equation_advanced.c
to handle unsupported modes gracefully.

Signed-off-by: Christian Gmeiner <cgmeiner@igalia.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38929>
 2026-01-24 21:44:00 +00:00
Christian Gmeiner
1eed6960c5 compiler/mesa/st: Move gl_advanced_blend_mode to util/blend.h 
Rename gl_advanced_blend_mode to pipe_advanced_blend_mode and move it
to src/util/blend.h so it can be shared between OpenGL and Vulkan
drivers.

This prepares for implementing VK_EXT_blend_operation_advanced by
providing a common enum for advanced blend modes across APIs.

Signed-off-by: Christian Gmeiner <cgmeiner@igalia.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38929>
 2026-01-24 21:44:00 +00:00
Daniel Schürmann
50bb16d8ea nir/lower_non_uniform_access: flag IF as always divergent taken
Some checks are pending
macOS-CI / macOS-CI (dri) (push) Waiting to run
Details
macOS-CI / macOS-CI (xlib) (push) Waiting to run
Details 
Totals from 849 (1.01% of 84383) affected shaders: (Navi48)

Instrs: 5411150 -> 5387935 (-0.43%)
CodeSize: 30782816 -> 30733420 (-0.16%); split: -0.21%, +0.05%
Latency: 44569512 -> 44282889 (-0.64%)
InvThroughput: 6822696 -> 6785429 (-0.55%); split: -0.55%, +0.00%
SClause: 159022 -> 159017 (-0.00%)
Branches: 174234 -> 152625 (-12.40%)
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39490>
 2026-01-24 12:08:40 +00:00
Samuel Pitoiset
bb4157ad31 spirv: Update the JSON and headers 
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39476>
 2026-01-23 16:07:30 +00:00
Connor Abbott
c7a60fca3c nir, ir3: Make ir3 GS varyings use a proper index 
Use a similar technique of reusing TCS-only varyings for VS/TES/GS.
This way outputs_written can be used to see whether GS lowering has
already been run.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39328>
 2026-01-21 20:54:15 +00:00
Connor Abbott
5abe3db518 nir/lower_clip: Correctly handle driver_location in VS lowering 
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39328>
 2026-01-21 20:54:15 +00:00
Connor Abbott
62085044a4 nir: Allow lower_clip_fs with lowered IO 
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39328>
 2026-01-21 20:54:14 +00:00
Connor Abbott
b8905ff2f0 nir/recompute_io_bases: Fix num_inputs with dual-slot VS inputs 
Use num_normal_inputs, which also includes dual-slot inputs.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39328>
 2026-01-21 20:54:14 +00:00
Connor Abbott
778ed567ed nir/recompute_io_bases: Fix handling of dual-source blending 
We forgot to increment num_outputs for dual-source outputs.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39328>
 2026-01-21 20:54:14 +00:00
Connor Abbott
32a4146992 nir: Fix recompute_io_bases with compact i/o arrays 
Follow the example of unlower_io_to_vars, so that when called within
that pass we don't incorrectly recalculate bases for backends that have
compact arrays.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39328>
 2026-01-21 20:54:14 +00:00
Connor Abbott
3208cc80b1 nir: Move is is_compact() out of unlower_io_to_vars 
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39328>
 2026-01-21 20:54:13 +00:00
Lionel Landwerlin
a19e949824 brw: move coarse_z computation to NIR 
So that we can print it easily with debug printfs

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38996>
 2026-01-21 16:00:52 +00:00
Lionel Landwerlin
98194dfa0b nir: add intrinsics for Z calculation in shaders with FSR 
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38996>
 2026-01-21 16:00:52 +00:00
Lionel Landwerlin
12be2a580c nir/compiler_options: add nir_load_pixel_coord 
And use it for nir_printf_fmt_at_px().

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38996>
 2026-01-21 16:00:50 +00:00
Daniel Schürmann
89b9fcb5e7 nir/opt_load_store_vectorize: delay aliasing test in try_vectorize_shared2() 
Checking for aliasing can be very expensive.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38659>
 2026-01-21 14:20:06 +00:00
Daniel Schürmann
598928d7e7 nir/loop_analyze: determine whether all control flow gets eliminated upon loop unrolling 
Totals from 17 (0.02% of 79839) affected shaders: (Navi48)

MaxWaves: 241 -> 243 (+0.83%); split: +5.81%, -4.98%
Instrs: 44198 -> 43786 (-0.93%); split: -8.19%, +7.26%
CodeSize: 230284 -> 226900 (-1.47%); split: -10.55%, +9.08%
VGPRs: 2152 -> 2524 (+17.29%); split: -3.90%, +21.19%
Scratch: 718848 -> 0 (-inf%)
Latency: 128977 -> 145720 (+12.98%); split: -2.12%, +15.10%
InvThroughput: 206804 -> 254250 (+22.94%); split: -0.32%, +23.27%
VClause: 1296 -> 1309 (+1.00%); split: -28.09%, +29.09%
SClause: 835 -> 833 (-0.24%)
Copies: 6284 -> 3630 (-42.23%); split: -44.51%, +2.28%
Branches: 1003 -> 961 (-4.19%)
PreSGPRs: 1003 -> 996 (-0.70%); split: -1.20%, +0.50%
PreVGPRs: 1510 -> 2130 (+41.06%)
VALU: 23577 -> 24309 (+3.10%); split: -6.26%, +9.37%
SALU: 5875 -> 5688 (-3.18%); split: -6.26%, +3.08%
VMEM: 3679 -> 3001 (-18.43%); split: -33.27%, +14.84%
SMEM: 1632 -> 1631 (-0.06%)
VOPD: 23 -> 24 (+4.35%)
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38659>
 2026-01-21 14:20:06 +00:00
Daniel Schürmann
4997d8fb1b nir/loop_analyze: determine for all ALU whether it can be constant-folded 
Totals from 16 (0.02% of 79839) affected shaders: (Navi48)

MaxWaves: 512 -> 464 (-9.38%)
Instrs: 11821 -> 17205 (+45.55%)
CodeSize: 60536 -> 86644 (+43.13%)
VGPRs: 732 -> 804 (+9.84%)
Latency: 68411 -> 39349 (-42.48%)
InvThroughput: 14217 -> 9306 (-34.54%)
VClause: 223 -> 302 (+35.43%)
SClause: 262 -> 317 (+20.99%)
Copies: 961 -> 696 (-27.58%); split: -39.23%, +11.65%
Branches: 182 -> 158 (-13.19%); split: -29.67%, +16.48%
PreSGPRs: 1210 -> 945 (-21.90%); split: -29.42%, +7.52%
PreVGPRs: 647 -> 633 (-2.16%)
VALU: 5112 -> 10857 (+112.38%)
SALU: 3215 -> 2335 (-27.37%); split: -30.67%, +3.30%
VMEM: 228 -> 349 (+53.07%)
SMEM: 567 -> 549 (-3.17%); split: -3.70%, +0.53%
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38659>
 2026-01-21 14:20:06 +00:00
Natalie Vock
30f6eacfad radv/rt: Call ahit/isec shaders 
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39314>
 2026-01-20 21:49:55 +00:00
Icenowy Zheng
b61dbc98fd nir/algebraic: fix Python-3.10-incompatible syntax
Some checks are pending
macOS-CI / macOS-CI (dri) (push) Waiting to run
Details
macOS-CI / macOS-CI (xlib) (push) Waiting to run
Details 
Using a string literal enclosed with the same type of quotation marks
with the outer f-string isn't supported on Python 3.10, which is
currently still with security maintainance.

This leads to syntax error when building Mesa with Python 3.10.

Fix this by alternating these string literals' quotation mark to '' (as
the outer f-string uses "").

Signed-off-by: Icenowy Zheng <zhengxingda@iscas.ac.cn>
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/14673
Reviewed-by: Valentine Burley <valentine.burley@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39372>
 2026-01-20 11:14:41 +00:00
Faith Ekstrand
2313bec66e nir: Expose the guts of nir_lower_blend as builder helpers 
Reviewed-by: Christoph Pillmayer <christoph.pillmayer@arm.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39367>
 2026-01-19 21:33:14 +00:00
Faith Ekstrand
d2c2d798f8 nir/lower_blend: Optimize trivial logic op cases 
There's no point in going to/from UNORM if we're just going to copy or
throw away the source.

Reviewed-by: Christoph Pillmayer <christoph.pillmayer@arm.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39367>
 2026-01-19 21:33:14 +00:00
Faith Ekstrand
68d22b5a2a nir/lower_blend: Move the format to nir_lower_blend_rt 
Reviewed-by: Christoph Pillmayer <christoph.pillmayer@arm.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39367>
 2026-01-19 21:33:14 +00:00
Faith Ekstrand
d6556a580f nir,pan: Add and implement a new store_tile_pan intrinsic 
Like we just did with load_tile_pan, this maps directly to ST_TILE in
the hardware.  This is more versatile and lets us do more of our
lowering in NIR.

Reviewed-by: Christoph Pillmayer <christoph.pillmayer@arm.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39367>
 2026-01-19 21:33:13 +00:00
Faith Ekstrand
11b6cd2f2c nir,pan: Rework the pafrost tile load intrinsic 
Instead of making it explicitly about outputs, this switchies it to
being a NIR version of LD_TILE.  It means we have to do a bit of work in
NIR and add a builder helper but the end result is something much more
versatile.

Reviewed-by: Christoph Pillmayer <christoph.pillmayer@arm.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39367>
 2026-01-19 21:33:13 +00:00
Faith Ekstrand
4189865347 nir: panfrost tile loads are always divergent 
Each lane refers to a different pixel.

Cc: mesa-stable
Reviewed-by: Christoph Pillmayer <christoph.pillmayer@arm.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39367>
 2026-01-19 21:33:13 +00:00
Georg Lehmann
70a951c3f3 nir/search: allow inexact patterns if denorms have to be flushed 
Patterns should ensure that they flush denorms with fcanonicalize.

Removing in between denorm flushing when fusing operations is explicitly
allowed unless those optimizations are generally disallowed by other
floating point math control flags.

Foz-DB Navi21:
Totals from 291 (0.35% of 82377) affected shaders:
Instrs: 138347 -> 137773 (-0.41%)
CodeSize: 751460 -> 748516 (-0.39%)
Latency: 1686466 -> 1686226 (-0.01%); split: -0.02%, +0.01%
InvThroughput: 270847 -> 269963 (-0.33%)
VClause: 2023 -> 2022 (-0.05%)
SClause: 5271 -> 5260 (-0.21%); split: -0.25%, +0.04%
Copies: 8929 -> 8912 (-0.19%)
VALU: 87108 -> 86552 (-0.64%)
SALU: 23460 -> 23443 (-0.07%)

Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39180>
 2026-01-19 16:11:29 +00:00
Georg Lehmann
442daeb54a nir/opt_algebraic: use fcanonicalize 
Mostly optimizations, some minor fixes but I don't
think they are worth backporting.

Foz-DB Navi21:
Totals from 7570 (9.21% of 82151) affected shaders:
MaxWaves: 204288 -> 204476 (+0.09%); split: +0.09%, -0.00%
Instrs: 4511439 -> 4500261 (-0.25%); split: -0.25%, +0.00%
CodeSize: 23727088 -> 23644388 (-0.35%); split: -0.35%, +0.00%
VGPRs: 290944 -> 290616 (-0.11%); split: -0.12%, +0.01%
SpillSGPRs: 1256 -> 1251 (-0.40%)
Latency: 16738072 -> 16726717 (-0.07%); split: -0.10%, +0.04%
InvThroughput: 3736856 -> 3716631 (-0.54%); split: -0.55%, +0.01%
VClause: 66150 -> 66156 (+0.01%); split: -0.05%, +0.06%
SClause: 93644 -> 93631 (-0.01%); split: -0.02%, +0.01%
Copies: 448816 -> 458584 (+2.18%); split: -0.05%, +2.22%
Branches: 139817 -> 139775 (-0.03%); split: -0.03%, +0.00%
PreSGPRs: 321922 -> 321900 (-0.01%); split: -0.01%, +0.00%
PreVGPRs: 239709 -> 238856 (-0.36%); split: -0.39%, +0.03%
VALU: 2595164 -> 2584250 (-0.42%); split: -0.43%, +0.01%
SALU: 839038 -> 838965 (-0.01%); split: -0.02%, +0.01%
VMEM: 137584 -> 137583 (-0.00%)

Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39180>
 2026-01-19 16:11:29 +00:00
Rhys Perry
625afb0d29 nir: add fcanonicalize 
v2(Georg Lehmann):
Always remove fcanonicalize if denorms must be neither flushed nor preserved.

Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39180>
 2026-01-19 16:11:29 +00:00
Georg Lehmann
43d998df84 nir: document that both input and output denorms have to be flushed 
This allows us to remove a * 1.0 or a - 0.0 if is_only_used_as_float.
We already rely on that.

Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39180>
 2026-01-19 16:11:28 +00:00
Georg Lehmann
d7e88c0ccd nir/constant_expressions: flush input denorms if denorms have to be flushed 
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39180>
 2026-01-19 16:11:28 +00:00