fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-18 20:18:06 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 2
207076 commits 66 branches 935 tags 952 MiB
Commit graph

6246 commits

Author SHA1 Message Date
Job Noorman
2490ecf5fc ir3: ingest global addresses as 64b values from NIR 
There are currently two places where we have to handle values that are
logically 64b: 64b atomics and 64b global addresses. For the former, we
ingest the values as 64b from NIR, while the latter uses 2x32b values.
This commit makes things more consistent by using 64b NIR values for
global addresses as well.

Of course, we could also go the other way around and use 2x32b values
everywhere, which would make things consistent as well. Given that ir3
doesn't actually have 64b registers, and 64b values are represented by
collected 2x32b registers, this could actually make more sense.

In the end, both methods are mostly equivalent and it probably doesn't
matter too much one way or the other. However, the reason I have a
slight preference for ingesting things as 64b is that it allows us to
use more of the generic NIR intrinsics, which use 1-component values for
64b addresses or atomic values. This commit already makes
global_atomic(_swap)_ir3 obsolete and I'm planning to create generic
intrinsics to support ldg.a/stg.a as well.

Signed-off-by: Job Noorman <jnoorman@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33797>
 2025-06-18 19:06:32 +00:00
Georg Lehmann
e9c886c331 nir/opt_intrinsic: fix inclusive scan rewrite with multiple uses 
Modifying the iterated list is a footgun, so just create a new instruction.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/13364
Fixes: 5c70a55bf3 ("nir/opt_intrinsics: optimize (exclusive_scan(op, a) op a) to inclusive scan")

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35577>
 2025-06-18 18:18:15 +00:00
Rhys Perry
ea0670dfb5 nir: simplify nir_addition_might_overflow 
nir_unsigned_upper_bound is good enough that this isn't needed anymore.

No fossil-db changes.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35514>
 2025-06-17 13:28:00 +00:00
Rhys Perry
f3b7ac730c nir/uub: improve ior/ixor with constant sources 
No fossil-db changes.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35514>
 2025-06-17 13:28:00 +00:00
Rhys Perry
ae6ad8977b nir/uub: improve iand with constant sources 
fossil-db (navi21):
Totals from 9 (0.01% of 79653) affected shaders:
Instrs: 11878 -> 11868 (-0.08%)
CodeSize: 61572 -> 61508 (-0.10%)
Latency: 44585 -> 44581 (-0.01%); split: -0.02%, +0.01%
InvThroughput: 9697 -> 9660 (-0.38%)
VALU: 8889 -> 8876 (-0.15%)
SALU: 1339 -> 1342 (+0.22%)

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35514>
 2025-06-17 13:27:59 +00:00
Rhys Perry
8ee5440073 nir/uub: improve ishl/imul with constant sources 
fossil-db (navi21):
Totals from 1 (0.00% of 79653) affected shaders:
Instrs: 1339 -> 1338 (-0.07%)
CodeSize: 7244 -> 7240 (-0.06%)
Latency: 19827 -> 19822 (-0.03%)
InvThroughput: 9913 -> 9911 (-0.02%)
SALU: 419 -> 418 (-0.24%)

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35514>
 2025-06-17 13:27:59 +00:00
Olivia Lee
88ac602cc2 panvk: implement shaderInputAttachmentArrayNonUniformIndexing 
Signed-off-by: Olivia Lee <olivia.lee@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35408>
 2025-06-13 19:02:19 +00:00
Christian Gmeiner
b30b87c096 nir/inline_uniforms: Convert to use nir_shader_intrinsics_pass(..) 
Signed-off-by: Christian Gmeiner <cgmeiner@igalia.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35463>
 2025-06-12 22:35:48 +02:00
Marek Olšák
fa2e7c3dfd nir: return progress from nir_group_loads, nir_inline_uniforms 
so that NIR_PASS is usable

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35315>
 2025-06-12 19:35:37 +00:00
Marek Olšák
0cbcb72869 nir/opt_vectorize_io: work around a 16-bit IO bug for RADV 
If nir_opt_vectorize_io isn't called, 16-bit IO is broken.
This is a workaround to keep RADV working and consume incorrect NIR
while other drivers consume correct NIR.

Hopefully this will be removed ASAP.

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35315>
 2025-06-12 19:35:37 +00:00
Marek Olšák
b636e5ca66 nir: add nir_clear_mediump_io_flag 
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35315>
 2025-06-12 19:35:36 +00:00
Marek Olšák
13005d5e4e nir/xfb_info: don't merge incompatible XFB outputs to fix mediump 
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35315>
 2025-06-12 19:35:36 +00:00
Marek Olšák
118c0e6991 nir/opt_vectorize_io: fix vectorizing 16-bit XFB 
Tested with mediump.

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35315>
 2025-06-12 19:35:36 +00:00
Marek Olšák
caddd67b8c nir/opt_vectorize_io: don't vectorize 16-bit IO to vec8 - it's illegal 
NIR represents low bits of 16-bit IO as a separate vec4, and high bits as
another separate vec4. There is no representation that allows vec8.

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35315>
 2025-06-12 19:35:35 +00:00
Marek Olšák
1f80ff5550 nir/opt_vectorize_io: convert bool merge_low_high_16_to_32 to an enum 
refactoring for the next commit

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35315>
 2025-06-12 19:35:35 +00:00
Marek Olšák
6270136b7d nir/opt_varyings: set prev_stage/next_stage if they are NONE and validate them 
Doing it here ensures that any linked shader will have the correct values
there.

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35315>
 2025-06-12 19:35:34 +00:00
Marek Olšák
e3d122ed7b nir/opt_varyings: completely exclude mediump from type changes 
It broke mediump XFB, which needs the correct type for the up-conversion.

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35315>
 2025-06-12 19:35:34 +00:00
Marek Olšák
aba7b0831c nir: add shader_info::prev_stage 
When lowering mediump to 16 bits, this will allow drivers to enable
the lowering only for certain pairs of stages, e.g. a driver can lower
mediump for VS->FS, but not GS->FS.

This could also be useful for other things.

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35315>
 2025-06-12 19:35:33 +00:00
Georg Lehmann
ad80b554f4 spirv: use feq for OpIsInf
Some checks are pending
macOS-CI / macOS-CI (dri) (push) Waiting to run
Details
macOS-CI / macOS-CI (xlib) (push) Waiting to run
Details 
This effectively reverts fcca6a83cd because feq was clarified to be ordered
when used with exact and without fast math flags.

It's common for HW to only have free abs for floating point instructions.

Foz-DB Navi21:
Totals from 63 (0.08% of 80065) affected shaders:
Instrs: 337027 -> 336667 (-0.11%); split: -0.12%, +0.02%
CodeSize: 1846752 -> 1845000 (-0.09%); split: -0.13%, +0.03%
Latency: 3401087 -> 3400633 (-0.01%); split: -0.04%, +0.03%
InvThroughput: 847299 -> 845939 (-0.16%); split: -0.19%, +0.03%
VClause: 7693 -> 7694 (+0.01%)
Copies: 45175 -> 45240 (+0.14%); split: -0.12%, +0.27%
PreSGPRs: 3555 -> 3553 (-0.06%)
PreVGPRs: 4565 -> 4564 (-0.02%)
VALU: 225473 -> 225245 (-0.10%); split: -0.13%, +0.03%
SALU: 44735 -> 44625 (-0.25%)

Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35437>
 2025-06-11 18:34:21 +00:00
Marek Olšák
f2c48652da nir: add shader_info::tess::tcs_*outputs_read_by_tes* 
Gather no_varying for AMD.

Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34780>
 2025-06-07 16:29:39 +00:00
Marek Olšák
a59464b6e3 radv,radeonsi: precompute and pass TCS per-vertex output stride via a user SGPR 
It's a stride of 1 output, which isn't 16. It's 16 * num_threads,
aligned to 256.

tcs_offchip_layout has 5 unused bits, so let's use them.

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34780>
 2025-06-07 16:29:39 +00:00
Marek Olšák
534b282573 ac/nir/tess: adjust memory layout of TCS outputs to have aligned store offsets 
There is a comment that explains it.

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34780>
 2025-06-07 16:29:38 +00:00
Mel Henning
d15b5fadbb nir/divergence_analysis: Update LCSSA comment 
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35271>
 2025-06-06 18:15:05 +00:00
Karol Herbst
33fb1eca3e nir/scale_fdiv: handle fp16 fdiv 
Not strictly scaling, but we upcast fo fp32, do the fdiv there and cast
back again.

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34053>
 2025-06-05 13:17:27 +00:00
Mike Blumenkrantz
208450fc57 nir/lower_to_scalar: fix opt_varying with output reads 
no_varying cannot be used to eliminate stores on locations which may
be subsequently read

Fixes: 0058989357 ("nir/lower_io_to_scalar: don't create output stores that have no effect")

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35325>
 2025-06-04 18:21:16 +00:00
Marek Olšák
c3034fa82c amd: replace most u_bit_consecutive* with BITFIELD_MASK/RANGE
Some checks are pending
macOS-CI / macOS-CI (dri) (push) Waiting to run
Details
macOS-CI / macOS-CI (xlib) (push) Waiting to run
Details 
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35346>
 2025-06-04 17:46:38 +00:00
Lionel Landwerlin
978933c015 nir/opt_algebraic: extend lowering for (i|u)bitfield_extract 
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35334>
 2025-06-04 16:28:39 +00:00
Georg Lehmann
1c4070f3e9 nir/opt_if: limit rewrite_uniform_uses iand recursion 
https://github.com/doitsujin/dxvk/issues/4970 has a shader
where unrolled loops caused large iand chains and if we don't
limit this  we won't finish compiling in reasonable time.

Cc: mesa-stable

Acked-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35312>
 2025-06-04 10:49:05 +00:00
Georg Lehmann
eaeaf9554d nir/opt_if: don't replace constant uses with other uniform values 
If constant folding wasn't run, this could replace constant uses with different
constants.

Additional, it could also create worse code for "if (subgroupXor(1) == 1)".

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/13281

Cc: mesa-stable

Acked-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35312>
 2025-06-04 10:49:05 +00:00
Samuel Pitoiset
226b0e28db nir: generalize bitfield insert/extract sizes 
Original patch from Alyssa Rosenzweig

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35209>
 2025-06-04 09:37:53 +00:00
Caio Oliveira
542836afe5 intel: Don't require dpas_intel src2 to match destination 
With upcoming configurations, the number of elements in the src2
slice might not match the destination.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35301>
 2025-06-03 21:31:23 +00:00
Rhys Perry
dd45bf5bce nir/load_store_vectorize: stabilize entry sort 
I think this was unlikely to cause issues, even if the qsort()
implementation is unstable.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35255>
 2025-06-03 09:45:01 +00:00
Rhys Perry
397920c16e nir: fix left shift of negative value in ibfe constant folding 
Fixes "left shift of negative value -128" with parallel_rdp/00f93a9497dfbb3b
and UBSan.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35255>
 2025-06-03 09:45:01 +00:00
Rhys Perry
78aae4b1ba nir: fix signed overflow in pack_half_2x16 constant folding 
Without this cast, the left shift is promoted to 'int'.

Fixes "left shift of 50432 by 16 places cannot be represented in type 'int'"
with horizon_zero_dawn/001064f580f8e3be and UBSan.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35255>
 2025-06-03 09:45:01 +00:00
Rhys Perry
6852538ba0 nir: fix unpack_unorm_2x16/unpack_snorm_2x16 constant folding 
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Backport-to: 25.0
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35255>
 2025-06-03 09:45:01 +00:00
Marek Olšák
bf2ed20eb9 nir: remove unused nir_io_semantics::invariant
Some checks are pending
macOS-CI / macOS-CI (dri) (push) Waiting to run
Details
macOS-CI / macOS-CI (xlib) (push) Waiting to run
Details 
Acked-by: Alyssa on IRC
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35256>
 2025-06-02 23:08:58 +00:00
Marek Olšák
44fcda9631 nir/opt_clip_cull_const: support GS 
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35256>
 2025-06-02 23:08:58 +00:00
Marek Olšák
6677d087c0 nir/xfb_info: add new fields to describe 16-bit XFB better 
for drivers that need this information

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35256>
 2025-06-02 23:08:58 +00:00
Marek Olšák
7b70b419b5 nir: always index SSA defs before printing 
This makes the output more readable.

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35256>
 2025-06-02 23:08:58 +00:00
Marek Olšák
cf94ae8544 nir: change the type of shader_info::patch_* fields to 32 bits 
Patch outputs only use 32 bits.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35256>
 2025-06-02 23:08:58 +00:00
Jesse Natalie
f0dde6ca7f nir_gather_output_deps: Fix incorrect enum in switch 
Cc: mesa-stable
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35247>
 2025-05-30 17:04:18 +00:00
Lionel Landwerlin
f0e18c475b intel: remove GRL/intel-clc 
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35227>
 2025-05-29 20:17:13 +00:00
Samuel Pitoiset
cecf6675be nir/lower_int64: add bitfield_extract lowering 
This will be used by RADV for ACO/LLVM.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35187>
 2025-05-29 08:45:40 +02:00
Alyssa Rosenzweig
d696b19dd0 nir/lower_int64: add bitfield_reverse lowering 
now that we can represent 64-bit bitfield_reverse in NIR, we need a lowering for
it as well.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35198>
 2025-05-28 16:29:30 +00:00
Alyssa Rosenzweig
c3fb0645d8 nir/lower_alu: compact bitcount lowering 
while in the area.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35198>
 2025-05-28 16:29:30 +00:00
Alyssa Rosenzweig
759dc70bde nir: generalize bitfield_reverse bit size 
No reason we can't reverse other bit sizes, we just need to generalize the
constant folding & bit size lowering.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35198>
 2025-05-28 16:29:30 +00:00
Marek Olšák
35c76bc7f7 nir/tcs_info: use range analysis to determine the range of tess levels 
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35195>
 2025-05-28 06:46:56 +00:00
Marek Olšák
24c3f30e4a nir/tcs_info: gather which patch outputs are only read/written by invoc 0 
Tested thoroughly by a shader test.

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35195>
 2025-05-28 06:46:56 +00:00
Marek Olšák
a3632d7d88 nir/tcs_info: gather for all patch outputs whether they're written by all invocs 
This substantially rewrites the pass. It also makes it easier to read.
Tested thoroughly by a shader test.

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35195>
 2025-05-28 06:46:56 +00:00
Lorenzo Rossi
2c0d0bad01 nak: Remove unused intrinsic image_load_raw_nv 
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34975>
 2025-05-28 01:47:19 +00:00