fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-24 21:28:10 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 1
217090 commits 66 branches 936 tags 996 MiB
Commit graph

19643 commits

Author SHA1 Message Date
Samuel Pitoiset
0cc4e16c70 ac/spm,radv,radeonsi: configure the SPM sample interval in common code 
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38577>
 2025-11-24 08:05:08 +00:00
Samuel Pitoiset
9a61eaa1e3 radv: remove the ability to create NULL devices with RADV_FORCE_FAMILY
Some checks are pending
macOS-CI / macOS-CI (dri) (push) Waiting to run
Details
macOS-CI / macOS-CI (xlib) (push) Waiting to run
Details 
On Linux, drm-shim is the replacement.

On Windows, the project to support a compile-only device has been
abandonned since a while, so it's fine to not allow creating NULL
devices for now.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38544>
 2025-11-24 07:44:49 +00:00
Yogesh Mohan Marimuthu
3ba6c9d0ac winsys/amdgpu: enable userq reg shadowing for gfx11.5
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: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36700>
 2025-11-23 19:44:07 +00:00
Yogesh Mohan Marimuthu
9beb668d8d winsys/amdgpu: fwm packet pre-emption for gfx 11.5 
gfx 11.5 uses f32 firmware. f32 firmware requires COND_EXEC
packet to flush the ring buffer when pre-emption occured.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36700>
 2025-11-23 19:44:06 +00:00
Dave Airlie
3eef0c0245 radv: add support for cooperative matrix per element operations.
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: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36992>
 2025-11-22 13:16:20 +10:00
Samuel Pitoiset
344040c367 radv: enable RADV_THREAD_TRACE_CACHE_COUNTERS on GFX12
Some checks are pending
macOS-CI / macOS-CI (dri) (push) Waiting to run
Details
macOS-CI / macOS-CI (xlib) (push) Waiting to run
Details 
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38489>
 2025-11-21 11:52:58 +00:00
Samuel Pitoiset
473118b6eb ac/spm: use hardware names for performance counters 
Much easier to read.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38489>
 2025-11-21 11:52:58 +00:00
Samuel Pitoiset
4c21a4846c ac/spm: adjust the granularity of SPM results on GFX12 
It's 1, only GFX11-11.5 uses units of segment.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38489>
 2025-11-21 11:52:58 +00:00
Samuel Pitoiset
f434c5c934 ac/spm: add cache counters configuration for GFX12 
This is for the cache counters prior to RGP 2.6.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38489>
 2025-11-21 11:52:58 +00:00
Samuel Pitoiset
da07f1ef3f radv: allocate the SQTT BO in GTT for faster readback 
Reading VRAM from CPU is very slow.

This is similar to the SPM BO, and generating RGP captures is now
way faster.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38551>
 2025-11-21 11:34:09 +00:00
Anna Maniscalco
3e01031f10 radv: consistently use the value in bytes for esgs_itemsize
Some checks are pending
macOS-CI / macOS-CI (dri) (push) Waiting to run
Details
macOS-CI / macOS-CI (xlib) (push) Waiting to run
Details 
Previosuly this value was in bytes for vs/tes and in dwords for gs.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38514>
 2025-11-20 16:45:37 +00:00
Anna Maniscalco
5e8885a339 radv: recalculate legacy_gs_info on bind 
Previously legacy_gs_info calculated based on
gs_info->legacy_gs_info.esgs_itemsize which is calculated based on gs
input varyings.

However, when using ESO vs/tes can have outputs not read by gs, which
leads to underestimating LDS usage.

Cc: mesa-stable
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38514>
 2025-11-20 16:45:37 +00:00
Pierre-Eric Pelloux-Prayer
9e76f5f2a2 radv: enable global BO list if vm_always_valid is supported 
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38529>
 2025-11-20 10:21:47 +00:00
Pierre-Eric Pelloux-Prayer
cf4c55a20f ac/info: get vm_always_valid support through ac_linux_drm 
For virtio it depends on the host support in virglrenderer.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38529>
 2025-11-20 10:21:47 +00:00
Pierre-Eric Pelloux-Prayer
f57993b71d ac/virtio: fix incorrect NULL check 
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38529>
 2025-11-20 10:21:47 +00:00
Pierre-Eric Pelloux-Prayer
51365585e2 ac/virtio: remove dead code 
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38529>
 2025-11-20 10:21:47 +00:00
Samuel Pitoiset
3889695e9f aco/tests: switch to drm-shim 
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38536>
 2025-11-20 09:53:29 +00:00
Samuel Pitoiset
b4121a30df amd/drm-shim: export a function that allows to select a different device 
To be used by ACO tests. Need to remove gnu_symbol_visibility for
exporting the symbol.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38536>
 2025-11-20 09:53:29 +00:00
Samuel Pitoiset
168a8d0b52 radv: fix RB+ for depth-only with unused attachments 
When there are no color outputs in the rendering state, but color write
enable/write aren't masked out (which seems legal with
VK_EXT_dynamic_rendering_unused_attachments), the driver must emit
CB_DISABLE to disable CB rendering completely.

Otherwise, if there is also a depth/stencil attachment in the rendering
state, CB0 is always set to 32_R for RB+. That means, the pixel shader
would still export fragments but to the previously bound color
attachment.

VKCTS is missing coverage.

Fixes: 4580293ab2 ("radv: implement RB+ depth-only rendering for better perf")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/14319
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38509>
 2025-11-20 07:37:17 +00:00
Marek Olšák
9e339f4b32 nir: rename nir_lower_indirect_derefs -> nir_lower_indirect_derefs_to_if_else_trees 
This describes better what it does.

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Acked-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38471>
 2025-11-20 05:42:11 +00:00
Marek Olšák
65837d8289 ac,radeonsi: remove gfx11 FW-based MCBP
Some checks are pending
macOS-CI / macOS-CI (dri) (push) Waiting to run
Details
macOS-CI / macOS-CI (xlib) (push) Waiting to run
Details 
It's too slow to be usable. User queues could replace it.

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38338>
 2025-11-20 03:31:47 +00:00
Georg Lehmann
fa66b670d4 aco/optimizer: reduce max alu_opt_info stack operands to 4
Some checks are pending
macOS-CI / macOS-CI (dri) (push) Waiting to run
Details
macOS-CI / macOS-CI (xlib) (push) Waiting to run
Details 
ALU instructions typically have a maximum of 3 operands, and even when combining
instructions, the peak count will not go above 4.

Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38150>
 2025-11-19 10:51:43 +00:00
Georg Lehmann
4da74eed96 aco/tests: test packed fma opts 
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38150>
 2025-11-19 10:51:43 +00:00
Georg Lehmann
1f0293be0d aco/optimizer: use new helpers for packed fma 
Foz-DB Navi48:
Totals from 374 (0.45% of 82419) affected shaders:
MaxWaves: 5476 -> 5480 (+0.07%)
Instrs: 2786653 -> 2784061 (-0.09%); split: -0.11%, +0.01%
CodeSize: 15163340 -> 15153460 (-0.07%); split: -0.08%, +0.01%
VGPRs: 46884 -> 46860 (-0.05%)
SpillVGPRs: 188 -> 189 (+0.53%)
Scratch: 3207936 -> 3208192 (+0.01%)
Latency: 27352681 -> 27350006 (-0.01%); split: -0.02%, +0.01%
InvThroughput: 5933554 -> 5932632 (-0.02%); split: -0.02%, +0.01%
VClause: 62355 -> 62359 (+0.01%); split: -0.03%, +0.04%
Copies: 290221 -> 289786 (-0.15%); split: -0.21%, +0.06%
Branches: 108566 -> 108569 (+0.00%); split: -0.01%, +0.01%
PreVGPRs: 40172 -> 40157 (-0.04%)
VALU: 1355753 -> 1353329 (-0.18%); split: -0.19%, +0.01%
SALU: 524836 -> 524831 (-0.00%); split: -0.01%, +0.01%
VMEM: 90948 -> 90950 (+0.00%)
VOPD: 10489 -> 10490 (+0.01%); split: +0.98%, -0.97%

Foz-DB Navi21:
Totals from 374 (0.45% of 82387) affected shaders:
MaxWaves: 4339 -> 4348 (+0.21%)
Instrs: 2255741 -> 2253554 (-0.10%); split: -0.10%, +0.00%
CodeSize: 12755276 -> 12744184 (-0.09%); split: -0.09%, +0.01%
VGPRs: 40376 -> 40352 (-0.06%)
Latency: 27357012 -> 27348737 (-0.03%); split: -0.07%, +0.04%
InvThroughput: 7213578 -> 7211136 (-0.03%); split: -0.07%, +0.04%
VClause: 62154 -> 62172 (+0.03%); split: -0.01%, +0.04%
Copies: 268204 -> 268048 (-0.06%); split: -0.22%, +0.16%
Branches: 107067 -> 107066 (-0.00%)
PreVGPRs: 37615 -> 37599 (-0.04%)
VALU: 1423326 -> 1421187 (-0.15%); split: -0.16%, +0.01%
SALU: 383388 -> 383390 (+0.00%); split: -0.00%, +0.00%

Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38150>
 2025-11-19 10:51:43 +00:00
Georg Lehmann
fec10ea3ea aco/optimizer: use new helpers for add16 opts 
Foz-DB Navi48:
Totals from 164 (0.20% of 82419) affected shaders:
Instrs: 145304 -> 145335 (+0.02%); split: -0.00%, +0.02%
CodeSize: 794156 -> 794280 (+0.02%); split: -0.00%, +0.02%
Latency: 1884349 -> 1884227 (-0.01%); split: -0.01%, +0.00%
InvThroughput: 350403 -> 350393 (-0.00%)

Foz-DB Navi21:
Totals from 164 (0.20% of 82387) affected shaders:
Instrs: 117416 -> 117414 (-0.00%)
CodeSize: 673328 -> 673312 (-0.00%)
Latency: 1896952 -> 1897094 (+0.01%); split: -0.00%, +0.01%
InvThroughput: 638536 -> 638556 (+0.00%); split: -0.01%, +0.01%
Copies: 14579 -> 14577 (-0.01%)
VALU: 65895 -> 65893 (-0.00%)

Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38150>
 2025-11-19 10:51:42 +00:00
Georg Lehmann
e8f5b9374b aco/optimizer: use new helpers to optimize mul(b2f(a), b) 
Foz-DB Navi48:
Totals from 979 (1.19% of 82419) affected shaders:
Instrs: 3630560 -> 3629463 (-0.03%); split: -0.03%, +0.00%
CodeSize: 19154176 -> 19147124 (-0.04%); split: -0.04%, +0.00%
Latency: 17700546 -> 17699505 (-0.01%); split: -0.01%, +0.01%
InvThroughput: 3143808 -> 3143254 (-0.02%); split: -0.02%, +0.01%
SClause: 76410 -> 76405 (-0.01%); split: -0.01%, +0.00%
Copies: 256544 -> 256554 (+0.00%); split: -0.02%, +0.02%
PreVGPRs: 40868 -> 40835 (-0.08%)
VALU: 2003291 -> 2002466 (-0.04%); split: -0.04%, +0.00%
SALU: 514000 -> 514006 (+0.00%)
VOPD: 3254 -> 3256 (+0.06%); split: +0.12%, -0.06%

Foz-DB Navi21:
Totals from 926 (1.12% of 82387) affected shaders:
MaxWaves: 21538 -> 21542 (+0.02%)
Instrs: 2984216 -> 2983187 (-0.03%); split: -0.04%, +0.00%
CodeSize: 16104112 -> 16097272 (-0.04%); split: -0.05%, +0.00%
VGPRs: 46864 -> 46848 (-0.03%)
Latency: 15678064 -> 15677099 (-0.01%); split: -0.01%, +0.00%
InvThroughput: 3779550 -> 3778230 (-0.03%); split: -0.04%, +0.01%
VClause: 81590 -> 81598 (+0.01%)
SClause: 70753 -> 70751 (-0.00%); split: -0.01%, +0.00%
Copies: 240446 -> 240466 (+0.01%); split: -0.01%, +0.02%
PreSGPRs: 51121 -> 51062 (-0.12%)
PreVGPRs: 38538 -> 38505 (-0.09%)
VALU: 1978847 -> 1977777 (-0.05%); split: -0.06%, +0.00%
SALU: 439184 -> 439212 (+0.01%)

Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38150>
 2025-11-19 10:51:42 +00:00
Georg Lehmann
f0e24284f5 aco/optimizer: create max3/min3/med3 with salu min/max 
Foz-DB Navi48:
Totals from 175 (0.21% of 82419) affected shaders:
Instrs: 465863 -> 465260 (-0.13%); split: -0.13%, +0.00%
CodeSize: 2362264 -> 2360744 (-0.06%); split: -0.07%, +0.00%
Latency: 1548501 -> 1548371 (-0.01%); split: -0.01%, +0.00%
InvThroughput: 227683 -> 227630 (-0.02%); split: -0.08%, +0.06%
Copies: 33646 -> 33648 (+0.01%)
PreSGPRs: 9996 -> 10004 (+0.08%)
VALU: 175836 -> 175850 (+0.01%)
SALU: 122094 -> 121621 (-0.39%); split: -0.39%, +0.00%

Foz-DB Navi21:
Totals from 1 (0.00% of 82387) affected shaders:
InvThroughput: 74 -> 76 (+2.70%)
VALU: 57 -> 58 (+1.75%)
SALU: 61 -> 60 (-1.64%)

Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38150>
 2025-11-19 10:51:42 +00:00
Georg Lehmann
d21734e024 aco/optimizer: use new helper functions to create med3 
Foz-DB Navi48:
Totals from 9659 (11.72% of 82419) affected shaders:
Instrs: 17301747 -> 17301735 (-0.00%); split: -0.00%, +0.00%
CodeSize: 93378108 -> 93378184 (+0.00%); split: -0.00%, +0.00%
Latency: 145441784 -> 145441791 (+0.00%); split: -0.00%, +0.00%
InvThroughput: 25768777 -> 25768778 (+0.00%)
Copies: 1370123 -> 1370124 (+0.00%)
VALU: 9705655 -> 9705656 (+0.00%)

Foz-DB Navi21:
Totals from 22 (0.03% of 82387) affected shaders:
Instrs: 27433 -> 27406 (-0.10%)
CodeSize: 146440 -> 146352 (-0.06%); split: -0.06%, +0.00%
Latency: 305857 -> 305806 (-0.02%); split: -0.02%, +0.00%
InvThroughput: 63634 -> 63580 (-0.08%)
VALU: 19109 -> 19082 (-0.14%)

Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38150>
 2025-11-19 10:51:42 +00:00
Georg Lehmann
6fc250fc06 aco/optimizer: use new helpers for min3/max3/minmax/maxmin 
Foz-DB Navi48:
Totals from 10453 (12.68% of 82419) affected shaders:
Instrs: 18676282 -> 18675798 (-0.00%); split: -0.00%, +0.00%
CodeSize: 100603268 -> 100603508 (+0.00%); split: -0.00%, +0.00%
Latency: 157036823 -> 157031708 (-0.00%); split: -0.00%, +0.00%
InvThroughput: 28049331 -> 28048776 (-0.00%); split: -0.00%, +0.00%
Copies: 1452464 -> 1452503 (+0.00%); split: -0.00%, +0.00%
PreVGPRs: 458422 -> 458413 (-0.00%); split: -0.00%, +0.00%
VALU: 10429583 -> 10429353 (-0.00%); split: -0.00%, +0.00%
SALU: 2628403 -> 2628416 (+0.00%); split: -0.00%, +0.00%
VOPD: 21738 -> 21744 (+0.03%); split: +0.04%, -0.01%

Foz-DB Navi21:
Totals from 889 (1.08% of 82387) affected shaders:
MaxWaves: 15641 -> 15639 (-0.01%); split: +0.01%, -0.03%
Instrs: 2505527 -> 2505489 (-0.00%); split: -0.01%, +0.01%
CodeSize: 13975300 -> 13976516 (+0.01%); split: -0.00%, +0.01%
VGPRs: 65584 -> 65576 (-0.01%); split: -0.02%, +0.01%
Latency: 37135606 -> 37132577 (-0.01%); split: -0.01%, +0.00%
InvThroughput: 10937032 -> 10935704 (-0.01%); split: -0.01%, +0.00%
VClause: 63136 -> 63140 (+0.01%); split: -0.01%, +0.01%
Copies: 256011 -> 256073 (+0.02%); split: -0.01%, +0.03%
PreSGPRs: 51804 -> 51809 (+0.01%)
PreVGPRs: 57905 -> 57890 (-0.03%); split: -0.03%, +0.00%
VALU: 1593523 -> 1593339 (-0.01%); split: -0.02%, +0.00%
SALU: 425116 -> 425134 (+0.00%); split: -0.00%, +0.01%

Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38150>
 2025-11-19 10:51:42 +00:00
Georg Lehmann
5d02eae052 aco/optimizer: add less agressive pattern matching option 
Still a bit more aggresive than the classic is_used_once,
but it should still prevent most regressions for patterns
that use min/max/mul as outer instruction.

Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38150>
 2025-11-19 10:51:42 +00:00
Georg Lehmann
2c05aa34aa aco/optimizer: create fma with s_mul_f32/f16 
Foz-DB Navi48:
Totals from 14473 (17.56% of 82419) affected shaders:
MaxWaves: 397738 -> 397720 (-0.00%); split: +0.00%, -0.01%
Instrs: 22133626 -> 21984649 (-0.67%); split: -0.68%, +0.01%
CodeSize: 117440104 -> 117111440 (-0.28%); split: -0.30%, +0.02%
VGPRs: 825820 -> 825928 (+0.01%); split: -0.01%, +0.02%
SpillSGPRs: 15496 -> 15512 (+0.10%); split: -0.19%, +0.29%
Latency: 152141755 -> 152058676 (-0.05%); split: -0.07%, +0.02%
InvThroughput: 25715152 -> 25681160 (-0.13%); split: -0.14%, +0.01%
VClause: 402752 -> 400798 (-0.49%); split: -0.53%, +0.04%
SClause: 587448 -> 586772 (-0.12%); split: -0.19%, +0.07%
Copies: 1650891 -> 1661495 (+0.64%); split: -0.14%, +0.78%
Branches: 541341 -> 541334 (-0.00%); split: -0.00%, +0.00%
PreSGPRs: 748235 -> 748332 (+0.01%); split: -0.03%, +0.04%
VALU: 11754090 -> 11755396 (+0.01%); split: -0.01%, +0.02%
SALU: 3659133 -> 3536435 (-3.35%); split: -3.36%, +0.01%
VOPD: 17201 -> 17083 (-0.69%); split: +0.05%, -0.74%

Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38150>
 2025-11-19 10:51:42 +00:00
Georg Lehmann
5abc961514 aco/optimizer: use new helpers to create fma 
Foz-DB Navi48:
Totals from 25949 (31.48% of 82419) affected shaders:
Instrs: 30904250 -> 30904153 (-0.00%); split: -0.00%, +0.00%
CodeSize: 164623100 -> 164604652 (-0.01%); split: -0.01%, +0.00%
Latency: 209402611 -> 209402684 (+0.00%); split: -0.00%, +0.00%
InvThroughput: 36622293 -> 36622236 (-0.00%); split: -0.00%, +0.00%
Copies: 2252080 -> 2251998 (-0.00%); split: -0.00%, +0.00%
VALU: 16831507 -> 16831382 (-0.00%); split: -0.00%, +0.00%
VOPD: 28252 -> 28295 (+0.15%)

Foz-DB Navi21:
Totals from 56269 (68.30% of 82387) affected shaders:
Instrs: 43751754 -> 43746463 (-0.01%); split: -0.01%, +0.00%
CodeSize: 233615096 -> 233576912 (-0.02%); split: -0.02%, +0.00%
VGPRs: 2445528 -> 2445520 (-0.00%)
Latency: 276776920 -> 276761183 (-0.01%); split: -0.01%, +0.00%
InvThroughput: 66406450 -> 66402214 (-0.01%); split: -0.01%, +0.00%
VClause: 902951 -> 902947 (-0.00%)
Copies: 3926260 -> 3926289 (+0.00%); split: -0.01%, +0.01%
VALU: 26924056 -> 26918783 (-0.02%); split: -0.02%, +0.00%
SALU: 6938335 -> 6938321 (-0.00%); split: -0.00%, +0.00%

Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38150>
 2025-11-19 10:51:42 +00:00
Georg Lehmann
1e2aea7461 aco/optimizer: add new helper functions for combining two instructions 
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38150>
 2025-11-19 10:51:42 +00:00
Georg Lehmann
87e168f223 aco/optimizer: make label_mad more generic 
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38150>
 2025-11-19 10:51:42 +00:00
Georg Lehmann
53f5e447db aco/optimizer: add extract_float helper 
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38150>
 2025-11-19 10:51:42 +00:00
Georg Lehmann
7eccf5c745 aco/optimizer: refactor insert 
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38150>
 2025-11-19 10:51:42 +00:00
Samuel Pitoiset
7c9e5b4c1c radv: remove unreachable code for prefetch in radv_cs_emit_cp_dma()
Some checks are pending
macOS-CI / macOS-CI (dri) (push) Waiting to run
Details
macOS-CI / macOS-CI (xlib) (push) Waiting to run
Details 
CP DMA prefetches are implemented with a separate function.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38449>
 2025-11-19 08:03:38 +00:00
Samuel Pitoiset
60d438e517 radv: always use MALL for CP DMA operations on GFX12 
CP DMA isn't coherent with L2 on GFX12, but {SRC,DST}_ADDR_TC_L2 means
MALL.

Only small buffers are using copy/fill CP DMA operations, so this
shouldn't have much effect.

Found by inspection.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38449>
 2025-11-19 08:03:38 +00:00
Samuel Pitoiset
b2a13ce92c radv/tests: require drm-shim and use it instead of RADV_FORCE_FAMILY 
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38507>
 2025-11-19 07:11:05 +00:00
Boris Brezillon
ea4d4d2a77 nir: Prepare nir_lower_io_vars_to_temporaries() for optional PLS lowering 
Rather than adding another boolean to optionally lower PLS vars, pass
the types we want to lowers through a nir_variable_mode bitmask.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37110>
 2025-11-18 20:25:42 +00:00
Natalie Vock
1243d575a5 aco/insert_nops: Consider s_setpc target susceptible to VALUReadSGPRHazard 
Some GPU hangs witnessed in the wild on RDNA4 in Control and Arc Raiders
seem to point towards closest-hit shaders reading a stale value for the
SGPR pair containing the currently-executing shader's address.

This SGPR pair was read by VALU in the preceding traversal shader,
making it susceptible to VALUReadSGPRHazard. Inserting
VALUReadSGPRHazard mitigations before accessing the s_setpc target seems
to fix the hang. We don't have conclusive proof that this is hazardous,
but given that all signs point towards it and we have a reasonably
simple workaround, let's roll with this for now to mitigate the hangs.

Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38290>
 2025-11-18 18:43:00 +00:00
Samuel Pitoiset
9f512d8f93 radv: advertise VK_EXT_custom_resolve
Some checks are pending
macOS-CI / macOS-CI (dri) (push) Waiting to run
Details
macOS-CI / macOS-CI (xlib) (push) Waiting to run
Details 
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38442>
 2025-11-18 17:03:13 +00:00
Samuel Pitoiset
91469bcc30 radv: implement VK_EXT_custom_resolve 
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38442>
 2025-11-18 17:03:13 +00:00
Dave Airlie
ad25196d35 radv: add support for cooperative matrix reductions.
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 add support to the lowering the reduction operations.

Thanks to Georg Lehmann for a lot of the ideas and optimising in
this.

Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38389>
 2025-11-17 23:33:59 +00:00
Georg Lehmann
3a175b54a4 aco,nir: support subdword v_permlane_b16 
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38389>
 2025-11-17 23:33:59 +00:00
Georg Lehmann
018f45f981 aco/insert_NOPs: remove redundant VALUReadSGPRHazard waits 
Mostly removes SALU->VALU waits if the VALU writes a sgpr.

Foz-DB GFX1201:
Totals from 18553 (22.51% of 82419) affected shaders:
Instrs: 27388414 -> 27321118 (-0.25%)
CodeSize: 145389276 -> 145118128 (-0.19%); split: -0.19%, +0.00%
Latency: 200288087 -> 200252583 (-0.02%); split: -0.02%, +0.00%
InvThroughput: 36311237 -> 36307369 (-0.01%); split: -0.01%, +0.00%

Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38445>
 2025-11-17 16:28:36 +00:00
Georg Lehmann
b1d730982e aco/insert_NOPs: remove redundant VALUMaskWriteHazard waits 
This removes a lot of VALU->SALU waits.

Foz-DB Navi31:
Totals from 8908 (10.84% of 82179) affected shaders:
Instrs: 17118986 -> 17084870 (-0.20%)
CodeSize: 91057212 -> 90919300 (-0.15%); split: -0.15%, +0.00%
Latency: 154044128 -> 154036848 (-0.00%); split: -0.00%, +0.00%
InvThroughput: 26608698 -> 26607933 (-0.00%); split: -0.00%, +0.00%

Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38445>
 2025-11-17 16:28:36 +00:00
David Rosca
3abb2707e2 radv/video: Fix coding used_by_curr_pic_lt_flag 
Fixes: d68a1fc0d4 ("radv/video: port hevc slice header encoding from radeonsi")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/14301
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38475>
 2025-11-17 11:51:08 +00:00
Samuel Pitoiset
8d4ba81ca8 radv: remove now unused SDMA helpers 
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38448>
 2025-11-17 11:29:24 +00:00
Samuel Pitoiset
a4e4f13c78 ac,radv: add ac_emit_sdma_copy_t2t_sub_window() 
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38448>
 2025-11-17 11:29:24 +00:00