radv: Update navi21 llvm fails

Acked-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25010>

radv: Don't advertise features requiring PS epilogs with LLVM

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25010>

ac/llvm: Use float types for float atomics

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25010>

ac/llvm: Use the correct return type for uadd_carry and usub_borrow

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25010>

ac/llvm: Fix typed loads with 16bit formats

For some reason, LLVM can't handle those. Emit a 32bit load and type
conversion instead,

Fixes: 22ca8c8 ("ac/llvm: Implement typed buffer load intrinsic.")
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25010>

radv/ci: Improve ray tracing skips

I didn't know they were regexes. This also excludes all "1048576" tests.
They build an acceleration structure with 1 primitive 1048576 times
which only warms up the Valve farm and doesn't accomplish anything else.

Acked-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24720>

radv: Perform multiple sorts in parallel

This was the last part that didn't scale with multiple infos. Reducing
the amount of barriers in this case improves DOOM Eternal performance by
50%. (Running with low resolution)

Reviewed-by: Friedrich Vock <friedrich.vock@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24720>

radv/radix_sort: Vendor the radix sort dispatch code

This needs to be done so we can optimize it for occpuancy when building
multiple acceleration structures in parallel. Changes to the original
code:

- Change // to /* */
- clang-format
- Replace vkCmd calls with calls to the driver entrypoints
- Add a light weight info struct
- Use radv_fill_buffer directly

Reviewed-by: Friedrich Vock <friedrich.vock@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24720>

radv: Remove dead radix_sort_vk_get_memory_requirements call

Reviewed-by: Friedrich Vock <friedrich.vock@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24720>

radeonsi/vcn: change max_poc to fixed value for hevc encoder.

problem: max_poc means the number of bits used in poc lsb
         in slice header, and it should not be related to GOP
 size. When large GOP size used, it could generate
 corrupted video, as the POC could not be correctly
 decoded.

solution: use fixed value of max_poc (16) for now.

Cc: mesa-stable
Reviewed-by: Boyuan Zhang <Boyuan.Zhang@amd.com>
Signed-off-by: Ruijing Dong <ruijing.dong@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25214>

ac/spm: initialize and set instance mapping for counters

This configures global, per-SE and per-SA counters with different
indexes. This is still unused because only for the first instance is
used by RADV/RadeonSI, but this will be changed.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25211>

ac/spm: rework how segment muxsel RAM are filled

This is more close to PAL and it will be easier to add GFX11 support
on top of it.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25211>

ac/spm: fix checking if the counter instance is valid

This should be compared against the number of global instances, and
there is also an off-by-one error.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25211>

ac/perfcounter: compute the number of global instances of TCP,SQ,GL1C and GL2C

This will be used by SPM.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25211>

ac/perfcounter: set the number of instances of GL1C to 4

According to PAL there is 4 GL1C quadrants. This will also be used
by SPM.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25211>

ac/gpu_info: init num_cu_per_sh from the kernel

This will be used to configure the number of instances of TCP.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25211>

ac/spm: add SPM block definition for GFX10-GFX10.3

Instead of using magic values.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25211>

ac/spm: remove useless SPM block setting for GFX9 and older GPUs

SPM is only implemented for GFX10+ on RADV/RadeonSI, although it's
technically possible on GFX9 but unused by RGP, so don't care.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25211>

radv,radeonsi: use AC_SPM_SEGMENT_TYPE_xxx instead of magic values

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25211>

radv: enable the PKT3 CAM bit for some SPM register writes

PAL does that.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25211>

radeonsi/vcn: Set H264/HEVC chroma sample location in bitstream

Reviewed-by: Boyuan Zhang <Boyuan.Zhang@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25078>

frontends/va: Parse chroma sample location in H264/HEVC SPS

Reviewed-by: Boyuan Zhang <Boyuan.Zhang@amd.com>
Reviewed-by: Sil Vilerino <sivileri@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25078>

ac/spm: add SPM counters configuration for GFX11

All SQ counters changed to SQ_WGP and the L2 miss changed too.
Sourced from PAL.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25175>

ac/perfcounter: add new SQ_WGP block for GFX11+

According to PAL, these SQ counters can be sampled at WGP granularity.
Some SPM counters captured for RGP are using this GPU block.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25175>

ac,radv,radeonsi: rework SPM counters configuration and share it

This should be easier to add GFX11 support.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25175>

issue_templates/Bug Report: fix outdated URL for GFXReconstruct

The URL moved.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25218>

aco/insert_exec_mask: Simplify WQM handling (2/2)

by calculating WQM requirements on demand.

No fossil-db changes.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25038>

aco/insert_exec_mask: Simplify WQM handling (1/2)

by using p_end_wqm as indicator for when to end WQM mode.

Totals from 10049 (13.12% of 76572) affected shaders: (GFX11)

MaxWaves: 301126 -> 301136 (+0.00%)
Instrs: 7061909 -> 7049272 (-0.18%); split: -0.21%, +0.03%
CodeSize: 37720684 -> 37664244 (-0.15%); split: -0.18%, +0.03%
VGPRs: 357204 -> 357180 (-0.01%); split: -0.13%, +0.12%
Latency: 62757830 -> 62827080 (+0.11%); split: -0.06%, +0.17%
InvThroughput: 8589248 -> 8589963 (+0.01%); split: -0.02%, +0.02%
VClause: 132541 -> 132547 (+0.00%); split: -0.03%, +0.03%
SClause: 322916 -> 322964 (+0.01%); split: -0.04%, +0.05%
Copies: 546446 -> 547657 (+0.22%); split: -0.13%, +0.35%
Branches: 189527 -> 188293 (-0.65%)
PreSGPRs: 332792 -> 332529 (-0.08%); split: -0.08%, +0.00%
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25038>

aco: insert a single p_end_wqm after the last derivative calculation

This new instruction replaces p_wqm.

Totals from 28065 (36.65% of 76572) affected shaders: (GFX11)
MaxWaves: 823922 -> 823952 (+0.00%); split: +0.01%, -0.01%
Instrs: 22221375 -> 22180465 (-0.18%); split: -0.26%, +0.08%
CodeSize: 117310676 -> 117040684 (-0.23%); split: -0.30%, +0.07%
VGPRs: 1183476 -> 1186656 (+0.27%); split: -0.19%, +0.46%
SpillSGPRs: 2305 -> 2302 (-0.13%)
Latency: 176559310 -> 176427793 (-0.07%); split: -0.21%, +0.14%
InvThroughput: 26245204 -> 26195550 (-0.19%); split: -0.26%, +0.07%
VClause: 368058 -> 369460 (+0.38%); split: -0.21%, +0.59%
SClause: 857077 -> 842588 (-1.69%); split: -2.06%, +0.37%
Copies: 1245650 -> 1249434 (+0.30%); split: -0.33%, +0.63%
Branches: 394837 -> 396070 (+0.31%); split: -0.01%, +0.32%
PreSGPRs: 1019139 -> 1019567 (+0.04%); split: -0.02%, +0.06%
PreVGPRs: 925739 -> 931860 (+0.66%); split: -0.00%, +0.66%

Changes are due to scheduling and re-enabling cross-lane optimizations.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25038>

aco: don't insert a copy when emitting p_wqm

Totals from 351 (0.46% of 76572) affected shaders: (GFX11)

Instrs: 709202 -> 709600 (+0.06%); split: -0.02%, +0.08%
CodeSize: 3606364 -> 3608040 (+0.05%); split: -0.01%, +0.06%
Latency: 3589841 -> 3590756 (+0.03%); split: -0.01%, +0.03%
InvThroughput: 463303 -> 463324 (+0.00%)
SClause: 28147 -> 28201 (+0.19%); split: -0.02%, +0.22%
Copies: 43243 -> 43204 (-0.09%); split: -0.24%, +0.15%
PreSGPRs: 21028 -> 21042 (+0.07%)
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25038>

aco: make p_wqm a marker instruction without Operands/Definitions

Totals from 28277 (36.93% of 76572) affected shaders: (GFX11)

MaxWaves: 833930 -> 833898 (-0.00%); split: +0.01%, -0.01%
Instrs: 21366950 -> 21353346 (-0.06%); split: -0.11%, +0.05%
CodeSize: 112855368 -> 112610508 (-0.22%); split: -0.24%, +0.03%
VGPRs: 1157748 -> 1158540 (+0.07%); split: -0.10%, +0.17%
SpillSGPRs: 2465 -> 2463 (-0.08%); split: -0.16%, +0.08%
Latency: 168339886 -> 168383646 (+0.03%); split: -0.10%, +0.12%
InvThroughput: 25164895 -> 25158376 (-0.03%); split: -0.08%, +0.06%
VClause: 347660 -> 346256 (-0.40%); split: -0.55%, +0.15%
SClause: 794460 -> 799521 (+0.64%); split: -0.33%, +0.97%
Copies: 1151908 -> 1148370 (-0.31%); split: -0.54%, +0.23%
Branches: 359447 -> 359437 (-0.00%); split: -0.01%, +0.00%
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25038>

aco: don't optimize cross-lane instructions across p_wqm

We will use p_wqm as a marker in the next step.

Totals from 8846 (11.55% of 76572) affected shaders: (GFX11)

Instrs: 7031274 -> 7072729 (+0.59%); split: -0.02%, +0.60%
CodeSize: 37060272 -> 37355244 (+0.80%); split: -0.01%, +0.80%
VGPRs: 402660 -> 398724 (-0.98%); split: -0.99%, +0.01%
Latency: 62231926 -> 62322311 (+0.15%); split: -0.01%, +0.15%
InvThroughput: 10341361 -> 10392589 (+0.50%); split: -0.00%, +0.50%
VClause: 105344 -> 105368 (+0.02%); split: -0.03%, +0.05%
SClause: 218330 -> 218469 (+0.06%); split: -0.07%, +0.14%
Copies: 378609 -> 377644 (-0.25%); split: -0.42%, +0.17%
Branches: 97218 -> 97207 (-0.01%); split: -0.01%, +0.00%
PreSGPRs: 307654 -> 307644 (-0.00%); split: -0.08%, +0.08%
PreVGPRs: 314744 -> 308650 (-1.94%)
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25038>

aco/insert_exec_mask: set Exact mode after p_discard_if when necessary

Fixes: 5e9df85b1a4504c5b4162e77e139056dc80accc6 ('aco: optimize discard_if when WQM is not needed afterwards')
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25038>

aco: remove fast path in insert_exec_mask's process_instructions

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25038>

aco: append p_logical_end after monolithic RT shaders

Fixes: bdec044c880ef38fad3641d306b9828e3c871003 ('aco: Do not fixup registers if there are no shader calls')
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25038>

ac,radeonsi: move vcn enc av1 default cdf file to common

This can be used by radv.

Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25196>

ac,radeonsi: move vcn enc structs to common

This just moves the header to make it easier to share with radv.

Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25196>

radv: emit relocation for mesh/task shaders

RGP requires shaders to be contiguous in memory.

Cc: mesa-stable
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25144>

radv: set THREAD_TRACE_MARKER_ENABLE for mesh/task draws

PAL does that.

Cc: mesa-stable
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25144>

ac/rgp: use correct API stage string for mesh/task shaders

This allows RGP to report the ISA.

Cc: mesa-stable
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25144>

radv: fix instruction timing on GFX11

GFX11 seems to operate on the last active CU.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25171>

radv: disable unsupported hw shader stages for RGP on GFX11+

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25171>

radv: set THREAD_TRACE_TOKEN_MASK.BOP_EVENTS_TOKEN_INCLUDE on GFX10.3+

PAL does that.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25171>

ci: Only look at file changes for MRs

Lists of files changed are only really relevant if we're pushing an MR.
We don't want to trigger 'user added .ci-farms-disabled/igalia' because
the user pushed a branch when this file was already present upstream.

Signed-off-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25205>

ci: Use container rules for containers

If we have diverging rulesets, we can have the base container jobs
possibly not being created, and the leaf container jobs then trying to
run, which ends up in us failing to create pipelines.

Signed-off-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25205>

ci: Always set user container jobs to manual

Don't require source changes to create the jobs, just create them
always.

Signed-off-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25205>

ci: Move farm-disable rules before anything else

For consistency, make our 'when: never' rules be the first rules to
match any job, either build or container.

Otherwise we could end up with some jobs having when: never and others
having when: manual.

Signed-off-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25205>

iris: modify Wa_14014414195 to use intel_needs_workaround

Only applies to DG2, this was fixed in MTL.

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Rohan Garg <rohan.garg@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25197>

radv/video: take db alignment into account when allocating images.

Make sure to take the db alignment into account when sizing the underlying
images.

Fixes a 360p sample from Lynne.

Cc: mesa-stable
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Lynne <dev@lynne.ee>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25168>

meson: Warn about side-effects from DRM for FD KMDs

Any build which requires Turnip to link to libdrm will cause a
runtime dependency which could be a compatibility hazard on
platforms such as Android which generally don't ship with libdrm.
This hazard could be unintentionally triggered and to avoid this
from happening, a warning is now issued when building FD with msm
alongside other backends or if DRM is enabled due to building a
Gallium driver using DRM.

Signed-off-by: Mark Collins <mark@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25005>

freedreno+meson: Add lua+libarchive+libxml from Meson WrapDB

These dependencies are required for building freedreno tools but on
platforms such as Android, these aren't included and would need to
be built from source. The easiest way to do this was using the
Meson WrapDB. Additionally, warnings were added when these deps are
missing as the decode tools will silently not be built otherwise.

Signed-off-by: Mark Collins <mark@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25005>

freedreno/decode: Support building replay for multiple KMDs

As `freedreno_kmds` can contain both `msm` and `kgsl`, this case
would just produce a build with just KGSL due to the precedence of
the check which is inconsistent behavior. Now `replay` is built for
all active KMDs with a suffix for the KMD (eg: replay-kgsl/replay-msm).

Signed-off-by: Mark Collins <mark@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25005>

freedreno: Only add drm/computerator when system_has_kms_drm

These are both MSM-DRM specific subprojects which depend on it and
cannot be built without, this change is required to allow for KGSL
freedreno tool builds to compile successfully.

Signed-off-by: Mark Collins <mark@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25005>

freedreno/rnn: Remove %n usage in fprintf

fprintf was utilized with %n to count the amount of characters it
outputted but this was redundant since it was at the end of the
statement and fprintf itself returns the amount of characters it
writes out.

It should be noted that %n is a compatibility hazard on Android as
it isn't supported on Android due to security reasons.

Signed-off-by: Mark Collins <mark@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25005>

freedreno/replay: Add "gpu_print" function for command streams

Allows to dump GPU memory, e.g.
 gpu_print(&ctx, cs, 0x4000086080, 4);

would print 4 dowrds from 0x4000086080:
 CP Log [0]:
         20002000 47340000 20012001 47340001

Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25005>

freedreno/replay: Add "print" instr to ir3 asm to be used in replay

Could be used in shader assembly from rddecompiler when overriding
the cmdstream.

Usage:
 print r1.x, r3.z

 First arguments is three consecutive registers used for internal print
 needs.
 Second arg is register to print.

Output:
 Shader Log Entries: 36
 [0] 00000004 0.0000
 [1] 00000000 0.0000
 [2] 00000000 0.0000
 [3] 00000000 0.0000
 [4] beadc429 -0.3394
 [5] beadc429 -0.3394

Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25005>

freedreno/rddecompiler: Update to handle a7xx

Also make rdcompiler compilable for Android.

Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25005>

freedreno/replay: Add limited support for KGSL

KGSL doesn't support IOVAs being set form userspace, so as a workaround
we pre-allocate a single buffer and hope that it always starts from the
same address.

Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25005>

anv: Remove i915_drm.h include from common code

With the previous commit we are now able to build Anv without
including i915_drm.h from common code.

This is important as avoids that i915 specific code is included in
common code by mistake.

Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Sagar Ghuge <sagar.ghuge@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25044>

anv: Move i915 handling of imported bos bo_flags

Imported bos already imported need special handling in i915.
That handling was moved to
anv_i915_gem_import_bo_alloc_flags_to_bo_flags() as the number of
imported bos is low.

Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Sagar Ghuge <sagar.ghuge@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25044>

anv: Move bo_alloc_flags_to_bo_flags() to backend

The bo_flags are i915 specific and should not be handled in common
code, so here adding it to backend as it is in the hot-path.

There still i915 bo_flags handling in anv_device_import_bo() that
will be handled in the next patch.

Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Sagar Ghuge <sagar.ghuge@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25044>

anv: Move i915 specific code from common anv_gem.c

Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Sagar Ghuge <sagar.ghuge@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25212>

anv: Move i915 specific gem_set_caching to backend

Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Sagar Ghuge <sagar.ghuge@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25212>

vk/enum2str: add more max enum vendors

Acked-by: Iván Briano <ivan.briano@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25210>

zink: Change zink_vertex_elements_hw_state::b.strides to VkDeviceSize

Currently zink_vertex_elements_hw_state::b.strides is an array of 32-bit
unsigned values, while CmdBindVertexBuffers2EXT expects an array of
VkDeviceSize (64-bit unsigned). Instead of converting them for a call,
just change the type of the array itself.

Fixes: 76725452 ("gallium: move vertex stride to CSO")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/9765
Signed-off-by: Sviatoslav Peleshko <sviatoslav.peleshko@globallogic.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25208>

aux/tc: fix renderpass tracking fb state clobber scenario

in a stream like:
* set fb state (A)
* flush
* set fb state (B)
* draw -> driver query
* flush

the "driver query" should return the tc info corresponding to the most
recent fb state (B). previously this would increment to C because
the flag for incrementing at the start of a batch was set

Fixes: 07017aa137b ("util/tc: implement renderpass tracking")

Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25206>

intel: Sync xe_drm.h

Sync xe_drm.h with commit e51e857ffad4 ("drm/xe/uapi: Remove useless max_page_size").
Most relevant changes are the removal of max_page_size from
drm_xe_query_mem_region and the typo fix in XE_QUERY_CONFIG_MIN_ALIGNMENT.

Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Ivan Briano <ivan.briano@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25162>

ci/zink: Add another conversion fail on a618

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25207>

ci/radeonsi: Skip more really slow tests

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25207>

freedreno: Include speedbin fallback in 740 chipid to fix probing

The kernel exposes CHIP_ID, which consists of:

[63:48] const 0 / reserved
[47:32] speedbin (default 0xffff)
[31:0 ] actual gpu chip ID

Fix the 740 chip id to make it probe correctly.
Keep the existing entry for the downstream KGSL driver.

Fixes: cd3719be7d09 ("freedreno: Fully define a730 and a740 device properties")
Reported-by: Neil Armstrong <neil.armstrong@linaro.org>
Signed-off-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25204>

aux/tc: Add ASSERTED to unreferenced release build variable

Fixes MSVC build error src/gallium/auxiliary/util/u_threaded_context.c(3184): error C4189: 'size': local variable is initialized but not referenced

Fixes: 51ad269198e ("aux/tc: handle stride mismatch during rp-optimized subdata")

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25203>

pvr: Use correct index when writing query availability data

The availability data would be written to a different location in
the user provided buffer depending on whether the query for a given
index was available. Fix this by using fixed indicies when writing
the query and availability data.

Fixes conformance failures seen in the
dEQP-VK.query_pool.occlusion_query.get_reset_* test group when
implementing VK_EXT_host_query_reset.

Fixes: 279c7c6d5ad ("pvr: Implement vkGetQueryPoolResults API.")
Signed-off-by: Vlad Schiller <vlad-radu.schiller@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25116>

panfrost: delete stale editorconfig file

We no longer use 8-space indent in panfrost, so let's remove the
needless .editorconfig, and instead inherit the 3-space indent from the
toplevel .editorconfig.

Fixes: 0afd691f296 ("panfrost: clang-format the tree")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25147>

zink: don't warn about missing scalarBlockLayout on v3dv

technically v3dv doesn't support the "full" feature, but it does support
enough of it to run zink without any issues

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25201>

ci/windows: drop build rules from test jobs

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25183>

ci/windows: add windows docker runner tags to .windows-docker-vs2019

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25183>

ci/windows: centralize definition of windows runners tags

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25183>

zink/ci: update list of expected failures for POLARIS10/NAVI10

Fixed since piglit has been updated to
f7db20b03de6896d013826c0a731bc4417c1a5a0.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25199>

ci: don't run everything just because a farm gets re-enabled

Let the other rules take care of this, as they did before
7cf13ea504e7e5be097e.

Fixes: 7cf13ea504e7e5be097e ("ci: skip containers & build jobs when disabling a farm")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25151>

v3dv: Enable VK API v1.2 for Android

Now as all features are in place, we can enable it.

Signed-off-by: Roman Stratiienko <r.stratiienko@gmail.com>
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14195>

v3dv/android: Add AHardwareBuffer support

The patch allows enabling Vulkan-based UI (SKIA, HWUI) on Android.
To enable - add 'TARGET_USES_VULKAN := true' into your device.mk file.

Passes:
 - dEQP-VK.api.external.memory.android_hardware_buffer.*
 - android.graphics.cts.BasicVulkanGpuTest
 - android.graphics.cts.MediaVulkanGpuTest

Camera preview window works.

Signed-off-by: Roman Stratiienko <roman.o.stratiienko@globallogic.com>
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14195>

v3dv: Split v3dv_image_init to use layout setting logic separately

In AHB importing flow, the format is known at vkCreateImage step,
but buffer layout itself is not yet known. Buffer layout and modifier
must be updated later in vkBindImageMemory.

Signed-off-by: Roman Stratiienko <r.stratiienko@gmail.com>
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14195>

v3dv: Use format stored in vk_image and vk_image_view after init

Both vk_image_init and vk_image_view_init are smarter about format
settings and respect AHB extension needs.

Signed-off-by: Roman Stratiienko <r.stratiienko@gmail.com>
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14195>

v3dv/android: Rework Android native buffer importing logic

Rework it to use the existing VkImageDrmFormatModifierExplicitCreateInfoEXT
logic to set the image's explicit layout.

The code turned out to be generic enough so it could be integrated into
vk_image.c in the future and removed from the v3dv code.

Signed-off-by: Roman Stratiienko <r.stratiienko@gmail.com>
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14195>

v3dv/android: Add a helper function to support explicit layouts

The function extracts buffer information from the gralloc and fills
the VkImageDrmFormatModifierExplicitCreateInfoEXT structure.

Signed-off-by: Roman Stratiienko <r.stratiienko@gmail.com>
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14195>

compiler/types: Move builtin type initialization to C

While both clang and gcc can handle designated initializers in C++,
MSVC only does with the C++20 support enabled.  So move the initialization
of builtins to a C file.

Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25006>

compiler/types: Add workaround to use builtin_type_macros.h in C

In most versions of C, bool is defined as _Bool, so the macro for bool
gets generated with the wrong name.

Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25006>

compiler/types: Make struct glsl_type visible to C code

Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25006>

compiler/types: Use a linear (arena) allocator for glsl_types

They are only deallocated in bulk all at once, which is exactly
the use case for the linear allocator.

Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25006>

compiler/types: Store builtin types directly as data

Remove constructors from glsl_type so it can be used as a
POD ("plain old data") struct, allowing the builtins to be
initialized directly in memory.

For other types, we now allocate them from glsl_type_cache's mem_ctx,
instead of using the global allocator.

As a side-effect of how the new helpers work, we can completely
create the mock key types for struct/interface lookup without
allocating any memory.

Note there's no `make_sampler_type` since all the sampler types
are created through direct initialization.

Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25006>

compiler/types: Move static asserts about glsl_type to a central place

Take it out of the way to reduce noise when reworking (and eventually
removing) the constructor code.

Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25006>

compiler/types: Simplify clearing the glsl_type_cache

Since now all the data referenced by it is allocated with the cache's
mem_ctx, it is sufficient to just free it, and then reset the cache
state to be ready for a next initialization if it happens.

Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25006>

compiler/types: Don't store a mem_ctx per type

These are used only by types created at runtime.  Since those will follow
the lifetime of the glsl_type_cache, we can use its mem_ctx for all the types.

Without a mem_ctx, there's nothing to be done in the destructor, so remove it.

Note some keys are calculated by building a mock type, so we need to create
a tmp_ctx in some cases.  We'll get rid of them in a later commit.

Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25006>

compiler/types: Use type cache mem_ctx for hash tables

Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25006>

compiler/types: Add a mem_ctx for the glsl_type_cache

Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25006>

compiler/types: Move local cache details to implementation file

Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25006>

compiler/types: Use designated initializer syntax to specify builtins

For now we use a temporary glsl_type_params struct, we will be able to
use the glsl_type directly once we make it a POD ("plain old data")
struct by getting rid of its constructors and destructors.

Note that since the name is statically allocated, there's no need to
strdup() it, deallocate it and also no need to have a mem_ctx.

Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25006>

compiler/types: Add extra level of macro to builtin_macros

This will make easier to transition from the macro calling a constructor
to the macro using designated initializers for a struct.

Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25006>

glsl: Don't create struct type builtins

Unlike for simpler types, struct types have a runtime cache, that's used to
ensure same type can be compared to same pointer.  The existing code was bypassing
it, potentially breaking that invariant.  One potential issue would be when
decoding/encoding types, the resulting type would be pointer-different than what
was stored.

This hasn't caused a visible issue, but the (incomplete) special handling for struct
builtins is in the way of other changes.

Change the code to use get_struct_instance(), and also only ever load those if the
parser need the types, since some of them are deprecated types that we might never
want to load.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25006>

glsl: Add missing glsl_types initialization to test_optpass

Will be needed when the builtin struct types will go through regular
initialization.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25006>