util: add mapping from Vulkan to Gallium R64 integer formats

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7234>

gallium: Fix NIR validation when lowering polygon stipple

The fmul operation takes the maximum number of components from either
of its operands. We only need to use 2 components from the fragment
coordinates.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7507>

gallium/util: do not pass undefined sample-count

We forgot to initialize the sample_count member here, leading to it
being undefined. This causes problems on MSVC when compiling in
debug-mode, where we get a run-time error for using an undefined
variable.

To avoid similar problems in the future if more fields are added,
let's initialize the whole struct to zero to start with. This also
allows us to remove a no-longer-needed zero-initialization.

Fixes: cf170616daa ("gallium: Add a util_blitter path for using a custom VS and FS.")
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7503>

aco: optimize v_and(a, v_subbrev_co(0, 0, vcc)) -> v_cndmask(0, a, vcc)

fossils-db (Vega10):
Totals from 7786 (5.70% of 136546) affected shaders:
SGPRs: 517778 -> 518626 (+0.16%); split: -0.01%, +0.17%
VGPRs: 488252 -> 488084 (-0.03%); split: -0.04%, +0.01%
CodeSize: 42282068 -> 42250152 (-0.08%); split: -0.16%, +0.09%
MaxWaves: 35697 -> 35716 (+0.05%); split: +0.06%, -0.01%
Instrs: 8319309 -> 8304792 (-0.17%); split: -0.18%, +0.00%
Cycles: 88619440 -> 88489636 (-0.15%); split: -0.16%, +0.01%
VMEM: 2788278 -> 2780431 (-0.28%); split: +0.06%, -0.35%
SMEM: 570364 -> 569370 (-0.17%); split: +0.12%, -0.30%
VClause: 144906 -> 144908 (+0.00%); split: -0.05%, +0.05%
SClause: 302143 -> 302055 (-0.03%); split: -0.04%, +0.01%
Copies: 579124 -> 578779 (-0.06%); split: -0.14%, +0.08%
PreSGPRs: 327695 -> 328845 (+0.35%); split: -0.00%, +0.35%
PreVGPRs: 434280 -> 433954 (-0.08%)

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7438>

spirv: Add support for SPV_EXT_shader_image_atomic_int64

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7509>

nir: Allow 64-bit image atomics

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7509>

compiler/types: Add 64-bit image types

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7509>

util,gallium: Add new 64-bit integer formats

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7509>

nir: Validate image atomic formats

GLSL requires that image atomics have formats and there are rules about
things matching properly.  We should enforce those in NIR unless we have
reason to do otherwise.

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7509>

nir: Print formats on image intrinsics as text

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7509>

spirv: Update headers and metadata from latest Khronos commit

This corresponds to 5ab5c96198f30804a6a29961b8905f292a8ae600
("Reserve additional loop control bit for Intel extension (NoFusionINTEL) (#175)") in
https://github.com/KhronosGroup/SPIRV-Headers.

Acked-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7509>

libgl-gdi: support building without softpipe

While we do need *some* fallback-driver, there's no good reason to
*always* require that to be Softpipe. LLVMpipe for instance does the job
just fine.

This makes the minimal build a bit smaller on Windows.

Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7506>

panfrost: Add missing Collabora copyright notices

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Acked-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7502>

pan/mdg: Add missing Collabora copyright notices

On a few of the older files.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Acked-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7502>

spirv: correct sematic-typo

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7504>

CI: Disable Panfrost T760

For some unknown reason, both RK3288 Chromebooks went catatonic over the
weekend. Disable them until we can get them fixed.

Signed-off-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7508>

radeon/vcn: Bitrate not updated when changing framerate

Issue: Encoding parameters not updated after changing FrameRate

Root Cause:
In radeon_enc_begin_frame, there is a parameter need_rate_control
which was enabled only if the bitrate is changed. Due to this the
radeon_enc_rc_layer_init was not updating the encoder parameters with new
framerate, peak_bits_per_picture_integer and avg_target_bits_per_picture

Fix:
Added the condition where we will check if there is a change in
other parameters and enable rate control. Eventually updating the
encoder parameters with new framerate and bitrate.

Signed-off-by: Krunal Patel <krunalkumarmukeshkumar.patel@amd.corp-partner.google.com>
Reviewed-by: Boyuan Zhang boyuan.zhang@amd.com
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7363>

v3dv/device: do nothing when asked physical device pci bus properties

When calling GetPhysicalDeviceProperties2 we were ignoring and logging
the structures for extensions not supported. But for the case of
VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PCI_BUS_INFO_PROPERTIES_EXT we
already know that we are not going to support it, so let's just do
nothing (not even logging) when passed.

Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7497>

v3dv/util: log debug ignored stype only on debug builds

Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7497>

v3dv/util: remove several logging functions

We already have vk_error to report errors, they add little specific
v3dv wrapping over a simple fprintf, and they are not used really
often.

Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7497>

radv: Fix exporting/importing multisample images.

Otherwise FMASK metadata segfaults and on import we disable it ...

CC: mesa-stable
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7358>

zink: add some spirv builder functions for barriers

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7486>

zink: add a quadop function in spirv_builder

this takes 4 operands like the unop/binop/triop functions we already have

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7486>

zink: use same function for all pipe_context::delete_*_state shader methods

we don't currently do anything special with any of these, so it'd be better
to not have the same code copy/pasted around

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7485>

zink: make physical device functions use a dynamic function pointers.

vkGetPhysicalDeviceFeatures2 and vkGetPhysicalDeviceProperties2 are not present on some MoltenVK versions.
VK_KHR_get_physical_device_properties2 exposes the KHR versions of the same functions.
These cannot be used via static linking, so we have to dynamically detect the loader version and then the extension
to work out which pointers to use.

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7447>

radv: Fix budget calculations with large BAR.

If we don't have a non-visible VRAM heap, we should be counting
our non-visible VRAM allocations to the visible-VRAM heap.

CC: mesa-stable
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6827>

radv: Skip tiny non-visible VRAM heap.

When I enable "Above 4G decoding" in my BIOS I still get 16 MiB of
non-visible VRAM on my 8G VRAM GPU ...

CC: mesa-stable
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6827>

docs: update calendar and link releases notes for 20.2.2

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7493>

dcs: Add sha256 sums for 20.2.2

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7493>

docs: add release notes for 20.2.2

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7493>

st/nine: Remove unnecessary NULL check.

resource cannot be NULL at this point since it has already been
dereferenced earlier.

Fix defect reported by Coverity Scan.

Dereference before null check (REVERSE_INULL)
check_after_deref: Null-checking resource suggests that it may be
null, but it has already been dereferenced on all paths leading to
the check.

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Axel Davy <davyaxel0@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7343>

zink: implement ARB_texture_query_lod

just needed hooking up the spirv function to the tex op

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7484>

ci/deqp: Switch to a new dEQP runner written in Rust.

I found the C++ runner hard to develop on, and we had stability issues and
outstanding feature needs that made me want something I felt good about
hacking on.  Thus, Rewrite It In Rust of the deqp runner.

The new runner includes:

- Skip lists don't reshuffle the test list.
- Known-flake handling without resorting to skip lists (fixing our main CI
  reliability issue on a3xx right now).
- Per-thread Vulkan shader caches should speed up VK CI runtime.
- Tracking of crashes separate from fails (so we can see progress on that
  front).
- Logging of deqp stderr spam (particularly assertion failures!) in the CI
  log.
- Integrated QPA filtering so we don't have bash perf issues for it.
- Logging of what caselist to go look at for a given error report (in red,
  so it's easier to find in your CI log).
- The code is 1/3 unit tests, and easy to extend for more coverage.
- Non-LAVA CI runs create a failures.csv in artifacts that you can check
  in as your deqp-*-fails.txt file.
- Test runtime is included in results.csv so you can debug how to speed up
  your CI job.
- Pretty summary at the end of the run of slow/flaky/failed tests.

Since this is a new runner with a different RNG, the test groups are
shuffled one more time.  This seems to result in some panfrost T720
stability issues (See its new deqp-panfrost-t720-flakes.txt), and one new
flake in freedreno a630.

Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7434>

ci/bare-metal: Reset colors at the end of a line of serial output.

We don't want the next line of our timestamp and other context to inherit
colors set by the serial command (visible with the new dEQP runner)

Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7434>

ci/bare-metal: Apply autopep8 to the bare-metal scripts.

Let's follow proper python formatting (easy now that vscode does it for
me)

Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7434>

turnip: Fix image size for 3D vkGetImageSubresourceLayout.

Fixes most subcases of dEQP-VK.image.subresource_layout.3d.* The remaining
failures appear to be in snorm, which 2D also fails on (and the blob
reports as not supported for this test).

We don't currently have these tests in CI, but they'll appear with
1.2.4.0.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7467>

spirv: fix GLSLstd450Modf/GLSLstd450Frexp when the destination is vector

We can't write to an individual component in a function_temp vector, so we
have to use vtn_variable_store() which does a load+insert+store.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3484
Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6231>

zink: For MoltenVk added vkFlushMappedMemoryRanges() to vkMapMemory() to fix empty mapped memory.

On MoltenVK/MacOS when mapping memory that should already have content it does not appear until flushed.
This noticably effects vertex attribute uploads to descrete devices.
Did also try to add the Coherent memory flag, which did work, until there the Coherent type could only be used for transfer usage only.

This is a known limitation of MoltenVK.
See https://github.com/KhronosGroup/MoltenVK/blob/master/Docs/MoltenVK_Runtime_UserGuide.md#known-moltenvk-limitations

Seen when using MoltenVK 1.0.121, 1.2.131

Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7470>

zink: have_triangle_fans support.

MoltenVK, at least upto 1.2.141, does not render triangle fans. This is reflected in the portability EXTX extension.
This code get the extensions properties and features and then sets the have_triangle_fans.
This extension is not avaiable on all systems, so an amout of the code has to be protected by the define VK_EXTX_PORTABILITY_SUBSET_EXTENSION_NAME.

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7457>

zink: add support to device info for macro guards and just VkPhysicalDevice*Features with out the have_.

Extends the flexability of the device info script.
Allows #if defined()/#endif guards around particular data, for platform or version specific structures.
Allows for feature structures to be retreved without having to have a have_ variable as well. Helps if there is more than one feature in the structure.

These changes help towards allowing the use of the portability set extensions.

Reviewed-by: Hoe Hao Cheng <haochengho12907@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7457>

nir/lcssa: consider loops with no back-edge invariant

Polaris:
Totals from 6233 (4.52% of 138014) affected shaders:
SpillSGPRs: 47860 -> 48976 (+2.33%)
CodeSize: 69764704 -> 69120700 (-0.92%); split: -0.97%, +0.04%
Instrs: 13801184 -> 13594107 (-1.50%)
Cycles: 1628800928 -> 1516137888 (-6.92%)
VMEM: 910459 -> 910208 (-0.03%); split: +0.00%, -0.03%
SMEM: 436625 -> 435194 (-0.33%); split: +0.06%, -0.38%
SClause: 534750 -> 534620 (-0.02%); split: -0.03%, +0.00%
Copies: 1587121 -> 1542867 (-2.79%); split: -2.81%, +0.03%
Branches: 545016 -> 509354 (-6.54%)
PreSGPRs: 618545 -> 619354 (+0.13%); split: -0.09%, +0.22%

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5924>

zink: Added inbuilt debug logging from the VK_LAYER_LUNARG_standard_validation layer.

Set the ZINK_DEBUG environment variable to 'validation' to automatically setup.
The debug util extnsion callback is used to capture information and logs the results to the error stream.

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7431>

zink: fix pNext chain for resource memory allocation

This only links the VkImportMemoryFdInfoKHR struct into the pNext
chain when VkExportMemoryAllocateInfo is also included, which was
the original behavior before the regression.

Fixes: 92022f2846e00 (zink: add VK_STRUCTURE_TYPE_WSI_MEMORY_ALLOCATE_INFO_MESA for WSI allocations)
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3755
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7474>

zink: return fail if create_instance fails

If create_instance() fails, returns to MESA as failed instead of contining in an bad state.

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7471>

docs/features.txt: VK_EXT_separate_stencil_usage not exposed on RADV

Signed-off-by: Anthoine Bourgeois <anthoine.bourgeois@gmail.com>
Reported-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Fixes: 3d58ab7576d ("docs/features: Minor update extensions support")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7466>

gallium: document convention for get_handle calls on multi-planar resources

There are resources that may have more planes than chained resources. The
frontend has no way of figuring out which (if any) chained resource is the
right one to call resource_get_handle with and until a (now reverted)
change to the dri frontend it just always called with the first resource.

The convention of calling with the first resource of a chain allows the
pipe driver, which has the necessary information of how resources and
planes map to each other for a specific format/modifier combination, to do
the necessary walking. Document this as the official calling convention
of this function.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7419>

etnaviv: rework ZSA into a derived state

The ZSA state is not fully self contained, as other states (mostly
shader using discard or writing depth information) have an influence
on whether we can use early Z test/write.

Rework the ZSA state into a derived state that gets updated whenever
a new ZSA or SHADER state is bound. This way we can automatically
enable/disable early Z as needed.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7396>

etnaviv: expose shader discard usage in etna_shader_variant

The information about a shader using discard/kill is interesting
to other parts of the driver, as depth states need to programmed
differently depending on this. As we don't want to deal with
NIR/TGSI differences in other parts of the driver, track this
usage in the common etna_shader_variant.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7396>

etnaviv: update headers from rnndb

Update to etna_viv commit c8ba5e0ba5da.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7396>

etnaviv: flush depth cache when changing depth config

Some depth config states changes require the depth cache to be
flushed, leading to a GPU hang if not done. As the conditions that
require the flush are not toally clear, better be safe than sorry
and always flush the cache on depth state changes.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7396>

etnaviv: emit RA_EARLY_DEPTH on dirty ZSA

The RA_EARLY_DEPTH is a depth state and so must be emitted on
dirty ZSA, instead of dirty SHADER.

Fixes: 785e2707b0d1 (etnaviv: Fix disabling early-z rejection on GC7000L)
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7396>

mesa/st: use a lock to protect access to variants when updating them

Multiple threads may access st_update_* function at same time. Issues
happen when the threads modify lists managed by shader compiler.

Issues were found with script that runs multithread tests 1000 times in
a row with MESA_GLSL_CACHE_DISABLE=1 set. Problems start when 2
simultaneous st_create_[vp|fp]_variant calls start to compile a new
shader variant for the same program and various nir passes use and
modify same exec_lists.

Example failure:
   deqp-egl: ../src/compiler/glsl/list.h:575: exec_list_validate: Assertion `node->next->prev == node' failed.

v2: instead of introducing new mutex, lock shared state

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7418>

nir/algebraic: distribute imul(iadd(a, b), c) when b and c are constants

This distributes imul(iadd(a, b), c) to iadd(imul(a, c), b * c)
when both b and c are constants. This might allow some compiler
backends to create more MADs.

For ACO, this allows to combine more DS additions.

fossilds-db (Vega10):
Totals from 673 (0.49% of 136546) affected shaders:
VGPRs: 44548 -> 44516 (-0.07%); split: -0.11%, +0.04%
CodeSize: 8301552 -> 8286220 (-0.18%); split: -0.19%, +0.01%
MaxWaves: 2731 -> 2735 (+0.15%); split: +0.26%, -0.11%
Instrs: 1642684 -> 1638725 (-0.24%); split: -0.24%, +0.00%
Cycles: 20846156 -> 20793444 (-0.25%); split: -0.25%, +0.00%
VMEM: 108870 -> 108106 (-0.70%); split: +0.03%, -0.73%
SMEM: 35718 -> 35674 (-0.12%); split: +0.22%, -0.34%
VClause: 20603 -> 20622 (+0.09%); split: -0.01%, +0.10%
SClause: 48527 -> 48539 (+0.02%)
Copies: 156735 -> 156742 (+0.00%); split: -0.05%, +0.05%
PreSGPRs: 43169 -> 43166 (-0.01%); split: -0.02%, +0.02%
PreVGPRs: 41369 -> 41330 (-0.09%)

shader-db results on Intel:
Ice Lake
total instructions in shared programs: 20027588 -> 20027446 (<.01%)
instructions in affected programs: 71766 -> 71624 (-0.20%)
helped: 70
HURT: 0
helped stats (abs) min: 1 max: 7 x̄: 2.03 x̃: 1
helped stats (rel) min: 0.10% max: 2.50% x̄: 0.29% x̃: 0.15%
95% mean confidence interval for instructions value: -2.42 -1.64
95% mean confidence interval for instructions %-change: -0.38% -0.20%
Instructions are helped.

total cycles in shared programs: 977525222 -> 977494323 (<.01%)
cycles in affected programs: 8884593 -> 8853694 (-0.35%)
helped: 56
HURT: 16
helped stats (abs) min: 2 max: 7852 x̄: 681.29 x̃: 400
helped stats (rel) min: <.01% max: 19.84% x̄: 2.79% x̃: 0.41%
HURT stats (abs)   min: 2 max: 1212 x̄: 453.31 x̃: 120
HURT stats (rel)   min: 0.05% max: 1.09% x̄: 0.32% x̃: 0.11%
95% mean confidence interval for cycles value: -802.75 -55.56
95% mean confidence interval for cycles %-change: -3.19% -1.01%
Cycles are helped.

total sends in shared programs: 1032273 -> 1032272 (<.01%)
sends in affected programs: 41 -> 40 (-2.44%)
helped: 1
HURT: 0

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7445>

clover: implements clSetContextDestructorCallback

Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7468>

clover: add empty cl 3.0 dispatch entries.

This just fills out the dispatch table blanks.

Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7468>

clover/spirv: hook up spir-v environment for 3.0

For now just use the 1.2 env.

Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7468>

clover/llvm: add 3.0 versioning.

Just adds the 3.0 versioning to the compiler interface.

Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7468>

clover: access 3.0 and deprecated 2.2 API

Adds the api defines to open up deprecaated 2.2 and new 3.0 APIs.

Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7468>

vtn/opencl: add ctz support

ctz is a CL2.0 opcode but 3.0 requires it as well so just add support
for it.

Tested against CTS integer_ops integer_ctz test.

(long line broken up)

Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7468>

CL: update CL headers to 3.0

This just updates the headers from Khronos.

Change the cl_mem initialisers, not sure what totally correct answer is.

Acked-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7468>

docs: Add MESA_pack_invert and ANGLE_pack_reverse_row_order

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3156>

mesa: Implement GL_ANGLE_pack_reverse_row_order

Identical to GL_MESA_pack_invert in effect, just need to check for a
different enum value for GLES vs GL. The spec claims that "OpenGL 1.5 or
OpenGL ES 1.0 are required", but ReadPixels isn't a thing for ES1 so we
only enable it for ES2+.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3156>

mesa: Enable GL_MESA_pack_invert unconditionally

This newly enables the extension for r100 and vieux. As far as I can
tell, that's safe to do. vieux's ReadPixels is just _mesa_readpixels,
which clearly already handles it correctly because the extension is
enabled for classic swrast. r100 has some custom acceleration paths
before falling down to _mesa_readpixels, which winds its way through to
r100_blit, which already has code to handle pack->Invert being set.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3156>

v3dv: Remove unsigned comparison to zero.

subpass_idx is of type uint32_t.

Fix defects reported by Coverity Scan.

Macro compares unsigned to 0 (NO_EFFECT)
unsigned_compare: This greater-than-or-equal-to-zero comparison of
an unsigned value is always true. subpass_idx >= 0U.

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7325>

nir: Handle ray-tracing intrinsics and storage classes in copy-prop etc.

We need to consider shader calls as potential writes to their payloads.
For other ray-tracing intrinsics, we may not have a shader payload
pointer and have to treat them more like a barrier.  We also need to
ensure that global and SSBO reads/writes aren't propagated across shader
call intrinsics.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6479>

spirv,nir: Add ray-tracing intrinsics

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6479>

nir,spirv: Add support for the ShaderCallKHR scope

It's currently entirely trivial.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6479>

spirv: Implement the new ray-tracing storage classes

The SPV_KHR_ray_tracing extension adds 6 new storage classes which is a
bit on the ridiculous side.  In order to avoid adding that many variable
modes to NIR, we make a few simplifying assumptions:

 1. CallableData and RayPayload data actually lives on the stack
    somewhere, presumably in the caller's stack.  We assume that these
    are no different from global variables and use nir_var_shader_temp
    for them.  We still need a separate storage class for the incoming
    variants but only so we can figure out which one the incoming one
    is and lower it to something useful.

 2. There's no difference between incoming CallableData and RayPaolad
    data.  We can use a single storage class for both.

 3. ShaderRecordBuffer data is just a global memory access.  This lets
    us avoid NIR variables entirely and just fetch the pointer via the
    shader_record_ptr system value and it's accessed using a 64-bit
    global memory pointer.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6479>

nir: Add new variable modes for ray-tracing

If we were desperate to reduce bits, we could probably also use
shader_in/out for hit attributes as they really are an output from
intersection shaders and read-only in any-hit and closest-hit shaders.
However, other passes such as nir_gether_info like to assume that
anything with nir_var_shader_in/out is indexed using vec4 locations for
interface matching.  It's easier to just add a new variable mode.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6479>

nir: Add intrinsics for object to/from world RT sysvals

These are a bit more tricky than most because they're matrix system
values.  We make the intentional choice here to not bother with allowing
indirect addressing of columns for these.  Since they're system values,
they may be magically constructed somehow or come from weird hardware so
it's easier on back-ends to just handle any indirects with bcsel.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6479>

nir/builder: Add a select_from_ssa_def_array helper

This is an operation we have to do already for nir_vector_extract and
I'm about to do something very similar for matrix columns.  Having a
more generic helper is useful.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6479>

spirv,nir: Add support for ray-tracing built-ins

Missing in this commit are NIR intrinsics for the ObjectToWorld and
WorldToObject built-ins.  Those are matrices and so they take a bit more
work and justify a separate commit.  For now, we add the enums and leave
the SYSTEM_VALUE <-> nir_intrinsic conversion commented out.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6479>

spirv: Add support for OpTypeAccelerationStructureKHR

For now, we assume its a 64-bit global pointer.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6479>

spirv: Pass the deref type to storage_class_to_mode for non-forward pointers

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6479>

spirv: Add a guard for OpTypeForwardPointer storage classes

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6479>

spirv: Remove a redundant vtn_fail_if

We already fail in these same cases in vk_desc_type_for_mode.  These
additional assertions are just extra code to update.

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6479>

spirv: Add Ray Tracing execution models

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6479>

spirv: Add basic plumbing for ray-tracing capabilities

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6479>

compiler: Add new Vulkan shader stages

This particular ordering makes them conveniently match
VkShaderStageFlagBits, which is a property we already take advantage
of in the previous shader stages.

Abbreviations are based on the ones used in glslangValidator.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6479>

tu: Make sure spirv_to_nir knows we support imageStorageWithoutFormat.

You have to set these flags along with the extension, or you get a bunch
of warnings from spirv-to-nir.

Fixes: e781cc702557 ("tu: Expose shaderStorageImage*WithoutFormat")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7465>

nir/clip_disable: handle 2x vec4 case

some drivers may have pre-lowered gl_ClipDistance to 2x vec4 to match hw
usage, so for those cases we'll be getting deref_var here and then components
will be stored to the deref at some point

fixes mesa/mesa#3480

Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6563>

nir/clip_disable: try for better no-op

we can just check the bits using clip_distance_array_size here to simplify
everything and more easily determine if we need to be running this pass

Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6563>

nir/clip_disable: write 0s instead of undefs for disabled clip planes

this should yield more reliable and ideally even correct results

Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6563>

iris: Move blit scissoring earlier.

There's no need to e.g. prepare_access() if the blit is a noop.

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7454>

anv: restrict number of subgroups per group

We are limited to 64 threads per dispatched group, regardless of what
num_cs_threads claims, so advertise that limit correctly.

Fixes (on TGL and up):
dEQP-VK.subgroups.size_control.compute.required_subgroup_size_min
and other *.required_subgroup_size_min tests.

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7453>

turnip: enable VK_EXT_image_drm_format_modifier

Add missing GetPhysicalDeviceImageFormatProperties2 logic for the extension
and enable it.

Also stop exposing optimal tiling for formats which are linear only, to
simplify dealing with those.

Passes dEQP-VK.drm_format_modifiers.*

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6940>

turnip: don't always fallback to linear for mutable formats

Use VkImageFormatListCreateInfo, and enable VK_KHR_image_format_list to
expose it. (and reorganize linear fallback code a bit)

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6940>

turnip: remove unnecessary/redundant tu_image fields

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6940>

turnip: remove useless tu_image asserts

Validation layer already catches these errors, so don't bother.

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6940>

turnip: LAYOUT_PREINITIALIZED is not different for optimal tiling

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6940>

turnip: don't implement CreateImage as two separate functions

Inline tu_image_create into tu_CreateImage.

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6940>

aco: Fix format string used when raising validation errors

Validation errors mention the pretty-printed instruction including
operands with the reserved % character, which caused vasprintf to
expect more format arguments than aco provided.

Fixes: c2b1978aa47c ("aco: rework the way various compilation/validation errors are reported")
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7442>

nir/deref: Fix a typo

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3754
Fixes: df51518dc5b "nir/opt_deref: Add a deref mode specialization..."
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7459>

ci: Distribute ADMGPU driver to LAVA as a module

As it needs firmware to probe, and we cannot bundle it within the kernel
image because it is incompatible with the GPL.

Currently we rebind the driver after boot but that's slow and fragile,
as unloads of DRM drivers aren't generally tested.

Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7420>

ci: Update dEQP skips and fails for Bifrost on G52

Runs are much more stable now with the new kernel, and lots of tests
do pass now.

Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7420>

ci: Update kernel for LAVA to 5.10-rc2 plus patches

Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7420>

util/threaded_context: use driver's buffer alignment for staging transfers

this coincidentally worked because radeonsi has a hardcoded value of 64, but
other drivers do not use this value and then things are subtly broken

Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7452>

v3d: Add GL_ARB_vertex_array_bgra support

This is done by adding support to PIPE_FORMAT_B8G8R8A8_UNORM, and
relying on the R/B swapping for vertex attributes implemented in the
compiler.

v2:
 - Simplify the loop (Iago)

v3:
 - Assert before derreferencing variable (Iago).

Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3078
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7196>

v3dv: mark the right bit to swap R/B vertex attributes

Now that the R/B swap mask for vertex attributes handles all the
attributes, ensure the right generic vertex attribute is marked.

Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7196>

v3d/compiler: extend swapping R/B support to all vertex attributes

So far the support for R/B swapping in vertex attributes were for the
generic attributes.

But there are cases like glSecondaryColorPointer() supporting BGRA
formats that require the R/B swapping to be also allowed in the
non-generic vertex attributes (in this case, in the COLOR1 attribute).

v2:
 - Don't split line (Iago)

Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7196>