tu: Clamp priority in DRM submitqueue creation

The kernel driver has a range of valid priority values that can
be supplied to it, submitting any priority value outside these
bounds will result in `-EINVAL`. To avoid this, the priority
value is now clamped to the range that the kernel supports.

Fixes: 0c6fbfca0c91ef012e8ab767a317c07f1f6dc5e6
Signed-off-by: Mark Collins <mark@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18389>

r300: allow presubtract when both ADD sources are negative

Current code doesn't handle this, however it is easy to make it work
by moving the negate to the presubtract source. Minor win in shader-db,
mostly with Unigine shaders.

Shader-db RV530:
total instructions in shared programs: 136382 -> 136236 (-0.11%)
instructions in affected programs: 9911 -> 9765 (-1.47%)
total temps in shared programs: 18939 -> 18942 (0.02%)
temps in affected programs: 37 -> 40 (8.11%)

Reviewed-by: Filip Gawin <filip@gawin.net>
Signed-off-by: Pavel Ondračka <pavel.ondracka@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18289>

virgl: Add some formats that the CTS uses

Otherwise running the CTS emits lots of warnings about
these formats missing in the drivers format table.

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18462>

egl: Relax locking

Now that we have the rwlock TerminateLock protecting us against
eglTerminate() yanking the rug from under us, drop the BDL across
calls to driver (or at least the main ones that can potentially
block).

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7039
Signed-off-by: Rob Clark <robdclark@chromium.org>
Acked-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18050>

egl: Introduce rwlock to protect eglTerminate()

eglTerminate() must be serialized against all other EGL calls.  But in
most cases, other EGL calls do not need to be serialized against each
other.  Which fits rather well with a rwlock.

One would be tempted to simply replace the existing BDL with a rwlock,
but several portability and debuggability limitations of the rwlock
implementation prevent that, as described in the TerminateLock comment
block.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Acked-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18050>

egl: Make RefCount atomic

Once we relax the locking, we will be doing _eglPutFoo() outside of the
big display lock.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Acked-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18050>

egl/dri2: Add display lock

In preperation of relaxing eglapi to not hold a lock across driver
calls, but instead only for protecting it's own state, add our own
lock to protect code paths that need locking or have not been audited
yet.  The blocking calls (ClientWaitSyncKHR) or critical path and/or
blocking (MakeCurrent, SwapBuffers*) are lockless, as they have already
been audited for thread safety.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Acked-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18050>

egl/dri2: Make ref_count atomic

In particular, MakeCurrent can be called on multiple threads in
parallel.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18050>

egl/wgl: Make ref_count atomic

Looks like wgl doesn't have much display state to protect.  But it's
ref_count should be atomic before we start removing locking from eglapi
to protect against MakeCurrent being called in parallel on multiple
threads.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Acked-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18050>

glsl: remove GLSL IR inverse comparison optimisations

As per 7d85dc4f350b GLSL IR is not smart enough to handle this
correctly for NANs.

Shader-db radeonsi (RX 6800):

Totals from affected shaders:
SGPRS: 26848 -> 26848 (0.00 %)
VGPRS: 13552 -> 13552 (0.00 %)
Spilled SGPRs: 134 -> 134 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Private memory VGPRs: 0 -> 0 (0.00 %)
Scratch size: 0 -> 0 (0.00 %) dwords per thread
Code Size: 635000 -> 630988 (-0.63 %) bytes
Max Waves: 5474 -> 5474 (0.00 %)

Shader-db iris (BDW):

total instructions in shared programs: 17538859 -> 17539018 (<.01%)
instructions in affected programs: 29369 -> 29528 (0.54%)
helped: 3
HURT: 126
helped stats (abs) min: 1 max: 1 x̄: 1.00 x̃: 1
helped stats (rel) min: 0.49% max: 0.49% x̄: 0.49% x̃: 0.49%
HURT stats (abs)   min: 1 max: 2 x̄: 1.29 x̃: 1
HURT stats (rel)   min: 0.27% max: 1.32% x̄: 0.61% x̃: 0.54%
95% mean confidence interval for instructions value: 1.13 1.33
95% mean confidence interval for instructions %-change: 0.54% 0.63%
Instructions are HURT.

total loops in shared programs: 4866 -> 4866 (0.00%)
loops in affected programs: 0 -> 0
helped: 0
HURT: 0

total cycles in shared programs: 858548230 -> 858548915 (<.01%)
cycles in affected programs: 1331737 -> 1332422 (0.05%)
helped: 0
HURT: 92
HURT stats (abs)   min: 2 max: 49 x̄: 7.45 x̃: 6
HURT stats (rel)   min: 0.01% max: 1.90% x̄: 0.12% x̃: 0.05%
95% mean confidence interval for cycles value: 5.72 9.17
95% mean confidence interval for cycles %-change: 0.05% 0.19%
Cycles are HURT.

Note: With the addition of "nir/comparison_pre: See through an inot to
apply the optimization", idr's shader-db results are:

All Broadwell and newer Intel platforms had similar results. (Ice Lake shown)
total instructions in shared programs: 19940805 -> 19940802 (<.01%)
instructions in affected programs: 582 -> 579 (-0.52%)
helped: 3 / HURT: 0

total cycles in shared programs: 858431633 -> 858431747 (<.01%)
cycles in affected programs: 4938 -> 5052 (2.31%)
helped: 0 / HURT: 3

All older Intel platforms had similar results. (Haswell shown)
total instructions in shared programs: 16715626 -> 16715670 (<.01%)
instructions in affected programs: 9496 -> 9540 (0.46%)
helped: 0 / HURT: 44

total cycles in shared programs: 881224396 -> 881232314 (<.01%)
cycles in affected programs: 600610 -> 608528 (1.32%)
helped: 6 / HURT: 44

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18006>

nir/comparison_pre: See through an inot to apply the optimization

This also prevents some small regressions in "glsl: remove GLSL IR
inverse comparison optimisations".

shader-db results:

All Sandy Bridge and newer Intel platforms had similar results. (Ice Lake shown)
total instructions in shared programs: 19941025 -> 19940805 (<.01%)
instructions in affected programs: 52431 -> 52211 (-0.42%)
helped: 188 / HURT: 6

total cycles in shared programs: 858451784 -> 858431633 (<.01%)
cycles in affected programs: 2119134 -> 2098983 (-0.95%)
helped: 183 / HURT: 12

LOST:   2
GAINED: 0

Iron Lake and GM45 had similar results. (Iron Lake shown)
total instructions in shared programs: 8364668 -> 8364670 (<.01%)
instructions in affected programs: 753 -> 755 (0.27%)
helped: 2 / HURT: 4

total cycles in shared programs: 248752572 -> 248752238 (<.01%)
cycles in affected programs: 87290 -> 86956 (-0.38%)
helped: 2 / HURT: 4

fossil-db results:

Skylake, Ice Lake, and Tiger Lake had similar results. (Ice Lake shown)
Instructions in all programs: 144909184 -> 144909130 (-0.0%)
Instructions helped: 6

Cycles in all programs: 9138641740 -> 9138640984 (-0.0%)
Cycles helped: 8

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18006>

nir: support loop unrolling with inot conditions

Ever since 4246c2869c3c and 7d85dc4f350b loop unrolling can no
longer depend on inot being eliminated from the loop
terminator condition so we need to be able to handle it.

This change avoids 292 loop unrolling regressions with shader-db
once the following patch is applied.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18006>

nir: update nir_is_supported_terminator_condition()

Ever since 4246c2869c3c and 7d85dc4f350b loop unrolling can no
longer depend on inot being eliminated from the loop
terminator condition so we need to be able to handle it.

Here we simply check to see if the inot contains a simple
terminator condition we previously handled. We also update
the previous users of this function to use a newly name
copy of the previous behaviour
nir_is_terminator_condition_with_two_inputs().

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18006>

amd/common: Disable DCC retile modifiers on RDNA1

Some claims of corruption, modifier-less Mesa already doesn't do
it. Since these modifiers have no purpose besides being displayed
lets just disable in Mesa.

Cc: mesa-stable
Tested-by: Michel Dänzer <mdaenzer@redhat.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18140>

amd/common: Don't rely on DCN support checks with modifiers.

Going to be a bad time if they disagree, which is bound to happen
sometimes. Not asserting and stuff tends to be a better experience
than crashing.

Cc: mesa-stable
Tested-by: Michel Dänzer <mdaenzer@redhat.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18140>

intel/pci_ids: Drop non-upstream dg2 pci-ids

These pci-ids should be included in
https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14523, since
these pci-ids will only be supported by kernels that support the
forked Linux uapi. (Note that !14523 will never be merged into
upstream Mesa.)

Ref: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/include/drm/i915_pciids.h?h=v6.0-rc3#n695
Fixes: 398a9be94b4 ("intel/dev: Enable remaining DG2 and ATS-M device IDs")
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18386>

venus: double the abort timeout

To avoid bumping abort timeout too much. This change also doubles the
busy wait cycles, which would further reduce unnecessary sleeps for
synchronous calls. Ultimately, after we fix the fencing and push all
roundtrip waiting to the renderer side as well as we fixing the abort
logic, we can live with busy wait alone here.

Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18472>

anv: pipelineStageCreationFeedbackCount is allowed to be 0

Fixes: 6601e5d6fc6 ("anv: implement VK_EXT_pipeline_creation_feedback")

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18451>

aco: Use plain VOPC for vcmpx when possible.

Foz-DB Navi21:
Totals from 66947 (49.62% of 134913) affected shaders:
CodeSize: 210383024 -> 210033376 (-0.17%)

Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18417>

radv: Deduplicate push constant structs

This patch adds a header that is shared between the accel struct build
kernels and the dispatch code.

Signed-off-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18376>

zink: fix sharedmem ops with bit_size!=32

* the rewrite_bo_access compiler pass already handles 64bit rewrites as-needed
* sharedmem access is not required to be 32bit

thus, this can use a similar methodology as ssbo/ubo vars to index based on bitsize
and handle operations through sized variables

Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18449>

ci: disable the freedreno farm.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18463>

v3dv: implement VK_EXT_shader_module_identifier

Passes `dEQP-VK.*.shader_module_identifier.*`

Signed-off-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18458>

Revert "Revert "Revert "ci: set venus on lavapipe to manual due to flakes"""

Now the flakiness might have been fixed for good.

This reverts commit e51c5a18ade928868623be0048fdd24ed158a42c.

Acked-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18454>

ci: Crosvm won't remove the control socket file on stop

When sending the stop command to a lingering Crosvm instance, the socket
file will remain and the next instance will fail to start.

Make sure the file is deleted before starting Crosvm with the same path.

Acked-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18454>

radv: only expose sparseResidencyImage3D on GFX9+

It's currently broken on Polaris10 and breaks running VKCTS entirely.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18457>

radv/ci: run vkcts on the two steam decks in parallel

We just added a new Steam Deck to our CI, which should allow us to
halve the execution time of a full VKCTS run from 1h20 to a more
reasonable 40 minutes.

Signed-off-by: Martin Roukala (né Peres) <martin.roukala@mupuf.org>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18432>

radv: compact MRTs to save PS export memory space

If there are holes between color outputs (e.g. a shader exports MRT1,
but not MRT0), we can remove the holes by moving higher MRTs lower. The
hardware will remap the MRTs to their correct locations if we remove
holes in SPI_SHADER_COL_FORMAT but not CB_SHADER_MASK. This is good for
performance because the hardware will allocate less space for color
MRTs.

This also allows to remove even more unused color exports because we no
longer need to force previous targets to be non-zero. Only SotTR seems
affected from our fossils db.

fossils-db (NAVI21):
Totals from 859 (0.64% of 134913) affected shaders:
VGPRs: 24328 -> 24216 (-0.46%)
CodeSize: 1433276 -> 1422576 (-0.75%)
Instrs: 255275 -> 253728 (-0.61%)
Latency: 1666836 -> 1661544 (-0.32%)
InvThroughput: 346038 -> 343406 (-0.76%)
Copies: 16520 -> 16506 (-0.08%)
PreSGPRs: 25934 -> 25920 (-0.05%)
PreVGPRs: 19903 -> 19662 (-1.21%)

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5786>

radv: gather MRTs that are written by the fragment shader

This will be used to filter color attachments without exports.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5786>

docs/zink: remove bptc from required formats for gl4.2

We now have lowering-code in the mesa state-tracker, meaning we no
longer need this feature.

Fixes: e4ff42684b9 ("mesa/st: enable bptc extension with fallback")
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Acked-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18434>

vc4: do not attempt to do deep tiled blits

We only copy a single layer, so let's not even try to support deep
blits here.

Reviewed-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Tested-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18427>

vc4: respect z-offset in tiled blits

Reviewed-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Tested-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18427>

v3d: do not pretend to fake rgtc-support

The is_format_support query doesn't pretent to have RGTC support, so
this doesn't seem like it ever did anything useful.

Reviewed-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18439>

intel/compiler: implement Wa_14014595444 for DG2

According to the workaround, we should setup MLOD as parameter
4 and 5 for the sample_b message.

v2: only SAMPLE_B, not SAMPLE_B_C (Lionel)

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18408>

anv: implement Wa_14015946265 for DG2

SOL unit issues, wa is to send PC with CS stall after SO_DECL.

v2: emit also in genX_gpu_memcpy (Lionel)

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18409>

iris: implement Wa_14015946265 for DG2

SOL unit issues, wa is to send PC with CS stall after SO_DECL.

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18409>

radv: Expose 3d sparse images.

Tested-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18165>

radv: Add 3d tile shapes for sparse binding.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18165>

radv: Add binding code for 3d sparse images.

GFX7-8 code is kinda expected. For GFX9 and GFX10 the entire
mipchain is duplicated by "layer" even though smaller mips also
have less layers.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18165>

asahi: Allocate new cmdbufs if out of space

Instead of crashing when we run out of space in the command buffer,
allocate a new buffer, jump to it with the STREAM_LINK command, and
use it to write new commands.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18421>

asahi: Handle Stream Link VDM commands

Jumps in the command streams, allowing us to chain ("link") command
buffers. Naming is from PowerVR, which contains an identical command.
PowerVR's has conditional jumps and function call support, it's likely
that AGX inherited this too but I haven't tested that. (Those might be
useful for conditional rendering and secondary command buffers
respectively?)

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18421>

asahi: Express VDM commands according to PowerVR

Piles of unknown bits go away, as we find they're either "field present"
bits or block types. And yep, the block type enum lines up between AGX
and RGX.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18421>

asahi: Annotate VDM/CDM commands as per PVR

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18421>

asahi: Make BO list growable

Back it by a simple dynamic array, ralloc'd off the batch (and make the
context/batch ralloc'd so stuff gets cleaned up).

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18421>

asahi: Dirty track everything

Now that we have fine grained state emit code, let's use it to reduce
driver overhead. Dirty tracking is delicate: while this seems to work,
I've also added an ASAHI_MESA_DEBUG=dirty option in debug builds
to disable the optimizations here for future debug.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18421>

asahi: Hoist constant PPP state to start of batch

This reduces how much we emit per draw.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18421>

asahi: Match PPP data structures with PowerVR

Looking at PowerVR's PPP definitions in tree in Mesa
(src/imagination/csbgen/), we find that AGX's "tagged" data structures
are actually sequences of state items prefixed by a header specifying
which state follows. Rather than hardcoding the sequences in which Apple's
driver chooses to bundle state, we need the XML to be flexible enough to
encode or decode any valid combination of state. That means reworking
the XML. While doing so, we find a number of fields that are identical
between RGX and AGX, and fix the names while at it (for example, the W
Clamp floating point).

Names are from the PowerVR code in Mesa where sensible.

Once we've reworked the XML, we need to rework the decoder.  Instead of
reading tags and printing the combined state packets, the decoder now
must unpack the header and print the individual state items specified by
the header, with slightly more complicated bounds checking.

Finally, state emission in the driver becomes much more flexible. To
prove the flexibility actually works, we now emit all PPP state (except for
viewport and scissor state) as a single PPP update. This works. After
this we can move onto more interesting arrangements of state for lower
driver overhead.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18421>

asahi: Don't use lower_wpos_pntc

Instead we can flip point coords with the object type. That means fewer
instructions without shader variants. Thanks, PowerVR ^_^

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18421>

asahi: Identify object type field via PowerVR

src/imagination/csbgen/rogue_ppp.xml STATE_ISPA bits 28. Looks like that
got split into two structs in AGX (with info duplicated?) but yeah I
have a lot to work with here.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18421>

asahi: Split RASTERIZER into constituent words

As done in the PowerVR driver.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18421>

fix(FTBFS): meson: raise C++ standard to C++17

LLVM 16 requires this since [0] and instead of just going all over the
various directories and adding overrides it should be easier to just
raise this globally. GCC supports C++17 since GCC 8 (2018), Clang since
version 5 (2017). Debian Buster (oldstable) has GCC 8.3 and Clang 7.0,
so all major distributions should have this and there shouldn't be an
issue with bumping.

[0] <https://github.com/llvm/llvm-project/commit/b4e9977fc18405d4a11cbaf1975bcadbf75920b8>

Closes: <https://gitlab.freedesktop.org/mesa/mesa/-/issues/7031>
Signed-off-by: Kai Wasserbäch <kai@dev.carbon-project.org>
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Reviewed-by: David Heidelberg <david.heidelberg@collabora.com>
Reviewed-by: Ganesh Belgur Ramachandra <ganesh.belgurramachandra@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17966>

venus: enable KHR_driver_properties on Android

Venus has a driver id now and Android cts has been patched.

Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18447>

venus: enable zink required extensions on Android

Below extensions are enabled:
- VK_KHR_external_memory_fd
- VK_EXT_external_memory_dma_buf
- VK_EXT_image_drm_format_modifier

Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18447>

venus: some clang format fixes

Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18447>

meson: add with_gallium_virgl to allow it as VA backend driver

So that no longer needs other driver such as radeonsi as backend
to enable VA.

Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18424>

anv: add a new NO_LOCAL_MEM allocation flag

We found a perf regression with 9027c5df4c51b ("anv: remove the
LOCAL_MEM allocation bit") which seems to be that we over subscribe
local memory, leading i915 to swap things in/out too much.

This change avoid putting buffers in local memory if they are not
allocated from a DEVICE_LOCAL heap.

Maybe we can revisit this later if i915 is better able to deal with
more buffers in local memory.

v2: Remove implicit_css from anv_bo when not in lmem (Ivan)

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes: 9027c5df4c51b ("anv: remove the LOCAL_MEM allocation bit")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7188
Reviewed-by: Ivan Briano <ivan.briano@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18395>

egl/kopper: Don't add EGL_SWAP_BEHAVIOR_PRESERVED_BIT configs

It's strictly inferior to EGL_EXT_buffer_age so apps shouldn't bother to
begin with, and we don't communicate the surface preservation state to
the backend so we don't handle it correctly in any case.

Acked-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18214>

wsi/x11: Defer clearing image busy flag for non-shm upload

The image is busy until xcb_put_image returns. This isn't a major worry
at the moment since we're doing the PutImage directly from
vkQueuePresent, but if we moved that to a worker thread the race window
would be a lot easier to hit.

Reviewed-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18214>

egl: Move an error check earlier in EGL_BUFFER_AGE_EXT query

I can't think of any good reason to call down to the driver and _then_
check whether your calling context was already bad.

Reviewed-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18214>

mesa: Remove PIPE_CAP_CS_DERIVED_SYSTEM_VALUES_SUPPORTED/lower_cs_derived.

We have fine NIR lowering for this (already called from mesa/st), no need
for a separate GLSL pass.

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18361>

tu: Trivially expose VK_EXT_rasterization_order_attachment_access

It is just a renamed VK_ARM_rasterization_order_attachment_access.

Zink depends on it to expose KHR_blend_equation_advanced_coherent

Passes GL tests via Zink:
 dEQP-GLES31.functional.blend_equation_advanced.*
 KHR-GLES31.core.blend_equation_advanced.*

Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18420>

tu: Use newly obtained magic reg values

Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18229>

freedreno: Add all variable magic regs to device-info tables

There are more magic regs which have different values between GPU
subgenerations than we specified.

The updated list and values where obtained by using libwrapfake
with v631 blob and dEQP-VK.draw.renderpass.basic_draw.draw.triangle_list.1
vk cts test.

Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18229>

freedreno: Name more *_DBG_ECO_CNTL regs

There is known pattern of *DBG_ECO_CNTL being right before
*_ADDR_MODE_CNTL, name such regs that we are sure about.

Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18229>

pvr: don't setup attachment tex state for images with input attachment bit unset

In this case the attachment tex state will be 0, which should make things easier
to debug.

Signed-off-by: Frank Binns <frank.binns@imgtec.com>
Reviewed-by: Rajnesh Kanwal <rajnesh.kanwal@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18320>

pvr: cube map tex state packing fixes

A VK_ERROR_FORMAT_NOT_SUPPORTED error was being returned when setting up the
image view tex state for images with the VK_IMAGE_USAGE_STORAGE_BIT and/or
VK_IMAGE_USAGE_INPUT_ATTACHMENT_BIT bit(s) set due to missing handling in
pvr_pack_tex_state(). Resolve this by handling these cases, while taking the
opportunity to simplify how the tex type is determined when packing
TEXSTATE_IMAGE_WORD0.

It was also found that the depth field in TEXSTATE_IMAGE_WORD1 was being set up
incorrectly, as it was relying on the image depth being 0 for 1D and 2D images,
but the image depth will always be 1 in these cases.

Partial fix for dEQP-VK.image.qualifiers.volatile.cube.r32f. This now goes from
failing to seg faulting when VK_FORMAT_R32_SFLOAT is added to the format table,
as the test is now getting further.

Signed-off-by: Frank Binns <frank.binns@imgtec.com>
Reviewed-by: Rajnesh Kanwal <rajnesh.kanwal@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18320>

pvr: set samplerAnisotropy to false

This now matches up with the value being returned for maxSamplerAnisotropy,
which is the unsupported limit of 1.

Fixes dEQP-VK.info.device_properties.

Signed-off-by: Frank Binns <frank.binns@imgtec.com>
Reviewed-by: Rajnesh Kanwal <rajnesh.kanwal@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18320>

pvr: Fix clang-format issue.

Signed-off-by: Rajnesh Kanwal <rajnesh.kanwal@imgtec.com>
Reviewed-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18436>

mesa/st: enable s3tc extensions with fallback

The VC4 and Lima Piglit failures seems to mostly fall in two camps:
1. The hardware lacks sRGB support, but the drivers decide to expose it
   nevertheless, with some varying level of emulation. This leads to some
   failures, probably because we're missing sRGB decoding somewhere.
2. The spec@ext_texture_compression_s3tc@compressedteximage fails,
   mostly due to the test not setting the mipfilter to nearest. With
   that fixed, the test passes on VC4, but still fails on Lima due to an
   a bit dodgy miplod bias in the driver.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Acked-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18180>

mesa/st: implement fallback for s3tc

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Acked-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18180>

mesa/main: add _mesa_unpack_s3tc

We treat sRGB formats as RGB, because we're unpacking to another sRGB
format.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Acked-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18180>

mesa: add format-helpers for s3tc

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Acked-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18180>

mesa/st: add context-flags for s3tc

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Acked-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18180>

util: allow unpacking less than a block from dxt

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Acked-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18180>

mesa/st: do not fallback to srgba formats for dxt

We don't do this for the non-sRGB DXT formats, and it doesn't really
make any sense when we're about to add fallback elsewhere.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Acked-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18180>

mesa/st: use _mesa_generate_mipmap for fallback-formats

For fallback formats, we need to update both the compressed and
uncompressed images. Let's fall back to _mesa_generate_mipmap that
already handles this.

Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Acked-by: Eric Engestrom <eric@igalia.com>
Fixes: e4ff42684b9 ("mesa/st: enable bptc extension with fallback")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18180>

Revert "Revert "ci: set venus on lavapipe to manual due to flakes""

This reverts commit 55e99a22f31e8cd0d58a2e419cfad212bea78fe7.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18433>

v3dv: ref pipeline layout earlier during pipeline init

If we fail to init the pipeline the callee will then destroy it
and if we had assigned the layout to the pipeline it will try to
unref it, so make sure we ref it right after assigning it.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7206
Fixes: dEQP-VK.api.object_management.alloc_callback_fail_multiple.graphics_pipeline
Fixes: 14dab6b10c ('v3dv: ref/unref pipeline layout objects')
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18426>

util/u_trace: Add locking for ctx_list

It could be modified from several threads, make it safe.

Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Reviewed-by: Yonggang Luo <luoyonggang@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18391>

vc4: drop dead qir_POW() helper

Dead code since 226bd9294541f65c91ca ("vc4: Use NIR lowering for sRGB decode.")

Signed-off-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18419>

v3dv: expose VK_EXT_border_color_swizzle

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18416>

v3dv: fix swap_rb and channel_reverse flags in image views

We were computing these from the final swizzle resulting from
combining the format swizzle and the view swizzle, but here we
want to use the format swizzle alone, which is the one we
use to define these properties in the format table.

Fixes CTS test fails with EXT_border_color_swizzle.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18416>

v3dv: expose VK_EXT_tooling_info

There is a stub for this implemented in the common code we can use
which reports 0 tools.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18410>

radeonsi: add and implement load_user_clip_plane abi

This is used by NGG culling for legacy user clip plane clip
which only happens in VS.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17455>

radeonsi: vs load input re-calculate vertex index after culling

After culling, vertices are repacked, we overwrite the original
vertex_id/instance_id, so vertex index for each thread need to
be re-calculated.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17455>

radeonsi: use replaced values when present

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17455>

ac/llvm: move culling replaced registers to abi

For RADV/Radeonsi llvm backend to use correct value after
culling.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17455>

radeonsi: implement cull nir intrinsics

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17455>

radeonsi: implement nir_intrinsic_load_viewport_xy_scale_and_offset

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17455>

radeonsi: implement nir_intrinsic_load_half_line_width_amd

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17455>

zink: lower discard/demote depending on derivatives

This should fix rendering in some games that incorrectly expect discard
to have demote behavior.

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18272>

zink: expose demote when supported

This enables GL_EXT_demote_to_helper_invocation on Zink.

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18272>

zink: hook up demote extension

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18272>

zink: add ntv support for is_helper_invocation

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18272>

zink: add spirv_builder support for is_helper_invocation

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18272>

zink: add ntv support for demote

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18272>

zink: add spirv builder function for demote

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18272>

v3dv/bo: reduce DEFAULT_MAX_BO_CACHE_SIZE to 64

The best way to tune this value is to test Vulkan
applications. Current somewhat big value (512), was obtained by
testing only vkQuake2. Additionally at that time the bo cache was the
first performance oriented improvement we implemented.

After more improvements were included, and retested with more
applications, the conclusion is that we can reduce the value. More
info on the issue that closes.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7090

Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18398>

Android.mk: Provide a custom entry name to ensure meson checks succeed

When meson tests for the compability of some linker flags, not having
the default entry symbol "_start" defined makes the check fail for
a warning unrelated to the linker flag meson is testing.

Since these tests normally just try to compile a C program with a main,
use "main" as the ficticious entry point.

Specifically, this makes "-Wl,--build-id=sha1" recognised as valid
and allows its use for drivers that rely on it.

This is a temporary workaround that is required until Android.mk is
changed to support both shared libraries and executables.

Cc: "22.0" "22.1" "22.2" mesa-stable
Reviewed-by: Roman Stratiienko <r.stratiienko@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17715>

Android.mk: Filter out --build-id=md5 linker flag

To avoid conflicts with meson-generated --build-id linker option

Cc: "22.0" "22.1" "22.2" mesa-stable
Reviewed-by: Roman Stratiienko <r.stratiienko@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17715>