panvk: Preload FB attachments when required

There are at least three situations where we need to preload FBs:

1. The attachment is flagged VK_ATTACHMENT_LOAD_OP_LOAD and has not been
   accessed in previous subpasses

2. The batch is implicitly split (e.g. too many jobs queued to the
   batch, wait/set events queued, ...)

3. The attachment has been written by a previous subpass

With those changes, we can get rid of panvk_emit_fb() and call
pan_emit_fbd() directly (fb_info is initialized when starting a subpass
and updated when an implicit split happens).

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12095>

panvk: Initialize the blend shader logic

The blitter logic rely on blend shaders when formats are not blendable.
We need to initialize the blend shader logic before we can use the
blitter.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12095>

pan/blit: Extend pan_preload_fb() to return emitted jobs

The vulkan driver needs to patch job headers when re-issuing batches.
Extend pan_preload_fb() so it can return the emitted tiler jobs.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12095>

pan/blit: Fix a NULL dereference in the preload path

The ZS view can be NULL if a stencil-only buffer is attached to the FB.

Fixes: 1de393fec539 ("panfrost: Fix ZS reloading on Bifrost v6")
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12095>

pan/bi: Allow passing RT conversion descriptors to fragment shaders

The Vulkan copy shaders sometimes need to partially write a texel and
issue a load on the FRAGOUT variable in that case, but they do know
the format of the tile buffer in advance in that case. Let's not add an
RT_CONVERSION sysval if we can avoid it.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12095>

pan/bi: Relax check on 8bit swizzles

Allow extracting components Y, Z or W from an 8bit vector.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12095>

radv: do not store meta shaders to the default shader disk cache

Meta shaders are already stored in a different shader cache file.
Storing them in two places wastes disk space and they are never
loaded from the default shader disk cache anyways.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12937>

wgl: fix 32 bits mingw exports

closes #5349

Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12940>

radv: do not use a different disk cache key for LLVM

The driver already adds a pipeline hash for LLVM which is redundant.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12938>

Revert "intel/dev: Add display_ver and set adl-p to 13"

This reverts commit c81acd365ed2a965741a828504d4a1498eaa3d3a.

Revert "iris: Disable I915_FORMAT_MOD_Y_TILED_GEN12* on adl-p/display 13"

This reverts commit 4961f4c50fe29e4b3e6145bfafae7e5026027d12.

iris: Disable I915_FORMAT_MOD_Y_TILED_GEN12* on adl-p/display 13

Cc: mesa-stable
Fixes: e435511b580 ("intel/dev: Add device info for ADL GT2")
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/12908>

intel/dev: Add display_ver and set adl-p to 13

Cc: mesa-stable
Fixes: e435511b580 ("intel/dev: Add device info for ADL GT2")
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/12908>

freedreno, turnip: Set TPL1_DBG_ECO_CNTL better

Match the blob better here. Note that the value of 0x1000000 for a650
comes from the Vulkan blob, and it's required to fix cubic filtering
even though the GLES driver doesn't set it (and doesn't support cubic
filtering).

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/5261
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12929>

freedreno/a6xx: Name TPL1_DBG_ECO_CNTL

This is a guess, but an informed guess, since every other block with a
known DBG_ECO_CNTL register has it at the very beginning immediately
followed by ADDR_MODE_CNTL.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12929>

lima: Remove depth near/far workaround

because this is fixed now.

Reviewed-by: Erico Nunes <nunes.erico@gmail.com>
Signed-off-by: Andreas Baierl <ichgeh@imkreisrum.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12804>

lima: Expose GL_EXT_clip_control

Reviewed-by: Erico Nunes <nunes.erico@gmail.com>
Signed-off-by: Andreas Baierl <ichgeh@imkreisrum.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12804>

radeonsi/test: use -t for deqp tests

deqp-runner added support for this.

Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12899>

radeonsi/test: don't require a folder name

Generate a temp one if the user didn't supply one.

Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12899>

radeonsi/sqtt: add si_se_is_disabled

Based on radv_se_is_disabled.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12899>

radeonsi/sqtt: export wave size and scratch size

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12899>

radeonsi/test: update expected results

These tests were fixed in piglit.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12899>

radeonsi/test: fix typo in the test script

glcts results were copied over deqp results.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12899>

radv: remove useless assertions in the SQTT path

The driver aborts when the chip class is older than GFX8.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12854>

radv: make the SQTT BO a resident buffer

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12854>

radv: replicate THREAD_TRACE_CTRL config when stopping SQTT

This seems missed and it might be important.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12854>

radv: Experimentally enable RT extensions.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

radv: Add caching for RT pipelines.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

radv: Add support for setting a dynamic stack size.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

radv: Combine all the parts together with a main loop for an RT pipeline.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

radv: Add ray traversal loop.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

radv: Add pass to lower anyhit shader into an intersection shader.

So we avoid having yet another shader calling loop. Such a thing
is not needed since neither shader types do recursion.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

radv: Add helper to parse raytracing stages.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

radv: Add helper to inline shaders into the main shader.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

radv: Add main loop variables.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

radv: Add scaffolding for RT pipeline compilation incl libraries.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

radv: Make some pipeline functions non-static.

Want to put the rt stuff in its own file. radv_pipeline.c is pretty
large already.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

radv: Add raytracing pipeline properties.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

radv: Add group info to pipeline.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

radv: Add pipeline type.

I want to keep pointers that need to be freed in the union and need
to figure out the type a destruction time. This seems more reliable
than checking the shader array in case we destroy mid-creation (i.e.
on failure).

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

radv: Add RT cache flushes.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

radv: Support nir_intrinsic_load_global_constant.

SPIR-V parsing can result in some direct constant usage for shader
records. Lower this early to a global based intrinsic so that it
doesn't interfere with the later 32-bit offset based constants
for scratch usage.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

nir: Support ray launch size in divergence analysis.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

aco: Add support for ray launch size.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

radv: Add support for ray launch size.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

nir: Add AMD rt intrinsics.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

radv: Implement NULL accel struct descriptor write.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

radv: Do more meta shader lowering.

Need this to clean up our generated RT pipeline.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

radv: Refactor some nir_channels usage to use nir_channel.

cleanup, nir_channels wasn't needed as these were only accessing a
single channel.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

aco: Implement call scope.

Since we do no repacking yet, just use invocation.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

radv: Modify load_sbt_amd intrinsic to get the descriptor.

That way we can get the address to the entry, which is needed for
some nir builtins because extra data in the entry can be used as
shader input.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

radv: Add bvh node definitions to a header.

So that we can avoid some magic numbers in the pipeline creation.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

radv: Add optimized CPU BVH builds.

This trivial sorting helps speed up bad cases (like in CTS significantly).

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

radv: Add more acceleration structure formats.

These are required ...

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

radv: Add accel struct build support for the object-to-world matrix.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

radv: Fix arrayOfPointers for instances in accel struct build.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

radv: Fix CPU AABB build.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12592>

meson: introduce option vmware-mks-stats controlling the instrumentations of gallium svga driver

The new boolean option controls the instrumentations of gallium svga driver for mksGuestStats.

Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Reviewed-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Charmaine Lee <charmainel@vmware.com>
(cherry picked from commit 57760b7fe4eeb80acc8d6cd8bf6ec609a11a11dc)
(cherry picked from commit 12aed00f08bd95afd605cab34c833e81a4957dbd)

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12873>

svga: enable DRM mks-stats via hooking to the corresponding DRM ioctls

SVGA DRM stat calls were situated but did not actually register with the mks-stats
mechanism due to absence of corresponding ioctls. The employed new ioctls in vmwgfx
are DRM_VMW_MKSSTAT_ADD and DRM_VMW_MKSSTAT_REMOVE, subject to version check.

Reviewed-by: Charmaine Lee <charmainel@vmware.com>
(cherry picked from commit be47c077cc927c27a8c36342b47697aa81719677)
(cherry picked from commit 0388afc67b830f6ab916d0839c33eb1d91d6353f)

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12873>

build: fix nine compilation with only zink enabled as a gallium driver

fixes #5360

Reviewed-by: Dylan Baker <dylan@pnwbakers.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12943>

zink: fix regex syntax from previous ci commit

argh

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12948>

zink: disable miplevel tests in ci completely for now

I still can't repro, and I'm running these regularly on multiple platforms,
so they're not going to get any worse for the time being

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12947>

etnaviv: add mov for direct depth store output from load input

If the fragment shader writes fragment depth from an ALU instruction,
the register allocator makes sure that the instruction is pointed to the
correct register and write mask (t0.__z_).
If there is no instruction emitted because the source is an input
load intrinsic, or if the source instruction does not support swizzle
and write mask, we have to add a mov instruction for this to work.

Fixes piglit test spec@glsl-1.10@execution@glsl-1.10-fragdepth.

Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12891>

nir_to_tgsi: Remove the abs on fcsel's bool src.

While the nir fcsel opcode specifies src0 != 0.0, as the comment says,
it's only ever used on bools-as-floats, so we know that src0 is
non-negative.  This saves an instruction per CMP on i915.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12921>

ci/freedreno: Update restricted trace sha1s.

Rendering changed slightly recently, but the diffs look fine.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12939>

ci/baremetal: Retry if our network device spontaneously fails.

Seen in https://gitlab.freedesktop.org/mesa/mesa/-/jobs/13824132.  It's
unlikely that graphics would kill the network, so just assume it's not our
fault and keep going.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12939>

ci/freedreno: Add some cubearray piglit flakes on a630 I noticed.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12939>

freedreno: Assert to check for the previous regression.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12942>

freedreno: Fix autotune regression since batch-cache rework.

I freed the key that autotune needed a little early.

Fixes: b2349a46715e ("freedreno: Move the batch cache to the context.")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12942>

turnip: Fix a6xx gen4 compute shaders

Port of 74d10525374 freedreno/a6xx: ("Fix a6xx gen4 compute shaders")

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/5354
Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12856>

turnip: Rast updates for a6xx gen4

Port of 219e12b7f39 ("freedreno/a6xx: Rast updates for a6xx gen3")

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12856>

turnip: Fix unitialized cs->device

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12856>

freedreno: Add info->a6xx.has_shading_rate

@flto noticed these registers seem to be related to GL_QCOM_shading_rate

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12856>

turnip: consider multiview_mask when clearing depth-stencil attachment

Otherwise only first layer is being cleared.

Would fix several VK_EXT_multi_draw tests:
 dEQP-VK.draw.multi_draw.overlapping.normal.max_draws.*.vert_only.multiview.no_offset

Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12913>

nir: Exclude non-generic patch variables from get_variable_io_mask.

These are I/O variables which are not going to be removed anyway.
However, get_variable_io_mask handles their location incorrectly.

Found using the GCC undefined behavior sanitizer.
Fixes the following error:

runtime error:
shift exponent 4294967258 is too large
for 64-bit type 'long unsigned int'

Closes: #5319
Fixes: cf5f8f55c3e25508fb975b263d6430a93442247a
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12719>

ac/nir: Fix match_mask to work correctly for VS outputs.

match_mask checks the intrinsic type and decides whether it's
per-patch or not. VS don't have per-patch outputs,
so this causes wrong behaviour there.

Found using the GCC undefined behavior sanitizer.
Fixes the following error:

runtime error:
shift exponent 18446744073709551584 is too large
for 64-bit type 'long unsigned int'

Closes: #5319
Fixes: bf966d1c1dd968116b8b547ca2739f5113caccb5
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12719>

mesa/st: Allow loops in GLSL when NIR is enabled, even if the HW can't.

The jump lowering enabled by EmitNoLoops breaks GLSL's loop unrolling on
various obviously unrollable loops, resulting in a lot of deqp-gles2 and
piglit failures.  NIR will help unroll whatever GLSL doesn't, so we can
trust the driver to apply that after GLSL's unrolling, so no need to ask
GLSL to lower all loops.

Fixes: #4979
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12917>

ci/i915g: Clarify failure happening in fbo-fragcoord2.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12917>

etnaviv: fix dirty bit check for baselod emission

Since baselod is stored in sampler state, not sampler view, we should
check the ETNA_DIRTY_SAMPLERS bit instead of ETNA_DIRTY_SAMPLER_VIEWS.

Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12916>

X11: Ensure that VK_SUBOPTIMAL_KHR propagates to user code

Commit 0245b825 switched from returning the error code VK_ERROR_OUT_OF_DATE_KHR
to returning the success code VK_SUBOPTIMAL_KHR. Prior to that commit, the error
code caused all code paths to fail immediately, but the success code does not.

Currently the success code is not recorded in some scenarios, resulting in a
result of VK_SUCCESS instead. This breaks applications that rely on the
result (per the spec) to trigger resizes.

This commit ensures that the proper VK_SUBOPTIMAL_KHR success code is set as a
sticky status (as comments indicate was intended), ensuring that it is
propagated to user code.

Fixes #5331

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12782>

radv: keep depth/stencil images compressed for TRANSFER_DST on compute

Only if the image is TC-compat HTILE because it can be decompressed
on compute for partial copies.

This should remove few depth/stencil decompressions for RAGE2 and Red
Dead Redemption 2 because they declare all images as concurrent but
never use the compute queue for them.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12452>

radv: add support for copying compressed depth/stencil images on compute

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12452>

radv: implement depth/stencil expand on compute

This works as long as the image is TC-compatible HTILE.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12452>

radv: rename radv_decompress_depth_stencil()

To radv_expand_depth_stencil() for consistency.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12452>

panfrost/ci: Skip the indirect_draw+XFB tests

We lack a dependency between the vertex job filling the indirect draw
buffers and the indirect draw compute job reading from these buffers,
leading to unreliable results (the tests pass if the vertex job is
done before the compute job starts, and fail otherwise). Let's disable
those tests until we sort it out.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12589>

panfrost: Fix collision in the indirect draw shader table

Min/max index search shaders are different for the !primitive_restart
and primitive_restart. We need to add entries for the primitive restart
cases otherwise we might retrieve a wrong shader from the cache.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12589>

panfrost: Fix indirect draws when vertex or instance count is 0

In that case we should just skip the vertex/tiler jobs as done in the
direct draw path.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12589>

panfrost: fix null deref when no color buffer is attached

Do not dereference color buffer #0 in the SFBD code path if no color buffer is
attached, as with depth-only attachments. Fixes a crash running glmark2 -b
shadow on Mali T720.

Signed-off-by: Italo Nicola <italonicola@collabora.com>
Fixes: c746747cb82 ("panfrost: fix GL_EXT_multisampled_render_to_texture regression")
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12927>

panfrost: Prepare shader helpers to per-gen XML

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12935>

panfrost: Prepare texture helpers to per-gen XML

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12935>

panfrost: Prepare pan_encoder.h to per-gen XML

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12935>

panfrost: Prepare scoreboard helpers to per-gen XML

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12935>

panfrost: Prepare pandecode to per-gen XML

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12935>

panfrost: Move panfrost_major_version() to gen_macros.h

So we can use this function in decode_common.c when transitioning to
per-gen XML. While at it rename the function pan_arch().

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12935>

panfrost: Prepare pan_cs helpers to per-gen XML

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12935>

panfrost: Prepare blend helpers to per-gen XML

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12935>

panfrost: Prepare blitter helpers to per-gen XML

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12935>

panfrost: RGB332_UNORM is not a valid texture format on v6+

Cc: mesa-stable
Fixes: c6bdd976e611 ("panfrost: Split out v6/v7 format tables")
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12508>

panfrost: Drop the R and T flags on SCALED formats

Sampling from SCALED textures / rendering to SCALED FBOs is a bit tricky
(requires extra int <-> float conversions in a few places).
mesa/st only use SCALED formats as vertex formats, and those formats
are optional in Vulkan, so let's drop the RENDER/TEXTURE flags to keep
things simple.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Suggested-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12508>

panfrost: RGB10_A2_SNORM is not a valid texture format on v6+

Cc: mesa-stable
Fixes: c6bdd976e611 ("panfrost: Split out v6/v7 format tables")
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12508>