nvk: Advertise VK_EXT_attachment_feedback_loop_layout

We don't need to do anything with the layout.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25808>

anv: uninitialize queues before utrace

We need to shut down the runtime queue threads before tearing down
anything else.

Gets rid of helgrind errors like this :

==212772== Possible data race during write of size 4 at 0xADCBFB0 by thread #1
==212772== Locks held: 1, at address 0x6B8F260
==212772==    at 0x8AC3EFF: simple_mtx_destroy (simple_mtx.h:97)
==212772==    by 0x8ACB24D: intel_ds_device_fini (intel_driver_ds.cc:603)
==212772==    by 0x6CBD4D4: anv_device_utrace_finish (anv_utrace.c:471)
==212772==    by 0x6C71577: anv_DestroyDevice (anv_device.c:3679)
==212772==    by 0x6B2F1E2: loader_layer_destroy_device (loader.c:4358)
==212772==    by 0x6B3F10B: vkDestroyDevice (trampoline.c:983)

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes: cc5843a573 ("anv: implement u_trace support")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/10010
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25805>

ci/freedreno: Generalize the implicit_unmap timeouts.

dedicated_alloc.image caused a job flake today.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25804>

anv: don't uninitialize bvh_bo_pool is not initialized

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes: 3e8d2617e1 ("anv: use buffer pools for BVH build buffers")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/10009
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25802>

ci/zink: Cut zink-lvp coverage in half.

It's taking 23 minutes of deqp-runner time.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25790>

ci/freedreno: Move skqp testing to a618.

The limozeen runners are quite under-committed, while we had 10 jobs
running for our 9 cheza runners, causing an unnecessary wait even when the
farms were idle.

I've removed the old 630 skqp bits, so we can just track what's being
tested on 618.  618 had skqp toml and xfails laying around alreaady.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25790>

ci/freedreno: Fold a630_egl into a630_gl.

The GL job was only taking 5 minutes in deqp, and egl is 2 minutes.  Saves
a boot, and gets us closer to starting all a630 jobs at the same time if
the farm was idle.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25790>

ci/iris: Drop parallel on kbl piglit to 2.

We have only 6 of these boards since one died in May, and 7 jobs allocated
to them.  So you ended up with a 5 minute delay on each pipeline with an
otherwise-idle farm while you waited for the first batch of jobs to
complete so you could get the last one started.  It turns out that piglit
was taking 3 minutes of runtime each, so we can just shard piglit 2 ways,
stay under runtime, and not over-allocate the farm

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25790>

ci/bare-metal: Default our boards to a 20-minute timeout for the whole job.

If the test time is 10 minutes, and boot is 1.5 minutes, that gives you
8.5 minutes of slop for intermittent costs like container download.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25790>

ci/bare-metal: Drop the 2 vs 1 exit code from poe_run.

This exit code was used for the intra-job retries on the other bare-metal
run scripts, but poe_run doesn't do that.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25790>

ci: Stop doing internal retries in bare-metal.

We have job-level retry on failure now, and will continue to need to in
order to work around fd.o infrastructure flakes.  If we stop doing retry
inside the job, then we can crank down the gitlab-level timeouts on test
jobs to be closer to our CI guidelines and avoid blocking a runner for an
hour when things go wrong (for example, cheza #16 failing to boot in a
recognized way and continuously looping due to the intra-job retry).
Plus, the job logs will be more readable when you don't have two boots in
one job, and we'll get the flakes surfaced in our monitoring dashboards.

If internal retries were really doing useful work we may see an increase
in flakes as a result of this.  I'm committing to turning off boards or
reducing coverage as necessary to handle this.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25790>

ci/zink: Only test half of piglit pre-merge on anv.

Unfortunately we're at 16 minutes of runtime for this job, which is much
too far past our guidelines.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25790>

zink: fix legacy depth texture rewriting for single component reads

if only a single component is read, this instruction can (and must) be
rewritten to use that component since depth sampling in vulkan is single-component

cc: mesa-stable

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25795>

zink: update pointer for GPL pipeline cache entry formats

this otherwise points to ctx->gfx_pipeline_state.rendering_formats,
which is broken for background compiles

cc: mesa-stable

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25795>

nir/lower_fragcolor: preserve location_frac

this otherwise breaks component-based outputs

cc: mesa-stable

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25794>

venus: Enable VK_EXT_graphics_pipeline_library behind debug flag

Until the performance issues get resolved, enable the extension
if `VN_DEBUG=gpl`.

Also enable the required (and trivial) VK_KHR_pipeline_library.

Signed-off-by: Lina Versace <linyaa@google.com>
Reviewed-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22419>

venus: Do pipeline fixes for VK_EXT_graphics_pipeline_library

Signed-off-by: Lina Versace <linyaa@google.com>
Reviewed-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22419>

venus: Refactor pipeline fixup into two stages

Function vn_fix_graphics_pipeline_create_infos() had two interleaved
phases: discovery of needed fixes, and application of those fixes.
Move the discovery phase into new function
vn_find_graphics_pipeline_create_info_fixes(). The two-phase approach
will be simplify the implementation of VK_EXT_graphics_pipeline_library.

No intended change in behavior.

Signed-off-by: Lina Versace <linyaa@google.com>
Reviewed-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22419>

venus: Renames for VkGraphicsPipelineCreateInfo fixes

Use a more consistent naming scheme for everything. Follow-up patches
will implement VK_EXT_graphics_pipeline_library, which will make the
code significantly more complex, in which better names will make the
code more readable.

Except for replacing `any_fix` with `erase.mask`, this patch does modify
any procedural code. It merely renames structs and fields.

Signed-off-by: Lina Versace <linyaa@google.com>
Reviewed-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22419>

venus: Add enum vn_pipeline_type

No change in behavior.  Prepares for storing more state in graphics
libraries for VK_EXT_graphics_pipeline_library.

Signed-off-by: Lina Versace <linyaa@google.com>
Reviewed-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22419>

venus: Use VkImageAspectFlags in vn_subpass

No intended change in behavior. This little improvement will help
vn_pipeline track its graphics state more accurately.

Signed-off-by: Lina Versace <linyaa@google.com>
Reviewed-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22419>

venus: Drop incorrectly-used always-true pipeline vars

When writing vn_fix_graphics_pipeline_create_info() long ago, I defined
some always-true dummy vars that represented some
VkGraphicsPipelineLibraryFlagsEXT. In the conditions that decide which
state to fixup, I used the dummy vars.

My intent was that this would ease the implementation of
VK_EXT_graphics_pipeline_library, because some of the GPL fixup logic
would already be present.  I was wrong. After I studied GPL more,
I discovered that the conditions were using the dummy vars incorrectly.
The incorrect usage produced no bugs, because the vars are always true.

Delete the dummy vars.

Signed-off-by: Lina Versace <linyaa@google.com>
Reviewed-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22419>

venus: Fix subpass attachments

We must check for VK_ATTACHMENT_UNUSED.

Signed-off-by: Lina Versace <linyaa@google.com>
Reviewed-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22419>

venus: Fix crash when VkGraphicsPipelineCreateInfo::layout is missing

With VK_EXT_graphics_pipeline_library, the layout may be omitted or
ignored in incomplete pipelines.

Signed-off-by: Lina Versace <linyaa@google.com>
Reviewed-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22419>

venus: Erase pViewports and pScissors in fewer cases

We should avoid erasing VkGraphicsPipelineCreateInfo when possible
because the erasure add cpu overhead. Do not erase pViewports if
viewportCount is 0. Do not erase pScissors if scissorCount is 0.

Signed-off-by: Lina Versace <linyaa@google.com>
Reviewed-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22419>

venus: Sync protocol for VK_EXT_graphics_pipeline_library

Signed-off-by: Lina Versace <linyaa@google.com>
Reviewed-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22419>

asahi: flush denorms on exact fmin/fmax

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25788>

ac: Add rudimentary implementation of printing SDMA IBs.

This isn't meant to be exhaustive and only prints the packets
and fields which were interesting to me during debugging the
RADV transfer queue implementation.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25769>

ac: Print IP type for IBs.

This makes it clearer what is what when reading IBs
for debugging purposes.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25769>

ac: Rename ac_do_parse_ib to parse_pkt3_ib.

This function only parses PKT3 packets, not other types.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25769>

ac: Add amd_ip_type argument to ac_parse_ib and ac_parse_ib_chunk.

The current parser only supports PKT3 packets, but not all HW IP
types use this format. This makes it possible to support more
than one format depending on the IP type.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25769>

ac: Clarify SDMA opcode defines.

Make the naming of these defines consistent with others.
And add a few new opcodes and related definitions.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25769>

ac/gpu_info: Add some SDMA related information.

SDMA only supports sparse resources on GFX9+.
SDMA only supports DCC and HTILE on GFX10+.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25769>

frontends/va: Implement vaMapBuffer2

Reviewed-by: Sil Vilerino <sivileri@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25620>

pvr: Enable VK_KHR_uniform_buffer_standard_layout

Signed-off-by: Vlad Schiller <vlad-radu.schiller@imgtec.com>
Reviewed-by: Matt Coster <matt.coster@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25708>

pvr: Enable KHR_image_format_list

Signed-off-by: Vlad Schiller <vlad-radu.schiller@imgtec.com>
Reviewed-by: Matt Coster <matt.coster@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25707>

pvr: Enable VK_EXT_scalar_block_layout

Signed-off-by: Vlad Schiller <vlad-radu.schiller@imgtec.com>
Reviewed-by: Matt Coster <matt.coster@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25706>

radeonsi/ci: update failures and flakes

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25774>

radeonsi/ci: enable GLES CTS in the runner

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25774>

radeonsi/ci: enable GTF tests in the runner

These are closed-source CTS tests we have always been building.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25774>

radeonsi/ci: update the runner for new build scripts

Needed for the new build scripts at:
  https://github.com/marekolsak/marek-build
which is needed for:
- updating deqp to main
- updating glcts to opengl-cts-4.6.3

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25774>

zink: ignore unacquired swapchain images during end-of-frame flush

> Use of a presentable image must: occur only after the image is returned by flink:vkAcquireNextImageKHR, and before it is released by flink:vkQueuePresentKHR.
> This includes transitioning the image layout and rendering commands.

fixes #10005

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25787>

ci: bump DEQP_FRACTION for some jobs

Otherwise it can't complete in a reasonable amount of time.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25476>

ci: bump the number of tests per group from 500 to 5000 for Vulkan drivers

Except for lvp which already forces 1.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25476>

ci: update CTS to vulkan-cts-1.3.7.0

This contains many new tests, especially shader object tests.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25476>

nvk: Advertise VK_EXT_primitive_topology_list_restart

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/9645
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25786>

nvk: Implement VK_EXT_image_sliced_view_of_3d

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/9637
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25785>

anv: track render targets & render area changes separately

The following instructions :
   - 3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP
   - 3DSTATE_VIEWPORT_STATE_POINTERS_CC
   - 3DSTATE_SCISSOR_STATE_POINTERS

do not care about the content/format/count of the render targets, only
the size of the render area and count of viewport/scissor.

By tracking render targets & render area we can reduce the emission of
those instructions.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25778>

anv: reuse local variable for gfx state

No functional change.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25778>

nvk: Advertise VK_KHR_workgroup_memory_explicit_layout

It's all done for us in NIR.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/9623
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25780>

radeonsi: Fix plane size in si_copy_multi_plane_texture

Size was wrong and also need to scale sbox xy.

Fixes: 4f047c9583a ("radeonsi: Copy all planes with multi-plane staging textures")

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25737>

radv/rt: Use nir_shader_instructions_pass for lower_rt_instructions

Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25187>

nvk: Enable VK_EXT_load_store_op_none

VK_ATTACHMENT_STORE_OP_NONE_EXT is already supported via VK_KHR_dynamic_rendering, logic dictates it shouldn't need much else for VK_ATTACHMENT_LOAD_OP_NONE_EXT.
CTS is passing with the following results on dEQP-VK.renderpass*.load_store_op_none.* tests
Test run totals:
  Passed:        78/110 (70.9%)
  Failed:        0/110 (0.0%)
  Not supported: 32/110 (29.1%)
  Warnings:      0/110 (0.0%)
  Waived:        0/110 (0.0%)

Requires !24596 to achieve parity with proprietary driver

Closes #9638

Signed-off-by: Sid Pranjale <sidpranjale127@protonmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25771>

llvmpipe: Compile a nop texture function for unsupported configurations

Fixes a crash in Quake II RTX. Maybe the texture sample codegen will be
more robust in the future so we do not need to work around its
limitations here.

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25617>

util: improve BITFIELD_MASK and BITFIELD64_MASK on clang

gcc is able to optimize away either the modulo or the logical and.  This
makes no difference to gcc.

clang is only able to optimize away the logical and.  This allows clang
to generate faster code for BITFIELD_MASK.

As for BITFIELD64_MASK, this also makes no difference to clang except it
fixes a compile error for BITFIELD64_MASK(64):

  error: shift count >= width of type [-Werror,-Wshift-count-overflow]

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/9989
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25757>

mesa: don't pass Infs to the shader via gl_Fog.scale

This is for GLSL versions where Infs are undefined.

It also helps piglit/glsl-fs-fogscale that breaks when we move the fragment
shader code into the vertex shader across interpolation.

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Acked-by: Jesse Natalie on IRC
Acked-by: Erico Nunes on Gitlab
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25391>

glsl: Remove lower_discard().

Replaced by the new NIR pass.

i915g results:
total instructions in shared programs: 510678 -> 510714 (<.01%)
total temps in shared programs: 30429 -> 30426 (<.01%)

rv370 results:
total instructions in shared programs: 737649 -> 737656 (<.01%)
instructions in affected programs: 82 -> 89 (8.54%)
total temps in shared programs: 112093 -> 112094 (<.01%)
temps in affected programs: 6 -> 7 (16.67%)

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24763>

nir: Flatten ifs with discards in nir_opt_peephole_select for HW without CF.

i915g and r300-r400 don't have if statements, and discards are all
nir_intrinsic_discard_if.  We can flatten those discards here, saving a
separate GLSL pass to try to do so.

i915g:
GAINED: shaders/closed/xcom-enemy-unknown/413.shader_test FS

rv370:
GAINED: shaders/closed/xcom-enemy-unknown/12.shader_test FS
GAINED: shaders/closed/xcom-enemy-unknown/122.shader_test FS
GAINED: shaders/closed/xcom-enemy-unknown/132.shader_test FS
GAINED: shaders/closed/xcom-enemy-unknown/145.shader_test FS
GAINED: shaders/closed/xcom-enemy-unknown/146.shader_test FS
GAINED: shaders/closed/xcom-enemy-unknown/19.shader_test FS
GAINED: shaders/closed/xcom-enemy-unknown/413.shader_test FS
GAINED: shaders/closed/xcom-enemy-unknown/415.shader_test FS

Closes: #9918
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24763>

i915: Make I915_DEBUG=fs log shaders that fail to link due to CF.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24763>

lavapipe: VK_EXT_nested_command_buffer

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25615>

zink: add automatic swapchain readback using heuristics

in cases where apps (stupidly) do swapbuffers->blitframebuffer, there's
no functional way to (legitimately) perform readback on the just-presented
vk image, which leads to the existing acquire+present loop dance

this adds a counter threshold which, when exceeded, begins copying the
scanout image for swapchains to provide local readback on images without
the massive perf penalty of roundtrips

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25754>

zink: fix readback_present locking

queue needs to always be locked and presentation fence should be waited on

cc: mesa-stable

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25754>

zink: use screen destructor for creation fails

this should be more robust

fixes https://github.com/pal1000/mesa-dist-win/issues/155

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25663>

zink: reorder glsl_type_singleton_init_or_ref call

this enables the screen destructor to be used on failure cases

cc: mesa-stable

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25663>

zink: don't destroy uninitialized disk cache thread

cc: mesa-stable

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25663>

zink: don't destroy any simple_mtx_t objects during screen destroy

these don't allocate anything on init, but they assert if destroy is
called without an init, which is annoying for error handling

cc: mesa-stable

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25663>

zink: add some checks to determine whether queue is init on screen destroy

cc: mesa-stable

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25663>

iris: Support parameter queries for main planes

In order to return correct offsets, strides and possibly other
parameters. This is relevant for formats like NV12 and P010 where
the second plane, when produced by the Intel VAAPI decoder, uses the
same FD like the first one, but with an offset.

Right now there are only two modifiers with well defined indices
of auxiliary planes for semi-planar formats according to the local
copy of drm_fourcc.h, `I915_FORMAT_MOD_Y_TILED_GEN12_MC_CCS` and
`I915_FORMAT_MOD_4_TILED_MTL_MC_CCS`, sharing the same layout.
Assume that future `MC_CCS` modifiers will get defined accordingly
for now.

Cc: mesa-stable
Tested-by: Nanley Chery <nanley.g.chery@intel.com>
Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/9952
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25603>

egl: error out if we can't find an EGLDevice in _eglFindDevice()

Follow up of "egl/drm: get compatible render-only device fd for kms-only
device". Now we can properly error out when we don't find the EGLDevice
in _eglFindDevice().

Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24825>

egl/drm: get compatible render-only device fd for kms-only device

Things have been working by accident for split display/render SoCs when
using the GBM platform.

The device fd given to the GBM platform may be associated with a
KMS-only device, so _eglFindDevice() should find nothing (because the
global EGLDevice list only has render-capable devices). The only thing
making it work is that we don't error out when we go through the
EGLDevice list and can't find the device we are looking for. We simply
return the last EGLDevice from the list.

This patch fixes that. Now we look for a compatible render-only device
for the KMS-only in the GBM platform.

Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24825>

loader: add loader_is_device_render_capable()

Add helper function to check if a device is render-capable.

Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24825>

kmsro: try to use only compatible render-capable devices

Since commit "pipe-loader: add
pipe_loader_get_compatible_render_capable_device_fd()", we have a
function to query a render-only driver that is available and is
compatible with the KMS-only device.

So start to use it, instead of selecting whatever render-only driver is
available to use.

Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24825>

dri: add queryCompatibleRenderOnlyDeviceFd() to __DRI_MESA extension

With this change we are able to query the render node fd of a
render-only device compatible with a given KMS-only device (at the
egl/dri2 level).

It uses pipe_loader_get_compatible_render_capable_device_fd(), which was
added in commit "pipe-loader: add
pipe_loader_get_compatible_render_capable_device_fd()".

Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24825>

pipe-loader: add pipe_loader_get_compatible_render_capable_device_fd()

pipe_loader_get_compatible_render_capable_device_fd() receives the fd of
a KMS-only platform device, find a compatible render-only device that is
available and returns the fd of its DRM render node.

This function will be helpful to fix a long standing issue that is
preventing us to add support for EGL_EXT_device_drm_render_node for
split display/render SoCs. And it will also help KMSRO to select a
render-only driver that we are sure that is compatible, because
currently KMSRO uses whatever render-only driver is available.

In sort, in the EGL GBM platform case, the GBM device may be created
with a KMS-only device. The information of what render driver will be
selected by KMSRO is not available before creating the EGLDevice global
list. Without this information we don't have a render node to use in the
EGL_EXT_device_drm_render_node query. We've tried to fix this before,
but failed. See [1-2].

For the moment, this function only works for platform KMS-only devices.
For other types of KMS-only devices, we'll need to add more heuristics.

[1] Detailed explanation of the issue:
https://gitlab.freedesktop.org/mesa/mesa/-/issues/5591

[2] Previous attempt to fix:
https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12796

Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24825>

loader: add driver list as parameter in loader_open_render_node_platform_device()

In a later commit in this series, we'll need to open the first supported
render-only platform device that we can find.

In order to avoid calling loader_open_render_node_platform_device()
multiple times (what is quite expensive), change this function to take a
driver list (instead of a single driver name) as parameter.

Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24825>

loader: rename loader_open_render_node() to loader_open_render_node_platform_device()

This function ignore devices that are not on the platform bus. So rename
it to better reflect that.

Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24825>

venus: properly expose KHR_external_fence/sempahore_fd

Summary:
1. hide those extensions for vtest since there's no export support
2. do block waiting upon export without implicit fencing support

Fixes: 68a478870d2 ("venus: expose KHR_external_fence/sempahore_fd extensions")
Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25633>

nvk: Implement VK_EXT_depth_bias_control

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/9628
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25765>

nvk: Re-sort device features

Vulkan versions, then KHR, then EXT, then vendor.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25765>

nvk: Add a nvk_cmd_buffer_dirty_render_pass() helper

This makes the dirtying work properly in the case where dynamic
rendering is being used.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25765>

anv: set ComputeMode.PixelAsyncComputeThreadLimit = 4

Heuristic-based optimization throttling CCS work (async compute).
Without throttling, background compute work consumes all threads,
deminishing performance gains by running dispatch in parallel with
3D work.

Optimization is heuristics based, meaning a workload might slow
down when using async compute.

Best value: PixelAsyncComputeThreadLimit = 4. On DG2, this
equates to a max CCS thread occupancy of 37.5%.

Reviewed-by: Sagar Ghuge <sagar.ghuge@intel.com>
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25508>

nir/algebraic: use only signed_zero_preserve_* for addition by 0 patterns, etc.

Some GLSL versions will set inf_preserve but not the other flags.
Additions by 0 only affect signed zeros.

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25392>

nir: split FLOAT_CONTROLS_SIGNED_ZERO_INF_NAN_PRESERVE_FP* flags

GLSL doesn't preserve NaNs, but it optionally preserves Infs.

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25392>

anv: fix debug string for PC flush

Signed-off-by: Rohan Garg <rohan.garg@intel.com>
Fixes: fc5cb54 ('anv: Add debug messages for DEBUG_PIPE_CONTROL')
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25690>

etnaviv: zero shared TS metadata block

Make sure to zero the shared TS metadata memory. Without this invalid
TS data may be mistaken as being valid, causing GPU hangs.

CC: mesa-stable
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <cgmeiner@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25751>

etnaviv: use correct blit box sizes when copying resource

The padded width/height is stored in samples, while the blit box
dimensions need to be specified in pixels. Use the unpadded
width/height of the resource levels to generate the blit box
dimensions used to copy a resource. The blit code already extends
those sizes to the padded sizes when necessary and possible.
Fixes crashes in some piglit tests with MSAA active.

CC: mesa-stable
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <cgmeiner@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25751>

broadcom/qpu: Remove duplicate variable opcode

Fix defect reported by Coverity Scan.

Evaluation order violation (EVALUATION_ORDER)
write_write_typo: In opcode = opcode = desc->opcode_first, opcode is
written twice with the same value.

Fixes: 3b20208f03f ("broadcom/qpu: add pack/unpack support for v71")
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25735>

intel/ds: provide names for different events of a timeline's row

Previously all items on a timeline row would have the same name. This
change uses the tracepoint names to put into the timeline instead.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Tested-by: Felix DeGrood <felix.j.degrood@intel.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25730>

u_trace: generate tracepoint name array in perfetto header

The driver glue doesn't have access to that information in a
centralized place. If you want to generate perfetto iid, you need
access to all names.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Tested-by: Felix DeGrood <felix.j.degrood@intel.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25730>

u_trace: generate tracepoint index parameter in perfetto callbacks

Useful to figure out what's the tracepoint name you're implementing.
We'll use this in the intel perfetto integration glue to index into an
array of perfetto iid.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Tested-by: Felix DeGrood <felix.j.degrood@intel.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25730>

anv: don't flush_llc on gen9

Fixes: 3d993e63bb59 ("anv: Enable barrier handling on video engines ")
Closes: mesa/mesa#9988

Signed-off-by: Hyunjun Ko <zzoon@igalia.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25762>

virgl: Fix logic for reporting PIPE_MIRROR_CLAMP

Fixes: 9efe50c83 (virgl: report MIRROR_CLAMP features better)

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25755>

intel-clc: avoid using spirv-linker.

There is not real need to use the spirv-linker here at all,
we can just read all the CL C files into one buffer, then compile
that buffer in a single pass.

This worksaround an issue seen with llvm17 and opaque pointers
and the spirv linker.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25759>

radeonsi/ci: update gfx1100 results

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25484>

winsys/amdgpu: don't send CP_GFX_SHADOW chunk if shadow address is not set

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25484>

radeonsi: disable register shadowing without SR-IOV to fix bad performance

Fixes: f71607c8d3c - radeonsi/gfx11: enable register shadowing by default

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25484>

ac/gpu_info: don't allow register shadowing with SR-IOV due to bad performance

This is only for gfx11.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25484>

docs: Mention 'meson devenv' in the pre-install test instructions

Also immunize ourselves against section renumbering.

Reviewed-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25686>

anv: advertise VK_KHR_global_priority_queue

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25758>

radv/rt: Reject hits within 10ULP of previous hits in emulated RT

This is an alternative workaround that fixes double hits on shared edges
failing some watertightness CTS tests.

Fixes: e034ba1c44 ("radv/rt: Miss rays that hit the triangle's v edge")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24093>