panfrost: Don't check alpha test in fs_required on Bifrost+

Alpha testing is only native on Midgard. Saves a few instructions.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15797>

intel/compiler: Fix sample_d messages on DG2

DG2 can only do sample_d and sample_d_c on 1D and 2D surfaces.  The
maximum number of gradient components and coordinate components should
be 2.  In spite of this limitation, the Bspec lists a mysterious R
component before the min_lod, so the maximum coordinate components is 3.

Fixes the following Vulkan CTS failures on DG2:

    dEQP-VK.glsl.texture_functions.texturegradclamp.isampler1d_fragment
    dEQP-VK.glsl.texture_functions.texturegradclamp.isampler2d_fragment
    dEQP-VK.glsl.texture_functions.texturegradclamp.sampler1d_fixed_fragment
    dEQP-VK.glsl.texture_functions.texturegradclamp.sampler1d_float_fragment
    dEQP-VK.glsl.texture_functions.texturegradclamp.sampler2d_fixed_fragment
    dEQP-VK.glsl.texture_functions.texturegradclamp.sampler2d_float_fragment
    dEQP-VK.glsl.texture_functions.texturegradclamp.usampler1d_fragment
    dEQP-VK.glsl.texture_functions.texturegradclamp.usampler2d_fragment

The Fixes: tag below is a bit misleading. This commit fixes some test
cases similar to ones fixed by the Fixes: commit.  I just want to make
sure this commit gets applied everywhere that commit was also applied.

Fixes: 635ed58e527 ("intel/compiler: Lower txd for 3D samplers on XeHP.")
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15781>

dzn: Pass a NULL ralloc context to dxil_create_validator()

instance is not allocated with ralloc, we can't pass it as a ralloc
context to dxil_create_validator().

Fixes: 09c2016d6be ("dzn: use dxil_validator")
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15787>

vulkan,docs: Add documentation for Vulkan dispatch

Acked-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15472>

vulkan,docs: Document vk_device

Acked-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15472>

vulkan,docs: Document vk_physical_device

Acked-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15472>

vulkan,docs: Document vk_instance

Acked-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15472>

vulkan,docs: Document vk_object_base

Acked-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15472>

vulkan: vk_object_base_init/finish have no unused parameters

Acked-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15472>

anv/formats: Relax usage checks if EXTENDED_USAGE_BIT is set

Reviewed-by: Ivan Briano <ivan.briano@intel.com>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14153>

android: Set max platform-sdk-version to 10000

platform-sdk-version::max require updating every time new Android releases.
It makes sense to set this value to some unreachable maximum value to avoid
such patches. Such approach will allow to use stable mesa3d without any
changes with newest Android as well.

As was suggested by Yiwei Zhang we can use value 10000 that corresponds to
CUR_DEVELOPMENT [1] definition which may be used by AOSP/master in between
of Android major releases.

[1]: https://developer.android.com/reference/android/os/Build.VERSION_CODES#CUR_DEVELOPMENT

Signed-off-by: Roman Stratiienko <r.stratiienko@gmail.com>
Reviewed-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15747>

panfrost: Split out image access tracking

This logic is not device-specific. It is pulled from our existing Bifrost image
implementation and will be reused for Valhall.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15795>

panfrost: Add helpers to set batch masks

This logic is not device specific and will be used for both Bifrost and Valhall.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15795>

panfrost: Disable PIPE_CAP_PRIMITIVE_RESTART on v9

Valhall removed the ability to set an explicit primitive restart index as
required by desktop OpenGL, in favour of fixed primitive restart indices only as
required by OpenGL ES.

Set the CAPs accordingly so that mesa/st lowers unusual primitive restart
indices at draw call time.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15795>

panfrost: Make alpha=0 NOP / 1 store Bifrost only

These fields were removed in Valhall in favour of a simpler overdraw mechanism.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15795>

panfrost: Move assign_vertex_buffer to pan_helpers

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15795>

panfrost: Hide AFBC on Valhall

The relevant data structures have been shuffled a bit. We need to wire up AFBC
for Valhall; however, that's out of scope for the initial bring up. Just hide it
so we can build.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15795>

panfrost: Adapt pan_shader.h for Valhall

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15795>

panfrost: Add shader_stage helper

For Valhall, which specifies these in the shader program descriptor.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15795>

panfrost: Add panfrost_make_resource_table helper

For Valhall drivers.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15795>

panfrost: Hide parts of pan_encoder.h for Valhall

These pertain to data structures that no longer exist.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15795>

panfrost: Control tiler memory usage

Ensure we don't hit OOM when rendering at 8192x8192 on Valhall by disabling
the smallest bin size of the hierarchy mask.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15795>

panfrost: Handle stencil texturing on Valhall

Use a Bifrost compatible path. It's not clear this is optimal but it passes the
tests and is no worse than what we do on Bifrost.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15795>

panfrost: Handle Valhall texturing

Surface descriptors have been replaced by plane descriptors, which facilitate
the intermediate layout of textures. This allows for more sophisticated handling
of texture compressions, of particular to interest to copy_image. However, it
requires a considerable amount of new logic to handle.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15795>

frontend/va: Create decoder once the max_references is updated

Issue: When a video is decoded where the max_references is updated the
decoder keeps using same old value. This results into green patches and
decoding is not proper.

Root Cause: The max_references is updated only once when the instance is
created for the first time.

Fix: Added a check along with the context->decoder to check if the
context->templat.max_references has changed. If yes, then we go ahead
and create the decoder again.

Reviewed-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Krunal Patel <krunalkumarmukeshkumar.patel@amd.corp-partner.google.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15750>

zink: ci updates

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15536>

zink: export PIPE_CAP_SEAMLESS_CUBE_MAP_PER_TEXTURE

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15536>

zink: run the cubemap -> array compiler pass if the shader key is set

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15536>

zink: handle nonseamless cube sampler binding

now when a cube is sampled with a nonseamless sampler, the relevant shader
stage is flagged for a variant update, the

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15536>

zink: create an array view for all cube samplerviews

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15536>

zink: set nonseamless hint for sampler states

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15536>

zink: handle shader key variants that have nonseamless cubemaps

this is just the program-based handling, which means finding the right
variant and/or creating new ones based on matching a 32bit mask indicating
which textures must be rewritten

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15536>

zink: specify struct member name when copying inline uniforms for gfx variants

avoid memory mismatch if inline uniform values aren't first member of struct

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15536>

zink: support nir_op_imod

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15536>

zink: add a gently mangled version of the d3d12 cubemap -> array compiler pass

this differs in that it doesn't handle images and also doesn't filter
based on sampler type, instead using a 32bit mask for determining which
samplers to rewrite

also zink doesn't use sampler derefs

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15536>

zink: rename a variable

no functional changes

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15536>

pan/va: Add LD_TILE.v3.f16 packing test

This tests the staging register behaviour.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15793>

pan/bi: Consider flow control in DCE

We don't want to remove instructions like `NOP.wait` on Valhall; this would be
tantamount to deleting barriers.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15793>

pan/bi: Don't lower vertex_id for malloc IDVS

Based on hardware behaviour, it appears vertex_id is zero-based with the legacy
geometry flow but not with the new malloc IDVS flow. Since the geometry flow is
per-shader (not per-machine), there's not a good way to communicate this to NIR.
Rather than trying to shoehorn this obscure detail into NIR, just do the
lowering ourselves instead of in NIR. It's not much more code anyway.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15793>

pan/bi: Report whether workgroups can be merged

This flag gates a Valhall hardware optimization for compute shaders.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15793>

pan/bi: Avoid masked writes for now

Our swizzle lowering optimizations depend on replication of scalar fp16. This
holds on Bifrost (at least for now), but not on Valhall which has proper support
for write masks. For now, enforce Bifrost-compatible behaviour as we do not make
use of the write masks on Valhall yet.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15793>

pan/bi: Generate LD_BUFFER on Valhall

Replace LOAD.ubo with LD_BUFFER since the .ubo segment doesn't exist on Valhall.
We could do this with a lowering pass instead but this is probably fine.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15793>

pan/bi: Make psiz variants

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15793>

pan/bi: Lower gl_PointSize to FP16 on Valhall

It is unclear if FP32 point sizes are supported on Valhall -- I can't get the
DDK to use them at any rate. Always lower them to FP16 and store them as FP16
for hardware use.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15793>

pan/bi: Force psiz to mediump

To match driver behaviour.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15793>

pan/bi: Set table for Valhall LD_ATTR

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15793>

pan/bi: Emit Valhall-style varying stores

Varying stores was changed in Valhall. Rather than using attribute descriptors
like on Bifrost and Midgard, on Valhall we store to memory directly with
hardware-allocated buffers. This requires a new implementation of store_output,
with special provisions for writing gl_PointSize from a position shader.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15793>

pan/bi: Emit Valhall-style varying loads

Memory-allocated IDVS requires special varying load instructions that take an
offset into the hardware-allocated varying buffer, as opposed to a varying slot.
Emit these instructions.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15793>

pan/bi: Track whether the malloc IDVS flow is used

This affects what instructions the fragment shader uses. Will be used for the
legacy geometry flow in blit shaders. Whether that is a good idea remains to be
seen, admittedly.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15793>

pan/bi: Handle Valhall texturing in helper analysis

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15793>

pan/bi: Emit Valhall texture instructions

Valhall uses an updated version fo the TEXC path. To avoid disrupting the
existing Bifrost code, add a new Valhall-specific texture path that generates
the new-style texture instructions.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15793>

pan/bi: Specialize BLEND emit for Valhall

Fewer arguments compared to Bifrost; the corresponding information is encoded in
a Valhall-specific blend shader prologue instead.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15793>

pan/bi: Waits before tilebuffer access on Valhall

On Bifrost, this is handled in the scheduler. Until we grow a Valhall scheduler,
add a NOP with the appropriate flow control. This is correct but carries a small
performance cost.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15793>

pan/bi: Fix spilling on Valhall

We need a slightly different idiom on Valhall, since the segment modifiers no
longer exist but we now have an immediate offset.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15793>

pan/bi: Mark LD_TILE as w=format

This tracks register usage more precisely for LD_TILE, which is an encoding
difference on Valhall.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15793>

panfrost: Add a table for images

For the default Valhall ABI.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15793>

radv: improve failure logging for amdgpu on init

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15771>

pan/va: Allow small constants in register pairs

They are zero extended 32->64-bit. Allow this. Noticed debugging spilling on
Valhall.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15756>

pan/va: Add flow control lowering pass

Something an instruction has two logic flow controls, namely wait + reconverge.
These are orthogonal -- we need to insert a NOP to handle this. Add a lowering
pass that works out flow control to replace the ad hoc previous va_pack_flow.

Fixes dEQP-GLES31.functional.ssbo.layout.single_basic_type.shared.lowp_vec3.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15756>

pan/va: Don't truncate slots

Causes BARRIER not to work.

Fixes: f45654af595 ("pan/va: Add packing routines")
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15756>

pan/va: Model image load instructions

These use the attribute pipe, the new versions of LD_ATTR_TEX, but reading
texture descriptors instead of attribute descriptors unlike their Bifrost
predecessors.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15756>

pan/va: Pack LEA_TEX_IMM

Mostly automatic.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15756>

pan/va: Add indirect LEA_{ATTR, TEX}

For parity with Bifrost. We might need these for images.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15756>

pan/bi: Model Valhall image loads

Like LD_ATTR_TEX.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15756>

intel/nir: don't report progress on rayqueries if no queries

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes: c78be5da300a ("intel/fs: lower ray query intrinsics")
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15769>

turnip: Implement VK_EXT_primitives_generated_query

Similar to pipeline statistics but done for a single counter.

We use REG_A6XX_RBBM_PRIMCTR_7 to get generated primitives
and not PRIMCTR_8 because PRIMCTR_7 counts pre-clipped prims
while PRIMCTR_8 counts them after clipping.

OpenGL spec for GL_PRIMITIVES_GENERATED says:
 "Subsequent rendering will increment the counter once for every
  vertex that is emitted from the geometry shader, or from the
  vertex shader if no geometry shader is present."

Passes tests:
 dEQP-VK.transform_feedback.primitives_generated_query.*

Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15746>

radv: mark all states declared dynamic at pipeline creation

It will be easier to merge dynamic states from libraries this way.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15764>

radv: add a new helper to determine if rasterization is enabled

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15764>

radv: fix dynamic raster discard with VK_EXT_depth_clip_control

Fixes: 43e83949dc6 ("radv: implement VK_EXT_depth_clip_control")
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15764>

zink: update ci list

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14541>

zink: ci fixup

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14541>

zink: handle zombie swapchains

inject a dummy, matching image that can be used until the frontend
catches up and rescues us from whatever is happening

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14541>

driconf: add override for Xwayland

zink needs this to avoid deadlocking on startup

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14541>

zink: export PIPE_CAP_DEVICE_RESET_STATUS_QUERY

this can work now

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14541>

zink: it's kopperin' time

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14541>

zink: add kopper api

Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14541>

zink: check whether clear is enabled before applying in unbind

this is implicit, but make it explicit

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14541>

zink: move variable decl up in unbind_fb_surface

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14541>

zink: pass index to unbind_fb_surface

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14541>

zink: move drirc handling up

this can modify instance creation

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14541>

zink: split surface creation more to allow disabling caching

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14541>

zink: add fail logging for drmPrimeFDToHandle

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14541>

zink: add VK_KHR_swapchain_mutable_format

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14541>

zink: use two submits for every queue submit

first one empty for now

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14541>

zink: change early returns in zink_blit to gotos

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14541>

zink: move blit src/dst decls up in function

no changes

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14541>

zink: move update_framebuffer_state() higher up in file

no changes

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14541>

zink: put screen param into flush queue global data

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14541>

zink: move flush queue init down a little further

no functional changes

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14541>

gallium: Learn about kopper

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14541>

glx: Learn about kopper

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14541>

egl: Learn about kopper

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14541>

kopper: Define the driver interface

The loader extension provides upcalls to get surface state (native
resource and size) into the driver. The driver extension is called by a
kopper-aware loader in preference to __DRI_SWRAST's createNewDrawable.

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14541>

meson: Define a HAVE_XXXX macro for every gallium driver we build

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14541>

st/manager: update framebuffer size if texture has been resized

zink/kopper can and does expect this when resizing swapchains,
so don't ignore it

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14541>

clc/tests: use dxil_validator

Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15751>

dzn: use dxil_validator

Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15751>

microsoft/spirv_to_dxil: use dxil_validator

This has one negative side-effect; we're no longer able to print
validation errors without dxcompiler.dll. I doubt that's a real problem,
but if it is, we should add this ability to dxil_validator instead of
having a second implementation here.

The reasons I didn't try adding this in the first place is:

1. This code seems a bit janky; it doesn't consult the "known"-variable
   to figure out if the encoding is OK, and it's lacking a fallback path
   in that case.
2. It seems unlikely that the compiler varies the encoding of the output
   in the first place; one of the two code-paths in here is probably
   untested.
3. Since dxil_validator leaves reporting to the call-site, we'd need to
   either add and output-encoding to the API (yuck), or re-encode the
   string to UTF-8 using WinAPI.

Right now, it seems questionable if fixing all of the above is worth it.

Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15751>

d3d12: use dxil_validator

Now that we have a shiny, new dxil validator interface, let's start
using it in the D3D12 gallium driver.

Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15751>

microsoft/compiler: add common dxil-validator API

This API is only available on Windows, which is the only OS where DXIL
validation is a requirement, and where DXIL.dll (and dxcompiler.dll) are
available.

Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15751>