radeonsi: fix uses_instanceid for merged mono shader stage

prev_shader.info.uses_instanceid is set by si_nir_lower_vs_inputs()
when attribute divisor is used.

Fixes: 894d92995cd ("radeonsi: monolithic VS emit prolog in nir directly")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/9042
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23101>

rusticl/device: add cl_khr_integer_dot_product ext

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23011>

rusticl/kernel: remove nir_lower_pack pass

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23011>

rusticl/clc add integer_dot_prod feature macros

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23011>

rusticl/api: add integer_dot_product api

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23011>

rusticl/types add ::new for cl_dev_idp_accel_props

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23011>

anv: drop duplicate checks when setting the compressed bit

We need compression tracking for full resolves and at the moment only
CCS_E has full resolves.

Signed-off-by: Rohan Garg <rohan.garg@intel.com>
Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22737>

aco/optimizer: allow DPP to use VOP3 on GFX11

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23059>

aco: use can_use_input_modifiers helper

Foz-DB GFX1100:
Totals from 80 (0.06% of 132657) affected shaders:
CodeSize: 504500 -> 503660 (-0.17%)
Instrs: 95033 -> 94824 (-0.22%)
Latency: 629695 -> 629235 (-0.07%)
InvThroughput: 97105 -> 97008 (-0.10%)
VClause: 1779 -> 1777 (-0.11%)
Copies: 3233 -> 3236 (+0.09%); split: -0.03%, +0.12%

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23059>

aco: use get_operand_size for dpp opt

This matters now that v_fma_mixlo_f16/v_fma_mixhi_f16 can use dpp.

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23059>

aco: add helper function for can_use_input_modifiers

Some instructions have restrictions that can't be expressed with the bitfield.

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23059>

aco/ra: use fmac with DPP/opsel on GFX11

Foz-DB GFX1100:
Totals from 2 (0.00% of 132657) affected shaders:
CodeSize: 8764 -> 8732 (-0.37%)

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23059>

aco/ra: use smaller operand stride for VOP3P with DPP

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23059>

aco: add withoutVOP3 helper

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23059>

aco/gfx11: use fmamk/fmaak with opsel

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23059>

aco: introduce helper to swap valu operands with modifiers

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23059>

gitlab-ci: add capture for i915 error state

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Acked-by: David Heidelberg <david.heidelberg@collabora.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23099>

ci/venus: add missing flakes

Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22888>

ci/gtest-runner: fix results reporting

Acked-by: Daniel Stone <daniels@collabora.com>
Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22888>

ci/amd: add radeonsi-raven-va-full job to cover all VA-API tests

Acked-by: Daniel Stone <daniels@collabora.com>
Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22888>

ci/amd: update VA-API expectations

Acked-by: Daniel Stone <daniels@collabora.com>
Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22888>

ci/gtest: improve the runner script

Acked-by: Daniel Stone <daniels@collabora.com>
Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22888>

ci: update libva to 2.18.1

Acked-by: Daniel Stone <daniels@collabora.com>
Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22888>

ci/rules: radeonsi VAAPI rules should include also VA-API targets

Acked-by: Daniel Stone <daniels@collabora.com>
Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22888>

ci/amd: re-enable VA-API testing

Acked-by: Daniel Stone <daniels@collabora.com>
Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22888>

nir: Fix incorrect comment

We have 1-bit bools these days.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23107>

nir: Remove stale TODOs

If nobody has added def-use lists for registers in all this time, it's probably
because we don't want them after all ;)

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23107>

pvr: Return VkResult from winsys buffer_map operation

This allows VK_ERROR_MEMORY_MAP_FAILED to propagate correctly.

Signed-off-by: Matt Coster <matt.coster@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23023>

pvr: Assorted cleanup

Signed-off-by: Matt Coster <matt.coster@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23023>

pvr: Use common physical device enumeration

The PowerVR IP does not contain display hardware - this means it is
always necessary to open two separate devices for render and display.
The try_create_for_drm callback is not suitable for this configuration,
so we use the enumerate callback instead.

The previous implementation did not check that the discovered display
device was compatible with the render device - this is corrected by
unifying the compatibility lists into pvr_drm_configs.

The pvr driver is not currently supported on systems which contain
multiple compatible render or display devices, so the enumerate callback
implementation returns the first discovered render device and its
compatible display device.

This change also removes the workaround for drmGetDevices2() required
after libdrm commit 8cb12a2528d795c45bba5f03b3486b4040fb0f45. The
upstream fix has been in releases of libdrm for over a year now, and
mesa requires reasonably a recent version which is new enough to
contain it.

Signed-off-by: Matt Coster <matt.coster@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23023>

pvr: Rename primary_{device,fd,path} to display_*

This makes it explicit that primary_fd is NOT just a handle to the
primary node on render_fd - rather it's a handle to the primary node
on a separate display device.

Signed-off-by: Matt Coster <matt.coster@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23023>

pvr: Drop pdevice from pvr_physical_device_get_supported_extensions()

This parameter doesn't make a lot of sense - it's currently unused, but
it can never be valid anyway since the return of this function is used
in the early initialization of pvr_physical_device.

Signed-off-by: Matt Coster <matt.coster@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23023>

pvr: Squeeze fd handling into winsys layer

This sets us up to make use of the common physical device initialization
code.

As well as lifting the fd handles out of the implementations, this
pushes creation and destruction of the fds into the winsys layer. In
order for this to make sense, the winsys object is now created *before*
each pvr_device or pvr_physical_device. If there's an error setting up
the winsys instance, there's no point in continuing to create either
one.

Also lifts alloc to the winsys layer since there's nothing special about
it in either implementation.

Signed-off-by: Matt Coster <matt.coster@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23023>

pvr: Fix incorrect error handling in pvr_render_ctx_switch_init()

Partially fixes: dEQP-VK.api.device_init
  .create_instance_device_intentional_alloc_fail.basic

Signed-off-by: Matt Coster <matt.coster@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23023>

pvr: Fix incorrect error return in pvr_ctx_sr_programs_setup()

Partially fixes: dEQP-VK.api.device_init
  .create_instance_device_intentional_alloc_fail.basic

Signed-off-by: Matt Coster <matt.coster@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23023>

pvr: Propagate errors as VkResults from ioctls through winsys

Partially fixes: dEQP-VK.api.device_init
  .create_instance_device_intentional_alloc_fail.basic

Signed-off-by: Matt Coster <matt.coster@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23023>

pvr: Return VkResult from pvr_winsys_create()

Previously this function simply returned NULL on any error, which
caused OOM errors to be suppressed as initialization failures.

Signed-off-by: Matt Coster <matt.coster@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23023>

radv: Use const keyword more.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23053>

ac: Use const keyword for some function arguments.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23053>

intel: infer scalar'ness locally for brw_vectorize_lower_mem_access

Signed-off-by: Rohan Garg <rohan.garg@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23098>

intel: update comments about non-existent function parameter

Signed-off-by: Rohan Garg <rohan.garg@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23098>

intel: drop unused is_scalar function parameter in brw_nir_apply_key

Signed-off-by: Rohan Garg <rohan.garg@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23098>

intel: infer scalar'ness locally for brw_postprocess_nir

Signed-off-by: Rohan Garg <rohan.garg@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23098>

pvr: Fix transfer image clearing PBE packmodes

The PBE input format must match format in clear color registers.
Fixes image clearing for following formats:
  - B4G4R4A4_UNORM_PACK16
  - A8B8G8R8_UNORM_PACK32
  - R5G6B5_UNORM_PACK16
  - A1R5G5B5_UNORM_PACK16
  - R8G8B8A8_SNORM
  - R8G8_UNORM
  - R8G8_SNORM
  - R8_UNORM
  - R8_SNORM
  - A2B10G10R10_UINT_PACK32 - only packmode U32 supported

For some of the norm formats the clear color register format was
changed from integer (pvr_float_to_sfixed) to float (pvr_float_to_f16).
This change was done to match the default PBE emit Norm settings.
An alternative way to fix this would have been to change the PBE
Norm setting.

Signed-off-by: Oskar Rundgren <oskar.rundgren@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23082>

pvr: Use idalloc as the allocator for sync prims

Previously we allocated the sync prim at the end of the block and
also freed from the end. This is problematic if things are freed
out of order and some new ones allocated within the frees.

This commits uses the idalloc to keep track of the sync prim block
allocations.

Signed-off-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23076>

pvr: Move pvrsrv sync prim code into new pvr_srv_sync_prim.{c,h}

Some setup for later on when we'll start keeping track of sync
prim allocation with a proper allocator.

Signed-off-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23076>

pvr: Add pvrsrvkm sync prim set bridge call

To reuse a sync prim we'll need to set it to 0 so adding the
bridge call for that.

Signed-off-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23076>

pvr: Simplify descriptor set layout dump separators

There's no need for macros here, and they make clang-format unhappy.

Signed-off-by: Matt Coster <matt.coster@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23028>

pvr: Add PVR_DEBUG=vk_desc option to dump descriptor set layouts

Prints on every call to vkCreate{DescriptorSet,Pipeline}Layout().

Previously, this could not be turned on or off at runtime, and was
dependent on DEBUG.

Signed-off-by: Matt Coster <matt.coster@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23028>

pvr: Remove bad assert in pvr_clear_attachments()

Fixes: dEQP-VK.renderpass.suballocation.formats.d16_unorm.input
  .dont_care.dont_care.clear
Fixes: dEQP-VK.renderpass.suballocation.formats.d32_sfloat.input
  .dont_care.dont_care.clear

Signed-off-by: Matt Coster <matt.coster@imgtec.com>
Reported-by: James Glanville <james.glanville@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23024>

pvr: Return correct pbe_accum_format size for A2B10G10R10_UINT_PACK32

Fixes: dEQP-VK.renderpass.suballocation.formats.a2b10g10r10_uint_pack32
  .input.dont_care.dont_care.clear

Signed-off-by: Matt Coster <matt.coster@imgtec.com>
Reported-by: James Glanville <james.glanville@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23024>

pvr: Set output_offset correctly in pvr_clear_color_attachment_static()

Fixes: dEQP-VK.renderpass.suballocation.formats.r16g16_sint.input
  .dont_care.dont_care.clear

Signed-off-by: Matt Coster <matt.coster@imgtec.com>
Reported-by: James Glanville <james.glanville@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23024>

intel/compiler: Postpone most int64 lowering to brw_postprocess_nir

Float conversions continue to be lowered early at the same time as
nir_lower_doubles, which we run early so we don't have to run it for
every shader key variant.  However, all other int64 lowering is now
done late, after nir_opt_load_store_vectorize(), allowing it to
comprehend basic arithmetic on 64-bit addresses.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23064>

nir: Add a variant of nir_lower_int64 for float conversions only

We'd like to postpone most int64 lowering until pretty late in the
process, because e.g. turning iadd@64 into (unpack + add-low + add-high
+ compare + b2i32 + repack) sequences makes it difficult for many
optimization passes to detect basic arithmetic patterns.  In particular,
nir_opt_load_store_vectorizer becomes unable to handle basic offset math
on 64-bit addresses.

We'd like to do double precision lowering earlier in the process,
however.  One snag is that nir_lower_int64's lower_2f and lower_f2 can
produce operations that may need lowering by nir_lower_doubles(), so
it's crucial to run those sets of lowering together.

To handle this, we make a new entrypoint that does nir_lower_int64
but skips everything except float conversions.  Note that the newly
produced instructions will still be lowered according to the full set
of int64 lowering options; this shouldn't be a huge deal.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23064>

pvr: Use the suballocator for queries

Fixes:
  dEQP-VK.api.object_management.max_concurrent.query_pool

Signed-off-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Reviewed-by: Luigi Santivetti <luigi.santivetti@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23071>

pvr: Fix vk_free() in vkCreateRenderPass2() error path

There was a missing vk_free() and one at the wrong place.

Fixes:
  dEQP-VK.api.object_management.alloc_callback_fail.render_pass

Signed-off-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23070>

pvr: Fix possible allocation of 0 size

Some tests had `input_count == 0` so the driver would attempt to
`vk_alloc()` a size of zero and error out. Add some checks to
prevent that.

Signed-off-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23070>

pvr: Fix missing invalidation of the command buffer

If the command buffer already had an error while recording, on
vkEndCommandBuffer() the command buffer wasn't being transitioned
to the invalid state.

Signed-off-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23069>

pvr: Remove custom status in command buffer

Some code still used our custom `status` field within the command
buffer. This could lead to unreliable error handling since we're
using the common vk_command_buffer handling.

This commit removes the field and changes the error paths to use
the common code instead.

Signed-off-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23069>

radv: fix a sync issue with primitives generated query and NGG/legacy

On RDNA1&2, the driver needs to support both NGG and legacy for
primitives generated query because we can't know that before starting
queries.

To get the query pool results, we check the availability bit wrote by
the SAMPLE_STREAMOUTSTATS packet but the GDS copy was emitted after,
which means the availability bit might be TRUE before the GDS copy is
actually done.

Fix this by emitting the GDS copy before to ensure the availability is
TRUE for both results.

This fixes recent updates in
dEQP-VK.transform_feedback.primitives_generated_query.* because the
tests no longer wait for the fence.

Cc: mesa-stable
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23080>

nir/lower_io: don't renumber VS inputs when not called from a linker

This fixes a Nine regression. The comment explains it.

The new varying linking code will set true here.

Fixes: d29dd333 - nir: assign IO bases in nir_lower_io_passes
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/8935

Reviewed-by: Timothy Arceri <tarceri@yahoo.com.au>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22859>

loader/dri3: temporarily work around a crash when front is NULL

Commit c65bde7b1e introduced a regression where under certain
circumstances `front` may be NULL, thus leading to a crash. It's not
currently known what exactly causes `front` to become NULL, nor we can
revert the offending commit, because there had been too many unrelated
changes that now depend on this commit.

So until someone comes up with a proper fix, let's add a workaround so
instead of crashing we just return from the function early.

This commit was tested with the bug `8982` and helps with the crash
with no other noticeable problems.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/8982
Fixes: c65bde7b1e ("frontend/dri: inline __DRIdrawable in dri_drawable, make __DRIdrawable opaque")
Cc: mesa-stable
Signed-off-by: Konstantin Kharlamov <Hi-Angel@yandex.ru>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23093>

r600/sfn: use correct FS output location if not all outputs are used

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/8973

Fixes 79ca456b4837b3bc21cf9ef3c03c505c4b4909f6
    r600/sfn: rewrite NIR backend

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23058>

r600/sfn: move kill handling fully to scheduling

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23058>

r600/sfn: add dependencies for kill instructions

GDS and image instructions have side effects, so they can't be
reordered with kill instructions.

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23058>

r600/sfn: move kill instruction test to alu instruction

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23058>

glsl: Delete dead intrinsics

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23041>

mesa/main: Exit early when trying to create an unsupported context API

Fixes: adbe8b6c17a ("mesa: optimize out _mesa_is_desktop_gl*() and _mesa_is_gles*() calls when not built")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/9038
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23068>

vtn: add more info to bitcast bit size error message

Reviewed-by: Caio Oliveira <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22932>

vtn: print spirv ids for type mismatch in bcsel

Reviewed-by: Caio Oliveira <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22932>

vtn: print spirv id for type mismatch error

Reviewed-by: Caio Oliveira <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22932>

vtn: add spirv index to type mismatch error for debugging

Reviewed-by: Caio Oliveira <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22932>

nir: Drop unused argument from nir_ssa_dest_init_for_type

Similar to nir_ssa_dest_init, but with fewer call sites to churn through.

This was done with the help of Coccinelle:

   @@
   expression A, B, C, D;
   @@

   -nir_ssa_dest_init_for_type(A, B, C, D);
   +nir_ssa_dest_init_for_type(A, B, C);

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23078>

nir: Drop unused name from nir_ssa_dest_init

Since 624e799cc34 ("nir: Drop nir_ssa_def::name and nir_register::name"), SSA
defs don't have names, making the name argument unused. Drop it from the
signature and fix the call sites. This was done with the help of the following
Coccinelle semantic patch:

    @@
    expression A, B, C, D, E;
    @@

    -nir_ssa_dest_init(A, B, C, D, E);
    +nir_ssa_dest_init(A, B, C, D);

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23078>

spirv: Improve the 'ID is the wrong kind of value' error messages

Include the expected and actual values in the errors -- since
very frequently we care about them to diagnose issues.

Since these helpers are meant to be inlined, also pull the
failure code out of the way into a separate function (not meant to
be inlined).  This way, extra calls to to_string will not harm
the existing client code size.  Verified this with GCC release build.

Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22977>

anv: mark images compressed for untracked layout/access

Most of the compressed writes are tracked by the driver, for
instances :
   - blorp writes
   - render target writes

But we don't have any tracking for storage images (which have gained
compression support on DG2+). So inspect the layout transition and
when we see a layout/access that can do writes outside of our driver
tracking, update the image state tracking.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: mesa-stable
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/8946
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22988>

ci/v3d: add flaking opengl 1.1@depthstencil-default_fb-drawpixels-float-and-ushort

Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23086>

vulkan/pipeline_cache: don't log warnings for client-invisible caches

Fixes: d3f06cf5ce0764b37a03a0f2bfbb109a4d75884d ('vulkan/pipeline_cache: don't log warnings for internal caches')
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22986>

docs/rusticl: add Contributing section

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23051>

spirv: Fix gl_spirv_validation when OpLine with strings is present

Fix issue by handling the OpString instructions when walking through
the preamble for validation.

The gl_spirv_validation() creates a vtn_builder() and walks the
instructions looking for a subset of the information.  However
our current way to walk the instructions will also perform tracking
of OpLine/OpNoLine, that may make references to OpString instructions
that were being previously ignored by gl_spirv_validation().
This would cause the parsing to fail.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/9004
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22973>

spirv: Extract vtn_handle_debug_text() helper

This will be later used by gl_spirv handling.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22973>

docs/venus: update vtest instructions

This is to reflect the latest steps after an issue with render server in
vtest has been fixed:
https://gitlab.freedesktop.org/virgl/virglrenderer/-/merge_requests/1115

Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23083>

iris: Set CS stall bit during HIZ_CCS_WT surface fast clear

It make sense to enable CS stall so that it guarantees that the fast
clear will start after tile cache flush has completed.

Signed-off-by: Sagar Ghuge <sagar.ghuge@intel.com>
Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23063>

anv: Set CS stall bit during HIZ_CCS_WT surface fast clear

It make sense to enable CS stall so that it guarantees that the fast
clear will start after tile cache flush has completed.

cc: mesa-stable
closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/9030

Fixes: e488773b ("anv: Fast clear depth/stencil surface in vkCmdClearAttachments"
Signed-off-by: Sagar Ghuge <sagar.ghuge@intel.com>
Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23063>

Uprev Piglit to d8c08d123fadb986e9a8a7887b922ff63fcff52e
https://gitlab.freedesktop.org/mesa/piglit/-/compare/536975d94a40cf76a69fcfa786c2513eccd0c989...d8c08d123fadb986e9a8a7887b922ff63fcff52e

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23017>

Add checks for NULL dxil_validator

Fixes https://gitlab.freedesktop.org/mesa/mesa/-/issues/8718

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22141>

anv: put private binding BOs into execlists

Not doing so all the reads/writes go to the scratch page on i915.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes: f9fa09ec92 ("anv/image: Add ANV_IMAGE_MEMORY_BINDING_PRIVATE")
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22957>

intel/dev: remove dg2 0x5698 pci id

0x5698 is now reserved for future use

From Matt Roper in linux
3a38be31ec82 ("drm/i915/dg2: Drop one PCI ID")

Fixes: 65c9a810eed ("intel/pci_ids: Add dg2 0x5698 pci-id")
Acked-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22985>

anv: drop duplicated nir_opt_dce passes

Signed-off-by: Rohan Garg <rohan.garg@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23075>

anv: use the common vulkan runtime to do the heavy lifting

Signed-off-by: Rohan Garg <rohan.garg@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23075>

zink: add feedback loop usage for swapchains

affects:
GTF-GL46.gtf30.GL3Tests.framebuffer_blit.framebuffer_blit_functionality_linear_filter_color_blit

cc: mesa-stable

Reviewed-by: Joshua Ashton <joshua@froggi.es>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22959>

vulkan/wsi: add feedback loop usage to swapchain caps if supported

this is needed by some apps

cc: mesa-stable

Reviewed-by: Joshua Ashton <joshua@froggi.es>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22959>

pvr: Fix deferred_control_stream_flags

Signed-off-by: James Glanville <james.glanville@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22958>

pvr: Fix seg fault on unused ds attachment

Fixes:
  dEQP-VK.renderpass.suballocation.formats.r16g16_sint.input
    .dont_care.dont_care.self_dep_clear

Signed-off-by: James Glanville <james.glanville@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22958>

pvr: Transfer image to buffer dest rect

In copy to buffer region, set dest rect to region size.
The rectangle must be block size adjusted in case of block compressed
format.

Fix test: dEQP-VK.api.copy_and_blit.core.image_to_buffer.regions

Signed-off-by: Oskar Rundgren <oskar.rundgren@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22958>

pvr: Transfer optimisation remove unused features from API

The transfer command "struct pvr_transfer_cmd" has support for
features not used by Vulkan: colour key, pattern, rop blit and
alpha blending

The whole "struct pvr_transfer_blit" can be removed. Also all code
related to transfer alpha blending can be removed.

This is an optimisation and doesn't fix any dEQP tests.

Signed-off-by: Oskar Rundgren <oskar.rundgren@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22958>

pvr: Transfer check valid source address mask

The mask should be inverted.

Fixes tests in dEQP-VK.api.copy_and_blit.* using the texel unwind workaround.

Signed-off-by: Oskar Rundgren <oskar.rundgren@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22958>

pvr: fix texel unwind workaround mappings

Fix the rectangle mappings used with texel unwind workaround (unaligned address).

Signed-off-by: Oskar Rundgren <oskar.rundgren@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22958>

pvr: Transfer remove byte unwind workaround

The byte unwind workaround can be used when source texture virtual
address doesn't meet HW requirements (is unaligned) and the pixel
format can't be changed i.e. destination is compressed. If
destination texture is not compressed the simpler texel extend
workaround can be used.

Currently byte unwind workaround has bugs so removing the
workaround fixes tests in
dEQP-VK.api.copy_and_blit.core.blit_image.simple_tests.* when they
instead use texel extend workaround.

Signed-off-by: Oskar Rundgren <oskar.rundgren@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22958>

pvr: Transfer support flipped rectangle mapping

The rectangles in "struct pvr_rect_mapping" are unsigned so a
flipped rectangle mapping isn't possible. Add new struct members
flip_x and flip_y to specify flipped mapping.

Add support for flipped rectangles in transfer copy blit path.
Support for flipped rectangles in the clip blit path is not done
in this change.

The new booleans are false by default because transfer command
"struct pvr_transfer_cmd" is zeroed on allocation in
pvr_transfer_cmd_alloc (pvr_blit.c).

Fixes:
  dEQP test case: dEQP-VK.api.copy_and_blit.core.blit_image.simple_tests
    .mirror_xy.nearest

Signed-off-by: Oskar Rundgren <oskar.rundgren@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22958>