tu: Add an extra storage descriptor for isam

Based on a workaround the blob does.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15288>

tu: Handle UBO/SSBO descriptors with different sizes

We reuse the otherwise-unused offset channel to represent the array
stride, so that reindexing works properly.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15288>

tu: Rewrite dynamic descriptor handling

We need to prepare for storage buffers having different sizes from
uniform buffers. This switches dynamic_offset_offset to have units of
bytes, the same as offset, and as a nice bonus we can more easily
combine the dynamic and non-dynamic paths in various different places.
This also entails rewriting the code that patches dynamic descriptors,
since we can no longer assume a linear mapping between indices in
dynamicOffsets and descriptor locations which the previous approach
heavily relied on.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15288>

zink: export indirect io pipe caps

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15327>

zink: fix unreachable() location in ntv streamout info

super annoying to debug otherwise

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15327>

zink: add DOUBLE glsl type for streamout export

not used yet but someday

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15327>

zink: add nir_var_function_temp support to ntv

I said I'd never do this, but here we are

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15327>

turnip: Enable VK_EXT_display_control using the common code.

It's all implemented now, so we can turn it back on.  Passes 15/16 tests
when X11 isn't running, and 1/16 when it is, with no failures in either
mode.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15351>

radv: fix returning empty drmFormatModifierTilingFeatures

From the Vulkan spec:
    "drmFormatModifierTilingFeatures is a bitmask of
     VkFormatFeatureFlagBits that are supported by any image created
     with format and drmFormatModifier. The returned
     drmFormatModifierTilingFeatures must contain at least one bit."

This fixes recent CTS dEQP-VK.drm_format_modifiers.*.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15383>

radv: remove VK_AMD_shader_info support

This extension is quite old and useless now.
VK_KHR_pipeline_executable_properties should be used instead.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15299>

radv: Expose VK_VALVE_descriptor_set_host_mapping for vkd3d only.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15363>

radv: Implement VK_VALVE_descriptor_set_host_mapping.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15363>

vk: Update xml and headers to 1.3.207.

Acked-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15363>

llvmpipe/fs: add missing depth_clamp key printing

Helps debugging shaders better.

Acked-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15385>

lavapipe: KHR_shader_integer_dot_product

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15384>

gallivm: implement nir_op_pack_32_4x8_split

just reusing existing helpers and llvm can optimize it for us

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15384>

lavapipe: maintenance4

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15378>

lavapipe: set maxBufferSize for maintenance4

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15378>

lavapipe: implement vkGetDevice*MemoryRequirements

for maint4

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15378>

lavapipe: ref/unref pipeline layouts for pipeline creation

required by maintenance4

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15378>

lavapipe: always clone shader nir for shader states

these become owned and freed by llvmpipe, so ensure that freeing
them there won't cause crashes

cc: mesa-stable

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15281>

panvk: Use the correct integer border colors

Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15382>

panvk: Rework texture, sampler, and image binding index calculation

This adds a new get_resource_deref_binding helper which decodes a
resource deref into set, binding, and index.  To make texture
instructions nicer, the index can optionally be split into immediate
and SSA parts.

Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15382>

panvk: Skip ZS setup if there is no depth/stencil attachment

Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15382>

panvk: Make panvk_image_view derive from vk_image_view

Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15382>

panvk: Make panvk_image derive from vk_image

Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15382>

Revert "glx: Fix build errors with --enable-mangling (v2)"

This reverts commit a27f2d991b1723c3349623401ce3c8f26dcdb28b.

As of a0829cf23b307ca44ab8 ("GL: drop symbols mangling support"), this
extra complexity isn't needed anymore.

Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2298>

Venus: add VN_CMD_ENQUEUE to vn_cmd_encode_memory_barriers

Signed-off-by: Igor Torrente <igor.torrente@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15344>

Venus: Adjust VN_CMD_ENQUEUE to set VN_COMMAND_BUFFER_STATE_INVALID

This improves the issue of a return inside the macro.

Signed-off-by: Igor Torrente <igor.torrente@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15344>

Venus: Add VN_CMD_ENQUEUE macro with vkCmd* common code

Several `vn_Cmd` share the same code to enqueue the command to the
command stream.

This adds a macro with this common code.

Signed-off-by: Igor Torrente <igor.torrente@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15344>

radv: export the pipeline hash via VK_KHR_pipeline_executable_properties

This will help to match RGP<->Fossilize pipelines.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15279>

radv,aco,ac/llvm: fix indirect dispatches on the compute queue on GFX7-10

Since neither PKT3_LOAD_SH_REG_INDEX nor PKT3_COPY_DATA work with compute
queues on GFX7-10, we have to load the dispatch size from memory in the
shader.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15064>

aco: split and recombine unaligned sgpr inputs

An example is the num_work_groups argument. Fixes invalid assembly with
func.compute.num-workgroups.basic.q0

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15064>

v3dv/ci: add test to skip list

Add test that it is a timeout in the CI, but otherwise it passes.

Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Reviewed-by: Jose Maria Casanova Crespo <jmcasanova@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15374>

panvk: Add support for texel buffers

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15334>

panvk: Add a dummy sampler for NIR tex operations that don't take one

In the NIR domain, some texture operations don't require a sampler, but
Bifrost/Midgard always want one. Let's add a dummy sampler to handle
that case.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15334>

panvk: Stop advertising Vulkan 1.1

We're nowhere close to even having Vulkan 1.0 working yet, there's no
reason to get too excited about 1.1.  It just means piles more test
crashes for features we're claiming to support but don't.  If we want to
enable more tests, we can turn on the extensions for those features once
we actually have them working.

Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15334>

Revert "ci: downgrade sphinx to v3.x"

The readthedocs theme now supports Sphinx 4.x, so there's no longer any
reason to stick with 3.x.

This reverts commit a545b6eda040ee40b5c36533d7ae378fc24fcf09.

Reviewed-by: Hoe Hao Cheng <haochengho12907@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15212>

panfrost: Optimise recalculation of max sampler view

Previously we always searched through 128 sampler views for set
sampler views, now we never look above the maximum updated view.

Fixes: 304851422a4 ("panfrost: Fix set_sampler_views for big GL")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15366>

panfrost: Don't initialise the trampolines array

PIPE_MAX_SHADER_SAMPLER_VIEWS is 128, so we just end up initialising a
kilobyte of memory for no reason, when usually only a couple of
sampler views are used.

Fixes: 53ef20f08d4 ("panfrost: Handle NULL sampler views")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15366>

radv: stop zeroing radv_sample_locations_state in barriers

This is useless because all fields should be correctly filled if the
pNext struct is found.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15164>

radv: move waiting for events to CmdWaitEvents2KHR()

CmdPipelineBarrier doesn't have events.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15164>

radv: remove unnecessary check in FreeCommandBuffers()

cmd_buffer->pool should never be NULL. Even if AllocateCommandBuffers()
fails, the successfully created cmdbuffers would have it set correctly.

From the Vulkan spec:

    "VUID-vkFreeCommandBuffers-pCommandBuffers-parent
     Each element of pCommandBuffers that is a valid handle must have
     been created, allocated, or retrieved from commandPool."

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15164>

radv: remove unnecessary NULL check in TrimCommandPool()

This function seems rarely used or maybe never but I noticed this.

From the Vulkan spec:

    "VUID-vkTrimCommandPool-commandPool-parameter
     commandPool must be a valid VkCommandPool handle".

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15164>

radv: remove useless check in radv_cmd_buffer_upload_data()

ptr shouldn't be NULL.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15164>

radv: fix compatibility with VK_IMAGE_CREATE_EXTENDED_USAGE_BIT

Some formats can be accepted if a compatible format is also supported
and VK_IMAGE_CREATE_EXTENDED_USAGE_BIT used.

Fixes new CTS dEQP-VK.image.extended_usage_bit_compatibility.*.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/6046
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15063>

radv: update inputs_read when lowering the view index

Otherwise inputs_read doesn't contain the information. This shouldn't
fix anything in practice because radv_shader_info gathers this from
the variable.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15337>

intel/compiler: Fix non-trivial designated initializer

Not supported by GCC 7.

src/compiler/nir/nir_builder_opcodes.h:14156:118:
sorry, unimplemented: non-trivial designated initializers not supported
src/intel/compiler/brw_mesh.cpp:515:7:
note: in expansion of macro ‘nir_store_per_primitive_output’

Reviewed-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Fixes: bc4f8c073a2 ("intel/compiler: inject MUE initialization")
Signed-off-by: Ernst Sjöstrand <ernstp@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15360>

radv: stop waiting for DMA to be idle for all transfer operations

Only copy operations actually use CP DMA.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15194>

Revert "llvmpipe: allow vertex processing and fragment processing in parallel"

This reverts commit ec8104c6b227421b3a21e9c0652e3050066bb169.
  llvmpipe: allow vertex processing and fragment processing in parallel

The commit breaks the the virglrenderer vtest environment used in the
virglrednerer CI and running wayland in virtualized environments.

Related: #6130
Related: #6110

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Acked-by: Dave Airlie <airlied@redhat.com>
Acked-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15338>

radv: fix indirect dispatches on the compute queue on GFX10.3+

For weird reasons, the COPY_DATA packet doesn't seem to copy anything
while on the compute queue. Instead, use PKT3_LOAD_SH_REG_INDEX which
seems to work as expected.

Note that LOAD_SH_REG_INDEX on the compute queue is only supported by
the CP on GFX10.3, so we need to implement a different solution (load
from the indirect BO in the shader) for older generations.

This should fix the Control RT GPU hang.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15053>

amd: add PKT3_LOAD_SH_REG_INDEX

It seems only available on GFX8+.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15053>

aco/ra: refactor collect_vars() to return a sorted vector

The vector of IDs is sorted with decreasing sizes,
and by increasing assigned registers.
This decouples register assingment from ssa IDs.

Totals from 12694 (9.41% of 134913) affected shaders: (GFX10.3)
VGPRs: 757864 -> 757848 (-0.00%); split: -0.00%, +0.00%
CodeSize: 72350540 -> 72348688 (-0.00%); split: -0.02%, +0.02%
MaxWaves: 237018 -> 237020 (+0.00%); split: +0.00%, -0.00%
Instrs: 13545494 -> 13544699 (-0.01%); split: -0.03%, +0.02%
Latency: 148539203 -> 148533292 (-0.00%); split: -0.01%, +0.00%
InvThroughput: 30319086 -> 30320382 (+0.00%); split: -0.01%, +0.01%
VClause: 326875 -> 327028 (+0.05%); split: -0.05%, +0.09%
SClause: 479833 -> 479837 (+0.00%); split: -0.00%, +0.00%
Copies: 862152 -> 860914 (-0.14%); split: -0.43%, +0.28%
Branches: 317775 -> 317777 (+0.00%)

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11526>

aco/ra: refactor find_vars() to return a vector

instead of std::set<>

No fossil-db changes.

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11526>

aco/ra: special-case get_reg_for_create_vector_copy()

This function implements separate handling for
p_create_vector during get_regs_for_copies().
This simplifies some code and lets more precisely select
swap instructions if possible.

Totals from 876 (0.65% of 134913) affected shaders: (GFX10.3)
VGPRs: 53312 -> 53336 (+0.05%)
CodeSize: 3792936 -> 3788160 (-0.13%); split: -0.15%, +0.03%
MaxWaves: 16084 -> 16078 (-0.04%)
Instrs: 707449 -> 706385 (-0.15%); split: -0.19%, +0.04%
Latency: 6288293 -> 6286677 (-0.03%); split: -0.03%, +0.01%
InvThroughput: 4264450 -> 4263671 (-0.02%); split: -0.02%, +0.00%
VClause: 18655 -> 18679 (+0.13%); split: -0.20%, +0.33%
Copies: 55397 -> 54353 (-1.88%); split: -2.45%, +0.57%
Branches: 12426 -> 12415 (-0.09%)

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11526>

aco/ra: count constant moves in get_reg_create_vector()

Also implements a correct_pos_mask to keep track which
operands are already at the right target position.
To mitigate some regressions, call get_reg_impl() less often.

Totals from 1229 (0.91% of 134913) affected shaders: (GFX10.3)
VGPRs: 60216 -> 59848 (-0.61%)
CodeSize: 3716496 -> 3711268 (-0.14%); split: -0.19%, +0.05%
MaxWaves: 27952 -> 28004 (+0.19%)
Instrs: 685983 -> 685035 (-0.14%); split: -0.20%, +0.06%
Latency: 6727587 -> 6725340 (-0.03%); split: -0.06%, +0.02%
InvThroughput: 9289043 -> 9289866 (+0.01%); split: -0.02%, +0.03%
VClause: 17730 -> 17740 (+0.06%); split: -0.25%, +0.30%
Copies: 54352 -> 53420 (-1.71%); split: -2.46%, +0.75%
Branches: 12122 -> 12121 (-0.01%)

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11526>

radv: rework the CS regalloc hang workaround

Move it to the pipeline creation to reduce computations in the hot path.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15162>

radv: fix the CS regalloc hang workaround on GFX6 and few GFX7 chips

RadeonSI uses a different terminology and info->blocks is actually the
number of threads, not the number of blocks (ie. info->grid).

Found by inspection.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15162>

zink: use dynamic rasterizer_discard state when possible

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15328>

zink: move dynamic state2 pipeline state to substruct in pipeline state

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15328>

zink: assert that the dynamic state array size is big enough

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15328>

radv/winsys: add support for queues without user fences.

The video queues don't have user fence support.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15257>

radv/winsys: add a ring level detection for ib bo usage.

The video rings can't use IB bos

v2: add use_ib flag to avoid a bunch of checks.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15257>

radv/winsys: add nop packets for uvd and vcn dec.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15257>

radv/winsys: complete ring/ip translations.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15257>

radeonsi/tests: update piglit baseline on vega20

current as of commit: 67314601947

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15273>

radv: Enable KHR_ray_query

Signed-off-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14565>

radv: Lower ray queries

Signed-off-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14565>

nir,spirv: Preserve ray_query_value

Signed-off-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14565>

radv: Move common code to seperate file

Signed-off-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14565>

radv: try and fix internal transfer queue mapping

The WSI code wants to remain generic and try and use vulkan APIs,
even though these queues aren't exposed through the API.

Add the transfer queue to the end of the queues.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Tested-by: Mike Lothian <mike@fireburn.co.uk>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15357>

zink: anv ci updates

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15361>

panfrost: Handle txs of cube arrays

We need to divide the array length by 6 to match what OpenGL expects.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15254>

pan/bi: Handle non-2D arrays

Handle arrays generically by using the last component of the coordinate source
as the array index. That works for both 2D arrays and cube arrays, fixing cube
arrays. Cube arrays were already handled correctly in core Panfrost code.

This code path is not tested in dEQP-GLES31 without exposing OES_cube_map_array,
which depends on OES_geometry_shader, which we don't have. Yet we do expose
PIPE_CAP_CUBE_ARRAY, so ARB_cube_map_array is exposed.

Disabling PIPE_CAP_CUBE_ARRAY would be an easy band-aid fix, but it's easy
enough to handle correctly.

dEQP-GLES31 passes with a hack enabling OES_cube_map_array [without geometry
shaders].

Also fixes 1D arrays on Bifrost for the same reasons.

Fixes: 70d6c5675d6 ("pan/bi: Emit TEXC with builder")
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15254>

panfrost: Emulate GL_CLAMP on Bifrost

Hardware support was removed with Midgard. Use mesa/st to emulate GL_CLAMP with
nir_lower_tex automatically (the Zink lowering), and disable GL_MIRROR_CLAMP
which isn't lowered correctly.

Fixes *texwrap* Piglit tests on G52.

Fixes: f9ceab7b237 ("panfrost: Fix CLAMP wrap mode")
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15253>

freedreno/ir3: Don't try re-swapping cat3 srcs

This can lead us to endless loops of "progress".. Note fixes commit
commit really just exposed an existing problem.

Fixes: 9c9e8c33498 ("nir: Reorder ffma and fsub combining")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/6133
Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15336>

freedreno/ir3: Remove unused define

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15336>

mesa: Easier shader capture for android

I've typed this patch a few times already.. lets just add some debug
code which can be easily switched on so I don't have to type it again.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15336>

crocus: force ignore_sample_mask_out on gen4/5 for precompile

on gen4/5 this won't ever be anything else, avoids some recompiles.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15315>

intel: add some missing debug recompile info.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15315>

bifrost: Constant fold after lower_explicit_io

nir_lower_explicit_io generates mul+add chains even for constants.  One
round of constant folding should get rid of these.  This fixes all of
the dEQP-VK.glsl.conversions.* tests on panvk.

GoGoGoGo'd-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15349>

mesa: Remove unused framebuffer validation

Probably unused since we deleted classic.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15255>

bifrost: Handle nir_op_frexp* and nir_op_ldexp

Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15352>

bifrost: Implement fine and coarse derivatives

We leave the undecorated ops as fine so we don't disturb panfrost.  For
coarse derivatives, we use a lane ID of 0 for the first lane and 1 or 2
for the second depending on axis.  This ensures that coarse derivatives
are quad-uniform.

Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15352>

bifrost: Simplify derivatives a bit

Instead of two magic ternary operations, define a new `axis` temporary
which is 1 for X and 2 for Y.  Then define everything else in terms of
this variable.  In particular, the mask operation we do on LANE_ID is a
mask so it makes more sense to use ~axis than 1/2 but in the other
order.

Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15352>

bifrost: Lower usub_borrow

Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15352>

util: fix msvc build

Fix msvc build regression after 0536b691338f2759b850f9ec94634033a5d1f9e1
reported by Prodea Alexandru-Liviu.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/6137
Fixes: 0536b691338 ("util: fix build with clang 10 on mips64")
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15355>

anv: Allow MSAA resolve with different numbers of planes

The Vulkan spec for VK_KHR_depth_stencil_resolve allows a format
mismatch between the primary attachment and the resolve attachment
within certain limits.  In particular,

    VUID-VkSubpassDescriptionDepthStencilResolve-pDepthStencilResolveAttachment-03181

    If pDepthStencilResolveAttachment is not NULL and does not have the
    value VK_ATTACHMENT_UNUSED and VkFormat of
    pDepthStencilResolveAttachment has a depth component, then the
    VkFormat of pDepthStencilAttachment must have a depth component with
    the same number of bits and numerical type

    VUID-VkSubpassDescriptionDepthStencilResolve-pDepthStencilResolveAttachment-03182

    If pDepthStencilResolveAttachment is not NULL and does not have the
    value VK_ATTACHMENT_UNUSED, and VkFormat of
    pDepthStencilResolveAttachment has a stencil component, then the
    VkFormat of pDepthStencilAttachment must have a stencil component
    with the same number of bits and numerical type

So you can resolve from a depth/stencil format to a depth-only or
stencil-only format so long as the number of bits matches.
Unfortunately, this has never been tested because the CTS tests which
purport to test this are broken and actually test with a destination
combined depth/stencil format.

Fixes: 5e4f9ea363a6 ("anv: Implement VK_KHR_depth_stencil_resolve")
Reviewed-by: Ivan Briano <ivan.briano@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15333>

zink: remove flake

this had already been resolved by the time the flake was added

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15350>

lavapipe: Use the auto-generated vk_enqueue_BeginRendering

Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15325>

vulkan/cmd_queue: Properly support non-array pointer members

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/5440
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15325>

radeonsi: add option to disable EFC

Signed-off-by: Thong Thai <thong.thai@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15196>

radeon: add EFC support to only VCN2.0 devices

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/5228
Signed-off-by: Thong Thai <thong.thai@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15196>

frontends/va: zero-copy efc

Signed-off-by: Thong Thai <thong.thai@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15196>

frontends/va: add encoder format conversion (EFC) support

Signed-off-by: Thong Thai <thong.thai@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15196>

gallium: add parameters for encoder format conversion (EFC) support

Signed-off-by: Thong Thai <thong.thai@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15196>

ci: update various ci result files

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12936>

mesa: check for valid internalformat with glTex[Sub]Image

This changes our error handling to be compatible with upcoming
specification change that unifies glTex[Sub]Image error handling
between OpenGL ES 2.0 vs ES 3.0+ specifications, see:

https://gitlab.khronos.org/opengl/API/-/issues/147

OpenGL ES 2.0.25 spec states:

   "Specifying a value for internalformat that is not one of
    the above values generates the error INVALID_VALUE. If
    internalformat does not match format, the error
    INVALID_OPERATION is generated."

This fixes following new tests:

    KHR-GLES31.core.compressed_format.*
    KHR-GLES32.core.compressed_format.*

v2: GL_INVALID_OPERATION -> GL_INVALID_VALUE in extension
    checks, remove (now overlapping) extension checks from
    _mesa_gles_error_check_format_and_type (Eric Anholt)

v3: take GLES version in to account in internalformat checks

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12936>

anv: silence compiler warning

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15241>

anv: make a couple of descriptor function private

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Rohan Garg <rohan.garg@intel.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15241>