anv: Use input assembly state only when pipeline has vertex stage

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13047>

anv: Set graphics pipeline active_stages earlier

So that we can use the active_stages in copy_non_dynamic_state() later.

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13047>

anv: Validate vertex related states only when VS is present

Mesh pipeline doesn't use those states, so we can't assert them
unconditionally for a graphics pipeline.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13047>

compiler/clc: Preserve OCL kernel arg type metadata on LLVM13

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13177>

compiler/clc: Null extensions should mean all supported, not all

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13177>

lima: use nir_shader_instructions_pass in lima_nir_split_load_input

Changes:
- nir_metadata_preserve(..., nir_metadata_block_index | nir_metadata_dominance)
  is called only when pass makes progress
- nir_metadata_preserve(..., nir_metadata_all) is called when pass doesn't
  make progress

Signed-off-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Vasily Khoruzhick <anarsoul@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13176>

wsi/x11: Fix a misunderstanding about how xcb_get_geometry works

The code here is well-intentioned, but the only way a GetGeometry
request can fail is if you name an invalid drawable. And if we did that,
either our internal state got corrupted, or - more likely - the user
destroyed the window. In either case there's nothing more we can do with
the surface, so report that it's been lost.

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13104>

freedreno: Handle PIPE_FORMAT_NONE buffers

Clover creates buffers with format=NONE.. which is not technically
incorrect.  Just treat this as r8_unorm so we know width0 is the size
in bytes.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13160>

freedreno: Handle cso==NULL in bind_sampler_states

This is a thing that comes up with clover.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13160>

freedreno/ir3: Remove used unused

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13160>

freedreno: Fix some indentation

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13160>

freedreno: Optimize no-op submits

In some cases we need to emit a no-op batch/submit, just to get a fence.
No need to emit all the boilerplate state-resture and flushing in this
case.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13160>

freedreno: Get shader variant msgs in perf debug output

We want FD_MESA_DEBUG=perf to also tell us about variants.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13160>

pan/mdg: Use the correct swizzle for condition moves

Fixes: 70072a20e00 ("pan/midgard: Refactor swizzles")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13173>

panvk: Support clearing ZS attachments

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13137>

panvk: Allow clear_attachment of RTs > 0

It's just a matter of skipping unused BLEND descriptors and emitting
the one we want to clear.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13137>

panvk: Pass the render target index to panvk_meta_clear_attachment()

The attachment index and render target don't necessarily match. We need
to explicitly pass the render target index to
panvk_meta_clear_attachment() so it can select the correct shader and
emit a valid RSD.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13136>

panvk: Fix wls_size retrieval

Fix a typo in wls_size assignment leading to a wrong wls_size.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13136>

panvk: Don't use panfrost_get_default_swizzle() on v7+

This helper produces invalid component order values on v7. Use the
MALI_RGB_COMPONENT_ORDER_xxx definitions instead.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13136>

panvk: Fix allocation of BOs bigger than the slab size

We can pick a BO from the pool if the BO size exceeds the pool size. In
that case, allocate a BO and save it in a separate array so we can free
it when the memory pool is reset or released.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13136>

radv: Add force_emulate_rt perftest option

Signed-off-by: Joshua Ashton <joshua@froggi.es>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12974>

radv: Enable raytracing extensions on older generations

We have shader-based bvh traversal for this on older generations now.

Signed-off-by: Joshua Ashton <joshua@froggi.es>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12974>

radv: Implement software emulation for intersect_ray

Signed-off-by: Joshua Ashton <joshua@froggi.es>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12974>

radv: Implement build_node_to_addr for GFX8 and below

Removes the nir_ prefix also given it no longer acts like a typical builder function.

Signed-off-by: Joshua Ashton <joshua@froggi.es>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12974>

radv: Do not pass result to insert_traversal_aabb_case

This is unused as it performs the tests itself.

Signed-off-by: Joshua Ashton <joshua@froggi.es>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12974>

radv: Remove assert in radv_rt_bind_tables

Not necessary anymore.

Signed-off-by: Joshua Ashton <joshua@froggi.es>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12974>

radv: determine the ES type (VS or TES) for GS earlier

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13170>

radv: remove useless loads_dynamic_offsets when emitting push constants

It's always TRUE if loads_push_constants is TRUE.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13170>

radv: remove redundant check of needs_multiview_view_index for PS

layer_input is always TRUE if needs_multiview_view_index is TRUE.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13170>

radv: move use of NGG to the graphics pipeline key

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/13099>

radv: move forcing VRS rates to the graphics pipeline key

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/13099>

radv: move forcing MRT output NaN fixup to the graphics pipeline key

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/13099>

radv: move forcing invariant geometry to the graphics pipeline key

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/13099>

radv: move forcing discard to demote to the graphics pipeline key

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/13099>

radv: constify radv_shader_info for radv_lower_{io_to_mem,ngg}()

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13134>

radv: remove unnecessary ac_nir_ngg_config output struct

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13134>

radv: move ngg passthrough determination earlier

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13134>

move: move ngg lds bytes determination earlier

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13134>

radv: move ngg early prim export determination earlier

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13134>

radv: move ngg culling determination earlier

Co-Authored-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13134>

radv: do not declare an extra user SGPR for sample positions and PS

This is part of the scratch buffer.

No fossils-db change on Sienna.

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/13140>

radeonsi: don't clear G_028644_OFFSET

Before 11d1309d827 this field was updated even when G_028644_PT_SPRITE_TEX was 0.

See https://gitlab.freedesktop.org/mesa/mesa/-/issues/5423

Fixes: 11d1309d827 ("radeonsi: restructure si_get_ps_input_cntl for future refactoring")
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13105>

radv: get the float controls execution mode from NIR for LLVM

No need to duplicate it. Though, I think it was already broken
for merged shaders, but it doesn't matter.

No CTS regressions anyways with LLVM.

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/12989>

radv: disable the DX10 diamond test for better line rasterization perf

Ported from RadeonSI. PAL also doesn't enable it for Vulkan.

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/13066>

lima/parser: add shader disassembly to dump

Reviewed-by: Vasily Khoruzhick <anarsoul@gmail.com>
Reviewed-by: Erico Nunes <nunes.erico@gmail.com>
Signed-off-by: Andreas Baierl <ichgeh@imkreisrum.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13138>

clc: let user specify the targetted SPIRV version

This version is given to the LLVM-SPIRV translator. On the SPIRV-Tools
side of things, we want to use the highest available version to be
sure to be able to parse back what was generated.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13113>

clc: print warnings/errors on their own line

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13113>

clc: add allowed extension for compile parameter

The LLVM-SPIRV translator can include a bunch of capabilities into the
generated SPIRV which is not what you always want. That include
internal Intel specific capabilities from the translator.

v2: Rename options
    Fixup checks (Jesse)

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13113>

zink: ensure fences are released before reusing them

at this point it's guaranteed that the cmdbuf has completed since the
timeline id has passed, but vulkan hasn't technically "released" the fence
until it's been waited upon, so cut down on some validation spam by waiting
here like in get_batch_state()

Acked-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13144>

zink: stop using VK_COMMAND_POOL_CREATE_RESET_COMMAND_BUFFER_BIT

the pool is reset anyway so this is unnecessary

Acked-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13145>

radv: Don't invalidate VCACHE after clear_htile_mask.

radv_src_access_flush sets all the required flags (which doesn't include VCACHE. The
flush after write is implicit. The invalidate happens for any user that needs it
with the radv_dst_access_flush).

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12274>

anv: honor INTEL_DEBUG=sync

Useful debug option for hangs.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Sagar Ghuge <sagar.ghuge@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13161>

radv: Fix Android build for common functions.

Fixes: 9fc16b66d0d ("radv: use common vkGetPhysicalDevice{Image}FormatProperties()")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/5328
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12764>

ac/surface: enable DCC image stores for all displayable DCC on gfx10.3

Co-authored-by: Joshua Ashton <joshua@froggi.es>
Signed-off-by: Joshua Ashton <joshua@froggi.es>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13153>

radeonsi: Use common DCC image store check

We need to keep RADV and RadeonSI on the same page about this due to modifiers.

Signed-off-by: Joshua Ashton <joshua@froggi.es>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13153>

radv: Use common DCC image store check

We need to keep RADV and RadeonSI on the same page about this due to modifiers.

Signed-off-by: Joshua Ashton <joshua@froggi.es>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13153>

ac/surface: Add helper for checking if a surface supports DCC Image stores

We need to keep RADV and RadeonSI on the same page about this due to modifiers.

Signed-off-by: Joshua Ashton <joshua@froggi.es>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13153>

iris: Delete the MI_COPY_MEM_MEM resource_copy_region implementation.

The MI_COPY_MEM_MEM version of resource_copy_region has known bugs:

- It's failing to set valid_buffer_range correctly
- It's missing iris_emit_buffer_barrier_for() for the
  source/destination, so there may be missing flushes.
- There are some bad interactions with the tile cache and VF using L3.

Even with those fixed, if you expand the "no more than 16 bytes"
restriction to allow copies up to 1024 bytes, then it starts failing
Piglit tests on Icelake.

We could probably fix this.  However, I had originally only measured a
0.689096% +/- 0.473968% (n=4) speedup in Shadow of Mordor's OpenGL
port, which is already fairly small, especially before adding missing
flushes.  Further, some of that likely came from not switching between
render and compute...which we'll soon be able to avoid thanks to BLOCS.

Folks were also worried that MI_COPY_MEM_MEM can't be pipelined, and
that stalling the command streamer may actually slow things down,
especially as the GPUs become more powerful.  We aren't really sure
about this, but it's another concern.

So, let's just get rid of this optimization.  It seemed like a good
idea at the time, but it's just causing issues for very little gain.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12863>

freedreno: Move the headergen2 test to be meson unit tests.

Now all the freedreno build-time testing is just "meson test -C build"

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6360>

freedreno: Move crashdec/cffdec tests to be meson unit tests.

Now they run automatically in parallel with other unit testing, rather
than needing a separate script and environment to run them.

Instead of doing shell script filtering afterwards, I just added a little
flag to suppress printing the path name.  Also dropped the "Parsing
<file>" in addition to "Reading <file>" in the tested script, since it's
redundant and baked the path name into the reference.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6360>

freedreno: Move afuc tests to meson unit tests.

Now they run automatically in parallel with other unit testing, rather
than needing a separate script and environment to run them.

Instead of doing shell script filtering afterwards, I just added a little
flag to suppress printing the path name.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6360>

freedreno: Reuse u_math.h instead of open coding uif().

Plus the old version had a comment with what conversion was being done
swapped!

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6360>

freedreno: Reuse u_math.h instead of open coding ALIGN/ARRAY_SIZE.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6360>

freedreno: Fix UBSan failures in cffdec's (uint8_t)x << 24

Types <32 bit get promoted to int32_t when you do expressions on them
(thus why (u8)x << 8 works at all), but shifting into the top bit of the
signed int is undefined behavior.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6360>

freedreno/afuc: Avoid ubsan warns about shifting to the top bit of 'int'

I think maybe it's being promoted to int due to the mismatched bitfield
sizes of the uint32_t values being referenced here?

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6360>

freedreno/rnndec: Avoid making 0-length variable length arrays.

ubsan hates it.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6360>

freedreno/rnndec: Fix use of undefined value_orig in the !ti case.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6360>

r300: implement forgotten tgsi's cases of textures

Fixes: d0c398a8 ("r300g: Use radeon compiler for fragment programs")

Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13135>

turnip: Expose a device name similar to the blob.

We add "Turnip" so that users (and vulkan.gpuinfo.org) can distinguish us
without requiring VK_KHR_driver_properties.  This will be a lot more
user-friendly than "FD618", though.

I made some little vk_asprintf helpers, because I figure other drivers
setting up deviceName's will want them too.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13041>

turnip: Match the blob's format for vendorID and deviceID.

This should hopefully cause us the least trouble with apps tuning for
device performance.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13041>

iris: Enable atomic operations on compressed surfaces

Let's not turn off compression for atomic operations since XeHPG
supports it.

Signed-off-by: Sagar Ghuge <sagar.ghuge@intel.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12966>

anv: support EXT_primitive_topology_list_restart

Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Tested-by: Ivan Briano <ivan.briano@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12628>

ac/surface: don't overwrite DCC settings for imported buffers

Fixes: 0f6251b31fc - ac/surface: use DCC compatible with image stores for < 4K resolutions

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13120>

lima: split_load_input: don't split unaligned vec2

Mali4x0 can't fetch unaligned vec2 (i.e. .yz), so don't split it.

Fixes: 6dd0ad66dedc ("lima/ppir: add NIR pass to split varying loads")
Reviewed-by: Erico Nunes <nunes.erico@gmail.com>
Signed-off-by: Vasily Khoruzhick <anarsoul@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13131>

panvk: remove feature checks from device creation

This is already handled by vk_device_init(); drivers no longer
need to do it themselves.

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12867>

lavapipe: remove feature checks from device creation

This is already handled by vk_device_init(); drivers no longer
need to do it themselves.

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12867>

v3dv: remove feature checks from device creation

This is already handled by vk_device_init(); drivers no longer
need to do it themselves.

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12867>

turnip: remove feature checks from device creation

This is already handled by vk_device_init(); drivers no longer
need to do it themselves.

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Hyunjun Ko <zzoon@igalia.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12867>

radv: remove feature checks from device creation

This is already handled by vk_device_init(); drivers no longer
need to do it themselves.

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12867>

anv: remove feature checks from device creation

This is already handled by vk_device_init(); drivers no longer
need to do it themselves.

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12867>

vulkan: provide common functions to check device features

v2: move checks to vk_device_init, ignore struct types from
    provisional extensions + lots of cleanups/fixes (Jason Ekstrand)
v3: squash in following patch from Jason:
    "vulkan: Restructure vk_physical_device_check_device_features()"
v4 (Jason): Fix a Windows build error

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12867>

radv: determine the VS output parameters in the shader info pass

This can be determined earlier instead of duplicating code in both
compiler backends.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13119>

radv: set export_clip_dists for the GS copy shader

This is needed for the next change to correctly compute the VS
output parameters from the shader info pass.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13119>

radv: Don't declare ngg_gs_state when there is no API GS.

This shader arg is only used when the pipeline has API GS,
so it is useless to declare it otherwise.

Also remove unused code from radv_declare_shader_args.
NGG is handled by the GS code path so these were never
actually used.

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/13129>

aco: Fix determining whether any culling is enabled.

Use 0xB instead of 0x00FFFFFF - this allows to jump over the culling
code when no actual culling is enabled but the ngg_cull_face_is_ccw
flag is set.

Fixes: 182d9b1e6072bec190cf0a52e9d93dbbdbaa850d
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/13129>

aco: Fix small primitive precision.

This is a mistake. It should use ngg_culling_settings
instead of ngg_gs_state.

Fixes: 182d9b1e6072bec190cf0a52e9d93dbbdbaa850d
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/13129>

radeonsi: implement draw_vertex_state for lower display list overhead

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13050>

radeonsi: separate VBO descriptor code into a new function (for future work)

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13050>

mesa: use pipe_vertex_state in vbo and st/mesa for lower display list overhead

Acked-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13050>

st/mesa: make setup_arrays more reusable for future display list support

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Acked-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13050>

st/mesa: add ST_PIPELINE_RENDER_NO_VARRAYS, for future display list support

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Acked-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13050>

gallium/util: add util_vertex_state_cache for deduplicating the states

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13050>

gallium/trace: add pipe_vertex_state support

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13050>

gallium/u_threaded: implement draw_vertex_state

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13050>

gallium: add pipe_vertex_state and draw_vertex_state for display lists

The main motivation is to improve the score of viewperf13/snx.

This new interface is designed to be optimal for display lists as implemented
by the vbo module. It has much lower CPU overhead in the frontend, threaded
context, and the driver.

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13050>

gallium/util: make pipe_vertex_buffer_reference safe for hashing dst

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13050>

util: import u_debug_refcnt, u_hash_table, u_debug_describe from gallium

to allow pipe_*_reference to be called in src/mesa/vbo.

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13050>

util: add util_popcnt_inline_asm

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13050>

radv: Add GPU serialization of acceleration structures.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12840>

radv: Add CPU serialization of acceleration structures.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12840>