pvr: Add clear attachment programs in device.

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/20055>

pvr: Add comments documenting large/normal clear words.

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/20055>

pvr: Add empty clear attachment usc programs.

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/20055>

pvr: Implement depth/stencil/depth+stencil attachment clear.

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/20055>

pvr: Fix typo in 'err_free_verices_buffer'.

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/20055>

pvr: Add pvr_clear_vertices_upload().

This function allows to easily setup and upload the vertices
required to clear a specific area. The vertices are
arranged in a triangle strip.

This will also be used in following commits implementing
vkCmdClearAttachments().

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/20055>

pvr: Move static clear state into pvr_clear.c .

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/20055>

pvr: Add layer count support to pvr_clear_vdm_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/20055>

pvr: Move clear VDM state into pvr_clear.h .

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/20055>

pvr: Add clear rta vert shader pds program.

The rta program will be used in following commits adding support
for vkCmdClearAttachments().

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/20055>

pvr: Add pvr_clear.{h,c} .

This moves some clear related functionality into a new
pvr_clear.{h,c} just to for better organisation and allow for
easier reusability.

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/20055>

pvr: Add multi layer passthough vert shader upload in device.

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/20055>

pvr: Change "ID" to "id" in instance_ID_modifier.

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/20055>

ci: Remove MESA_ARM_BUILD_TAG environment variable

Its value is already the same as MESA_IMAGE_TAG so no need to duplicate it.

Signed-off-by: Corentin Noël <corentin.noel@collabora.com>
Reviewed-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20488>

ci: Bump crosvm and virglrenderer versions

Update virglrenderer and crosvm to the latest version on time.

Signed-off-by: Corentin Noël <corentin.noel@collabora.com>
Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20488>

r600/sfn: make sure we return a non-negative number of registers

If a shader doesn't use any register and only ssa values we might
end up with zero minimum registers, and because a unsigned is
returned that goes wrong.

Fixes: 565816dfa15214abbeef9a9d94e44f30507ca4d7
    r600/sfn: Set minimum required registers based on array allocation

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/8008

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20516>

radv: rework generating the PS epilog key

Generating a PS epilog key will also be used when compiling PS epilogs
on-demand. This introduces a new helper that generates it.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20485>

radv: simplify removing unused color exports

If CB_TARGET_MASK (color write mask) is 0 for a given MRT, this implies
that the color format is 0 because the driver compacts MRTs.

No fossils-db changes.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20485>

docs/freedreno: Extract debug tooling docs and improve gpu dbg docs

Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20492>

radv/winsys: set has_3d_cube_border_color_mipmap for null winsys

Without this, NIR->LLVM will set level_zero to false, crashing compilation
of some GFX11 shaders with LLVM (image_gather4_c_o is not supported, while
image_gather4_c_lz_o is).

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/20483>

docs/freedreno: Extract LRZ docs from tu_lrz

Most of the docs describe HW and are not specific to Turnip.

Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20491>

radv: fix multiple resolves in the same subpass

If there is multiple resolves, the driver shouldn't always select the
fragment path because it doesn't work for all images.

Fixes dEQP-VK.pipeline.monolithic.multisample.misc.*

Cc: 22.3 mesa-stable
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19986>

glsl: Remove bit_count lowering

As far as I can tell, every driver that supports GLSL 1.30 or
GL_EXT_gpu_shader4 (and therefore also enables support for
GL_MESA_shader_integer_functions) also sets the NIR lower_bit_count
flag.

Reviewed-by: Emma Anholt <emma@anholt.net>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20323>

glsl: Remove bitfield_reverse lowering

As far as I can tell, every driver that supports GLSL 1.30 or
GL_EXT_gpu_shader4 (and therefore also enables support for
GL_MESA_shader_integer_functions) also sets the NIR
lower_bitfield_reverse flag.

Reviewed-by: Emma Anholt <emma@anholt.net>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20323>

glsl: Remove bitfield_extract and bitfield_insert lowering

As far as I can tell, every driver that supports GLSL 1.30 or
GL_EXT_gpu_shader4 (and therefore also enables support for
GL_MESA_shader_integer_functions) also sets some subset of the various
NIR lower_bitfield_extract and lower_bitfield_insert flags.

v2: Declaration of 'result' still needs to be added to the IR. Noticed
by marge.

v3: Fix 'git rebase --autosquash' putting the v2 fix in the wrong
place. I've never seen that happen before. :(

Reviewed-by: Emma Anholt <emma@anholt.net> [v1]
Reviewed-by: Matt Turner <mattst88@gmail.com> [v1]
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20323>

nir: Don't allow conflicting bitfield lowering passes

Reviewed-by: Emma Anholt <emma@anholt.net>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20323>

intel/compiler: Enable lower_bitfield_extract_to_shifts and lower_bitfield_insert_to_shifts for pre-Gfx7

GLSL IR opcodes generated for bitfieldExtract and bitfieldInsert are
lowered by lower_instructions.  4dff3ff005b ("nir/opt_algebraic:
Optimize open coded bfm.") adds an optimization that can rematerialize
nir_op_bfm that was prevented by the GLSL IR lowering.

It appears that every piece of hardware, except older Intel GPUS, that
has real integers (i.e., lower_bitops is not set) also sets
lower_bitfield_extract_to_shifts and lower_bitfield_insert_to_shifts.

Reviewed-by: Emma Anholt <emma@anholt.net>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Fixes: 4dff3ff005b ("nir/opt_algebraic: Optimize open coded bfm.")
Closes: #7874
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20323>

util: Drop the ENUM_PACKED macro

We have both PACKED and ENUM_PACKED macros which expand to the same
thing.  PACKED was based on a meson check for function attributes while
ENUM_PACKED appears to be a legacy gallium thing which was based on
defined(__GCC__).  This changes the one use of ENUM_PACKED to PACKED and
deletes ENUM_PACKED.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20412>

ci: Update the skqp testing docs and retire the old runner script.

Reviewed-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20070>

ci/intel: Switch skqp testing over to deqp-runner.

The skqp runner gets us parallel execution, automatic caselist handling,
nice reports, and the same xfail/flake handling you know and love from
deqp and piglit.

And, now that we have flake handling, we can turn the tests back on!

Reviewed-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20070>

ci/amd: Switch raven skqp testing over to deqp-runner.

The skqp runner gets us parallel execution, automatic caselist handling,
nice reports, and the same xfail/flake handling you know and love from
deqp and piglit.

Reviewed-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20070>

ci/freedreno: Switch skqp testing to using deqp-runner.

The skqp runner gets us parallel execution, automatic caselist handling
(which would have prevented a recent regression due to some skqp tests
having been forgotten in the checked in caselists), nice reports, and the
same xfail/flake handling you know and love from deqp and piglit.

Reviewed-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20070>

glx: Remove the GetProcAddress special case for indirect rendering

Some GL entrypoints would be aliased in an API sense but have different
GLX protocol. The only one that matters to us is EXT_texture_object,
which is the pre-GL-1.1 API. We're just going to drop support for that
and assume you have 1.1 or better, since 1.0 + EXT_texture_object is a
vanishingly rare combo at this point.

Acked-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20300>

glx: Only compute client GL extensions for indirect contexts

This is sort of a spiky way to do it, but the effect is to send the
appropriate SetClientInfo twice for indirect screens, where the second
one fills in the GL extensions. We can get away with this because the
only place the string is used is when the server computes the reply for
glGetString(GL_EXTENSIONS), which never matters for direct contexts.

Acked-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20300>

glx: Require GLX 1.3

GLX is a means to the end of direct rendered GL, really. Our indirect
protocol support has been largely untouched forever, anyone who wants it
can find it in amber. We're not going to drop or intentionally break it
(indirect support), but we're also not going to try super hard to
preserve its quirks anymore.

xserver has typically supported GLX 1.4 since 2009 (xserver 1.8, ad5c0d9e)
and unconditionally since 2016 (xserver 1.19, 36bcbf76). Assuming GLX
1.3 internally will let us fix some GLX drawable lifetime issues.

Acked-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20300>

glx: Remove pointless GLX_INTEL_swap_event paranoia

It's not our job to filter this out, it's the server's job to not send
events that haven't been selected for. We'll still throw the event away
if we don't have any client-side state for it though.

Acked-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20300>

glx: Drop GLX_MESA_{pixmap_colormap,release_buffers} stubs

Whatever compatibility purpose these served has long since passed.

Acked-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20300>

glx: Replace FreeB 2.0 text with SPDX-License-Identifier: SGI-B-2.0

Acked-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20300>

glx: Remove dead declarations from <GL/glx.h>

MESA_swap_control is defined in glxext.h now. MESA_swap_frame_usage was
removed in Mesa 7.9 in 2010. The other two were never specified or
implemented.

Acked-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20300>

include: Sync <GL/glxext.h> with Khronos

Acked-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20300>

r300: don't convert to constant swizzles when translating from TGSI

We currently convert it twice for unknown reasons, first when
translating from TGSI and later in constant folding. Not only is this
unnecessary, the first translation doesn't check for non-native
swizzles, so removing it actually saves few instructions and gains
a single Unigine shader for R300 at the expense of few more constant
loads and temps.

Also fixes few dEQPs because we could previously generate code like
TEX temp[1], none.01__, 2D[0];
and the native swizzle rewrite pass was not ready for it.

RV370 shader-db:

total instructions in shared programs: 84441 -> 84436 (<.01%)
instructions in affected programs: 63 -> 58 (-7.94%)
helped: 4
HURT: 0

total temps in shared programs: 12398 -> 12400 (0.02%)
temps in affected programs: 10 -> 12 (20.00%)
helped: 1
HURT: 3

total consts in shared programs: 79081 -> 79090 (0.01%)
consts in affected programs: 12 -> 21 (75.00%)
helped: 0
HURT: 7

GAINED: shaders/tropics/465.shader_test FS

No shader-db change with RV530.

Signed-off-by: Pavel Ondračka <pavel.ondracka@gmail.com>
Reviewed-by: Filip Gawin <filip@gawin.net>
Tested-by: Filip Gawin <filip@gawin.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20484>

r300: allow copy propagate of RC_FILE_NONE reads to TEX instructions

Texturing instructions can't read from constant sources, however this
can work when the constant was transformed to constant swizzles and
hence RC_FILE_NONE.

Prevents a regression in single Unigine Tropics shader that uses
constant (0.5,0.5) as a TEX coordinate in a next patch. We now
convert to constant swizzles twice, first when translating from TGSI
and than in constant folding. If we disable the first conversion
rc_transform_tex will emit a mov from constant to temporary. With this
patch, copy propagate will clean it up later.

Signed-off-by: Pavel Ondračka <pavel.ondracka@gmail.com>
Reviewed-by: Filip Gawin <filip@gawin.net>
Tested-by: Filip Gawin <filip@gawin.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20484>

r300: don't copy propagate constant swizzles to KIL on R300

Transforming
0: MOV temp[1], -none.1111;
1: KIL temp[1];
 to
0: KIL -none.1111;

Doesn't work on R300 while it works just fine with R500.

Prevents a regression when we enable the copy propagate of RC_FILE_NONE
to texture instructions in the next commit.

Signed-off-by: Pavel Ondračka <pavel.ondracka@gmail.com>
Reviewed-by: Filip Gawin <filip@gawin.net>
Tested-by: Filip Gawin <filip@gawin.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20484>

aco: Use v_mov_b16 on GFX11.

Foz-DB GFX1100:
Totals from 4684 (3.47% of 134913) affected shaders:
CodeSize: 41086444 -> 41043476 (-0.10%)
Instrs: 8176019 -> 8175995 (-0.00%)
Latency: 83792071 -> 83792023 (-0.00%)
InvThroughput: 10311371 -> 10311369 (-0.00%)

Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20369>

CI/Windows: Use waffle instead of freeglut for piglit

Reviewed-by: Sil Vilerino <sivileri@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20466>

CI/Windows: Update piglit for Waffle fix

Reviewed-by: Sil Vilerino <sivileri@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20466>

docs/freedreno: Add bindless/bindful descriptor docs

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20456>

freedreno/registers: Cleanup bindless-base regs

Make it clear that the low two bits of the 64b address is it's own
bitfield.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20456>

freedreno/registers: Fix bo fields with low != 0

We need to add the missing left-shift.  And a right-shift is negative!

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20456>

freedreno/decode: Improved reg64 decoding

This also (other than for an a5xx hack) gets rid of relying on
type0_reg_vals which isn't updated in all paths.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20456>

freedreno/decode: Add rnn_reginfo_free() helper

Simplify things a bit.. and fix a few places that just leaked the
rnndecaddrinfo.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20456>

freedreno/registers: Fix reg64 support

The maximum "high" position depends on 32b vs 64b registers.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20456>

ci: Update mingw and vs2019 libva build dependency to libva/releases/tag/2.17.0

Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20498>

radv: Use the correct pipeline layout for LBVH IR generation

Fixes: 5ba950e ("radv: Switch to new LBVH implementation.")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20481>

nv30: Fix an offset for vbos being applied to a buffer twice

Similar to 1387d1d4, this offset was being applied twice (once in
translate_generic, and once when the buffer is mapped).

This fixes 7972, which was initially thought to be an endianness
specific issue.

CC: mesa-stable
Tested-by: Filip Gawin <filip@gawin.net>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20468>

ci: update venus-lavapipe test expectations

Remove fixed push descriptor tests from expected failures.

Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Reviewed-by: Corentin Noël <corentin.noel@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20471>

venus: properly ignore the sampler for immutable sampler

This was found while debugging venus-lavapipe ci failure. It's a real
bug though no tests have caught this yet, but fixing this would regress
venus-lavapipe non-templated push tests if without the dependent lvp
fix. The sampler in the descriptor write can be garbled if the binding
has immutable samplers.

cc: mesa-stable

Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Reviewed-by: Corentin Noël <corentin.noel@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20471>

lvp: properly ignore sampler write for immutable sampler

The issue is hidden due to a overly relaxed cts:
dEQP-VK.binding_model.shader_access.primary_cmd_buf.with_push*
that doesn't scrub the sampler from descriptor writes for immutable
samplers. The issue is exposed via venus-lavapipe ci because venus must
ignore the potentially garbled sampler. This change aligns the
VK_DESCRIPTOR_TYPE_SAMPLER path with the
VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER path by removing a false check
against the provided sampler from push since the sampler can be null. An
alternative is to also check against !binding->immutable_samplers there.

Test: venus-lavapipe with venus push descriptor support

cc: mesa-stable

Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Acked-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20471>

spirv2dxil: Support linking multiple shaders

This probably could/should be split up into multiple commits, but
it's simpler to make this a monolithic change.

This change inlines a bunch of logic from spirv_to_dxil into the
spirv2dxil tool so that linking can be done on the nir shaders.
Probably the linking functionality should be exposed in the lib/dll
form too, which means that a helper for freeing intermediate nir
would be needed too. That's TODO for now.

The tool now requires arguments to be in-order, and once a filename
is encountered, will use the previous arguments to compile the shader.
If multiple graphics shaders are passed, they're linked as if they
were forming a pipeline together.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20440>

spirv2dxil: Rename and move prep helper

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20440>

CI/Windows: Use deqp-runner for D3D12 piglit

Acked-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20454>

vulkan/runtime: Preserve pNext when upgrading to synchronization2 structs

The functions that upgraded VkFooMemoryBarrier to VkFooMemoryBarrier2
dropped the pNext pointers. It loses VkSampleLocationsInfoEXT, and may
lose additional structs too if VkFooMemoryBarrier receives further
extensions in the future.

Signed-off-by: Chad Versace <chadversary@chromium.org>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20477>

radv: determine the gfx scratch size at pipeline bind time

This doesn't need to be in the draw path.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20299>

radv: dirty all dynamic states when beginning a new cmdbuf

Sounds safer to not rely on other cmdbuf states.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20299>

radv: dirty states when beginning a cmdbuf instead of when a pipeline is bound

To reduce CPU overhead of radv_emit_graphics_pipeline().

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20299>

radv: move emitting the strmout buffer in CmdDrawIndirectByteCountEXT()

This doesn't need to be in the generic draw path because only one
draw command uses it.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20299>

radv: flush DFSM on CB_TARGET_MASK changes when it's emitted

To avoid performing the same check twice and to emit it at the right
place.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20299>

frontends/va: Update state var frame_num disregarding cap check

The frame_num variable must be updated for encode entrypoint disregarding
the outcome of the PIPE_VIDEO_CAP_REQUIRES_FLUSH_ON_END_FRAME cap check

fixes: 229c6f79a660e5c7999ffc94e1fb514692df3b6a ("frontends/va: Implement vaSyncBuffer")

Reviewed-by: Ruijing Dong <ruijing.dong@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20490>

radeonsi,radv/llvm: fix amdgpu-color/depth-export with epilogs

The main shader wouldn't use ac_build_export(), and the discard exit would
have no export.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Fixes: 1174ab6d56e ("ac/llvm: use amdgpu-color-export/amdgpu-depth-export")
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Tested-by: Mikhail Gavrilov <mikhail.v.gavrilov@gmail.com>
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7991
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20482>

postprocess: move the definition of pp_filters into **/pp_init.c

An LTO-friendly move.

Fixes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7881

Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20374>

ci: build test LTO

Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20374>

zink: use EXT_descriptor_buffer with ZINK_DESCRIPTORS=db

this should be bug-free, as it passes cts/piglit/gaming on multiple drivers,
but since it's new, it stays behind an env var for at least one release

Acked-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20489>

zink: move some descriptor data into a substruct

no functional changes

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20489>

Revert "zink: remove descriptor-mode selection infrastructure"

this would've been in-use, but khronos changes while I was on vacation
blocked a merge

This reverts commit 3f371d4e940509c73fa19c4e50ae319e75636eb0.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20489>

anv: implement Wa_14015814527 for task shaders

After using task shader, we need to emit a zero URB state and a
nullprim (empty pipe control) before rendering with primitives.

After this, a normal URB state needs to be returned, this will
happen when pipeline batch is emitted during pipeline switch.

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20334>

nir: basic tests for nir_opt_shrink_vectors

Signed-off-by: Pavel Ondračka <pavel.ondracka@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20213>

nir: fix shrinking of load_const for large vectors

Specifically when shrinking load_const with number of components
> 5, if the final number of components is not allowed (for example 8->6)
it would report false for progress even if we actually did some
reshuffling and also it would skip on the rewrite of the readers.

Signed-off-by: Pavel Ondračka <pavel.ondracka@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20213>

nir: remove duplicate alu channels in nir_opt_shrink_vectors

This will clean code like:
   vec3 32 ssa_8 = frcp ssa_7.www
   vec3 32 ssa_9 = fmul ssa_7.xyz, ssa_8
into
   vec1 32 ssa_8 = frcp ssa_7.w
   vec3 32 ssa_9 = fmul ssa_7.xyz, ssa_8.xxx

This helps r300 driver because we can only do single channel for math
ops at a time, so the first version would result in three frcp
instructions. The nir_opt_shrink_vectors comments even claim the pass
should be doing this, however it actually does it only for nir_op_vecx
instructions, so extend this for generic alu instructions.

RV530 shader-db:
total instructions in shared programs: 135032 -> 133707 (-0.98%)
instructions in affected programs: 46121 -> 44796 (-2.87%)
helped: 452
HURT: 26
total temps in shared programs: 17051 -> 17033 (-0.11%)
temps in affected programs: 1509 -> 1491 (-1.19%)
helped: 91
HURT: 30

12.02->12.08 (+0.5%) fps gain in Unigine Sanctuary (n=5) with RV530

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7051
Signed-off-by: Pavel Ondračka <pavel.ondracka@gmail.com>
Reiewed-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20213>

pan/bi: Move Bifrost specific C code to src/compiler/bifrost

The goal is to make files at the root of src/compiler/ apply to both Bifrost and
Valhall, while ISA-specific code (e.g. instruction packing) code goes in
compiler/bifrost/ or compiler/valhall/. This is what Valhall is already doing,
the Bifrost specific stuff was just grandfathered in.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20455>

pan/bi: Remove standalone compiler

This functionality is now available on Linux with drm-shim + shader-db, and I
suspect the version bundled here is broken anyway. Strictly this drops
Windows/macOS support for the known-broken frontend to the shader compiler but I
can't say I'm terribly worried about that.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20455>

pan/bi: Rename panfrost/bifrost -> panfrost/compiler

This is the compiler for both Bifrost and Valhall, and presumably future
Mali GPUs too. Give it a more generic name so we can use the bifrost/ path for
something a bit more specific.

For historical reasons the compiler's name is still "bifrost" and uses the
prefix `bi_`. I think that's ok in the same way that i915 in the kernel supports
way more than just i915.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20455>

hasvk: Add layer with work-around for Doom 64 texture corruption

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7817
Signed-off-by: Sviatoslav Peleshko <sviatoslav.peleshko@globallogic.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19502>

anv: Add layer with work-around for Doom 64 texture corruption

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7817
Signed-off-by: Sviatoslav Peleshko <sviatoslav.peleshko@globallogic.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19502>

radv: Add an app layer driconf and use it for Metro Exodus

To make adding more application layers easier.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20439>

radv: Clean up entrypoints generation

This should make it easier to add new tracing and application layers.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20439>

radv: Use multiple dispatch tables for layers

Every layer has its own dispatch table that it can use to call down the
layer stack. This allows us to use RRA and RGP tracing simultaneously.
Using application layers with tracing should work as well.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20439>

radv: Move dispatch table init into a separate function

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20439>

vulkan: Allow passing NULL dispatch tables to vk_device_init

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20439>

r600: Don't merge alu groups with variable length dot using t-slot

Since the variable length dot must stay in its slot configuration
do not try to merge the group with the previous group when an op may be
moved to the t slot, because this may lead to breaking the multi-slot
operation.

Fixes:  357e5fac9953b26eedc8819ab528b981be6e1b69
   r600/sfn: Use variable length DOT on Evergreen and Cayman

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20451>

r600/sfn: Set minimum required registers based on array allocation

In the rare case that after register allocation the highest directly
accessed register index is below the highest value used for an
indirectly accessed array we have to ensure that the shader allocates
enough registers to account for these indices that are not seen by the
assembler.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7966
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20451>

r600: enable ARB_gl_spirv

76 out of 86 piglits pass.
Some fail because SSBOs are only supported for FS and CS on r600, but
the piglits try to use SSBOs with VS, and there are piglits that try to
bind SSBO at 8, and only 0-7 is supported as binding point.

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20451>

r600: Fix early exit when setting SSBOs

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20451>

r600/sfn: Fix FS primid input slot

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20451>

r600/sfn: Fix warning for mixed use of enum and integer

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20451>

r600/sfn: pre-evaluate allowed dest mask in Alu instructions

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20451>

r600/sfn: move handling of legacy math rules to assembler

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20451>

st/mesa: Fix free of non-shareable shaders on context destroy

On drivers that do not set PIPE_CAP_SHAREABLE_SHADERS,
st_destroy_program_variants() may reach st_save_zombie_shader()
which accesses st->zombie_shaders.mutex.
Destroying st->zombie_shaders.mutex before destroying program variants
may result in an invalid access in a multiple context scenario for
those drivers.
Move the mutex destroy call to after program variants destroy so that
it doesn't hit a deadlock on context destroy.

Signed-off-by: Erico Nunes <nunes.erico@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20458>

ac/llvm: use amdgpu-color-export/amdgpu-depth-export

These are necessary to use the correct export target on GFX11:
https://reviews.llvm.org/D128185

Fixes artifacts on Lara in Rise of the Tomb Raider benchmark and hair in
The Witcher 3 (classic).

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20357>

radv: Decouple radv_before_taskmesh_draw from radv_before_draw.

radv_before_taskmesh_draw will no longer call radv_before_draw and
instead implement the necessary functionality on its own.

radv_before_draw will no longer have to emit mesh shader descriptors.

As a result, both functions should have a lower CPU overhead now.

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/18829>

radv: fix missing initialization of radv_resolve_barrier::dst_stage_mask

Otherwise, this value is unitialized when read in
radv_ace_internal_barrier().

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7909
Fixes: 4c6f83006d4 ("radv: Synchronization for task shaders.")
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20351>