panvk: catch unsupported arch in the panvk_physical_device_init

Suggested-by: Boris Brezillon <boris.brezillon@collabora.com>
Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24201>

panvk: architecture isn't invalid, just unsupported

When we fail, tell users clearly why.

Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24201>

gallium/auxiliary/vl: Set vertex element src_stride in vl_deint_filter

Fixes: 76725452239 ("gallium: move vertex stride to CSO")

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24926>

aco/spill: skip p_branch in process_block

Fixes compilation of a Dead by Daylight shader.

fossil-db (gfx1100):
Totals from 58 (0.04% of 133461) affected shaders:
Instrs: 319824 -> 319421 (-0.13%); split: -0.13%, +0.00%
CodeSize: 1711260 -> 1708744 (-0.15%); split: -0.15%, +0.00%
SpillSGPRs: 2567 -> 2459 (-4.21%)
Latency: 3274930 -> 3274921 (-0.00%); split: -0.00%, +0.00%
InvThroughput: 924106 -> 924105 (-0.00%); split: -0.00%, +0.00%
Copies: 41883 -> 41757 (-0.30%); split: -0.31%, +0.00%
Branches: 9144 -> 9146 (+0.02%)

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Cc: mesa-stable
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/9599
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24896>

ci/panfrost: add G52 simple_tests.partial_image_pot_same_format_noclear flake

Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24922>

ci/freedreno: add another a530 flake

Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24922>

ci/virgl: flakes in functional.draw_buffers_indexed group

Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24922>

util: add radeonsi workaround for Nowhere Patrol

Cc: mesa-stable
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24919>

aco: fix emitting TCS epilogs end on GFX9+

With merged shaders, the long-jump should be emitted inside the
divergent if (ie. only for TCS invocations) and other non TCS
invocations should just end the program.

This fixes a bunch of failures with CTS by forcing TCS epilogs on
RDNA2.

Not sure how RadeonSI will handle that but maybe doing the merged
wave info thing in epilogs would help.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24832>

radv: remove the pipeline dependency for emitting VGT_GS_MODE

For shader object.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24865>

v3dv: re-enable sync_fd import/export on the simulator

On commit 29588fe11667 we re-enable sync_fd import/export. But only
with the real hw, because at that time there were wrong CTS tests
(that were calling vkSetEvent after submission) that needed to be
fixed.

Since this commit:
https://github.com/KhronosGroup/VK-GL-CTS/commit/717c051d4bcc9b71f13bc6b223e9926dcf9b7bd5

Those tests are fixed. That fix has been on CTS releases for some
time. So we can enable it on the simulator too.

With this change the pattern dEQP-VK.api.external.semaphore.sync_fd*
goes from 2 Passed/10 Not Supported to 12 Passed on the simulator.

Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24688>

radv: fix emitting TCS epilogs if TES and GS are linked on GFX9+

TES would be NULL because everything is merged to GS.
Found by inspection.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24890>

radv: small cleanups in radv_emit_patch_control_points()

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24890>

radv: rename tcs_shader to tcs in radv_emit_tcs_epilog_state()

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24890>

zink: remove sync TODO

after investigating, this is pointless and won't ever generate any value

fixes #9016

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24886>

zink: simplify some image barrier conditionals

no functional changes

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24886>

zink: make image barrier init functions void return

the return value was never used

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24886>

zink: reset unordered flags for image barriers on non-matching batch access

this allows more reordering when the first barrier in a new cmdbuf can
be reordered after previous ordered access exists

KHR-GLES3.copy_tex_image_conversions.required.texture2d_cubemap_negz:
before - ordered 68
after - ordered 16

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24886>

zink: force-reset unordered flags for buffer barriers on non-matching batch access

this should allow slightly better reordering

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24886>

vulkan/wsi/wayland: fix unset present_mode

chain->base.present_mode is unset at this point, ie. it's
zero-initialized. VK_PRESENT_MODE_IMMEDIATE_KHR happens to be 0,
so the WSI will attempt to use tearing-control on compositors that
don't support it.

Signed-off-by: Simon Ser <contact@emersion.fr>
Fixes: 5ceba97c2e18 ("vulkan/wsi/wayland: add support for IMMEDIATE")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24885>

zink: fix optimal_keys warning message

needs more newlines

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24894>

zink: be consistent with ds3 state resetting for blits

handle no-stipple case

Fixes: 122ffb0c888 ("zink: unset line stipple ds3 state flags when stipple not available")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24894>

zink: break out ds3 state resetting

no functional changes

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24894>

vulkan: Add trace points for more Vulkan waiting functions.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24799>

frontends/va: checking va version for av1enc support

need to ensure va version >= 1.16 to support av1enc.

Reviewed-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Ruijing Dong <ruijing.dong@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24901>

venus: expose at least one cached memory type

Kernel makes every mapping coherent. If a memory type is truly
incoherent, it's better to remove the host-visible flag than silently
making it coherent. However, for app compatibility purpose, when
coherent-cached memory type is unavailable, we emulate the first cached
memory type with the first coherent memory type.

Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24875>

d3d12: Fix H264 interlaced decode

Have to set the interlaced field of the surface before
end_frame is called in the pipe codec object so the info
is available to the frontend/va, instead of setting it
directly in end_frame like before.

Fixes: 578e10e1571 ("frontends/va: Alloc interlaced surface for interlaced pics")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24909>

d3d12: Fix Map/Unmap of YUV resources

Restore transfer box original size after temporal per plane
dimension calculation.

Currently the returned transfer object on Map will have the
size (usually downsampled) of the latest plane instead of
the overall resource size. Then on unmap, when flushing the
changes the received transfer box has the wrong dimensions
and only partial data is flushed.

Fixes: 12a4f2c1328 ("frontends/va: Also map VAImageBufferType for reading")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24909>

vk/wsi/x11: Remove dead code

Fix defect reported by Coverity Scan.

Logically dead code (DEADCODE)
dead_error_line: Execution cannot reach this statement: return VK_ERROR_SURFACE_LOS....

Fixes: fb9f697fbb8 ("vk/wsi/x11: move surface alpha check from get_caps to creation")
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24802>

nv50: Remove unused value

Fix defect reported by Coverity Scan.

Unused value (UNUSED_VALUE)
assigned_pointer: Assigning value from &nv50->vtxbuf[b] to vb here, but
that stored value is overwritten before it can be used.

Fixes: 76725452239 ("gallium: move vertex stride to CSO")
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24801>

driconf: add a workaround for Rainbow Six Extraction

CC: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24784>

nv/codegen: Delete copy and assign

Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24904>

nv/codegen: Change copy-constructor call to assign

This almost certainly intends to call the user-definied assignment
operator here instead of the automatically generated copy constructor.

Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24904>

nir/algebraic: Remove redundant pack / unpack lowering patterns

No shader-db or fossil-db changes on any Intel platform.

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24900>

nir/builder: Add nir_extract_i8_imm and nir_extract_u8_imm helpers

v2: Fix problems with 16-bit src0.

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24899>

hasvk/tests: Propagate failures to gtest

Acked-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24355>

hasvk/tests: Link a single hasvk_tests binary using gtest

Acked-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24355>

hasvk/tests: Refactor state_pool_test_helper to not use macros for parametrization

Acked-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24355>

anv/tests: Propagate failures to gtest

Acked-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24355>

anv/tests: Link a single anv_tests binary using gtest

Acked-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24355>

anv/tests: Refactor state_pool_test_helper to not use macros for parametrization

Acked-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24355>

ci/panfrost: we have enough device, parallelize Vulkan tests

Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24873>

ci/panfrost: add G52 flakes

Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24873>

lavapipe: further limit accurate_a0 hack

With lavapipe the previous change to only enable the hack when there's
no textures bound doesn't work anymore, since we don't have that
information when using the texture handles.
So add another random state restriction which is sufficient to pass
tests. (This really needs a better long term solution.)

Reviewed-by: Brian Paul <brianp@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24887>

zink: add a618 flake

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24892>

r300: there is no limitation on presubtract source file

RV530 shader-db:
total instructions in shared programs: 128840 -> 128803 (-0.03%)
instructions in affected programs: 1085 -> 1048 (-3.41%)
helped: 37
HURT: 1
total presub in shared programs: 7670 -> 7751 (1.06%)
presub in affected programs: 328 -> 409 (24.70%)
helped: 0
HURT: 81
total temps in shared programs: 16926 -> 16939 (0.08%)
temps in affected programs: 182 -> 195 (7.14%)
helped: 6
HURT: 19
total cycles in shared programs: 193751 -> 193729 (-0.01%)
cycles in affected programs: 3088 -> 3066 (-0.71%)
helped: 33
HURT: 5

Reviewed-by: Filip Gawin <filip.gawin@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24830>

r300: move power of two multipliers down

RV530 shader-db:
total instructions in shared programs: 128864 -> 128840 (-0.02%)
instructions in affected programs: 1260 -> 1236 (-1.90%)
helped: 21
HURT: 2
total presub in shared programs: 7682 -> 7670 (-0.16%)
presub in affected programs: 77 -> 65 (-15.58%)
helped: 12
HURT: 0
total omod in shared programs: 386 -> 403 (4.40%)
omod in affected programs: 3 -> 20 (566.67%)
helped: 0
HURT: 14
total temps in shared programs: 16948 -> 16926 (-0.13%)
temps in affected programs: 280 -> 258 (-7.86%)
helped: 20
HURT: 2
total cycles in shared programs: 194101 -> 193751 (-0.18%)
cycles in affected programs: 3422 -> 3072 (-10.23%)
helped: 25
HURT: 5

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/6855
Reviewed-by: Filip Gawin <filip.gawin@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24830>

r300: convert x * 2 into x + x for presubtract

total instructions in shared programs: 128859 -> 128864 (<.01%)
instructions in affected programs: 931 -> 936 (0.54%)
helped: 0
HURT: 5
total presub in shared programs: 7635 -> 7682 (0.62%)
presub in affected programs: 208 -> 255 (22.60%)
helped: 0
HURT: 17
total cycles in shared programs: 194124 -> 194101 (-0.01%)
cycles in affected programs: 1671 -> 1648 (-1.38%)
helped: 9
HURT: 1

Reviewed-by: Filip Gawin <filip.gawin@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24830>

r300: implement bias presubtract

RV530 shader-db:
total instructions in shared programs: 129468 -> 128859 (-0.47%)
instructions in affected programs: 34432 -> 33823 (-1.77%)
helped: 362
HURT: 56
total presub in shared programs: 5411 -> 7635 (41.10%)
presub in affected programs: 2069 -> 4293 (107.49%)
helped: 8
HURT: 468
total temps in shared programs: 16918 -> 16944 (0.15%)
temps in affected programs: 2022 -> 2048 (1.29%)
helped: 73
HURT: 79
total lits in shared programs: 3555 -> 2913 (-18.06%)
lits in affected programs: 2346 -> 1704 (-27.37%)
helped: 479
HURT: 0
total cycles in shared programs: 194675 -> 194124 (-0.28%)
cycles in affected programs: 62939 -> 62388 (-0.88%)
helped: 343
HURT: 84

Also dEQP-GLES2.functional.shaders.random.trigonometric.fragment.15
now fits into the instruction limit on RV370.

Reviewed-by: Filip Gawin <filip.gawin@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24830>

r300: exit early in presubtract is not supported

Reviewed-by: Filip Gawin <filip.gawin@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24830>

r300: reorder for easier presubtract 1-x pattern recognition

It is much easier to just add a simple late algebraic pass than actually
trying to teach the backend to recognize all the different patterns.

RV530 shader-db:
total instructions in shared programs: 129643 -> 129468 (-0.13%)
instructions in affected programs: 17665 -> 17490 (-0.99%)
helped: 176
HURT: 39
total presub in shared programs: 4912 -> 5411 (10.16%)
presub in affected programs: 1651 -> 2150 (30.22%)
helped: 0
HURT: 287
total temps in shared programs: 16904 -> 16918 (0.08%)
temps in affected programs: 812 -> 826 (1.72%)
helped: 25
HURT: 37
total cycles in shared programs: 194771 -> 194675 (-0.05%)
cycles in affected programs: 28096 -> 28000 (-0.34%)
helped: 146
HURT: 41

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/9364
Reviewed-by: Filip Gawin <filip.gawin@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24830>

zink: pass KERNEL shaders through successfully

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24868>

rusticl: fixes for zink shader images

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24868>

rusticl/device: _MAX_CONST_BUFFER0_SIZE is unsigned

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24868>

rusticl: add debug option to sync every event

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24868>

radv,aco: remove unused clip/cull distances variables

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24883>

aco: add support for compiling {VS,TES}+GS separately on GFX9+

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24862>

aco: ensure to initialize exec manually for non-monolithic {VS,TES}/GS on GFX9+

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24862>

radv: preserve shader arguments for non-monolithic {VS,TES}/GS on GFX9+

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24862>

radv: always declare some arguments for non-monolithic {VS,TES}/GS shaders

When compiling VS/TES and GS separately, we can't know if the GS will
need shader queries, so we have to always declare this argument.
Similar story for the view index.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24862>

iris: Check prog[] instead of uncompiled[] for BLORP state skipping

Huge thanks to Tapani Pälli for debugging this issue, figuring out
what was going wrong, proposing fixes, and walking me through where
things were going off the rails.

BLORP always disables tessellation and geometry shaders.  Our handling
tried to look at ice->shaders.uncompiled[] to determine whether the next
draw needed those shaders.  If not, we can leave BLORP's residual state
that disabled those stages in place, and skip looking at it.

Unfortunately, predicting the future is a bit fraught, in part due to
the uncompiled[] and prog[] arrays being slightly out of sync at times.

Consider the following case:

1. Draw with tessellation shaders in place

   => uncompiled[TES] and prog[TES] will both point at valid shaders.

2. Gallium calls pipe->bind_tes_state(NULL).

   => This makes uncompiled[TES] point at NULL, and flags
      IRIS_STAGE_DIRTY_UNCOMPILED_TES.

      Because iris_update_compiled_shaders() hasn't happened yet,
      uncompiled[TES] is NULL but prog[TES] has the stale TES from
      the previous draw still.

3. BLORP operations happen

   => BLORP sees uncompiled[TES] == NULL and decides that tessellation
      is off for the upcoming draws.  So it skips flagging tess state.

4. Gallium calls pipe->bind_tes_state(shader from step #1).

   => uncompiled[TES] points at the original shader.
      IRIS_STAGE_DIRTY_UNCOMPILED_TES gets flagged again.

5. Draw again

   => This calls iris_update_compiled_shaders(), which sees that
      a TES is bound, and calls iris_update_compiled_tes().  But
      because the same shader was bound as before, the program it
      comes up with is identical to the one already bound at
      ice->shaders.prog[TES].  So, it thinks it doesn't have to
      flag any tessellation state dirty because it was already
      set up for the last draw.

This random unbind and rebind between draws leads to a situation
where, at step #3, BLORP thinks it can skip flagging tessellation
state (nothing is bound), and at step #5, normal state handling
thinks it can skip flagging tessellation state (nothing changed
since last time).  So nobody does, and things break.

This unbind appears to be happening when st_release_variants()
decides it wants to free some shaders.  Then a rebind happens to
put back the actual shader for the draw.  So, it's not theoretical.

To fix this, we change BLORP to look at ice->shaders.prog[] rather
than uncompiled[].  This is equivalent to thinking about the previous
draw, rather than the next.  If the last draw had tessellation off,
then BLORP's disabling was a no-op, and the GPU is still in the same
state as the previous draw.  This is more reliable than predicting
the future.

Cc: mesa-stable
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/8308
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/9678
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24880>

venus: set deviceMemoryReport feature

VK_EXT_device_memory_report is implemented in venus driver side, which
has a feature struct. So we must enable it after setting features for
renderer extensions. This change also includes tiny format fixes.

Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24881>

venus: Use the common GetPhysicalDeviceFeatures2 implementation

Reviewed-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24881>

vulkan/wsi/x11: Implement capture hotkey using the keymap

This way, we can avoid opening another connection. The capture key is
changes to F1 because F12 has issues on Wayland. (After pressing F12,
all keys become unresponsive, refocussing the window fixes it)

Fixes: 291fa05  ("vulkan/wsi/x11: Capture traces using a hotkey")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/9578
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24710>

aco: add support for compiling VS+TCS separately on GFX9+

The VS will just jump to the TCS.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24697>

aco: ensure to initialize exec manually for VS as LS on GFX9+

When VS and TCS are compiled separately with shader object on GFX9+.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24697>

aco: disable shared VGPRs for non-monolithic shaders on GFX9+

For unmerged shaders on GFX9+, we would need to jump to the second
shader part which means shared VGPRs can't be enabled.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24697>

radv: preserve shader arguments for non-monolithic VS/TCS on GFX9+

This is more robust than re-creating the function signature in ACO.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24697>

ac: allow to mark shader arguments as preserved

These arguments would be used by ACO to generate a function signature
for merged shaders automatically.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24697>

radv: add a new shader argument for non-monolithic shaders PC

This will be used to jump from VS to TCS/GS or TES to GS on GFX9+.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24697>

radv: always declare some arguments for non-monolithic VS/TCS shaders

For separate VS/TCS compilation on GFX9+, the TCS might be using push
constants but not the VS and we can't know this information when
compiling the VS. Similar logic for the other arguments.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24697>

radv: force indirect descriptor sets for non-monolithic shaders

When VS and TCS are compiled separately on GFX9+, we can't know how
many descriptor sets are used for both stages and the function
arguments must match.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24697>

radv: do not inline push constants for non-monolithic shaders

It's hard to implement this because the function arguments must match
when eg. VS or TCS are compiled separately on GFX9+.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24697>

radv: use info->uses_view_index directly when declaring shader arguments

No need for a separate variable.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24697>

radv: add radv_shader_info::is_monolithic

This will be used to implement shader object on GFX9+.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24697>

anv/video: send h264 scaling list in raster order

ITU spec defines the H264 ScalingList{4x4,8x8} in zig-zag order, but
Intel HW wants raster order.

Reviewed-by: Lynne <dev@lynne.ee>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24572>

radv/video: send h264 scaling list in raster order

ITU spec defines the H264 ScalingList{4x4,8x8} in zig-zag order, but
AMD HW wants raster order.

Reviewed-by: Lynne <dev@lynne.ee>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24572>

util/vl: extract gallium vl scanning data to shared code

Vulkan video on both ANV and RADV need these data to make converting
from zig-zag to raster order easier.

Reviewed-by: Lynne <dev@lynne.ee>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24572>

anv/video: use vk_video_derive_h264_scaling_list

Reviewed-by: Hyunjun Ko <zzoon@igalia.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Lynne <dev@lynne.ee>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24572>

radv/video: use vk_video_derive_h264_scaling_list

Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Lynne <dev@lynne.ee>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24572>

vulkan/video: add helper to derive H264 scaling lists

The H264 spec defines a complicated way of layering PPS scaling lists
on top of SPS scaling lists. The details can be found in 7.4.2.1
(seq_scaling_matrix_present_flag semantics) and 7.4.2.2
(pic_scaling_matrix_present_flag semantics).

Both ANV and RADV need to derive the final scaling lists sent to HW
using this logic in order to handle H264 scaling lists correctly.

Reviewed-by: Hyunjun Ko <zzoon@igalia.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Lynne <dev@lynne.ee>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24572>

venus: add no_sparse debug option to disable sparse resource support

Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24877>

util/driconf: add workarounds for the Chronicles of Riddick

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24567>

zink: fix rewrite_read_as_0 filtering

Fixes: 9e42553ca8d ("zink: use lowered io (kinda) for i/o vars")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24874>

nouveau/mme: Fix a compile warning

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24840>

nvk: Plumb no_prefetch through to the DRM back-end

Instead of using bit 23 of nvk_cmd_push::range for this, pass it as a
separate bool.  This lets us use the actual kernel flag with the new
UAPI.

Reviewed-by: Danilo Krummrich <dakr@redhat.com>
Tested-by: Danilo Krummrich <dakr@redhat.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24840>

drm-uapi: Sync nouveau_drm.h

From https://cgit.freedesktop.org/drm-misc/

    commit 443f9e0b1ab5e3b95abf8606097d13e30e2f2413
    Author: Danilo Krummrich <dakr@redhat.com>
    Date:   Wed Aug 23 20:15:34 2023 +0200

        drm/nouveau: uapi: don't pass NO_PREFETCH flag implicitl

Reviewed-by: Danilo Krummrich <dakr@redhat.com>
Tested-by: Danilo Krummrich <dakr@redhat.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24840>

util/rb-tree: Fix typo in comment

Trivial.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24856>

util/rb-tree: Return the actual first node from rb_tree_search

Previously rb_tree_search would return the first node encountered, but
that may not be the first node that would be encoutnered by, say,
rb_tree_foreach.

Two test cases are added.

v2: Add more curly braces. Suggested by Faith.

Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24856>

ci/iris: add GL46.arrays_of_arrays_gl.SizedDeclarationsPrimitive timeout

Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24870>

tu: Pass real size of prime buffers to allocator

The msm driver reserves the actual DMABUF size in the memory map, while
TU can request smaller memory chunk to be allocated. This potentially
can lead to a situation when next allocation IOVA will be in the middle
of the address space which is reserved for the DMABUF. Pass the
`real_size' to TU allocator instead, so that kernel and userspace have
the same picture of memory allocations.

Cc: mesa-stable
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24861>

treewide: Also handle struct nir_builder form

Via Coccinelle patch:

    @def@
    typedef bool;
    typedef nir_builder;
    typedef nir_instr;
    typedef nir_def;
    identifier fn, instr, intr, x, builder, data;
    @@

    static fn(struct nir_builder* builder,
    -nir_instr *instr,
    +nir_intrinsic_instr *intr,
    ...)
    {
    (
    -   if (instr->type != nir_instr_type_intrinsic)
    -      return false;
    -   nir_intrinsic_instr *intr = nir_instr_as_intrinsic(instr);
    |
    -   nir_intrinsic_instr *intr = nir_instr_as_intrinsic(instr);
    -   if (instr->type != nir_instr_type_intrinsic)
    -      return false;
    )

    <...
    (
    -instr->x
    +intr->instr.x
    |
    -instr
    +&intr->instr
    )
    ...>

    }

    @pass depends on def@
    identifier def.fn;
    expression shader, progress;
    @@

    (
    -nir_shader_instructions_pass(shader, fn,
    +nir_shader_intrinsics_pass(shader, fn,
    ...)
    |
    -NIR_PASS_V(shader, nir_shader_instructions_pass, fn,
    +NIR_PASS_V(shader, nir_shader_intrinsics_pass, fn,
    ...)
    |
    -NIR_PASS(progress, shader, nir_shader_instructions_pass, fn,
    +NIR_PASS(progress, shader, nir_shader_intrinsics_pass, fn,
    ...)
    )

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Acked-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24852>

treewide: Use nir_shader_intrinsic_pass sometimes

This converts a lot of trivial passes. Nice boilerplate deletion. Via Coccinelle
patch (with a small manual fix-up for panfrost where coccinelle got confused by
genxml + ninja clang-format squashed in, and for Zink because my semantic patch
was slightly buggy).

    @def@
    typedef bool;
    typedef nir_builder;
    typedef nir_instr;
    typedef nir_def;
    identifier fn, instr, intr, x, builder, data;
    @@

    static fn(nir_builder* builder,
    -nir_instr *instr,
    +nir_intrinsic_instr *intr,
    ...)
    {
    (
    -   if (instr->type != nir_instr_type_intrinsic)
    -      return false;
    -   nir_intrinsic_instr *intr = nir_instr_as_intrinsic(instr);
    |
    -   nir_intrinsic_instr *intr = nir_instr_as_intrinsic(instr);
    -   if (instr->type != nir_instr_type_intrinsic)
    -      return false;
    )

    <...
    (
    -instr->x
    +intr->instr.x
    |
    -instr
    +&intr->instr
    )
    ...>

    }

    @pass depends on def@
    identifier def.fn;
    expression shader, progress;
    @@

    (
    -nir_shader_instructions_pass(shader, fn,
    +nir_shader_intrinsics_pass(shader, fn,
    ...)
    |
    -NIR_PASS_V(shader, nir_shader_instructions_pass, fn,
    +NIR_PASS_V(shader, nir_shader_intrinsics_pass, fn,
    ...)
    |
    -NIR_PASS(progress, shader, nir_shader_instructions_pass, fn,
    +NIR_PASS(progress, shader, nir_shader_intrinsics_pass, fn,
    ...)
    )

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Acked-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24852>

ci: switch to 6.4 kernel, improving Adreno 660 reliability

Second argument is to leave Linux 6.3 which is EOL.

Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24815>

ci/freedreno: There is only one King of Town.

Drop -r1, which got removed in the Linux 6.4.

Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24815>

ci/piglit: add extra space on top to prevent single quote getting into URL

Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24815>

zink: update some compute caps

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24859>

zink: fix global stores

We have to cast the value if the type doesn't match.

Fixes: ddc5c304899 ("zink: handle global and scratch vars")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24859>

zink: fix source type in load/store scratch

Fixes: ddc5c304899 ("zink: handle global and scratch vars")
Signed-off-by: Karol Herbst <git@karolherbst.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24859>

zink: use Aligned with global load/store ops

this is required by spec

Fixes: ddc5c304899 ("zink: handle global and scratch vars")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24859>