glsl/standalone: Pull program create/destroy out to a public function.

For reuse with unit tests.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21886>

freedreno/a6xx: Restore mode

We set the "mode" to 1 for CS because we want CP_SET_DRAW_STATE to
immediately execute the state groups.  But in the 3d path, we don't
restore the value in the sysmem path.  This was causing GPU faults
on 7c3 and presumably other a6xx gen4 things.  But somehow not on
a6xx gen1.

Let's just set it as part of initial state restore where we are
ensuring that the GPU is in a sane state.

Fixes: dec49ec50a2 ("freedreno/a6xx: Move CS state to PROG state group")
Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22080>

ci/docs: start documenting ci_run_n_monitor.py

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22079>

intel/fs: restore message layout changes for cube array

This reverts commit bc04e2dacac that handled the change as a WA while
this is about a new feature, change done in message layout. Patch also
changes the original comment to not refer to Wa but bspec page.

Fixes: bc04e2dacac ("intel/fs: use generated helpers for Wa_1209978020 / Wa_18012201914")
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Mark Janes <markjanes@swizzler.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22068>

Move combining clip and cull optimization before linking

As far gl_nir_link_glsl fills xfb data we should do it after lowering clip and cull in order to get correct locations.

Fixes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7152
Signed-off-by: Oleksii Bozhenko <oleksii.bozhenko@globallogic.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21625>

egl: don't expose swrast device if swrast is not built

This fixes piglit/egl_ext_device_base without swrast.

Reviewed-by: Yogesh Mohan Marimuthu <yogesh.mohanmarimuthu@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20951>

egl: reorder code in _eglQueryDevicesEXT, add *swrast variable

This is a preparation for the next commit. No change in behavior.

Reviewed-by: Yogesh Mohan Marimuthu <yogesh.mohanmarimuthu@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20951>

ci: deduplicate compiler wrappers

Signed-off-by: Eric Engestrom <eric@igalia.com>
Suggested-by: Michel Dänzer <mdaenzer@redhat.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21997>

ci: always use the -Werror wrapper

Signed-off-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21997>

ci: add linker wrapper for clang

Signed-off-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21997>

ci: centralize detection of ccache in link-werror wrapper

Signed-off-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21997>

zink/ci: add a test to the fails list

KHR-GL46.multi_bind.dispatch_bind_image_texture has been failing on
both Navi10 and VanGogh, so let's document that.

Zmike says he could not reproduce the fails on a newer version of
glcts, so the next release should address this issue.

Signed-off-by: Martin Roukala (né Peres) <martin.roukala@mupuf.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22055>

radv/ci: update the navi10 expectations

As part of https://gitlab.freedesktop.org/tanty/mesa-valve-ci/-/jobs/38416444,
we saw the following flakes:

 - dEQP-VK.dynamic_rendering.primary_cmd_buff.basic.2_cmdbuffers_resuming
 - dEQP-VK.dynamic_rendering.primary_cmd_buff.basic.contents_secondary_2_primary_cmdbuffers_resuming
 - dEQP-VK.pipeline.fast_linked_library.extended_dynamic_state.two_draws_static.topology_line

And the following failure (seen 4/4 times in the run):

 - dEQP-VK.draw.dynamic_rendering.primary_cmd_buff.linear_interpolation.offset_min_2_samples,Fail

Samuel told me that these are usual flakes, so let's document all of them.

Signed-off-by: Martin Roukala (né Peres) <martin.roukala@mupuf.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22055>

anv,hasvk: cleanup unused enum

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22070>

zink: track whether zsbuf is unused

zink_is_zsbuf_used updates dynamically, so its return may not reflect
the actual state of the current renderpass

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22054>

panfrost: fix strict-aliasing violations when packing fb ptrs

Compilers are free to make the assumption that pointers don't violate
strict aliasing. If that assumption is incorrect, as it is with the
framebuffer pointer packing code here, the job can fail.

This depends heavily on the compiler and optimization levels, so it's
hard to reproduce, but it did happen for at least two users running with
-O2 on gcc.

Fixes: 67cbbf941751 ("panfrost: Use framebuffer pointer XML")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/8627
Signed-off-by: Italo Nicola <italonicola@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21991>

lima/ci: restore swap buffers egl tests

With the weston idle-time issue fixed, these no longer timeout and
can be enabled again.

Signed-off-by: Erico Nunes <nunes.erico@gmail.com>
Reviewed-by: Vasily Khoruzhick <anarsoul@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22012>

nir/range_analysis: use perform_analysis() in nir_analyze_range()

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21381>

nir/range_analysis: use perform_analysis() in nir_unsigned_upper_bound()

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21381>

nir/range_analysis: add helpers for limiting stack usage

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21381>

nir/range_analysis: add missing masking of shift amounts

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Fixes: 72ac3f60261 ("nir: add nir_unsigned_upper_bound and nir_addition_might_overflow")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21381>

util/dynarray: allow an initial stack allocation to be used

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21381>

intel/isl: Support Yf/Ys/Tile-64 in isl_surf_get_image_offset_sa

All that's really needed here is to handle the array offsetting by using
an Z or array offset instead of the Y offset.

This patch originally changed get_image_offset_sa_gfx9_1d(), but since
we only use linear with the 1d case, it was dropped.

Rework:
 * Jordan: Include ISL_TILING_64 as well
 * Jordan: Drop change to get_image_offset_sa_gfx9_1d as
   recommended by Nanley

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21113>

radv: use common GetBufferMemoryRequirements2()

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22047>

v3d/ci: group dEQP-GLES3.functional.texture.specification.teximage2d_pbo.* flakes and add another one

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22063>

ir3: Use umod_imm

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22010>

radv: Use umod_imm

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22010>

v3d,v3dv: Use udiv_imm/umod_imm

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22010>

blorp,anv,hasvk: Use umod_imm

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22010>

nir/builder: Add nir_umod_imm helper

Like nir_udiv_imm, we can do a similar power-of-two trick. It's also really
convenient.

v2: Assert reasonable bounds on the modulus (Faith).

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl> [v1]
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com> [v1]
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22010>

nir: optimize i2f(f2i(fsign))

Foz-DB Navi10:
Totals from 3013 (2.23% of 134906) affected shaders:
VGPRs: 138068 -> 136964 (-0.80%); split: -0.80%, +0.00%
CodeSize: 10476416 -> 10391800 (-0.81%)
MaxWaves: 79118 -> 80088 (+1.23%)
Instrs: 1963227 -> 1945003 (-0.93%)
Latency: 24734883 -> 24649279 (-0.35%); split: -0.39%, +0.05%
InvThroughput: 6366777 -> 6334735 (-0.50%); split: -0.50%, +0.00%
VClause: 36845 -> 36882 (+0.10%); split: -0.26%, +0.36%
SClause: 59249 -> 59273 (+0.04%); split: -0.25%, +0.29%
Copies: 108570 -> 108501 (-0.06%); split: -0.19%, +0.13%
PreSGPRs: 105371 -> 105862 (+0.47%)
PreVGPRs: 117675 -> 116625 (-0.89%); split: -0.89%, +0.00%

Reviewed-by: Emma Anholt <emma@anholt.net>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22003>

pvr: Don't allocate/upload 0 size coeff programs

Fixes assert seen in dEQP-Vk.api.smoke.triangle

Signed-off-by: Jarred Davies <jarred.davies@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22051>

pvr: fix clang-format issue

Signed-off-by: Frank Binns <frank.binns@imgtec.com>
Reviewed-by: Luigi Santivetti <luigi.santivetti@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22056>

aco: drop leftover variable

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/8652
Fixes: 44fdd2ebcb27 ("aco: end reduce tmp after control flow, when used within control flow")

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22011>

ci/etnaviv: Get the gc2000_piglit manual job mostly working.

It was timing out, and the xfails were rather stale.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22039>

ci/hasvk: Update some xfails from the 8-sample fast clear disable.

Fixes: e509afacf3fb ("hasvk: Disable non-zero fast clears for 8xMSAA images")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22039>

ci/crocus: Update expectations from VK CTS 1.3.5.0.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22039>

ci/iris: Update more manual job xfails from the Wayland build change.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22039>

radv: implement VK_KHR_map_memory2

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22032>

radv: pass radv_shader to radv_shader_need_indirect_descriptor_sets()

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21878>

radv: pass radv_ray_tracing_pipeline to radv_rt_pipeline_compile()

The base.base is redundant but this will allow us to move the shaders
array out of radv_pipeline.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21878>

radv: rework dumping shaders when a GPU hang is reported

Preliminary work for moving the shaders array outside of radv_pipeline.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21878>

radv: pass radv_shader to radv_dump_shader_stats()

Preliminary work for moving the shaders array outside of radv_pipeline.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21878>

radv: use a separate compute path in radv_flush_constants()

radv_get_shader() will be a radv_graphics_pipeline only function.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21878>

radv: pass shader/base_reg to radv_emit_userdata_address()

Preliminary work for moving the shaders array outside of radv_pipeline.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21878>

radv: pass shader/base_reg to radv_emit_inline_push_consts()

Preliminary work for moving the shaders array outside of radv_pipeline.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21878>

radv: pass shader/base_reg to radv_emit_descriptor_pointers()

Preliminary work for moving the shaders array outside of radv_pipeline.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21878>

radv: replace radv_lookup_user_sgpr() by radv_get_user_sgpr()

radv_get_user_sgpr() no longer relies on radv_pipeline which is
another step for moving the shaders array outside of it.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21878>

radv: stop using radv_get_shader_shader() for task shaders

radv_get_shader() should only be used for VS or TES, no need to add
another indirection for task shaders. While we are at it, rename
compute_shader to task_shader.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21878>

gallium: Fix u_stream_outputs_for_vertices with QUADS

Per the spec. This helper is only used in nv50 and panfrost, the latter is known
to have a completely broken transform feedback implementation and I'd be
unsurprised if the same is true for nv50. So unsurprising that compatibility
profile interaction was missed.

This is part of the Piglit ext_transform_feedback-tessellation quads puzzle.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Acked-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22013>

docs: remove old thanks-article

This article is so out of date it's of no real use any more, and
updating it seems quite pointless. Let's just move it to the graveyard,
and forward any readers to the version we have in the amber-release.

Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22028>

docs: move developers article to main website

This isn't content that relates directly to a specific release of Mesa,
and it's also quite out-of-date. Let's move it to the main mesa website
instead, where we have an updated version.

Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22028>

ac/nir/ngg: Rename repacked variables to clarify their name.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21995>

ac/nir/ngg: Create separate variable for repacked rel_patch_id.

This should be handled separately from the other repacked
variables, because it doesn't use a dword.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21995>

ac/nir/ngg: Remove some superfluous variables.

These mostly existed because of the long name of the state variable
and are not really necessary.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21995>

ac/nir/ngg: Rename state variables to "s".

It was a bad idea to give them a long name, let's correct it.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21995>

ac/nir/ngg: Move divergence analysis call to analyze_shader_before_culling.

It actually belongs there.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21995>

ac/nir/ngg: Split some functions out of save_reusable_variables.

To make the function a little easier to read and understand.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21995>

ac/nir/ngg: Rename saved_uniform to reusable_nondeferred_variable.

We are planning to reuse more than just uniforms later,
hence let's clarify the name of these.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21995>

ac/nir/ngg: Remove usused lds_es enum values.

These were meant to explain the LDS layout, but
the actual LDS usage is better explained by:
ngg_nogs_get_culling_pervertex_lds_size().
Also add some comments there.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21995>

docs/zink: clean up requirements-language

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22053>

docs/zink: mention vk1.2 mirror-clamp feature option

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22053>

ac/llvm: fix build with LLVM 17

This builds with LLVM 12 -> 17 and a running a simple app seems to work.

I couldn't test LLVM 11 because meson fails with:

    Looking for a fallback subproject for the dependency llvm (modules:
    bitwriter, engine, mcdisassembler, mcjit, core, executionengine,
    scalaropts, transformutils, instcombine, amdgpu, bitreader, ipo, native)

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/8297
Cc: mesa-stable
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22021>

vulkan: depend idep_vulkan_runtime_headers on vk_physical_device_features.h

Fixes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/8643
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21958>

vulkan: fix building with python3.8

Fixes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/8640
Reviewed-by: Ricardo Garcia <rgarcia@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21958>

radv: add support for caching PS epilogs

For PS epilogs created at link time because libraries are still not
cached.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21897>

radv: make radv_shader_part_create() non-static

It will be used when creating shaders from the cache.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21897>

radv: allow to return the PS epilog binary to the pipeline

To add it to the shaders cache.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21897>

radv: upload prologs/epilogs as part of radv_shader_part_create()

Don't need to duplicate this code.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21897>

radv: store the total radv_shader_part_binary size

Similar to radv_shader. This will be used for the shaders cache.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21897>

radv: store spi_shader_col_format to radv_shader_part_binary

For PS epilogs in the cache.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21897>

radv: stop storing the binary as part of radv_shader_part

It's unnecessary to keep a pointer to the binary.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21897>

radv: remove set but never used num_preserved_sgprs

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21897>

iris: implement occlusion query related Wa_14017076903

Fixes artifacts on some games that relied on occlusion query
results when no PS or depth buffers are bound.

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/21721>

anv: implement occlusion query related Wa_14017076903

Fixes artifacts on some games that relied on occlusion query
results when no PS or depth buffers are bound.

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/21721>

v3d: implement NV_conditional_render extension

The hardware doesn't support native conditional rendering, so it is
implemented by software.

Code borrowed from Freedreno and Panfrost.

Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Signed-off-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17373>

v3d: include offset as part of streamout target

When dealing with multiple Transform Feedback buffers, each of them
needs to have their own offset, so when resuming from one to another we
know exactly were to continue adding primitives.

Fixes "spec@arb_transform_feedback2@change objects while paused (gles3)"
piglit test.

Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Reviewed-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17373>

v3d: cache pipe query results

As the BO storing the results is destroyed after getting the query
results, store the results in case requesting the results again.

Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Signed-off-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17373>

anv: report shader max dispatch width in pipeline props

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/22014>

intel/compiler: report max dispatch width statistic

Most tools looking at shader stats assume that there is only a single
resulting binary shader out of a single input. On Intel HW this is not
always the case. So having a statistic on each variant that reports
the maximum dispatch width helps showing improvement on a single
shader in terms of how large we manage to compile it.

For shaders that can be compiled in multiple SIMD width (like fragment
shaders), this will report the maximum dispatch width in the
statistics of each variants.

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/22014>

broadcom/compiler: track pending ldtmu count with each TMU lookup

And use this information when scheduling QPU to avoid merging
a new TMU request into a previous ldtmu instruction when doing
so may cause TMU output fifo overflow due to a stalling ldtmu.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22044>

docs: fixup broken indentation

This extra indentation caused these to be inside <blockquote>-tags,
which obviously isn't what we want.

Reviewed-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21927>

docs: escape a few more strings

This makes it more obvious that these are not just plain english words.

Reviewed-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21896>

docs: fixup broken envvar-role syntax

This needs a leading colon, otherwise it renders as "envvar:" followed
with environment variable using the c-expression role.

Fixes: 4c19426fd6a ("docs/envvars: add missing mesa disk cache envvars")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21896>

radv: enable fullyCoveredFragmentShaderInputVariable on GFX9+

For vkd3d-proton Tier 3.

It's passing CTS and vkd3d-proton tests.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/8311
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21497>

radv: implement fullyCoveredFragmentShaderInputVariable

1 means INNER_COVERAGE.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21497>

radv: enable SAMPLE_COVERAGE_ENA if the fully covered built-in is used

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21497>

radv: lower nir_intrinsic_load_fully_covered

The sample coverage VGPR input would be the inner coverage and 0 means
it's uncovered.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21497>

spirv,nir: add support for SpvBuiltInFullyCoveredEXT

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/21497>

spirv: add SpvCapabilityFragmentFullyCoveredEXT

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/21497>

radv: initialize cmd_buffer upload list earlier

When cs allocation fails in radv_create_cmd_buffer,
radv_destroy_cmd_buffer is called before returning
VK_ERROR_OUT_OF_HOST_MEMORY. At that point, the upload list is not
initalized yet, so SIGSEGV will occur when trying to iterate through the
upload bo list. Initialize the upload list earlier to avoid this.

Signed-off-by: Benjamin Cheng <ben@bcheng.me>
Reviewed-by: Tatsuyuki Ishi <ishitatsuyuki@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22016>

zink: add ZINK_DEBUG=flushsync

this disables the submission thread

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22040>

zink: don't update fbfetch in db mode if inputAttachmentDescriptorSize==0

turnip gets away with this somehow, so silence harmless validation errors

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22040>

zink: rework handling of unordered->ordered write buffer barriers

this improves handling for barriers that originate from a write in the
unordered cmdbuf, adding tracking to resources to better determine access
in the unordered cmdbuf and then utilizing that to generate a single split
memory barrier added at the end of the unordered cmdbuf for all the buffers
written to on that cmdbuf

the next step will be to also merge the read access down onto the end-of-cmdbuf
barrier so that all stream upload-type functionality becomes a single barrier

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22027>

zink: always set sampler layouts when unbinding fb images while rp optimizing

this may or may not always be accurate, but in the cases where it is, it should
avoid some renderpass splitting, and in the cases where it isn't, there may
already be issues

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22027>

zink: only run post-fb-unbind layout stuff if the resource isn't being destroyed

if refcount==1, the resource won't be used any further, and all the code
in this block is only for resources that will be used as samplers,
so don't run any of it

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22027>

zink: add an assert to ensure zsbuf invalidation doesn't break rendering

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22027>

zink: force unordered_write=false when binding image descriptors

there's no way to link up image layouts between the unordered cmdbuf
and the main one, so if an op is promoted to unordered after an image
is used as a descriptor, the layout will be broken

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22027>

zink: fix unordered access for image descriptors

previously the unordered access flags would be set before the deferred
barrier was added, which would guarantee no descriptor barriers could
be deferred and thus terminate renderpasses any time a new descriptor
was bound that was both an image and needed a layout change

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22027>

zink: add batch refs for framebuffer surfaces on bind and ref update

if this scenario occurs:

* bind fb on ctx A
* draw
* flush + change context to B
* read fb on ctx B
* delete ctx A

then a dead batch write will be left on the fb bo

cc: mesa-stable

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22027>