docs: expand mobile-menu without js

Without javascript support, a user can't expand the menu. So let's
leave it always-expanded in that case.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24721>

radv/amdgpu: add support for submitting external IBs with the chained path

External IBs are currently only used for DGC. With the chained path,
these IBs will only be used to workaround missing IB2 packet on the
compute queue, which is rare enough to care about chaining inside CS.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24207>

radv/amdgpu: allow to execute external IBs on the compute queue

IB2 isn't supported on ACE, so external IBs should be submitted as IB1.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24207>

Revert "radv/amdgpu: skip adding per VM BOs for sparse during CS BO list build"

This reverts commit 51caece74cd54e99a5f89521e8feacd894781ca2.

Cc: mesa-stable
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24774>

Revert "radv/amdgpu: workaround a kernel bug when replacing sparse mappings"

This workaround was added temporarily but it can actually cause
stuttering in some games like Forza Horizon 5.

The kernel fix
(https://lists.freedesktop.org/archives/amd-gfx/2023-June/094648.html)
landed in some stable kernels (5.15.121+, 6.1.40+ and 6.4.5+). Sadly,
older stable kernels don't have it, so you might experiment random GPU
hangs in games that use sparse mapping. Please ensure your kernel is
up-to-date for the best experience.

This reverts commit 9b00867327c2b266fcdebcef8bc7e7497eaab06b.

Cc: mesa-stable
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/9443
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24774>

etnaviv: unbreak cmdline compiler

Fixes the following assert:
  etnaviv_compiler: ../src/compiler/glsl_types.cpp:1219: static const glsl_type* glsl_type::get_array_instance(const glsl_type*, unsigned int, unsigned int): Assertion `glsl_type_users > 0' failed.

Signed-off-by: Christian Gmeiner <cgmeiner@igalia.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24753>

v3d/qpu: fix type of function argument

Fixes: 05c7d9715b8a419fd6fb ("broadcom: Add V3D 3.3 QPU instruction pack, unpack, and disasm.")
Signed-off-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24787>

v3dv: fix shader stage name in error message

Fixes: 60145629a2bdcd4e7835 ("v3dv: initial CreateGraphicsPipeline/DestroyPipeline implementation")
Signed-off-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24787>

v3dv: fix copy/pasted type of `sample`

And use the type in functions instead of a generic `uint32_t` to make it
easier to notice the wrong type.

Fixes: 47e02a2ef11237cf4c2c ("v3dv: add a fast path for vkCmdClearAttachments")
Signed-off-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24787>

v3dv: fix VK_PIPELINE_ROBUSTNESS_{BUFFER,IMAGE}_BEHAVIOR_DEVICE_DEFAULT_EXT copy/paste typo

Fixes: 24d9a80247605ac2c237 ("v3dv: implement VK_EXT_pipeline_robustness")
Signed-off-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24787>

docs: Add documentation for gpuvis.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22505>

vulkan: Add CPU tracing for vkWaitForFences.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22505>

egl,venus,vulkan,turnip,freedreno: Update CPU trace init to init more than perfetto.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22505>

util/perf: Add gpuvis integration.

Initial integration, still needs the init functions to be changed
across the codebase.

For the context usage https://github.com/mikesart/gpuvis/pull/82 is
needed to display it correctly in gpuvis.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22505>

radeonsi: fix templated si_draw_rectangle callback for Navi14

Navi14 is the only gfx10 chip that doesn't enable NGG.

Fixes: cd7e20f51388 ("radeonsi: specialize si_draw_rectangle using a C++ template")

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24759>

radeonsi/gfx11: pass attribute ring addr via SGPR instead of memory for blits

This removes the scalar memory load from blit vertex shaders.

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24759>

radeonsi: don't pass gl_Layer to PS for blit shaders

This will be required by the next commit to prevent hangs. See the comment.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24759>

radeonsi: simplify/merge emit_shader_ngg functions

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24759>

radeonsi: improve the heuristic when to use Wave32 for compute shaders

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24759>

radeonsi: rename uses_subgroup_info to uses_tg_size

that's the name of the SGPR

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24759>

radeonsi: allow setting any index in radeon_set_sh_reg_idx

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24759>

ac/surface: add radeon_surf::u::gfx9::uses_custom_pitch

so that we don't try to guess when the pitch is overridden

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24759>

ac/surface: trivial non-functional changes

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24759>

ac: add a standalone IB parser program

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24759>

ac: update gfx11 shadowed register tables

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24759>

ac: improve the IB parser

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24759>

ac: change offsets of DMA_DATA dwords to prevent reg offset conflicts

Use non-multiples of 4.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24759>

ac: minor updates to packet documentation and definitions

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24759>

ac: document ac_shader_args::gs_vtx_offset

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24759>

ac: implement AMD_FORCE_FAMILY properly, remove SI_FORCE_FAMILY

This sets radeon_info for the forced family correctly.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24759>

Revert "ac: don't call ac_query_pci_bus_info from ac_query_gpu_info"

This reverts commit a48642400be86a17d460ce2786988ec67f423f01.

Instead, add a new parameter require_pci_bus_info to control the behavior.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24759>

ac/gpu_info: add the /dev/dri/ filename into radeon_info

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24759>

nir/algebraic: collapse ALU opcodes sourcing NaN

Undef will be replaced by NaN whenever it leads to elimination of FP
instructions. This implements the elimination part.

Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24792>

nir: fix constant evaluation of fddx/fddy sourcing Inf & NaN constant

A derivative of NaN is NaN.

Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24792>

nir/opt_dead_cf: Remove if branches with undef condition.

Treat them as if the undef were false.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24792>

nir/passthrough_gs: Fix array size

Triangle strips with adjacency have 6 vertices input, so we need an array big
enough for all 6 vertices to avoid overflow. Fixes passthrough GS generated for
KHR-GLES31.core.draw_indirect.basic-mode-*-triangle*adj*.

Fixes: ea14579f3dc ("nir: handle primitives with adjacency")
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Antonino Maniscalco <antonino.maniscalco@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24780>

nir/passthrough_gs: Correctly set vertices_in

If the input primitive has adjacency, the output primitive will have fewer
vertices than the input. For example, if we input TRIANGLE_STRIPS_ADJACENCY, we
need to set vertices_in = 6 even though we'll output TRIANGLE_STRIPS with
vertices_out = 3. Respect that, in order to correctly handle adjacency inputs.

Fixes: ea14579f3dc ("nir: handle primitives with adjacency")
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Antonino Maniscalco <antonino.maniscalco@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24780>

nir/lower_helper_writes: Consider bindless images

These need to be handled like other image ops.

Fixes KHR-GLES31.core.shader_image_load_store.basic-allTargets-atomicFS on Asahi
with bindless image access forced.

Fixes: 586da7b3290 ("nir: Add nir_lower_helper_writes pass")
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Italo Nicola <italonicola@collabora.com>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24778>

nv/codegen: Use nir_lower_clip

Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24653>

zink: fix format in zink_make_{image,texture}_handle_resident()

`ds->db.format` is a `pipe_format`, while `buffer_infos[handle].format` is
a `VkFormat`; the conversion from one to the other was missing.

Fixes: 99ba529feed6f9917a44 ("zink: implement descriptor buffer handling of bindless texture")
Signed-off-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24790>

vc4: drop duplicate .lower_ldexp

It's set 3 lines above already.

Fixes: 2a33ea95d66cd7ba83e2 ("glsl: Retire ldexp lowering in favor of the nir lowering flag.")
Signed-off-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24786>

egl: Advertise EGL_MESA_gl_interop extension if support present

Signed-off-by: Sylvain Munaut <tnt@246tNt.com>
Acked-by: Antonio Gomes <antoniospg100@gmail.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24573>

glx: Advertise GLX_MESA_gl_interop extension if support present

Signed-off-by: Sylvain Munaut <tnt@246tNt.com>
Acked-by: Antonio Gomes <antoniospg100@gmail.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24573>

glx: Remove MESA_depth_float_bit from enum

This is not used anywhere. Been there from the beginning of this
file but with no related code in glxextensions.c

Signed-off-by: Sylvain Munaut <tnt@246tNt.com>
Acked-by: Antonio Gomes <antoniospg100@gmail.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24573>

egl: Export the MESA GL Interop functions through eglGetProcAddress

Signed-off-by: Sylvain Munaut <tnt@246tNt.com>
Acked-by: Antonio Gomes <antoniospg100@gmail.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24573>

glx: Export the MESA GL Interop functions through glXGetProcAddress

Signed-off-by: Sylvain Munaut <tnt@246tNt.com>
Acked-by: Antonio Gomes <antoniospg100@gmail.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24573>

glx: Add missing MesaGLInteropGLXFlushObjects

All the plumbing was added in 8d55fb54b15381d7fa950f448e5c67d0556088db
but for some reason the top level access method was not ...

Signed-off-by: Sylvain Munaut <tnt@246tNt.com>
Acked-by: Antonio Gomes <antoniospg100@gmail.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24573>

include: Fix the PFN declarations to be pointers as they should

Broken by b5f9820d905a275bc01bbffa9b4927ec11286f8d back in 2016.

Signed-off-by: Sylvain Munaut <tnt@246tNt.com>
Acked-by: Antonio Gomes <antoniospg100@gmail.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24573>

egl: bump extension string length

We've actually been over the 1000 char limit for a while, but we didn't
have a driver in CI that enabled enough extensions to notice it.

If all currently supported extensions are enabled, we need 1441 chars.
Let's bump the string to 2048 to have a little bit of margin.

Signed-off-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24781>

agx: Lower fquantize2f16

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24616>

nir: Lower fquantize2f16

Passes dEQP-VK.spirv_assembly.*opquantize*.

Unlike the DXIL lowering, this should correctly handle NaNs. (I belive Dozen has
a bug here that is masked by running constant folding early and poor CTS
coverage.) It is also faster than the DXIL lowering for hardware that supports
f2f16 conversions natively. It is not as good as a backend implementation that
could flush-to-zero in hardware... but for a debug instruction it should be more
than good enough.

It might be slightly better to multiply with 0.0 to get the appropriate zero,
but NIR really likes optimizing that out ...

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24616>

gtest: backport ansi color fix

Fixes ubsan runs for Mesa3D.

Adds prevents from returning
nullptr by choosing default color.

Upstream PR: https://github.com/google/googletest/pull/4322

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/9404
Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24783>

ci/freedreno: reuse freedreno_gl_file_list instead of re-definining it

Fixes: 9d442b459a43264c2899 ("ci/freedreno: handle disabling farm properly for each FD/Collabora farm")
Signed-off-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24779>

r300: add dEQP baseline for RV370 with forced swtcl

Reviewed-by: Filip Gawin <filip.gawin@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24769>

r300: don't abort on flow control when using draw for vs

It can handle it just fine. Around 250 dEQPs go from Skip to Pass.

Fixes: 1021e2b946b18739b65b575b0770e6158ba05592
Reviewed-by: Filip Gawin <filip.gawin@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24769>

r600: better tracking for vertex buffer emission

Fixes: 76725452 (gallium: move vertex stride to CSO)

Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24754>

r600: store the mask of buffers used by a vertex state

Fixes: 76725452 (gallium: move vertex stride to CSO)

Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24754>

radeonsi/vcn: Fix leaking fences in decode

Unref fence used in destroy.

Don't store the fence reference in picture->fence, instead
keep it in the driver. Because only the last fence will
now be valid, check the fence pointer in get_decoder_fence.

Reviewed-by: Leo Liu <leo.liu@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24677>

nir: Use nir_shader_intrinsic_pass() a few places

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Acked-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24760>

nir: Fix metadata in nir_lower_is_helper_invocation

It does not preserve everything.  It adds and removes instructions and
even adds a variable.

Fixes: f17b41ab4f01 ("nir: add lowering pass for helperInvocationEXT()")
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Acked-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24760>

nir: Add nir_shader_intrinsics_pass

Like instructions_pass but specialized to intrinsics. More ergnomic for this
extremely common case, and possibly a bit faster by avoiding the extra function
call on non-intrinsics.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Acked-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24760>

lavapipe: Advertise AMDX_shader_enqueue

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24512>

lavapipe: Implement AMDX_shader_enqueue commands

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24512>

lavapipe: Implement exec graph pipelines

Just a collection of compute shaders that can enqueue each other.

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24512>

lavapipe: Add lvp_pipeline_type

A boolean is not enough to support exec graph pipelines.

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24512>

spirv: Implement SPV_AMDX_shader_enqueue

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24512>

spirv: Update headers and grammer JSON

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24512>

nir: Add shader enqueue data structures and handling

There are two new variable modes:
- nir_var_mem_node_payload
- nir_var_mem_node_payload_in

Also add a few more intrinsics and some shader info.

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24512>

vulkan Add enqueue entrypoint for CmdDispatchGraphAMDX

The generyted one doesn't copy deep enough.

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24512>

vulkan: Allow beta extensions for physical device properties

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24512>

vulkan: Allow beta extensions for physical device features

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24512>

bin: Update spirv sources

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24512>

radeonsi: modify algorithm of skipping holes of sparse bo

Modify current algorithm of skipping holes of sparse bo to cover the
following using situations:
1. The whole sparse buffer is uncommitted.
2. More than one page that in the tail of sparse buffer are uncommitted.

Signed-off-by: Julia Zhang <julia.zhang@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24535>

radeonsi: limit CP DMA to skip holes in sparse bo

CP DMA on gfx9 can't handle the hole in sparse buffer. The fix skip
sparse bo hole so that arb_sparse_buffer-buffer-data &&
arb_sparse_buffer-commit pass

Signed-off-by: Flora Cui <flora.cui@amd.com>
Signed-off-by: Julia Zhang <julia.zhang@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24535>

anv: emitting 3DSTATE_PRIMITIVE_REPLICATION is required on Gen12+

This change helps fix the following tests on future platforms:
  - func.multiview
  - dEQP-VK.fragment_shading_rate.renderpass2.monolithic.multiviewsrlayered.dynamic.attachment.noshaderrate.keep.replace.1x1.samples1.vs
  - anything else that uses multiview

Signed-off-by: Rohan Garg <rohan.garg@intel.com>
Reviewed-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/24746>

dri: Use RGB internal formats for RGBX formats

These formats do not contain alpha channel, so their internal formats
should reflect that.

Fixes: bf576772 ("dri_util: add driImageFormatToSizedInternalGLFormat function")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/9429
Signed-off-by: Sviatoslav Peleshko <sviatoslav.peleshko@globallogic.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24580>

ci/freedreno: another batch of Adreno 530 flakes

Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24772>

radv: stop copying if VS or TES uses the InvocationID built-in

It's only allowed in TCS or GS which means the src shader stage
value is always FALSE.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24691>

radv: simplify declaring VS specific input SGPRs

stage/previous_stage are actually useless.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24691>

radv: remove unused param from radv_pipeline_init_multisample_state()

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24691>

radv: remove radv_cmd_buffer::cached_vertex_formats

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24770>

radv: fix emitting TCS epilogs for GFX6-9

The number of SGPRs need to be adjusted.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24747>

radv: add missing comment about TCS_OFFCHIP_LAYOUT_LSHS_VERTEX_STRIDE

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24745>

radv: reduce TCS_OFFCHIP_LAYOUT_NUM_PATCHES to 6-bits

RADV clamps the number of tess patches to 40.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24745>

nir: Drop nir_push_if_src()

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24765>

nir: Drop nir_instr_rewrite_src()

Replace all its remaining users with nir_src_rewrite().

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24729>

nir: Drop most uses if nir_instr_rewrite_src()

Generated by the following semantic patch:

    @@
    expression I, S, D;
    @@

    -nir_instr_rewrite_src(I, S, nir_src_for_ssa(D));
    +nir_src_rewrite(S, D);

Acked-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24729>

nir: Drop nir_instr_rewrite_src_ssa()

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24729>

nir: Drop most uses of nir_instr_rewrite_src_ssa()

Generated with the following semantic patch:

    @@
    expression I, S, D;
    @@

    -nir_instr_rewrite_src_ssa(I, S, D);
    +nir_src_rewrite(S, D);

Acked-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24729>

nir: Drop nir_if_rewrite_condition()

Use nir_src_rewrite() instead.  In a couple of cases, we can even drop a
switch on whether or not it's an if source.

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24729>

nir: Add and use a nir_instr_init_src() helper

This helper exists for a very tiny set of use-cases but it's better to
have the helper live in nir.c than hand-roll it elsewhere.

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24729>

nir: Add a nir_instr_clear_src() helper and use it

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24729>

nir/opt_undef: Don't rewrite a bcsel to mov

Technically, it's possible because bcsel has more sources than mov.
However, it's not worth the pain of trying to get it right.

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24729>

nir: Take a nir_def * in nir_phi_instr_add_src()

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24729>

nir: Take a nir_def * in nir_tex_instr_add_src()

NIR bits were hand-typed.  Driver updates done through the following
semantic patch:

    @@
    expression T, ST, D;
    @@

    -nir_tex_instr_add_src(T, ST, nir_src_for_ssa(D));
    +nir_tex_instr_add_src(T, ST, D);

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24729>

ci/etnaviv: update ci expectation

Signed-off-by: Christian Gmeiner <cgmeiner@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24751>

nv50: limit max code uploads to 0x8000

I have no idea why a bigger size doesn't work, the hardware doesn't
complain, but it turns out that uploading big shaders still causes issues
with the old limit. *shrug*

Fixes: 7f63d2ebdbc ("nv50: fix code uploads bigger than 0x10000 bytes")
Signed-off-by: Karol Herbst <git@karolherbst.de>
Reviewed-by: M Henning <drawoc@darkrefraction.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24758>

iris,crocus: drop unnecessary DEBUG_NO_CCS/NO_HIZ checks

Now isl_surf_supports_ccs helper handles DEBUG_NO_CCS check and
isl_surf_get_hiz_surf handles DEBUG_NO_HIZ, so we don't
need to check it everywhere.

Signed-off-by: Sagar Ghuge <sagar.ghuge@intel.com>
Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24731>

anv,hasvk: drop unnecessary DEBUG_NO_CCS/NO_HIZ checks

Now isl_surf_supports_ccs helper handles DEBUG_NO_CCS check and
isl_surf_get_hiz_surf handles DEBUG_NO_HIZ, so we don't
need to check it everywhere.

Signed-off-by: Sagar Ghuge <sagar.ghuge@intel.com>
Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24731>

intel/isl: Enable INTEL_DEBUG=noccs/nohiz in ISL helpers

Let's enable INTEL_DEBUG=noccs in isl_surf_supports_ccs helper and
INTEL_DEBUG=nohiz in isl_surf_get_hiz_surf helper.

Signed-off-by: Sagar Ghuge <sagar.ghuge@intel.com>
Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24731>