radv: Use GLSL matrices for instance transforms in BVH.

Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18692>

radv: Use proper matrices for instance nodes.

Converts both wto and otw matrices to be full row-major 4x3 matrices.

Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18692>

radv: Remove aabb bounds from instance nodes.

I need space ...

Furthermore, this only gets used during the build, and we can eat
the cost of generating the AABB a second time there.

Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18692>

radv: Translate the BVH copy shader to glsl from nir_builder.

Much easier to change.

Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18692>

radv: Hardcode root node id.

Optimizes code a tiny bit, and avoid the hack of encoding the root
node id in the low bits of the BLAS address in an instance node.

This is needed to adjust serialization/deserialization as the
instance address there has to be the base address, so this avoids
some wrapping/unwrapping.

Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18692>

radv: Fix GLSL BDA struct alignment and use pointer arithmetic SIZEOF.

Use pointer arithmetic from GL_EXT_buffer_reference2 to replace explicit
struct sizes. For that to work we also need to fix BDA alignment by setting
it to the smallest possible value (default is 16).

We could also replace the INDEX macro with a[b] operators, but that's actually
a change in behavior because a[b] always uses 64bit arithmetic.

Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18765>

intel/l3: Use L3 full-way allocation setting for gfx12.5 (DG2, MTL)

For now we use an empty set of L3 config settings on DG2 & MTL, which
will cause the L3 programming to set L3FullWayAllocationEnable.

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18770>

d3d12: Fix usage of %ld with PRIu64 instead for cross-platform build

Reviewed-by: Yonggang Luo <luoyonggang@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18777>

zink: PIPE_CAP_SHAREABLE_SHADERS

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18689>

zink: use screen param for zink_shader_free()

the context is no longer used

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18689>

zink: use program's context for cache deletion

this forces deletion of the program from its owner context under lock

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18689>

zink: add locking for program caches

these are context-based, so destroying them from another context needs locks

locks are not currently able to be contended

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18689>

zink: don't try to delete programs from the cache if stages have been modified

this is guaranteed to fail anyway

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18689>

zink: use screen params for program reference functions

these never used the context anyway

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18689>

zink: track the owner context for gfx programs

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18689>

zink: track a mask of gfx stages remaining in gfx programs

this provides a hint to improve program deletion

cc: mesa-stable

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18689>

freedreno: Disable 8bpp_ubwc on a6xx gen2

Fixes text corruption in VSCode on a680.

Signed-off-by: Joshua Ashton <joshua@froggi.es>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18779>

radeon/vcn: process invalid frame rate in encoding

Just in case app sending zeros in frame rate,
which could cause zero division, it will be treated
as fixed value, 30fps.

Signed-off-by: Ruijing Dong <ruijing.dong@amd.com>
Reviewed-by: Boyuan Zhang <boyuan.zhang@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18737>

radeonsi/gfx11: don't set VERTS_PER_SUBGRP to 0

It seems slower.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Fixes: 25a66477d02 ("radeonsi/gfx11: register changes")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18758>

winsys/amdgpu: init 'r' before using it

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Fixes: 471c82d21ef ("winsys/amdgpu: flatten huge if and reorder code in amdgpu_cs_submit_ib")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18758>

iris: implement Wa_14016118574

After each 3DPRIMITIVE, we need to send a dummy post sync op if point or
line list was used or if had only 1 or 2 vertices per primitive.

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/18746>

anv: implement Wa_14016118574

After each 3DPRIMITIVE, we need to send a dummy post sync op if point or
line list was used or if had only 1 or 2 vertices per primitive.

v2: add missing _3DPRIM_POINTLIST_BF (Lionel)

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/18746>

zink: correct emulation of luminance-alpha formats

util_format_luminance_to_red returns a red-alpha format, not a red-green
format, which is what we need in Vulkan. So let's also call
emulate_red_alpha so we can finish up with a valid format.

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18596>

zink: fix luminance/luminance-alpha emulation

We need to check the pipe_resource's format, not the texture-view
format. We're trying to overcome differences in the physical memory
layout, not the interpretation of the view.

But since we're now looking at the resource-format, we also need to take
the view-swizzle into account, because the state tracker likes to do
blits between LA and RA formats.

This had little effect in practice for luminance formats, but for
luminance-alpha formats things went kinda sideways. But luckily,
we don't actually use those formats yet, triggering state-tracker
emulation instead.

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18596>

zink: refactor swizzle-clamping code

It's going to be a lot easier to work on this if we can use loops
instead of open-coding. So let's move the swizzles into arrays, process
them, and then stick it back into their place at the end.

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18596>

zink: fixup indent

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18596>

zink: fix luminance clears

We should clear the red and green channels for luminance alpha formats,
and leave the other channels alone.

It doesn't really matter much what we're filling them in with, and the
rest of the code is kinda confused, and either does 0 or 1 a bit
randomly. Randonly pick 1.0 for the alpha and 0 for the rest.

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18596>

zink: emulate latc formats with rgtc

util_format_luminance_to_red returns PIPE_FORMAT_NONE for LATC formats,
because there's no red-alpha variant of it, only red-green.

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18596>

docs: add an amber article

This gives a brief introduction to the Amber branch, and where to
find documentation for it.

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18648>

radv/ci: add piglit testing with Zink on NAVI10

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Martin Roukala (né Peres) <martin.roukala@mupuf.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18315>

zink/ci: update list of flakes for Piglit with RADV on NAVI10

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Martin Roukala (né Peres) <martin.roukala@mupuf.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18315>

zink/ci: update list of expected failures with RADV on NAVI10

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Martin Roukala (né Peres) <martin.roukala@mupuf.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18315>

zink/ci: skip GLX tests with RADV

Can't work with GBM.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Martin Roukala (né Peres) <martin.roukala@mupuf.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18315>

zink/ci: skip arb_shader_clock with RADV

They are too random.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Martin Roukala (né Peres) <martin.roukala@mupuf.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18315>

zink/ci: skip ext_external_objects from Piglit with RADV

This subset randomly hangs.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Martin Roukala (né Peres) <martin.roukala@mupuf.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18315>

intel/nir/rt: fixup alignment of memcpy iterations

Not sure if fixes anything because it's always 16 at least, but this
is more correct.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Ivan Briano <ivan.briano@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17396>

intel/fs: fixup a64 messages

And run algebraic when either int64 for float64 are not supported so
those don't end up in the generated code.

Cc: mesa-stable
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Ivan Briano <ivan.briano@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17396>

intel/nir/rt: store ray query state in scratch

Initially I tried to store ray query state in the RT scratch space but
got the offset wrong. In the end putting this in the scratch surface
makes more sense, especially for non RT stages.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes: c78be5da300ae3 ("intel/fs: lower ray query intrinsics")
Reviewed-by: Ivan Briano <ivan.briano@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17396>

intel/nir/rt: change scratch check validation

It's very unfortunate that we have the RT scratch being conflated with
the usual scratch. In our implementation those are 2 different buffers.

The usual scratch access are done through the scratch surface state
(delivered through thread payload), while RT scratch (which outlives
thread dispatch with shader calls) is its own buffer.

So checking the NIR scratch size makes no sense as we can have normal
scratch accesses completely unrelated to RT scratch accesses.

This change switches the validation by looking at whether the scratch
base pointer intrinsic is being used (which is what we use/abuse to
implement RT scratch).

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes: c78be5da300ae3 ("intel/fs: lower ray query intrinsics")
Reviewed-by: Ivan Briano <ivan.briano@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17396>

intel/nir/rt: fix ray query proceed level

Initially the level is world (top level), then it's whatever level the
potential hit is.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes: c78be5da300ae3 ("intel/fs: lower ray query intrinsics")
Reviewed-by: Ivan Briano <ivan.briano@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17396>

intel/nir/rt: remove ray query mem hit writes at initialization

This will not even be read by HW.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Ivan Briano <ivan.briano@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17396>

intel/nir/rt: spill/fill the entire ray query data

We need the traversal stack to saved/restored along with mem hits.
Total spill/fill is 256bytes.

We can potentially optimize this but we have to be very careful about
what state the query is in.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes: c78be5da300ae3 ("intel/fs: lower ray query intrinsics")
Reviewed-by: Ivan Briano <ivan.briano@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17396>

intel/nir/rt: fixup generate hit

This function copies the potential hit from its memory location to the
committed hit location. A couple of fields got their bit offset wrong.

Fixes some CTS tests in dEQP-VK.ray_query.*

v2: Copy primitive/instance leaf pointers

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes: 046571479028 ("intel/nir/rt: add more helpers for ray queries")
Reviewed-by: Ivan Briano <ivan.briano@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17396>

meson: Improve detection of qsort_r().

Instead of trying to guess the interface style by solely checking
the OS name, check if a fake test program can be built and linked.
This will give more accurate result when FreeBSD and other systems
moved to the interface based on GNU qsort_r().

Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18527>

r600: Make NIR the default shader IR

By default sb is disabled, to enable it one can
set
  R600_DEBUG=nirsb

TGSI can be enabled by using
  R600_DEBUG=use_tgsi

In this case "sb" will be enabled, but one can use
the TGSI backend without sb by setting

  R600_DEBUG=use_tgsi,sb

Closes: #6858

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18753>

r600: Add slot to ALU disassambly

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18753>

lavapipe: use slow clear path if conditional rendering.

the ->clear path checks render condition which isn't necessary.

Fixes dEQP-VK.conditional_rendering.draw.condition_host_memory_expect_noop_rp_clear.draw*

Reviewed-by: Brian Paul <brianp@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18686>

radv: rework and rename radv_make_buffer_descriptor()

Pass the VA directly and rename to radv_make_texel_buffer_descriptor().
This helper will be re-used for future work.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18750>

radv: remove unnecessary radv_buffer_view::vk_format

It's not really useful.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18750>

radv: pass a VkSampler to write_sampler_descriptor()

For future work, and remove useless radv_device parameter.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18750>

broadcom/compiler: don't apply robust buffer access to shared variables

This feature is only concerned with buffers bound through a descriptor
set. We are still keeping the code for this (disabled by default) since
it may be useful for debugging some scenarios.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18744>

broadcom/compiler: handle shared stores with robust buffer access

For some reason we supported all shared intrinsics but this one.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18744>

broadcom/compiler: fix robust buffer access

Our implemention was bogus, it was only putting a cap on the offset
based on the aligned buffer size and this doesn't ensure the access
to the buffer happens within its valid range.

I think the only reason we have been passing the tests is that we
align all buffers sizes to 256B and the tests create buffers with a
size that is smaller than that (like 64B). When get the size of the
buffer from the shader, we get the actual bound range (so 64B in this
case) and by capping to that we don't ensure the access will stay
within that range, but we ensure it will stay within the underlying
memory bound to the buffer (256B), and this is fine by the spec,
however, I think if the actual buffer range was the same as the
underlying allocation we would fail the tests.

A valid behavior for robust buffer access on an out-of-bounds access
is to return any valid bytes within the buffer, so we can just
make that offset 0.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18744>

vbo/dlist: keep buffers used in loopback_vertex_list() mapped.

When display list loopback path is hit the major performance
drop during glCallList() happens due to constantly mapping and
unmapping the buffer.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17285>

gallium/u_threaded: add an option for unsychronized create_fence_fd

And enable it for radeonsi. This may be needed by Android.

Reviewed-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18731>

zink: handle culldistance xfb outputs like clipdistance

fixes:
spec@ext_transform_feedback@builtin-varyings gl_culldistance

cc: mesa-stable

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18724>

glsl: fix gl_CullDistance xfb linking

this matches the behavior for gl_ClipDistance

ref #7152

Fixes: 0c8492cd3b2 ("glsl: fix location for array subscript")

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18724>

spirv: Make a helper function to check glslang/shaderc

When deciding on workarounds that apply to GLSLang we should also check
for ShaderC (when using glslang).  Add a helper to do that and reuse in
the three places we check for it.

This patch will now consider shaderc for wa_glslang_cs_barrier workaround.

Suggested by Timur Kristóf.

Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18442>

spirv: Call vtn_emit_ret_store() only when handling return branch

No need to call it at the end of every block, since we already
know and identified the blocks terminating with returns.

Suggested by Jason Ekstrand.

Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18442>

spirv: Add workaround to ignore OpReturn after OpEmitMeshTasksEXT

Fixes: 7d1bcf1f55b ("spirv, nir: Handle EmitMeshTasksEXT opcode.")
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18442>

spirv: Treat EmitMeshTasksEXT as a terminating instruction.

glslang had a bug and didn't actually emit it as a terminating
instruction, therefore this went unnoticed thus far.  See
https://github.com/KhronosGroup/glslang/issues/3020.

Modified by Caio Oliveira to emit the corresponding NIR intrinsic
right before the halt.

Fixes: 7d1bcf1f55b ("spirv, nir: Handle EmitMeshTasksEXT opcode.")
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18442>

spirv: Pass vtn_block into vtn_emit_branch()

This will be used later to handle branches instructions that have
parameters in them, e.g. SpvOpEmitMeshTasksEXT.  No behavior change in
this commit.

Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18442>

nir/lower_task_shader: Don't fail adding a launch when last instruction is a jump

Fixes: 8aff8d3dd42 ("nir: Add common task shader lowering to make the backend's job easier.")
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18442>

gallivm: LLVM opaque pointers: use lp_build_array_get_ptr2() in lp_bld_nir_soa.c

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18721>

gallivm: LLVM opaque pointers: add lp_build_array_get[_ptr]2 in lp_bld_struct.c

i.e. variants taking an explicit LLVM type

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18721>

gallivm: LLVM opaque pointer fixes in lp_bld_sample.c

load_mip()

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18721>

gallium/draw: fix LLVM opaque pointer warnings in store_aos()

Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18720>

gallium/draw: fix LLVM opaque pointer warnings in draw_tcs_llvm_emit_store_output()

Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18720>

gallium/draw: fix LLVM opaque pointer warnings in draw_tcs_llvm_emit_fetch_output()

Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18720>

gallium/draw: fix LLVM opaque pointer warnings in draw_tcs_llvm_emit_fetch_input()

Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18720>

gallium/draw: fix LLVM opaque pointer warnings in draw_gs_llvm_epilogue()

Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18720>

gallium/draw: fix LLVM opaque pointer warnings in draw_gs_llvm_end_primitive()

Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18720>

gallium/draw: fix LLVM opaque pointer warnings in draw_gs_llvm_fetch_input()

Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18720>

gallium/draw: fix LLVM opaque pointer warnings in generate_clipmask()

Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18720>

gallium/draw: fix LLVM opaque pointer warnings in store_clip()

Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18720>

anv: add missing wokraround for texture cache invalidate

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18743>

ir3/analyze_ubo_ranges: Account for reserved consts

We weren't accounting for the reserved consts when calculating how much
we can upload. This led to assertion failures later if we pushed too
much.

Fixes: d3b7681df28 ("tu: ir3: Emit push constants directly")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18757>

tu: Set textures_used for input attachments correctly

We were accidentally multiplying by 2 twice. Noticed by inspection.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18757>

tu: Fix maxPerStageDescriptorUpdateAfterBindInputAttachments

We need this to be the same as maxPerStageDescriptorInputAttachments.

Fixes: d9fcf5de55a ("turnip: Enable nonuniform descriptor indexing")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18757>

d3d12: Fix texture array DPB encode path

Mark D3D12_VIDEO_ENCODER_PICTURE_CONTROL_FLAG_USED_AS_REFERENCE_PICTURE when applicable

Reviewed-by: Giancarlo Devich <gdevich@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18763>

d3d12: Add debug verbosity for output reconstructed picture

Reviewed-by: Giancarlo Devich <gdevich@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18763>

d3d12: Fix out of bounds index check for texture array DPB

Reviewed-by: Giancarlo Devich <gdevich@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18763>

d3d12: Relax slice caps requirements for video encode

The check in d3d12_has_video_encode_support was returning no encode support unless
the driver supported D3D12_VIDEO_ENCODER_FRAME_SUBREGION_LAYOUT_MODE_UNIFORM_PARTITIONING_SUBREGIONS_PER_FRAME
when it should also return support when the driver reports only full frame support
(ie. D3D12_VIDEO_ENCODER_FRAME_SUBREGION_LAYOUT_MODE_FULL_FRAME)

Reviewed-by: Giancarlo Devich <gdevich@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18763>

zink: delete some descriptor caching cruft

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18664>

zink: delete more sw conditional render handling

I forgot to delete this when I deleted the draw path handling

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18664>

zink: remove batch usage from sampler states

this was useless

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18664>

zink: remove batch tracking/usage from view types

instead of incurring all the overhead of tracking lifetimes for these,
it makes more sense to just let them delete whenever and then store
the vk object onto its parent image/buffer to be deleted when that gets
freed, as the parent object's lifetime will always be >= the view's lifetime

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18664>

zink: split out bind_sampler_states hook for drivers without EXT_non_seamless_cube_map

this incurs some overhead and is seldom used, so avoid having it in the codepath
at all if it won't be used

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18664>

zink bump batch state limit a bit

this should yield better throughput

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18664>

zink: rework batch tracking for resources

this uses a revised version of radeonsi's buffer list array to track
resources for lifetime/usage management, massively improving performance
by avoiding hash table lookups

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18664>

zink: defer more semaphore destruction

these have noticeable overhead, so handle them in the submit thread

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18664>

zink: rework flush_frontbuffer to always flush

previously this would no-op unacquired swapchains and then try to
execute a copy from the last-presented swapchain image to the current
image to avoid presenting garbage data

instead, just present garbage like the app requested

fixes #7165

Tested-By: Matti Hamalainen <ccr@tnsp.org>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18730>

zink: check format compatibility up front when creating framebuffer surface

this avoids an assert in create_ivci while enforcing correct behavior

cc: mesa-stable

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18446>

egl: Rewrite eglSetBlobCacheFuncsANDROID to use the standard macros

As the deleted comment hints, you just need the macro to expand to

    return /* void */;

for things to work.

Reviewed-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18713>

egl: Factor out the eglTerminate write-lock pattern

Reviewed-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18713>

egl: Untangle some spaghetti around _eglLookupDisplay

Only eglapi needs any of this.

Reviewed-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18713>

egl: Promote _eglLock/UnlockDisplay to internal API

This reads a bit more cleanly than egl_lock/egl_unlock I think.

Reviewed-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18713>

egl: Fix a bunch of maybe-uninitialized warnings

Fedora's gcc 12 is apparently not smart enough to deduce that these are
always initialized, so set some clear default values.

Reviewed-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18713>

Update 22.2.0.rst

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18726>

docs: Add calendar entries for 22.2 release.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18726>