v3dv: add ASTC formats to get_compatible_tlb_format

CTS doesn't seem to hit this, but they're all 128bit formats so this
should be right

Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11336>

v3dv: divide by block size in copy_image_blit

This handles compressed formats with non-4x4 blocks, like ASTC.

Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11336>

v3dv: add the unswizzled RGBA4444 format

If we're supporting the R/B swapped one we might as well support the one that
isn't.

Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11336>

r600/sfn: Clean up some ALU lowering and move code

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11348>

r600/sfn: Don't read return values of atomic ops that are not used

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11348>

r600/sfn: Drop method for emit_atomic_add, it is handled in generic code

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11348>

r600/sfn: don't read back unused image atomic result values

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11348>

r600/sfn: don't designates initializers, since they are c++20

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11348>

iris: Avoid abort() if kernel can't allocate memory

When the system doesn't have enough memory, GNOME Shell may be crashed
by iris:
gnome-shell[1161]: iris: Failed to submit batchbuffer: Cannot allocate memory
gnome-shell[1161]: GNOME Shell crashed with signal 6

So don't abort() when kernel can't allocate memory to avoid crashing the
entire desktop.

Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11178>

v3dv: expose VK_KHR_shader_non_semantic_info

This is entirely implemented in the SPIR-V frontend.

Relevant CTS tests:
dEQP-VK.spirv_assembly.instruction.compute.non_semantic_info.*

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11440>

disk_cache: use UTIL_QUEUE_INIT_SCALE_THREADS

Instead of spawning 4 threads when the cache is created,
spawn 1 and let u_queue grow the number of threads if
needed.

I wrote this patch because when running piglit's quick_shader
profile I had lots of samples in disk cache threads - mostly
in native_queued_spin_lock_slowpath kernel function.

Since these tests shouldn't really stress the cache, I assumed
it was caused only by thread creations.

After writing the patch and redoing the measurement, I got an
improvement but I still more hits in the same function for
shader_runner:$disk0 thread so something was wrong.

After digging more, I found out that my shader cache index was
corrupted: the on-disk size was 29MB but the index reported it
was way more than 1GB. So each disk cache thread was spending
a lot of time trying to evict files. Given that my cache had
a really low count of files, the LRU method based on randomly
generating subfolder names failed, so evicting was very slow.

Now that my cache index is fixed, the disk cache threads are
mostly idle but I still think it makes sense to grow the
number of threads instead of spawning 4 at the program start.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11296>

util/u_queue: add UTIL_QUEUE_INIT_SCALE_THREADS flag

This flag allow to create a single thread initially, but set
max_thread to the request thread count.

If the queue is full and num_threads is lower than max_threads,
we spawn a new thread to help process the queue faster.

This avoid creating N threads at queue creation time.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11296>

util/u_queue: move function definition up

Will be used by the next commit.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11296>

radeonsi: delay sample_pos_buffer creation until first use

And use pipe_buffer_create_with_data instead of doing it
manually.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11296>

vbo: delay vbo_exec_vtx_map call

Instead of doing vbo_exec_vtx_map during initialization,
defer it until the first actual user.

v2: move init to vbo_exec_wrap_upgrade_vertex (Emma Anholt)

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11296>

nvc0/ir: Initialize CodeEmitterNVC0 member progType in constructor.

Fix defect reported by Coverity Scan.

Uninitialized scalar field (UNINIT_CTOR)
uninit_member: Non-static class member progType is not initialized
in this constructor nor in any functions that it calls.

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11350>

mesa: execute glFlush asynchronously if no image has been imported/exported

This improves viewperf performance and it shouldn't break synchronization
with external clients when it's indirectly implied by glFlush.

This should not break the cases described in:
    https://gitlab.freedesktop.org/mesa/mesa/-/issues/4903

Acked-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11341>

mesa: move _mesa_notifySwapBuffers into the x11 swrast driver

It has no other use and no relevance to DRI drivers despite the name.

Acked-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11341>

mesa: add gallium flush_flags param into ctx->Driver.Flush

Acked-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11341>

st/mesa: move the st_flush_bitmap_cache call into st_flush

Acked-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11341>

st/mesa: fix an incorrect comment in st_context_flush

Acked-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11341>

glthread: change when glFlush flushes asynchronously

This fixes the flushing with external textures.

We don't know if we need to flush synchronously with multiple contexts,
so I removed that.

Acked-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11341>

anv: Fix dynamic primitive topology for tess on Gfx7.x too

Commit 24342e499bc58348b257716f629dccca3c1b0833 changed how primitive
topology is handled on Gfx8+ but missed updating the Gfx7.x code.

As a result, tests which previously used topologies like PATCHLIST_3
instead started using bogus ones like LINESTRIP_ADJ.  This caused a
GPU hangs in a bunch of Vulkan conformance tests involving tessellation.

This fixes those hangs.

Fixes: 24342e499bc ("anv: fix dynamic primitive topology for tess")
Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11434>

Add default driver selections for RISC-V

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11346>

zink: always defer image descriptor barriers

this is simpler and guaranteed to be accurate

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11393>

zink: use fake buffer barriers for descriptors

GL requires explicit glMemoryBarrier calls for shader synchronization
and only calls that map/copy buffers get implicit sync, so we don't actually
need barriers for any of these cases, only the state needs to be updated

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11393>

zink: check actual mem props to determine if resource object is coherent

this was correct for what it has been used for until now, but it will no
longer be correct going forward

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11397>

zink: key alloc cache on heap index, not heap flags

this is a bit more sane

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11397>

zink: avoid caching visible vram allocations

the visible vram heap is potentially going to be limited in size, so avoid
caching these allocations since that locks them to a given allocation size

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11397>

zink: change a bunch of sparse buffer resource checks to host-visible checks

(sparse buffer) is a subset of !host-visible, and !host-visible is actually the
more correct check to be using

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11397>

zink: collapse host_visible and non-coherent alignment alloc cases

* buffers can use normal mem prop checking to determine host_visible setting
* sparse buffers are never coherent, so this case can be dropped from the conditional

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11397>

compiler/glsl: Use mutex lock while freeing up mem_ctx

builtin_builder::~builtin_builder() and builtin_builder::release()
are running into race condition. This leads lightsmark to crash at
the end because both calls ralloc_free which mutates the arguments state

This patch fixes lightsmark2008 crash

Fixes: e4da8b9c331cc3a ("mesa/compiler: rework tear down of builtin/types")

Reviewed-by: Charmaine Lee <charmainel@vmware.com>
Reviewed-by: Neha Bhende <bhenden@vmware.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Tested-by: Neha Bhende <bhenden@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11385>

zink: mark some functions inline

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11429>

ir3/ra: Fix array parallelcopy confusion

With array registers, there are two num's we care about:

1. The base num that the whole array starts at (->array.base)
2. The num that the instruction uses, plus possibly an indirect offset
   (->num or ->array.offset)

For parallel copies we always copy the whole array, so (2) is irrelevant
here. For phis and parallel copies inserted for phis, we used
assign_reg() which assigned ->array.base, but we forgot about this when
constructing our own parallel copies for live range splitting, just
setting ->num instead. The parallel copy lowering was also inconsistent
here, using ra_reg_get_num() (which looks at ->array.base for arrays)
for sources but looking at ->num directly for destinations. This makes
everything use ->array.base consistently.

While we're here, make sure to remove IR3_REG_SSA from liveout copies to
make sure printing works correctly.

Fixes: 0ffcb19 ("ir3: Rewrite register allocation")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11422>

ir3: Improve printing of array parallelcopies/phis

Normally something with IR3_REG_ARRAY doesn't have a register assigned,
but we keep IR3_REG_ARRAY for parallel copies after RA because we need
to know the appropriate size. We want to see the register assigned for
these when printing the RA result before parallel copies are lowered.
The register is in ->array.base in this case, so initialize it to
INVALID_REG and print ->array.base if it's been assigned to something,
similar to ->num in the normal case.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11422>

shader_enums: change VERT_BIT back to the 32-bit shift

This reverts 0e2566a8. The warning is fixed differently.

Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10993>

gallium/pb: change alignment to 32 bits

This partially reverts 4a3f0444. The warning is fixed differently.

Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10993>

radeonsi: remove -Wstrict-overflow=0 since it doesn't seem to be needed

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11384>

radeonsi: remove the chip_class dimension from the draw_vbo array

We don't use/initialize draw_vbo callbacks for other generations anymore.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11384>

radeonsi: compile si_state_draw.cpp for each gfx generation separately

It makes compilating faster.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11384>

radeonsi: move a few functions from si_state_draw.cpp into si_gfx_cs.c

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11384>

venus: remove workarounds for multi-planar format interop

Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Reviewed-by: Chia-I Wu <olvaffe@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11282>

anv: fix build errors after commit 8b7ff78

Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Reviewed-by: Roman Stratiienko <r.stratiienko@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11373>

radv: fix build errors after commit 8b7ff784

Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Reviewed-by: Roman Stratiienko <r.stratiienko@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11373>

zink: ci updates

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11391>

zink: support more RGBX formats

the base formats are supported, so these should fall into place naturally

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11391>

zink: also nope out of any dst alpha blends for rgbx formats

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11391>

zink: smash dstAlphaBlendFactor to ZERO for RGBX attachments

this is a no-op

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11391>

zink: add a more direct check for rgbx formats in create_sampler_view hook

really the point of this is to clamp void channels for any permutation of rgbx
where all channels are the same (e.g., both rgbx8 and rgbx16), so the previous
helper isn't inclusive enough

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11391>

util/queue: add a global data pointer for the queue object

this better enables object-specific (e.g., context) queues where the owner
of the queue will always be needed and various pointers will be passed in
for tasks

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11312>

radeonsi: set desc[3] of all buffer descriptors at context creation

instead of at bind time. When we unbind, we shouldn't memset the last
element.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11340>

radeonsi: remove no-op unref in si_set_constant_buffer

buffer is already NULL here

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11340>

radeonsi: restructure si_set_sampler_views for faster unbinding trailing slots

si_set_sampler_view updates only one slot. This rewrites it to update
multiple slots.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11340>

radeonsi: don't clear register fields in si_set_mutable_tex_desc_fields

They are always initialized to 0.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11340>

radeonsi: use the restrict keyword to set sampler view descriptors faster

All places must set restrict for the same pointer.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11340>

radeonsi: check is_buffer once instead of 4 times in si_set_sampler_view_desc

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11340>

zink: do not check buffer-format for usage-bits

Buffers are created without a format in Vulkan, and we always pass in
R8_UNORM for them in Gallium. It's the view-formats we should have
checked, if anything.

But that's orthogonal to this. We shoudn't keep checking R8_UNORM
capabilities for buffers, all it's going to do is trigger asserts.

Fixes: 00dc0036b ("zink: flatten out buffer creation usage flags codepath")
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11416>

zink: drop repeated usage-bit

We already set this bit unconditionally right before, no point in
repeating it.

Fixes: 00dc0036b ("zink: flatten out buffer creation usage flags codepath")
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11416>

zink: remove unused function

There's no call-sites to this function, so let's drop it.

Fixes: e4e20556d62 ("zink: switch to memory barriers instead of actual buffer barriers")
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11421>

zink: fixup signedness of subtraction

I'm not even going to pretend that I grok this code, but since we take
the abs value, it's pretty obvious that we meant to use a signed value
here. So let's cast the two operands to int before we subtract.

This was noticed by the following clang warning:

---8<---
../src/gallium/drivers/zink/zink_context.c:3284:14: warning: taking the
absolute value of unsigned type 'unsigned int' has no effect
[-Wabsolute-value]
      last = abs(reads - writes) > UINT32_MAX / 2 ? MIN2(reads, writes) : MAX2(reads, writes);
             ^
---8<---

Fixes: 0c1fe392e8d ("zink: implement a tc is_resource_busy hook")
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11421>

zink: move queue init to screen creation

this is a race condition

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11398>

zink: set subdata hook as PIPE_MAP_ONCE

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11401>

zink: update pipe_screen::num_contexts

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11394>

anv: fix dynamic primitive topology for tess

this needs to use the pre-converted topology using tess state patch control points

Fixes: f6fa4a80000 ("anv: add support for dynamic primitive topology change")

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11332>

zink: correct type of flags to flush

This type is unsigned in the prototype, so this produces a warning on
MSVC.

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11380>

zink: use alloca instead of hard-to-size vlas

These variable-length arrays are hard to size statically, and VLAs
aren't supported by MSVC. So let's use alloca instead.

Reviewed-by: Hoe Hao Cheng <haochengho12907@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11380>

zink: use max-descriptor define

This avoids compile errors on MSVC due to VLAs.

Reviewed-by: Hoe Hao Cheng <haochengho12907@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11380>

zink: introduce a define for max descriptors per type

We know what this max is in the compiler, let's move that out into
zink_descriptors.h, so we can reuse the constant.

Reviewed-by: Hoe Hao Cheng <haochengho12907@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11380>

zink: fix more initializer styles

Empty initializer lists are a GCC extension, let's use the syntax that
compilers like MSVC also supports.

Reviewed-by: Hoe Hao Cheng <haochengho12907@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11380>

zink: drop some more vla usage

MSVC doesn't like them, so let's get rid of more of them.

Reviewed-by: Hoe Hao Cheng <haochengho12907@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11380>

zink: add missing compiler-dependency

Without this, we depend on something else previously in the build to
have built these.

Fixes: ed2fb809 ("zink: introduce vk_dispatch_table")
Reviewed-by: Hoe Hao Cheng <haochengho12907@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11380>

zink: drop paranoid code

If we ever get here, res->dt is non-NULL. No need to assert, and no need
to check twice.

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11414>

zink: do not unmap dt-buffers twice

Seems I missed that we already did an unconditional unmap here, and
forgot to remove it. Whoops.

Fixes: 5159f406d ("zink: use gallium api to copy to display-target")
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11414>

zink: split stencil ref changes to separate dirty flag

the values here are for the cmdbuf, not the pipeline, so they should
always be updated regardless of what the current dsa state uses

Reviewed-by: Hoe Hao Cheng <haochengho12907@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11396>

zink: add update flag for dsa state change

reduce overhead by avoiding unnecessary updates

Reviewed-by: Hoe Hao Cheng <haochengho12907@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11396>

zink: add update flag for rasterizer state change

this can be used to avoid updating related dynamic states too frequently

Reviewed-by: Hoe Hao Cheng <haochengho12907@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11396>

zink: handle nir_op_pack_64_2x32

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11392>

v3dv: remove sRGB blending workaround

This reverts commits 4c15131b1d36874b643aed8407933ccd8675841d and
1cf36797bf9d63b33c405d85ce720dcd5e75a47c.

Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11377>

v3dv: clamp srgb render targets

Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11377>

Revert "v3dv: allow creating uncompressed views from compressed images and vice versa"

This reverts commit b32a48c7e26bffdc696d3cceef0305e2c4731c96.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11415>

v3dv: don't support VK_IMAGE_CREATE_BLOCK_TEXEL_VIEW_COMPATIBLE_BIT

The hardware doesn't support this naturally and we need to do a lot
of nasty stuff in the driver to almost make it work.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11415>

nv50/ir: fix surface lowering when values get shared accross operations

With nir I encountered the case where the same value can be written to from
multiple surface operations. This caused some weird messups with the unions
as the def.rewrite operations caused unrelated instructions to get new their
value replaced as well.

In order to replace def.rewrite, we have to create a new temp value, write
to that one instead and move to the original value.

Fixes: 869e32593a9 ("gm107/ir: fix loading z offset for layered 3d image bindings")
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11053>

st/mesa: always report the max samples as supported

This is required by GL. Doing this even when the backend driver does not
support it leads to creating attachments which are not renderable. This
is not ideal, but does pass tests. This covers scenarios such as missing
MSAA for 128-bit formats on Sandybridge, and missing 8x MSAA on 128-bit
formats on NVIDIA Tesla generation boards.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11372>

zink: stop sanitizing primitive_restart flag in draw info

this is a 1bit struct member, sanitizing is pointless

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11408>

zink: move batch decl to top of draw_vbo

this may still change during descriptor updating, but having the pointer
earlier is helpful

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11408>

zink: remove unnecessary draw checks

I don't know how/why these got here but they aren't needed

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11408>

zink: use u_live_shader_cache

let the duplicated shader madness end

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11403>

zink: unify gfx shader create callbacks

these are identical

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11403>

zink: mark bind_stage() as inline

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11403>

zink: simplify zink_program_has_descriptors()

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11407>

zink: remove return types from program update functions during draw

also remove some null checks; these add overhead and the app is going to
explode anyway if they fail

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11406>

zink: slightly refactor program updating during draw

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11406>

zink: destroy lazy descriptor pools during batch reset when unused

this should keep resource usage a bit lower

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11404>

zink: track active use counts for descriptor layouts

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11404>

zink: cache descriptor update templates along with layout

this simplifies a bunch of code and just makes more sense given that the
lifetimes should match

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11404>

crocus: disable Z16

This is disabled on i965 for all pre-gen8, so do the same here.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11409>

crocus: fixup render aux usage function.

This got misaligned somewhere in development, should fix the
regressions vs i965 in tesseract as seen on phoronix benchmarks.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11409>

zink: make batch_usage_matches take a batch state param

no functional changes, just some unwinding in some cases

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11399>

zink: make batch_usage_set take a batch state param

no functional changes

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11399>

zink: remove unnecessary conditionals in resource batch tracking

this is no longer an atomic op

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11399>