ac/gpu_info: use hw_ip::ip_discovery_version to set IP versions

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17411>

ac/gpu_info: rework and extend device info to be more informative

This is the result with AMD_DEBUG=info:

Device info:
    name = NAVI23
    marketing_name = AMD Radeon RX 6600
    num_se = 2
    num_rb = 8
    num_cu = 28
    max_gpu_freq = 2750 MHz
    max_gflops = 9856 GFLOPS
    l0_cache_size = 16 KB
    l1_cache_size = 128 KB
    l2_cache_size = 2048 KB
    l3_cache_size = 32 MB
    memory_channels = 8 (TCC blocks)
    memory_size = 8 GB (8192 MB)
    memory_freq = 14 GHz
    memory_bus_width = 128 bits
    memory_bandwidth = 224 GB/s
    clock_crystal_freq = 100000 KHz
    IP GFX     10.3  queues:1
    IP COMP    10.3  queues:4
    IP SDMA     5.2  queues:2
    IP VCN_DEC  3.0  queues:1
    IP VCN_ENC  3.0  queues:1
    IP VCN_JPG  3.0  queues:1

It might not be 100% correct with other chips.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17411>

ac/gpu_info: remove tabs

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17411>

ac/gpu_info: rename info fields to num_cu, memory_bus_width, memory_freq_mhz

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17411>

llvmpipe: keep context list and use to track resource usage.

Just check across all contexts if a resource is referenced.

Fixes: 6bbbe15a783a ("Reinstate: llvmpipe: allow vertex processing and fragment processing in parallel")
Acked-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17702>

llvmpipe: make last_fence a screen/rast object not a context one.

When a flush happens the per-context setup is used to hold the fence
for the last scene sent to the rasterizer. However when multiple
contexts are in use, this fence won't get returned to be blocked on.

Instead move the last fence to the rasterizer object, and return
that instead as it should be valid across contexts.

Fixes gtk4 bugs on llvmpipe since overlapping vertex/fragment.

Fixes: 6bbbe15a783a ("Reinstate: llvmpipe: allow vertex processing and fragment processing in parallel")
Acked-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17702>

intel/dev: Enable remaining DG2 and ATS-M device IDs

Mostly Matt Roper's kernel patch commit message:

The device IDs here are associated with DG2 add-in cards. We need to
wait for some additional functionality (e.g., small BAR recovery) to
land before we're ready to upstream these.

Ref: https://patchwork.freedesktop.org/patch/483381/?series=103098&rev=1
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12706>

v3dv: only check binning sync for semaphores for the first CL job

Semaphore waits in a command buffer only affect the first jobs we execute
in each hardware queue since jobs in the same queue are serialized against
each other. Binning syncs in particular, only affect CL jobs.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17594>

v3d,v3dv: add a common v3d_hw_prim_type helper

We had this replicated in both drivers.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17594>

v3dv: remove unnecesary FIXME comment

We only want to cleant BCL barrier flags if we consume a BCL barrier.
For example, if the client records a barrier for an index buffer
it should apply to the next draw call that uses an index buffer
which may not be the current draw call but one coming after it.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17594>

zink: enable unordered blit/resolve ops

can't do swapchain or u_blitter ops, but these should be fine

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17667>

zink: make get_cmdbuf() public

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17667>

zink: expand unordered_exec

track read+write

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17667>

zink: move buffer/image functions down in file

no functional changes

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17667>

zink: promote image/buffer copy operations to unordered

if the src and dst for these operations can both be promoted, then the
entire operation can be promoted to potentially avoid splitting renderpasses

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17667>

zink: add a second resource param to get_cmdbuf() for unordered promotion

no functional changes

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17667>

zink: rework unordered_exec handling

previously this would opportunistically promote barriers to the unordered
cmdbuf only if a renderpass was active or there was no access, which
was the wrong approach

instead, opportunistically promote barriers to the unordered cmdbuf
any time it's possible to do so, which is when one of these conditions is true:
* when there is no access to the resource on the current cmdbuf
* when the only access to the resource is in the unordered cmdbuf

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17667>

zink: explicitly unset unordered_exec when resource can't be unordered

ensure that subsequent commands can't be promoted

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17667>

zink: zink_resource::unordered_barrier -> unordered_exec

this is a bit more clear

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17667>

zink: add ZINK_DEBUG=noreorder

this will be useful for debugging

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17667>

zink: add doc for ZINK_DEBUG=compact

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17667>

zink: convert ZINK_DEBUG to enum

more readable/debuggable

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17667>

zink: stop ORing unordered barrier access

this doesn't make any sense and is confusing

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17667>

intel/dev: Determine the amount of free vram using small BAR uapi

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16739>

intel/dev: Use i915 region probed_cpu_visible_size when non-zero

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16739>

iris/bufmgr: Add I915_GEM_CREATE_EXT_FLAG_NEEDS_CPU_ACCESS for vram mappable buffers

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16739>

anv: make use of the new smallbar uAPI

Instead of having 2 VkMemoryType pointing to the same VkMemoryHeap, we
have each VkMemoryType with VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT (one
host visible, the other not) point to its own VkMemoryHeap. For the
local heap that is host visible, we'll use the
I915_GEM_CREATE_EXT_FLAG_NEEDS_CPU_ACCESS flag at GEM BO creation.

When the smallbar uAPI is not available we fallback to a single heap
and do not use I915_GEM_CREATE_EXT_FLAG_NEEDS_CPU_ACCESS.

v2: Handle probed_cpu_visible_size == probed_size (Matthew)

v3:
 * Jordan: Use region info from devinfo

v4: Also make the vram host visible heap as local (Ken)

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16739>

drm-uapi/i915_drm.h: Update from drm-next (2022-07-22)

git://anongit.freedesktop.org/drm/drm 417c1c1963549e9a48b83ada59d90258e38c6594

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16739>

lavapipe: fix shader access stage conversion from mesa -> pipe

ok ok but this time it's fixed and works I promise

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17758>

r600/sb: fall back to unoptimized shader if RA fails

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Filip Gawin <filip@gawin.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17755>

radv: expose disjoint image support

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16510>

radv: implement disjoint memory for multiplane images

For descriptor binding, we need to allow up to three buffers to be
referenced by any image.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16510>

radv: consolidate descriptor set buffer counts

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16510>

radv: refactor image binding into struct

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16510>

util/blitter: respect PIPE_TEXTURE_RECT

if this isn't supported, don't use rect-related sampling

cc: mesa-stable

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17714>

ci/virgl: Make the VM-based virgl tests report to #virgl-ci, too.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17730>

gallivm: Optimize single-invocation SSBO stores.

The CTS does a lot of 1x1x1 compute shaders (all that stuff like
dEQP-GLES31.functional.shaders.builtin_functions.precision.mul.highp_compute.scalar)
which finish with store_ssbos.  Instead of doing the invocation loop in
that case (which LLVM has to later unroll), just emit the single
invocation's store.

Fixes timeouts running
dEQP-GLES31.functional.ssbo.layout.random.all_shared_buffer.36, which does
a spectacular number of SSBO stores in a long 1x1x1 compute shader.
Reduces runtime of on llvmpipe from 66s to 29s locally, and virgl from
1:38 to 43s.  virgl
dEQP-GLES31.functional.ssbo.layout.random.nested_structs_arrays_instance_arrays.22
goes down to 7 seconds.

Fixes: #6797
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17730>

gallivm: Refactor a bit of UBO/SSBO range checking.

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17730>

mesa: Mark render-to-texture as unsafe if there's no pipe_resource

Let's be slightly more defensive here.  If a texture image doesn't have
an associated pipe_resource allocated, then render_texture() will pass
that along to _mesa_update_renderbuffer_surface(), which will crash on a
NULL pointer dereference.  So, if there isn't a pipe_resource, then we
should just skip this altogteher.

Today, this isn't an issue, because each gl_texture_image always
allocates a pipe_resource up front.  On a branch of mine, I prototyped
some improvements to the compressed texture fallback handling, where it
would defer resource allocation, examine the source image's block data,
and dynamically select a format based on that, then allocate it later.

With that prototype in place, we saw crashes the Android "My Talking
Tom" series of games, which appear to be attaching ASTC textures to a
framebuffer color attachment.  That FBO would be incomplete anyway, as
ASTC textures aren't renderable, but we got into a situation where the
render-to-texture code was crashing due to the lack of pt before it
could properly signal that it was incomplete and bailing.

Technically, we don't need this now, but I figure that being defensive
won't hurt and this would probably save whoever encounters such an issue
in the future a bunch of frustrating debugging.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17508>

tu: Do not dereference descriptorSetLayout in push descriptors tmpl

Fixes crash when capturing with RenderDoc.

From VK spec:

 descriptorSetLayout [...] This parameter is ignored if templateType
 is not VK_DESCRIPTOR_UPDATE_TEMPLATE_TYPE_DESCRIPTOR_SET.

Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17751>

vulkan: Include self-dep info in rendering continues

Fixes: 3a204d5cf370 ("vulkan/render_pass: Add a better helper for render pass inheritance")
Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17748>

vulkan: Append subpass structures to VkRenderingInfo last

If we don't append subpass->self_dep_info last, other __vk_append_struct()
calls will update its pNext chain which lives in the subpass which
should be treated as immutable.  This is easily fixable by just making
it the last thing we append to the chain.

Fixes: 7e11cdc77a6a ("vulkan/render_pass: Pass sample locations to barriers")
Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17748>

intel/fs: Use canonical form for "work around" tags

Trivial.  Also clean up some weird whitespace.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17605>

intel/fs: Eliminate "masked" and "per slot offset" URB messages

All of this information can be inferred from the sources.

v2: Fix "error: unused variable 'opcode'" detected by marge-bot.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17605>

intel/fs: Don't pass flags to lower_urb_read_logical_send or lower_urb_write_logical_send

...because the flags can be inferred from the sources.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17605>

intel/fs: Make logical URB read instructions more like other logical instructions

No shader-db changes on any Intel platform

Fossil-db results:

Tiger Lake
Instructions in all programs: 156926440 -> 156926470 (+0.0%)
Instructions hurt: 15

Cycles in all programs: 7513099349 -> 7513099402 (+0.0%)
Cycles hurt: 15

Ice Lake and Skylake had similar results. (Ice Lake shown)
Cycles in all programs: 9099036492 -> 9099036489 (-0.0%)
Cycles helped: 1

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17605>

intel/fs: Make logical URB write instructions more like other logical instructions

The changes to fs_visitor::validate() helped track down a place where I
initially forgot to convert a message to the new sources layout.  This
had caused a different validation failure in
dEQP-GLES31.functional.tessellation.tesscoord.triangles_equal_spacing,
but this were not detected until after SENDs were lowered.

Tiger Lake, Ice Lake, and Skylake had similar results. (Ice Lake shown)
total instructions in shared programs: 19951145 -> 19951133 (<.01%)
instructions in affected programs: 2429 -> 2417 (-0.49%)
helped: 8 / HURT: 0

total cycles in shared programs: 858904152 -> 858862331 (<.01%)
cycles in affected programs: 5702652 -> 5660831 (-0.73%)
helped: 2138 / HURT: 1255

Broadwell
total cycles in shared programs: 904869459 -> 904835501 (<.01%)
cycles in affected programs: 7686744 -> 7652786 (-0.44%)
helped: 2861 / HURT: 2050

Tiger Lake, Ice Lake, and Skylake had similar results. (Ice Lake shown)
Instructions in all programs: 141442369 -> 141442032 (-0.0%)
Instructions helped: 337

Cycles in all programs: 9099270231 -> 9099036492 (-0.0%)
Cycles helped: 40661
Cycles hurt: 28606

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17605>

aco: Add a note about G16 without A16 to README-ISA.md.

Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16979>

radv: Fold 16bit tex sources.

Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16979>

aco: Support 16bit sources for texture ops.

Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16979>

aco: Add G16 opcodes.

Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16979>

tu: Fix resolving d32s8 into s8 on fast path

The code assumed that if the source was d32s8 then the destination would
also be d32s8, in particular that depth_base_addr/stencil_base_addr
would also be filled out. Move the destination and source handling into
two different ifs with different conditions.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17684>

tu: Support resolving D24S8 -> S8

This was missed when we added support for VK_KHR_depth_stencil_resolve.
There is a similar feature where the stencil aspect of a D24S8 can be
copied "tightly" in CopyImageToBuffer, but it used the texture swizzle
and so required the 3d path. To get it to work with the 2D path, which
is required for resolves, we have to instead use the A8_UNORM format,
which works for texture sampling even for tiled images. We also have to
reuse the pre-existing image views because subpass resolves work on
image views rather than images, whereas before the fixup was applied
while creating the image view. This means threading through the
corresponding "opposite" format through setup, src, and dst functions,
doing the fixup there (through some shared helpers), and then getting
every user to specify the right format. As a bonus, we no longer need to
force the 3d path for the CopyImageToBuffer and CopyBufferToImage
special cases.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17684>

llvmpipe: remove unused function

Fixes: a1846e39269 ("llvmpipe: merge all rasterizer-bind functions")
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17735>

ir3: fix tess param allocation

primitive_param takes up 2 vec4's.  Remove an align that I don't
understand.

The align upset

  Test case 'dEQP-VK.subgroups.ballot_broadcast.graphics.subgroupbroadcast_vec4'..
  deqp-vk: ../src/freedreno/ir3/ir3_nir.c:1039:
  void ir3_setup_const_state(nir_shader *, struct ir3_shader_variant *, struct ir3_const_state *):
  Assertion `constoff <= ir3_max_const(v)' failed.

with an older version (android11-tests-dev branch) of deqp-vk.  This is
because ir3_nir_opt_preamble uses the function for the worst case but
the function fails to replace the align by the worst case.

No regression with dEQP-VK.*tess*.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17570>

ir3: increment constoff right after it is assigned

Minor improvement to readability.  No real change.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17570>

ir3: remove unused patch_vertices_in

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17570>

ir3: fix output_loc size

It was off-by-one.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17570>

ir3: copy req_local_mem for MESA_SHADER_KERNEL

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17570>

ir3: update ir3_const_state comment

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17570>

lavapipe: Use the vk_graphics_pipeline_state

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17669>

lavapipe: Use shader_info::uses_sample_shading

This is more accurate because it's computed directly in spirv_to_nir and
takes even unused SampleID and SamplePos builtings into account.

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17669>

lavapipe: advertise VK_EXT_multisampled_render_to_single_sampled

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17705>

lavapipe: VK_EXT_multisampled_render_to_single_sampled

this works by doing a replicate at renderpass start and an in-place
resolve on renderpass end

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17705>

vulkan: hook up VK_EXT_multisampled_render_to_single_sampled

this adds support for the extension to common renderpass

Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17705>

llvmpipe: add a shortcut for blitter resolves

if this is only going to do a sample0 resolve, the functionality is
equivalent to just copying the first sample, and in llvmpipe terms,
this just means doing a direct copy at offset=0

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17705>

lavapipe: fix stencil clears

this uses the value from the stencil attachment

Fixes: 8068c68b1ff ("lavapipe: Delete render passes")

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17705>

llvmpipe: allow resource_copy_region to perform replicate operations

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17705>

llvmpipe: remove weird clamping for format support with sample counts

this breaks u_blitter

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17705>

lavapipe: set image access on pipeline bind, not descriptor bind

this is where the access originates, so set it here

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17739>

lavapipe: use 64bit masks for shader access

ensure that all descriptors are tracked for this

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17739>

lavapipe: use correct offset calculation for flagging ssbo/image write access

var->data.binding has no relation to the actual index of the resource,
which must be separately calculated based on the layout

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17739>

lavapipe: make binding/layout helper functions static inline

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17739>

mesa: Remove Purgeable fields from texture/buffer/renderbuffer objects

These are unused since !17425.

Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17745>

gallium/hud: do not use texture-rect for font

RECT textures used to be required to be supported by drivers. But since
the state-tracker learned how to lower these to 2D textures, some
drivers no longer support them.

While we have lowering in place for this, lowering it involves some
needless overhead. So let's just use a 2D texture instead of a RECT
texture.

Because having two versions and switching between them is more
complicated than it needs to be, let's just always use a 2D texture.
Similarly, let's just always multiply the reciprocal here, so we don't
have to test for PIPE_CAP_TGSI_DIV first.

Cc: mesa-stable
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17707>

lavapipe: move vertex input dynamic states into vertex input block

these states were erroneously assigned to the pre-rasterization stage
for pipeline libraries when they instead belong to the vertex input stage

cc: mesa-stable

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17738>

zink: add a ci flake

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17722>

lavapipe: remove some skips

these should all be fast enough now

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17722>

lavapipe: implement extreme uniform inlining

this adds a mangled variation on nir_inline_uniforms that enables inlining
from any uniform buffer in order to try inlining every possible load

if the shader is too small or the ssa_alloc delta from inlining is too small,
then inlining is disabled for that shader to avoid pointlessly churning
the same shaders for no gain

with certain types of shaders, the speedup is astronomical

before:
dEQP-VK.graphicsfuzz.cov-int-initialize-from-multiple-large-arrays (4750.76s)

after:
dEQP-VK.graphicsfuzz.cov-int-initialize-from-multiple-large-arrays (0.505s)

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17722>

ci/freedreno: 3 pixel change in Raven restricted trace

Acked-by: Rob Clark <robdclark@chromium.org>
Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17732>

wsi/x11: Synchronously check for error from xcb_present_pixmap

Yes this is a round trip, but X_PresentPixmap is not itself a blocking
operation, it just instructs the server to do the next presentation at
some time. More importantly, if _we_ don't catch the presentation error,
xlib's error queue will, and the calling code is certainly not prepared
to handle errors from Present.

Forcing the round trip here is also a bit more correct semantically.
This is the end of the Vulkan client part of the present queue, and the
X_PresentPixmap request transfers the queue operation to the server, so
we should not return until we are sure the handoff has happened.

Fixes some flakiness with piglit@glx-visuals-* with zink+radv.

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17685>

zink: fix EXT_color_write_enable check

Checking for the extension isn't enough, we also need to check for the
feature-bit.

Fixes: 49a20e0981c ("zink: start a unified driver workarounds struct")
Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17709>

zink: warn about missing image2DViewOf3D feature

While we're at it, let's store zink_screen in a temp varaible for reuse.

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17709>

docs/zink: document required VK_EXT_border_color_swizzle

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17709>

docs/zink: document required VK_KHR_swapchain_mutable_format

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17709>

docs/zink: correct name of extension

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17709>

radeonsi/vcn/enc: enable cabac entropy mode

Set cabac entropy mode if enabled.

v2: add extra check on radeon driver side, disable cabac if profile is
baseline or extended.

Signed-off-by: Boyuan Zhang <boyuan.zhang@amd.com>
Acked-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Ruijing Dong <ruijing.dong@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16113>

frontends/va: get entropy mode info for h264 encode

Get entropy mode and cabac init idc from VAAPI interface.

Signed-off-by: Boyuan Zhang <boyuan.zhang@amd.com>
Acked-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Ruijing Dong <ruijing.dong@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16113>

vl: add cabac init idc for h264 encode

Add interface to store cabac init idc value.

Signed-off-by: Boyuan Zhang <boyuan.zhang@amd.com>
Acked-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Ruijing Dong <ruijing.dong@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16113>

pvr: Add support for input attachment descriptor.

Signed-off-by: Rajnesh Kanwal <rajnesh.kanwal@imgtec.com>
Reviewed-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17719>

pvr: Add support for storage and uniform texel descriptor.

Signed-off-by: Rajnesh Kanwal <rajnesh.kanwal@imgtec.com>
Reviewed-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17719>

pvr: Add support for image descriptor.

Signed-off-by: Rajnesh Kanwal <rajnesh.kanwal@imgtec.com>
Reviewed-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17719>

pvr: Add support for combined image and sampler descriptor.

Signed-off-by: Rajnesh Kanwal <rajnesh.kanwal@imgtec.com>
Reviewed-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17719>

pvr: Add support for sampler descriptor.

Signed-off-by: Rajnesh Kanwal <rajnesh.kanwal@imgtec.com>
Reviewed-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17719>

pvr: Move binding related checks in common code.

Signed-off-by: Rajnesh Kanwal <rajnesh.kanwal@imgtec.com>
Reviewed-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17719>

pvr: Implement vkCreateBufferView and vkDestroyBufferView.

Signed-off-by: Rajnesh Kanwal <rajnesh.kanwal@imgtec.com>
Reviewed-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17719>

pvr: Use common vk_buffer implementation.

Signed-off-by: Rajnesh Kanwal <rajnesh.kanwal@imgtec.com>
Reviewed-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17719>

pvr: Init and store default sampler in pvr_device.

Signed-off-by: Rajnesh Kanwal <rajnesh.kanwal@imgtec.com>
Reviewed-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17719>

llvmpipe: clean up bresenham implementation

This moves some common code into shared locations, limits the scope of
some variables, switches some booleans for bools, and cleans up some
whitespace.

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17653>

llvmpipe: consolidate legacy_points and multisample state

This just makes the code a bit easier to follow.

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17653>