intel/tools: drop wrappers around mmio regs macros

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16664>

intel/tools: clean up mmio regs definitions

Each unit has the same regs at the same offsets.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16664>

intel/tools: add support for GEM_CREATE_EXT in intel_dump_gpu

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16664>

radv: Replace nir_vector_extract with nir_channel

Signed-off-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16650>

radv: Add and use radv_bvh_node_type

We use magic numbers for node types in many
places of the RADV code. This commit adds an enum
for the node types and uses it to replace magic
numbers in the acceleration structure build code.

Signed-off-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16650>

lima: enable PIPE_CAP_BUFFER_MAP_PERSISTENT_COHERENT

It looks like we don't need to do anything special to enable it.

Piglit spec@arb_buffer_storage@bufferstorage-persistent skip: 16 =>
skip: 12, pass: 4. The rest require GL_ARB_shader_image_load_store

Reviewed-by: Erico Nunes <nunes.erico@gmail.com>
Signed-off-by: Vasily Khoruzhick <anarsoul@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16660>

lima: Implement texture_barrier

Improves x11perf -copywinwin100 from ~1520/sec to ~3700/sec

Reviewed-by: Qiang Yu <yuq825@gmail.com>
Reviewed-by: Erico Nunes <nunes.erico@gmail.com>
Signed-off-by: Vasily Khoruzhick <anarsoul@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16658>

ci/lava: Fix Gitlab Section markers

LAVA is mangling the escape codes from ANSI during log fetching from the
target device, making the gitlab section markers from deqp, for example,
to not work, inputting noise into the log.

This commit makes the simplest fix which is to replace the mangled
characters to the fixed ones.

This approach is error-prone, since it may unwittingly replace a genuine
log that resembles the mangled escape code. But this solution should
suffice until we get a proper fix from LAVA team itself.

Signed-off-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16520>

ci/lava: Fix colored LAVA outputs

LAVA is mangling the escape codes from ANSI during log fetching from the
target device, making the colored lines from deqp, for example, to not
work, inputting noise into the log.

This commit makes the most straightforward fix which is to replace the
mangled characters to the fixed ones.

This approach is error-prone since it may unwittingly replace a genuine
log that resembles the mangled escape code. But this solution should
suffice until we get a proper fix from LAVA developers itself.

Fixes: #5503

Signed-off-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16520>

nir: Fixes MSVC compiling error about unused variable `_`

Fixes: 836470d4331 ("nir: allow NIR_PASS(_, )")

MSVC compiling error:
```
nir/nir_lower_shader_calls.c
../mesa/src/compiler/nir/nir_lower_shader_calls.c(1188): error C2220: the following warning is treated as an error
../mesa/src/compiler/nir/nir_lower_shader_calls.c(1188): warning C4101: '_': unreferenced local variable
../mesa/src/compiler/nir/nir_lower_shader_calls.c(1190): warning C4101: '_': unreferenced local variable
```

Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16666>

nir/algebraic: Add two more pack/unpack rules

Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16591>

d3d12: add missing cmath includes

If we want to use the std:: version of the math functions, we need to
include <cmath>.

Fixes: b171a6baa21 ("d3d12: Add video encode implementation of pipe_video_codec")
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16636>

aco: only add/subtract low bits of program addresses

fossil-db (Sienna Cichlid):
Totals from 4007 (2.47% of 162293) affected shaders:
Instrs: 3733239 -> 3728018 (-0.14%)
CodeSize: 20770340 -> 20749456 (-0.10%)
Latency: 46883958 -> 46872764 (-0.02%); split: -0.02%, +0.00%
InvThroughput: 10550392 -> 10548698 (-0.02%); split: -0.02%, +0.00%

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16460>

aco/tests: add test for p_constaddr with a non-zero offset

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16460>

aco: fix p_constaddr with a non-zero offset

Seems this broke a while ago and we never noticed.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Fixes: 0af7ff49fde ("aco: lower p_constaddr into separate instructions earlier")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16460>

radv: introduce radv_graphics_pipeline_info and split existing info structs

I should have introduced radv_graphics_pipeline_info earlier. More
states will be added later.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16643>

lima/ppir: merge branch nodes in simple cases

In some simple cases, we can merge the branch operation with an existing
instruction and avoid the creation of an empty new instruction just for
the branch node.

Signed-off-by: Erico Nunes <nunes.erico@gmail.com>
Reviewed-by: Vasily Khoruzhick <anarsoul@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16163>

lima/ppir: optimize branch comparison when possible

The branch operation supports comparisons, so it may be possible to
merge a previous comparison operation with the branch operation.
There are several restrictions to do it at this stage, but it may save
instructions in many cases.

Signed-off-by: Erico Nunes <nunes.erico@gmail.com>
Reviewed-by: Vasily Khoruzhick <anarsoul@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16163>

docs/ci: fix literalinclude of caching setup files

Fix the syntax to show the files in the generated sphinx, otherwise
they just don't show at all.

Signed-off-by: Erico Nunes <nunes.erico@gmail.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15599>

pvr: use UINT64_C for 64-bit constant

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/6528

Signed-off-by: Jan Palus <jpalus@fastmail.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16630>

dzn: Use the correct ICD path on Win32

Similar to commit c6cddd2e17e ("lavapipe: Use the correct ICD path on
Win32"), but applied to dozen.

Suggested-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16634>

spirv: Fix windows build

Looks like MSVC doesn't like VLAs:

src/compiler/spirv/spirv_to_nir.c(3879): error C2057: expected constant expression
src/compiler/spirv/spirv_to_nir.c(3879): error C2466: cannot allocate an array of constant size 0
src/compiler/spirv/spirv_to_nir.c(3879): error C2133: 'srcs': unknown size

so let's use a static array size.

Fixes: 87d74311985 ("spirv: Use nir_vec_scalars() to simplify matrix transpose.")
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16632>

dzn: Use the vk_sync_binary abstraction

D3D12 fences are capable of handling binary operations, but the
current dzn_sync implementation doesn't match vk_sync expectations
when sync objects are used to back semaphores. In that case, the wait
operation is supposed to set the sync object back to an unsignaled
state after the wait succeeded, but there's no way of knowing what
the sync object is used for, and this implicit-reset behavior is not
expected on fence objects, which also use the sync primitive.
That means we currently have a semaphore implementation that works
only once, and, as soon as the semaphore object has been signaled it
stays in a signaled state until it's destroyed.

We could extend the sync framework to pass an
implicit-reset-after-wait flag, but, given no one else seems to
need that, it's probably simpler to drop the binary sync
capability and rely on the binary-on-top-of-timeline emulation provided
by the core.

Fixes: a012b219640e ("microsoft: Initial vulkan-on-12 driver")
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16629>

radv: split radv_pipeline into radv_{graphics,compute,library}pipeline

Instead of using a union in radv_pipeline, this introduces new
structures for graphics, compute and library pipelines which inherit
from radv_pipeline. This will ease graphics pipeline libary implem.

There is still no radv_raytracing_pipeline because RADV actually
uses a compute pipeline for everything but it could be introduced
later when necessary.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16603>

radv: cleanup using device/physical_device during pipeline creation

This removes a TON of pipeline->device occurences which were a mess.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16603>

radv: move gs_table_depth to the physical_device

This is not a logical device property.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16603>

radv: remove unused radv_get_conservative_raster_mode() helper

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16603>

radv: remove couple of useless pCreateInfo pointers in the emit path

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16603>

radv: rename radv_pipeline_generate_XXX() to radv_pipeline_emit_XXX()

Shorter and clearer.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16603>

radv: constify radv_pipeline in more radv_pipeline_generate_XXX() helpers

These functions should only emit registers and not change the pipeline.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16603>

virgl/ci: Update virglrenderer

Update virglrenderer to the latest version on time.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16627>

docs/asahi: Add initial driver docs

Emma asked for some docs on how to use wrap for when wrap gets landed.
This should do :-)

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Add wrap library

Add a library that wraps the key IOKit entrypoints used in the macOS
UABI for AGX. Our wrapped routines print information about the kernel
calls made and dump work submitted to the GPU using agxdecode. This code
has two major use cases:

1. Debugging Mesa, particularly around the undocumented macOS
   user-kernel interface. Logs from Mesa may compared to Metal to check
   that the UABI is being used correcrly.

2. Reverse-engineering the hardware, using this as glue to get at the
   "interesting" GPU memory.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Guard against encoder overflows

We don't seem to hit this yet, but it's possible so add an assert to
rule it out for failing tests.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Allocate a larger segment list

This ought to be dynamically allocated but because this is macOS kernel
UABI only gunk, let's just hack around it...

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Handle RECT textures as 2D

Rectangle textures are just 2D textures with unnormalized coordinates,
but we already handle unnormalized coordinates in the sampler state. So
we just need to alias RECT and 2D.

Fixes GALLIUM_HUD.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Fix pipeline decoding

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Improve render target dumping

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Sync some more unknown fields

I don't understand these structs well. Good news, though-- Asahi Lina
confirmed this is all software-defined crap in the macOS UABI.
Unfortunately, they seem to correspond to physical registers.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Drop some unknowns

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Identify suballocated size

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Identify suballocation mechanism

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Decode resource allocation requests

We already know much of this structure, let's pretty print to simplify
our traces.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Decode resource allocation responses

We already know much of this structure, let's pretty print to simplify
our traces.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Align segment list unknowns with Metal

No observable change.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Compute segment list length

Quiesces agxdecode error introduced in previous commit.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Validate segment list length

This is easy now!

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Split unknown field in segment list header

Seen as 0x8000.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Add segment list pretty printer

Validate all the new expectations and print all the fields. This should
make differences between the drivers obvious, I hope.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Sync some names with Project Zero

This should clarify a few things I didn't get independently
investigating the interface. Of coruse, I got other pieces... the sum of
the parts is better :-)

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Identify partial render pipeline

Needed if a partial render is incurred from overflowing the parameter
buffer (too much geometry).

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Implement polygon offset

This is pretty simple now that the hardware is understood. The hardware
interfaces parallels that of scissors, so the scissor path is reused
with minor modifications to accommodate the new functionality.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Add a depth bias structure

This is simple and corresponds directly to the Metal inputs. However,
the alignment is a bit tricky, so let's add formal XML for it.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Identify depth bias enable bit

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Identify "set depth bias index" field

Grouped together with the "set scissor index" field, which is natural.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Identify depth bias array

"Inspired" by PowerVR code, by comparison.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Note unknown field

...used internally with visibility tests, together with a weird
vertex/compute-like shader used to zero the visibility buffer.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Relax Draw "command" check

Other values in the lower byte seen with multiple draws and
visibility testing.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Add XML used for visibility tests

Aka occlusion queries. There is an annoying limitation in the hardware
(reflected in Metal) that only a single buffer may be bound per render
pass, with the per-draw settings merely specifying an offset.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Dump "unknown 4a" structure

A few fields are known at least, let's decode them correctly instead of
falling back to a hex dump.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Identify more depth pipeline fields

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Account for every byte of the cmdbuf

This adds the remaining XML. I don't know how much of this is correct,
but it nominally accounts for every byte. So there shouldn't be more
surprises in the command buffer after this.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Handle 3rd deflake address

Context switching is so broken. Just trying to get closer. Adding some
XML here to see if we're missing something else obvious.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Flush on partial clears

Not the fastest thing, but should be conformant. (Panfrost to this day
has this effective behaviour...)

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Pass depth/stencil clear values to firmware

These need to be format-packed, but as we only support Z32F_S8 right
now, that's trivial.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Handle uncompressed Z32F depth buffers

This uses a subset of the depth/stencil infrastructure we built out to
support writing back tiled, uncompressed Z32F depth buffers to memory.
Texturing from this format is already supported.

This gets glmark2 -bshadow working.

v2: Fix partial renders

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Don't store to unbound render targets

Otherwise we have a state leak with depth-only attachments, as in
glmark2 -bshadow.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Validate depth/stencil formats

Don't go advertising formats we don't actually support.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Use z24_in_z32f lowering

Thanks Dave!

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

asahi: Run lower_blend for all fragment shaders

Unify the paths so that colour masks are respected even when regular
blending is off (or when logic ops are used!) Fixes

dEQP-GLES2.functional.color_clear.masked_rgb

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16512>

turnip: fix assert for max xfb outputs

this is a counter, not an index, so use <=

cc: mesa-stable

Reviewed-by: Hyunjun Ko <zzoon@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16620>

turnip: disable UBWC for SNORM formats

In copy_format, we treat snorm as unorm to avoid clamping.  But snorm
and unorm are UBWC incompatible for special values such as all 0's or
all 1's.  Disable UBWC for snorm.

For reference, I dumped the first byte of an UBWC blocks and it was

  color      UNORM  SNORM
  all black  0x01   0x31
  all white  0x0d   0x11

@flto clarified that bit 4 is unset for fast clear encoded blocks.  It
looks like fast clear is not used for SNORM.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/6480
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16534>

egl: dereference XCB drawable pointers

eglCreatePlatformWindowSurface[EXT] and
eglCreatePlatformPixmapSurface[EXT] should be passed (xcb_window_t *)
and (xcb_pixmap_t *), so we must dereference these types before using
them as drawables.  We already do something similar with X11 drawable
pointers.

Signed-off-by: Jeffrey Knockel <jeff@jeffreyknockel.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16269>

clc: disable opaque pointers until they are supported

Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16479>

clc: fix compiler features_macro CTS Test

Even with that alone we can't pass the test, as LLVM enables some
extensions based on the SPIR target we choose.

Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16479>

clc: speed up compilation by not relying on opencl-c.h

This depends on LLVM change: https://reviews.llvm.org/D125401

Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16479>

clc: drop parsingComplete check

This relies too much on the properties of the SPIRV-LLVM-Translator and is
required to load SPIR-Vs found in the OpenCL CTS.

Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16479>

clc: parse localSize and localSizeHint

Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16479>

turnip: fix tu6_pack_border_color for z24

The value should be at the bottom 24 bits, not at the top.

dEQP-VK.pipeline.sampler.* still passes.  This fixes most of
dEQP-GLES31.functional.texture_border_clamp.formats.*depth* on angle.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16570>

panfrost: Move patched_s out of the pan_blitter_views struct

The struct is returned from a function, so in debug builds the address
may change after returning, and pointers to patched_s will be broken.

Pass the pointer to the patched stencil view as a parameter to
pan_preload_get_views to avoid this.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16343>

panfrost: Increase the limit for blend shader variants

Qt uses blend constants to set text colour, this will allow more
colours onscreen before thrashing happens.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16343>

panfrost: Copy blend constant into variant even when reusing it

Otherwise future lookups will match searches for the old constant.

Fixes: bbff09b9521 ("panfrost: Move the blend shader cache at the device level")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/6355
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16343>

zink: use dynamic rendering (most of the time)

this is a simplified version of the renderpass infrastructure which
tracks rendering info on the context and updates it incrementally to
try and reduce cpu overhead

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16476>

zink: split out scissor flagging from framebuffer updating

this is kinda similar but not actually the same thing

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16476>

zink: lift some renderpass mechanics up a level in the api

(should be) no functional changes

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16476>

zink: track whether there's any transient attachments bound

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16476>

zink: break out zs rt attrib analysis into util function

(should be) no functional changes

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16476>

zink: break out color rt attrib analysis into util function

(should be) no functional changes

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16476>

zink: fix indentation

no functional changes

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16476>

zink: pass rt attrib to zink_render_pass_attachment_get_barrier_info()

no functional changes

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16476>

zink: remove renderpass state flag for swapchain

this is redundant and can be merged onto the rt attrib

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16476>

zink: break up swapchain renderpass update to util function

no functional changes

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16476>

zink: break out sample location updating to separate function

no functional changes

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16476>

zink: use local variable for getting rt barrier info

no functional changes

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16476>

zink: remove fb rebind check during renderpass begin

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16476>

zink: use global image rebind counter for dmabuf export

this is a bit less flimsy and handles more than just framebuffer rebinds

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16476>

zink: move renderpass handling to zink_render_pass.c

no functional changes

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16476>

zink: use zink_batch_no_rp to end renderpasses

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16476>

zink: always use zink_batch_rp to init renderpass

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16476>

zink: rename a fb surface variables

no functional changes, just being more consistent

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16476>