nouveau/nir: Enable nir_opt_move/sink.

NIR load_consts/inputs tend to happen together at the top of the program.
In the TGSI backend the loads got emitted at use time, while the NIR
backend was emitting the loads at load intrinsic time.  By sinking the
intrinsics, we can greatly reduce register pressure.

nv92 NIR results:

total local in shared programs: 2024 -> 2020 (-0.20%)
local in affected programs: 4 -> 0
total gpr in shared programs: 790424 -> 735455 (-6.95%)
gpr in affected programs: 215968 -> 160999 (-25.45%)
total instructions in shared programs: 6058339 -> 6051208 (-0.12%)
instructions in affected programs: 410795 -> 403664 (-1.74%)
total bytes in shared programs: 41820104 -> 41660304 (-0.38%)
bytes in affected programs: 7147296 -> 6987496 (-2.24%)

Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15542>

pvr: Use vk_common_GetDeviceQueue API.

Removes pvr_GetDeviceQueue implementation. As we are now
using the common vk_queue structure as a base for pvr_queue,
we can use vk_common_GetDeviceQueue implementation instead.

Signed-off-by: Rajnesh Kanwal <rajnesh.kanwal@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15574>

dzn: Compile-test the driver

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Acked-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14766>

microsoft: Initial vulkan-on-12 driver

This is Dozen, the Vulkan on DirectX 12 driver. Not to be confused with
DirectEggs.

This is an early prototype, and not meant to be upstreamed as-is.

Co-Authored-by: Boris Brezillon <boris.brezillon@collabora.com>
Co-Authored-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Co-Authored-by: Louis-Francis Ratté-Boulianne <lfrb@collabora.com>
Co-Authored-by: Jesse Natalie <jenatali@microsoft.com>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Acked-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14766>

microsoft/spirv_to_dxil: Add missing ralloc_free

Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Acked-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14766>

vulkan/util: Make STACK_ARRAY() work for arrays of pointers

And add an explicit cast on the pointer returned by malloc() to
make C++ happy.

Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14766>

vulkan/image: Make MSVC C++ compiler happy

Fix 'error C4576: a parenthesized type followed by an initializer
list is a non-standard explicit type conversion syntax' errors by
declaring an actual variable and returning it in
vk_image_view_subresource_range().

All those MSVC/c++ related-constraints are quite annoying to be honest,
but it looks like the D3D12 headers have been updated to plain C
recently, which will allow us to write the driver in C, and hopefully
get all this sort of issues behind us.

Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14766>

zink: add anv cts skips from waiver

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15578>

zink: update radv fails

this should be the exact current baseline

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15576>

pvr: Implement vkCreateSampler and vkDestroySampler APIs.

Implements vkCreateSampler and vkDestroySampler APIs.

Also fixes maxSamplerLodBias value from 15.0f to 16.0f
as it's the max supported by our hardware.

Also changing maxSamplerAnisotropy from 16.0f to 1.0f to
temporarily disable anisotropy as we are missing software
support for it.

Signed-off-by: Rajnesh Kanwal <rajnesh.kanwal@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15557>

vulkan/util: Get rid of VK_OUTARRAY_MAKE()

Get rid of VK_OUTARRAY_MAKE() so people don't get tempted to
use it and produce code that doesn't compile with MSVC, which
doesn't support typeof().

Suggested-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15522>

vulkan/wsi: Stop using VK_OUTARRAY_MAKE()

We're trying to replace VK_OUTARRAY_MAKE() by VK_OUTARRAY_MAKE_TYPED()
so people don't get tempted to use it and make things incompatible with
MSVC (which doesn't support typeof()).

Suggested-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15522>

vulkan/device_select: Stop using VK_OUTARRAY_MAKE()

We're trying to replace VK_OUTARRAY_MAKE() by VK_OUTARRAY_MAKE_TYPED()
so people don't get tempted to use it and make things incompatible with
MSVC (which doesn't support typeof()).

Suggested-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15522>

venus: Stop using VK_OUTARRAY_MAKE()

We're trying to replace VK_OUTARRAY_MAKE() by VK_OUTARRAY_MAKE_TYPED()
so people don't get tempted to use it and make things incompatible with
MSVC (which doesn't support typeof()).

Suggested-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Chia-I Wu <olvaffe@gmail.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15522>

panvk: Stop using VK_OUTARRAY_MAKE()

We're trying to replace VK_OUTARRAY_MAKE() by VK_OUTARRAY_MAKE_TYPED()
so people don't get tempted to use it and make things incompatible with
MSVC (which doesn't support typeof()).

Suggested-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15522>

anv: Stop using VK_OUTARRAY_MAKE()

We're trying to replace VK_OUTARRAY_MAKE() by VK_OUTARRAY_MAKE_TYPED()
so people don't get tempted to use it and make things incompatible with
MSVC (which doesn't support typeof()).

Suggested-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15522>

pvr: Stop using VK_OUTARRAY_MAKE()

We're trying to replace VK_OUTARRAY_MAKE() by VK_OUTARRAY_MAKE_TYPED()
so people don't get tempted to use it and make things incompatible with
MSVC (which doesn't support typeof()).

Suggested-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15522>

turnip: Stop using VK_OUTARRAY_MAKE()

We're trying to replace VK_OUTARRAY_MAKE() by VK_OUTARRAY_MAKE_TYPED()
so people don't get tempted to use it and make things incompatible with
MSVC (which doesn't support typeof()).

Suggested-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15522>

v3dv: Stop using VK_OUTARRAY_MAKE()

We're trying to replace VK_OUTARRAY_MAKE() by VK_OUTARRAY_MAKE_TYPED()
so people don't get tempted to use it and make things incompatible with
MSVC (which doesn't support typeof()).

Suggested-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15522>

intel/fs: Fix IsHelperInvocation for the case no discard/demote are used

Use emit_predicate_on_sample_mask() helper that does check where to
get the correct mask depending on whether discard/demote was used or
not.

Fixes: 45f5db5a84a ("intel/fs: Implement "demote to helper invocation"")
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15400>

intel/fs: Initialize the sample mask in flags register when using demote

Without this change, a check for "is helper invocation" could read
uninitialized values.

Fixes: 45f5db5a84a ("intel/fs: Implement "demote to helper invocation"")
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15400>

nir: Use const for nir_shader_get_entrypoint(..)

nir_shader_get_entrypoint(..) should not modify the passed nir_shader
object. Enforce this by marking shader paramenter as const.

Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Reviewed-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Mihai Preda <mhpreda@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15362>

mesa/st: add special casing for pointsize constant updating during validate

the previous method of using affected_states to trigger constant updates
was ineffectual in the scenario where a ubo pointsize was needed on
the first time a non-precompiled shader was used after being the not-last
vertex stage:

* have vs+gs -> gs precompiles with pointsize lowering -> gs constants get updated
* remove gs -> vs was precompiled without pointsize lowering -> vs constants broken

now just do a quick check as in st_atom_shader.c and set the flag manually to
ensure the update is done correctly every time

cc: mesa-stable

fixes #6207

fixes (radv):
KHR-GL46.texture_cube_map_array.image_op_fragment_sh
KHR-GL46.texture_cube_map_array.sampling
KHR-GL46.texture_cube_map_array.texture_size_fragment_sh
KHR-GL46.constant_expressions*

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15570>

zink: update anv icl ci list

praise be to dynamic state fixes!

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15569>

zink: flush clears before toggling color write

ensure these sync up onto the expected buffers

Fixes: 3892c133811 ("zink: add an alternate path for EXT_color_write_enable usage")

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15568>

zink: fix up color_write_enable workaround

this needs to only swizzle to dummy surfaces if it's the workaround,
not just if color_write_enable is active

Fixes: 3892c133811 ("zink: add an alternate path for EXT_color_write_enable usage")

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15568>

freedreno/ci: Update a306 expectations

These have started to flakey UnexpectedPass somewhere along the way.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14900>

pipe-loader: Try loading freedreno for virtgpu device

Freedreno will check if the virtgpu supports the pass-thru context, and
if not will bail, falling back to virgl.

TODO this requires that virgl is also enabled in the mesa build, even if
it is not needed.. maybe there is a better way to handle this?

Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14900>

freedreno/drm: Add virtio backend

Add a new backend to enable using native driver in a VM guest, via a new
virtgpu context type which (indirectly) makes host kernel interface
available in guest and handles the details of mapping buffers to guest,
etc.

Note that fence-fd's are currently a bit awkward, in that they get
signaled by the guest kernel driver (drm/virtio) once virglrenderer in
the host has processed the execbuf, not when host kernel has signaled
the submit fence.  For passing buffers to the host (virtio-wl) the egl
context in virglrenderer is used to create a fence on the host side.
But use of out-fence-fd's in guest could have slightly unexpected
results.  For this reason we limit all submitqueues to default priority
(so they cannot be preepmted by host egl context).  AFAICT virgl and
venus have a similar problem, which will eventually be solveable once we
have RESOURCE_CREATE_SYNC.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14900>

freedreno/drm: Reorder device destroy

Call backend specific cleanup fxn earlier.  This is needed if the
backend has things like bo's to delete, otherwise the handle_table
will already be destroyed causing problems in bo_del()

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14900>

freedreno/drm: Extract out "softpin" submit/ringbuffer base class

We are going to want basically the identical thing, other than
flush_submit_list, for virtio backend.  Now that we've moved various
other dependencies into the base classes, extract out an abstract base
class for submit/ringbuffer.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14900>

freedreno/drm: Move ring_pool slab parent to base

Prep to move most of sp submit/ringbuffer to something that can be
re-used by virtio backend.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14900>

freedreno/drm: Move bo idx to base

The virtio backend will want this too, and it will make it easier to
share most of the submit/ringbuffer implementation with the virtio
backend.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14900>

freedreno/drm: Move submit_queue to base

The virtio backend will want this too.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14900>

freedreno/drm: Avoid CPU_PREP ioctl if bo is idle

With userspace fences, if we know definitely that the buffer is idle
(which implies that it is not shared with other processes, etc), then
skip the ioctl.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14900>

freedreno/drm: Add fd_bo_upload()

There are some buffers that we mmap just to write to them a single time.
Add the possibility of the drm backend to provide an alternate upload
path to avoid these mmap's.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14900>

freedreno/drm: Add FD_BO_SHARED hint

With the virtio backend we will need to pass an extra flag when
allocating buffers that will be shared cross-device (such as with
virtio-wl for passing between host and guest)

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14900>

freedreno/drm: Add FD_BO_NOMAP hint

Add a hint for buffers that we won't need to mmap.  With the virtio
backend, virglrenderer needs to create a dmabuf fd for mapping into
the host, which we want to avoid when possible.

Low hanging fruit is to use this hint for anything tiled/ubwc.  There
are probably more bo's that can be flagged as such.

TODO add fd_bo_upload() for memcpy to bo.. this would be useful for
uploads, for example, shaders which we just write once and never touch
again.. for virtio this could be implemented with a TRANSFER_TO_HOST
ioctl.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14900>

freedreno/drm: Rework bo creation path

Decoupling handle and fd_bo creation simplifies things for "normal" drm
drivers, avoiding duplication for the create vs import paths.  But this
is awkward for the virtio backend when wants to do multiple things in
the same guest<->host round trip.

So instead, split the paths in the interface backend and move the code
sharing for the two different paths into the msm backend itself.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14900>

freedreno/drm: Add fd_device_open() helper

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14900>

freedreno/drm: Split msm backend into subdir

Let's keep things a bit better organized when we add a new backend.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14900>

zink: update radv ci

bunch of these I just removed by mistake

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15567>

venus: update protocol to remove redundant decoders

Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15537>

venus: Increase the base sleep of vn_relax

Based on profiling, these 10us sleeps are behaving closer to ~60us
and causing higher-than-necessary CPU overhead. 120us seems like a
good balance between latency management and overhead.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15565>

lavapipe: Delete render passes

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15431>

c11: Fixes unused parameter warnings

Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15553>

util: Fixes unused parameter warnings

The compiler warning:
```
../src/mesa/util/u_thread.h: In function 'util_thread_get_time_nano':
../src/mesa/util/u_thread.h:239:34: warning: unused parameter 'thread' [-Wunused-parameter]
  239 | util_thread_get_time_nano(thrd_t thread)
      |                           ~~~~~~~^~~~~~
```

Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15553>

zink: update resource layout in copy_scanout

This transitions the resource to TRANSFER_SRC_OPTIMAL, but
never updates the res->layout field, so subsequent transitions
are wrong and throw validation errors.

UNASSIGNED-CoreValidation-DrawState-InvalidImageLayout(ERROR / SPEC): msgNum: 1303270965 - Validation Error: [ UNASSIGNED-CoreValidation-DrawState-InvalidImageLayout ] Object 0: handle = 0x6660f60, type = VK_OBJECT_TYPE_COMMAND_BUFFER; | MessageID = 0x4dae5635 | vkQueueSubmit(): pSubmits[0].pCommandBuffers[0] command buffer VkCommandBuffer 0x6660f60[] expects VkImage 0x2f000000002f[] (subresource: aspectMask 0x1 array layer 0, mip level 0) to be in layout VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL--instead, current layout is VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL.

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15545>

lavapipe: Use vk_image_subresource_layer/level_count

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15535>

lavapipe: Use vk_image_view

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15535>

vulkan: Add a vk_image_view_subresource_range helper

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15535>

vulkan/runtime: Add vk_cmd_queue.h to idep_vulkan_runtime_headers

If we don't do that, meson might start compiling source files
including vk_command_buffer.h which in turn includes vk_cmd_queue.h
before this file is even generated, and we end up with errors like
that https://gitlab.freedesktop.org/mesa/mesa/-/jobs/20157936#L1119.

Fixes: 6bd8a3c7e434 ("vulkan/runtime: Add a vk_cmd_queue object to vk_command_buffer")
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15546>

venus: add VK_EXT_extended_dynamic_state2 extension

Implements all the necessary code in the device initialization
and extension functions.

Signed-off-by: Igor Torrente <igor.torrente@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15428>

pvr: ci: Initial freedesktop CI integration

This patch adds the PowerVR driver to the following shared builds:

 * debian-arm64
 * debian-clang
 * debian-release
 * debian-vulkan
 * fedora-release

It also adds the associated "tools" to these builds:

 * debian-arm64-build-test
 * debian-release (already uses -Dtools=all)
 * fedora-release

And removes them from these builds by expanding -Dtools=all:

 * debian-gallium

Signed-off-by: Matt Coster <matt.coster@imgtec.com>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15284>

pvr: Gate offline compiler build behind -Dtools=imagination

This matches the behavior of e.g. panfrost's bifrost_compiler target.

Signed-off-by: Matt Coster <matt.coster@imgtec.com>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15284>

pvr: do not use fallthrough for unreachable code

unreachable() doesn't lead to executing the code that follows it,
neither in debug nor release builds. So falling through doesn't make any
sense.

This fixes a compile-error on clang.

Let's move the default-block to the end to make it clearer that there's
no intended fallthrough.

Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15531>

pvr: do not use fallthrough for unreachable code

unreachable() doesn't lead to executing the code that follows it,
neither in debug nor release builds. So falling through doesn't make any
sense.

This fixes a compile-error on clang.

Let's move the default-block to the end to make it clearer that there's
no intended fallthrough.

Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15531>

anv: don't store sample location sample count

This information should match the current pipeline sample count.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15310>

anv: fix dynamic sample locations on Gen7/7.5

3DSTATE_MULTISAMPLE should be baked into the pipeline if not dynamic.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes: 27ee40f4c9d8 ("anv: Add support for sample locations")
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15310>

anv: use local dynamic pointer more

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15310>

anv: fix color write enable interaction with color mask

Color writes & color masks occupy the same fields in the BLEND_STATE
structure. So we need to store color mask (which are not dynamic) on
the pipeline to merge that information with color writes.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes: b15bfe92f7f8 ("anv: implement VK_EXT_color_write_enable")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/6111
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15310>

anv: fix VK_DYNAMIC_STATE_COLOR_WRITE_ENABLE_EXT state

First, there is a problem if you do the following

 vkCmdSetColorWriteEnableEXT(attachmentCount = 8)
 vkCmdBindPipeline(GFX, with attachmentCount = 4)
 vkCmdDraw()
 vkCmdBindPipeline(GFX, with attachmentCount = 8)
 vkCmdDraw()

Because in the dynamic state emission code we rely on the first
pipeline to figure the number of BLEND_STATE entries to prepare. This
is wrong, we should fill all entries so that the dynamic state works
regardless of the number of attachments in the pipeline. With regard
to the dynamic values, we should retain enable/disable values that do
not concern the current pipeline.

Second, 3DSTATE_WM was not always reemitted when the pipeline changed.
But since it is not emitted as part of the pipeline, this results in
inconsistent state being programmed.

Third, we end up disabling the fragment stage completely in some
cases. And that is programming the pipeline inconsistently and
triggering a hang on TGL.

v2: Fix comment (Tapani)

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes: b15bfe92f7f8 ("anv: implement VK_EXT_color_write_enable")
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15310>

anv: fix dynamic state emission

The problem is that we missed looking at pipeline changes. Pipelines
hold bits of dynamic states and when it changes we might need to
reemit a packet.

v2: fix comment (Tapani)
    Add missing anv_cmd_buffer_needs_dynamic_state() use (Tapani)

Cc: mesa-stable
Fixes: 505d176a8ebf ("anv: disable baked in pipeline bits from dynamic emission path")
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15310>

anv: allow baking of 3DSTATE_DEPTH_BOUNDS in pipeline batch

If it's not dynamic.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15310>

docs: add a minimal docs page for radv

RADV is the last driver in Mesa that doesn't have a webpage somewhere
that we can link to from mesa3d.org. So let's make a minimal docs article
for it where we link to relevant information elsewhere.

Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15494>

virgl: Update virgl_protocol and use the provided constants

Allow for better readability of the code.

Signed-off-by: Corentin Noël <corentin.noel@collabora.com>
Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14473>

Revert "ci: Disable windows-vs2019"

This reverts commit 04b80489d575dab7ed2cd3714d60d32e91e26589.

Acked-by: Daniel Stone <daniels@collabora.com>
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Acked-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15518>

lavapipe: Don't use VK_OUTARRAY_MAKE()/vk_outarray_append()

MSVC doesn't support typeof(). If we want to keep compiling radv on
windows we need to use the typed variants of those macros.

Fixes: dc8fdab71efd ("lavapipe: Use VK_OUTARRAY for GetPhysicalDeviceQueueFamilyProperties[2]")
Acked-by: Daniel Stone <daniels@collabora.com>
Acked-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/15518>

radv: Don't use VK_OUTARRAY_MAKE()/vk_outarray_append()

MSVC doesn't support typeof(). If we want to keep compiling radv on
windows we need to use the typed variants of those macros.

Fixes: 10d69d5f0bcc ("radv: fix returning empty drmFormatModifierTilingFeatures")
Acked-by: Daniel Stone <daniels@collabora.com>
Acked-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/15518>

amd: Fix ac_gpu_info.c compilation on windows

Fixes: 75a783ea73e8 ("ac: Query the amdgpu MEC firmware version.")
Acked-by: Daniel Stone <daniels@collabora.com>
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Acked-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15518>

aco: Fix an MSVC warning

'warning C4804: '<<': unsafe use of type 'bool' in operation'

Fixes: 9934c8676132 ("aco: use v_fma_mix to combine mul/add/fma input conversions")
Acked-by: Daniel Stone <daniels@collabora.com>
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Acked-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15518>

virgl/ci: Add support for dEQP GL vtest-ing

Provide the 'virpipe-on-gl' job for running dEQP GL tests on
virpipe/vtest.

The rationale is to help reducing the regressions in Virglrenderer CI
due to sharing the Mesa CI infrastructure.

Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Acked-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15490>

ci/freedreno: Increase console timeout for perf jobs

Piglit is very sparse in its status output and downloads of big traces
can take a while.

Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15527>

ci/iris: Increase console timeout for perf jobs

Piglit is very sparse in its status output and downloads of big traces
can take a while.

Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15527>

iris: don't synchronize BO for batch decoding

We don't need to go to the kernel to synchronize the BO we want to
decode with INTEL_DEBUG=bat, mostly because we'll decode what was
written by the driver in the batch.

This also works around an issue in the simulation environment.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes: 9ac81f189050 ("iris: decoder fixes")
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15517>

i915: Use the sin/cos lowering in nir_opt_algebraic.py

It's nearly identical.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15539>

ci/freedreno: Update checksum for GolfWithYourFriends trace

The MR below changed the rendering slightly and the checksum isn't valid
any more:

"ir3, turnip, freedreno: Shader preambles"

https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13148

Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15526>

zink: lower txp for array textures

this is also illegal according to spirv spec

fixes #6177

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15510>

zink: lower txp for cube and ms textures

this is illegal according to spirv spec

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15510>

nir/lower_tex: add txp lowering option for arrays

this is illegal in vulkan, so zink needs to be able to lower these

Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15510>

gallium: Remove unused macro PIPE_ARCH_SSSE3

After removal inline function `_mm_shuffle_epi8`, the
macro `PIPE_ARCH_SSSE3` have no need anymore, remove it too.

Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14077>

llvmpipe: Revise u_sse.h to remove unused _mm_shuffle_epi8 inline function

error log:
```
[2/43] Compiling C object src/gallium/drivers/llvmpipe/libllvmpipe.a.p/lp_setup_tri.c.obj
FAILED: src/gallium/drivers/llvmpipe/libllvmpipe.a.p/lp_setup_tri.c.obj
"cc" "-Isrc/gallium/drivers/llvmpipe/libllvmpipe.a.p" "-Isrc/gallium/drivers/llvmpipe" "-I../../src/gallium/drivers/llvmpipe" "-I../../src/gallium/include" "-Isrc/gallium/auxiliary" "-I../../src/gallium/auxiliary" "-Iinclude" "-I../../include" "-Isrc" "-I../../src" "-Isrc/compiler/nir" "-I../../src/compiler/nir" "-Isrc/util" "-I../../src/util" "-IC:/CI-Tools/msys64/clang64/include" "-fvisibility=hidden" "-fcolor-diagnostics" "-Wall" "-Winvalid-pch" "-std=c11" "-O0" "-g" "-DPACKAGE_VERSION=\"22.0.0-devel\"" "-DPACKAGE_BUGREPORT=\"https://gitlab.freedesktop.org/mesa/mesa/-/issues\"" "-DHAVE_WINDOWS_PLATFORM" "-DHAVE_SURFACELESS_PLATFORM" "-DUSE_ELF_TLS" "-DUSE_TLS_BEHIND_FUNCTIONS" "-DENABLE_ST_OMX_BELLAGIO=0" "-DENABLE_ST_OMX_TIZONIA=0" "-DEGL_NO_X11" "-DDEBUG" "-DHAVE___BUILTIN_BSWAP32" "-DHAVE___BUILTIN_BSWAP64" "-DHAVE___BUILTIN_CLZ" "-DHAVE___BUILTIN_CLZLL" "-DHAVE___BUILTIN_CTZ" "-DHAVE___BUILTIN_EXPECT" "-DHAVE___BUILTIN_FFS" "-DHAVE___BUILTIN_FFSLL" "-DHAVE___BUILTIN_POPCOUNT" "-DHAVE___BUILTIN_POPCOUNTLL" "-DHAVE___BUILTIN_UNREACHABLE" "-DHAVE___BUILTIN_TYPES_COMPATIBLE_P" "-DHAVE_FUNC_ATTRIBUTE_CONST" "-DHAVE_FUNC_ATTRIBUTE_FLATTEN" "-DHAVE_FUNC_ATTRIBUTE_MALLOC" "-DHAVE_FUNC_ATTRIBUTE_PURE" "-DHAVE_FUNC_ATTRIBUTE_UNUSED" "-DHAVE_FUNC_ATTRIBUTE_WARN_UNUSED_RESULT" "-DHAVE_FUNC_ATTRIBUTE_WEAK" "-DHAVE_FUNC_ATTRIBUTE_FORMAT" "-DHAVE_FUNC_ATTRIBUTE_PACKED" "-DHAVE_FUNC_ATTRIBUTE_RETURNS_NONNULL" "-DHAVE_FUNC_ATTRIBUTE_ALIAS" "-DHAVE_FUNC_ATTRIBUTE_NORETURN" "-DHAVE_FUNC_ATTRIBUTE_VISIBILITY" "-DHAVE_UINT128" "-D_WINDOWS" "-D_WIN32_WINNT=0x0A00" "-DWINVER=0x0A00" "-DPIPE_SUBSYSTEM_WINDOWS_USER" "-D_USE_MATH_DEFINES" "-DUSE_SSE41" "-DUSE_GCC_ATOMIC_BUILTINS" "-DHAS_SCHED_H" "-DHAVE_CET_H" "-DHAVE_STRTOF" "-DHAVE_STRTOK_R" "-DHAVE_QSORT_S" "-DHAVE_ZLIB" "-DHAVE_ZSTD" "-DHAVE_COMPRESSION" "-DLLVM_AVAILABLE" "-DMESA_LLVM_VERSION_STRING=\"13.0.0\"" "-DLLVM_IS_SHARED=1" "-DDRAW_LLVM_AVAILABLE" "-DMESA_EXECMEM" "-DVK_USE_PLATFORM_WIN32_KHR" "-Werror=implicit-function-declaration" "-Werror=missing-prototypes" "-Werror=return-type" "-Werror=empty-body" "-Werror=incompatible-pointer-types" "-Werror=int-conversion" "-Wimplicit-fallthrough" "-Wno-missing-field-initializers" "-fno-math-errno" "-fno-trapping-math" "-Qunused-arguments" "-fno-common" "-Wno-microsoft-enum-value" "-Werror=format" "-Wformat-security" "-Werror=thread-safety" "-ffunction-sections" "-fdata-sections" "-pthread" "-D_FILE_OFFSET_BITS=64" "-D__STDC_CONSTANT_MACROS" "-D__STDC_FORMAT_MACROS" "-D__STDC_LIMIT_MACROS" "-Werror=pointer-arith" "-Werror=gnu-empty-initializer" -MD -MQ src/gallium/drivers/llvmpipe/libllvmpipe.a.p/lp_setup_tri.c.obj -MF "src/gallium/drivers/llvmpipe/libllvmpipe.a.p/lp_setup_tri.c.obj.d" -o src/gallium/drivers/llvmpipe/libllvmpipe.a.p/lp_setup_tri.c.obj "-c" ../../src/gallium/drivers/llvmpipe/lp_setup_tri.c
In file included from ../../src/gallium/drivers/llvmpipe/lp_setup_tri.c:37:
In file included from ../../src/gallium/drivers/llvmpipe/lp_setup_context.h:38:
In file included from ../../src/gallium/drivers/llvmpipe/lp_setup.h:31:
In file included from ../../src/gallium/drivers/llvmpipe/lp_jit.h:40:
In file included from ../../src/gallium/auxiliary/gallivm/lp_bld_limits.h:37:
In file included from ../../src/util/u_cpu_detect.h:41:
In file included from ../../src/util/u_thread.h:35:
In file included from ../../include/c11/threads.h:64:
In file included from ../../include/c11/threads_win32.h:58:
In file included from C:/CI-Tools/msys64/clang64/x86_64-w64-mingw32/include/windows.h:69:
In file included from C:/CI-Tools/msys64/clang64/x86_64-w64-mingw32/include/windef.h:9:
In file included from C:/CI-Tools/msys64/clang64/x86_64-w64-mingw32/include/minwindef.h:163:
In file included from C:/CI-Tools/msys64/clang64/x86_64-w64-mingw32/include/winnt.h:1555:
In file included from C:/CI-Tools/msys64/clang64/lib/clang/13.0.0/include/x86intrin.h:15:
In file included from C:/CI-Tools/msys64/clang64/lib/clang/13.0.0/include/immintrin.h:37:
C:/CI-Tools/msys64/clang64/lib/clang/13.0.0/include/tmmintrin.h:582:1: error: redefinition of '_mm_shuffle_epi8'
_mm_shuffle_epi8(__m128i __a, __m128i __b)
^
../../src/gallium/auxiliary/util/u_sse.h:159:1: note: previous definition is here
_mm_shuffle_epi8(__m128i a, __m128i mask)
^
1 error generated.

```

Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14077>

nir/legalize_16bit_sampler_srcs: Don't guess source type.

Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14895>

nir/fold_16bit_sampler_conversions: Fix src type mismatches.

Gitlab: https://gitlab.freedesktop.org/mesa/mesa/-/issues/5996
Fixes: fb29cef8 ("nir: add many passes that lower and optimize 16-bit input/outputs and samplers")
Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14895>

nir/fold_16bit_sampler_conversions: Fix dest type mismatches.

Gitlab: https://gitlab.freedesktop.org/mesa/mesa/-/issues/5996
Fixes: fb29cef8dda ("nir: add many passes that lower and optimize 16-bit input/outputs and samplers")
Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14895>

nir/fold_16bit_sampler_conversions: Don't fold dest upcasts.

This is not a valid optimization.

Fixes: fb29cef8dda ("nir: add many passes that lower and optimize 16-bit input/outputs and samplers")
Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14895>

Revert "ci: Convert generate-env.sh to a POSIX compliant script"

This reverts commit 9904ea2c76e246629fab8bda455eec07aa84f3f9 since it is
not able to properly escape all special characters (i.e. [']).

The POSIX conversion is not needed anymore, as we have made 'bash'
available in LAVA rootfs.

Reported-by: Daniel Stone <daniels@collabora.com>
Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15524>

ci: Make bash available in LAVA rootfs

Ensure 'bash' shell interpreter is available in LAVA rootfs since it is
going to be a dependency requirement from several scripts, e.g.
generate-env.sh.

Suggested-by: Daniel Stone <daniels@collabora.com>
Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15524>

fix(clover): FTBFS: Added missing include for ConstantInt for LLVM 15

With LLVM 15 the include of llvm/IR/Constants.h is required for
ConstantInt.

This commit fixes an FTBFS.

Cc: mesa-stable
Signed-off-by: Kai Wasserbäch <kai@dev.carbon-project.org>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15232>

docs: fixup breakage in release-calendar

Seems the branch was accidentally changed to no longer match the
release. Let's fix that.

Fixes: 9ba636cdc76 ("docs: update calendar and link releases notes for 21.3.8")
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15528>

Revert "intel/fs: handle interpolation modes for at_sample and at_offset too"

This reverts commit 5afbb0e7308de59dce8f5d3afde2ffc79d552ed7.

Closes: #6198
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15534>

pvr: zero-initialize variable

Because of incomplete codepaths further down in this function, we end up
trying to use this variable without initialization. Let's initialize it
to zero, just to keep the compiler happy.

This avoids a compile-time warning, which would get treated as an error
on CI.

Reviewed-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15525>

pvr: use a helper to translate stencil-ops

This gives us a single place to handle the conversion, and fixes a
compilation warning by adding an explicit cast.

Reviewed-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15525>

pvr: use a helper to translate compare-ops

This gives us a single place to handle the conversion, and fixes a
compilation warning by adding an explicit cast.

Reviewed-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15525>

microsoft/compiler: remove phi-value limit

There's no guarantee that we don't have more than 128 PHI values either,
so let's stop asuming so.

We do this by changing dxil_phi_set_incoming to dxil_phi_add_incoming,
which lets us add more incoming phi-values to the current one instead of
setting a new set of them.

This also lets us reduce these stack-arrays a bit, down to something
much more reasonable.

Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15519>

microsoft/compiler: ralloc incoming phi-values

Reserving 127 incoming values for every phi instruction is neither
robust nor memory efficient.

Let's ralloc this array instead when filling it. This way, we only pay
for what we use here.

Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15519>

pan/va: Add atomic instructions

Equivalent to their Bifrost counterparts, with a much more sensible encoding.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15515>

pan/va: Allow omitting staging registers

It's not usually valid, but sr_count == 0 is encodable and used for the
non-RETURN variant of ATOM1. Allow dis/assembling this syntax.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15515>

pan/va: Allow forcing staging flags to read-write

Required for the correct encoding of atomics.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15515>

nir: rename nir_src_is_dynamically_uniform to nir_src_is_always_uniform

As this function doesn't check for any control-flow
dependence, it only returns true for statically
(or globally) uniform values.
The same holds true for is_binding_dynamically_uniform()
in nir_opt_gcm().
Rename to better reflect that property.

Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14994>