zink: rework descriptor pool overflow

previously this would just destroy and recreate pools, but it's much
smarter to instead store the pools and just reuse them when needed

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18065>

zink: use a single allocation for zink_descriptor_layout_key

this is slightly more smart

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18065>

zink: use a dynarray instead of hash table for tracking pools

this may use a little more memory, but having direct access to the pool
without needing a lookup in a (potentially big) hash table is definitely
worth it

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18065>

zink: add an id member for zink_descriptor_pool_key

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18065>

zink: require VK_KHR_descriptor_update_template

it's been long enough, and there's no technical reason why drivers
shouldn't have this implemented

ZINK_DESCRIPTORS environment variable is preserved for future use

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18065>

aco: test for one and_savexec opcode in try_optimize_branching_sequence

A situation where it doesn't match is probably not possible, so this
probably doesn't fix anything.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18077>

aco: improve vcc check for instructions between exec_val and exec_copy

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18077>

aco: remove val_and_copy_adjacent

If this is true, then the only instruction the loops visit is
p_logical_end and the loops are no-ops.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18077>

aco: test branch opcode if removing it in try_optimize_branching_sequence

We shouldn't remove a p_cbranch_nz branch in this situation.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Fixes: b731be2e96c ("aco: Remove branch instruction when exec is constant non-zero.")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18077>

aco: fix re-write of uses of exec_val's lo/hi half

The isConstant() check isn't useful. If it's a constant, then the
physReg() check will fail.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Fixes: baab6f18c91 ("aco: Optimize branching sequence during SSA elimination.")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18077>

aco: fix assembly of vopc_sdwa writing exec

We would assemble an instruction writing vcc instead.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Fixes: 5ffc73896f3 ("aco/assembler: Fix v_cmpx with SDWA.")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18077>

aco: rename is_cmp to is_fp_cmp

The old name is no longer accurate.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18077>

aco: fix consecutive exec writes when finding exec_copy instruction

This can happen with transitions to exact:
 s2: %0:exec = p_parallelcopy %622:s[0-1]
 s2: %625:s[0-1],  s1: %624:scc,  s2: %0:exec = s_and_saveexec_b64 %141:vcc, %0:exec
 s2: %626:s[12-13] = p_cbranch_z %0:exec BB2, BB1

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Fixes: 410eff4d2f3 ("aco: Fix optimizing branching sequence with s_and_saveexec.")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18077>

ci: rename GitHub CI workflow from `ci` to `macos` to be a bit more explicit about what it covers

Reviewed-by: Yonggang Luo <luoyonggang@gmail.com>
Acked-by: Yurii Kolesnykov <root@yurikoles.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18057>

pvr: Implement vkResetDescriptorPool API.

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/18060>

docs/radv: advertise VK_KHR_global_priority

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18078>

radv: Remove workarounds for D3D9 feedback loops

RADV supports VK_EXT_attachment_feedback_loop_layout now,
and DXVK is using since 63d582a6e7ab63e18503d1238101705a818f5ad0 which
will be in DXVK 2.0, so no need for these anymore.

Older versions of DXVK will still be affected by the issue however.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17920>

tu: Make KGSL wait IOCTLs interrupt-safe

`tu_WaitForFences` and `tu_GetFenceStatus` ended up using `ioctl` instead of
`safe_ioctl` and as a result were responsible for handling `EAGAIN`/`EINTR`
but they weren't handled correctly in wait calls which led to dropping the
wait if the thread received a signal without automatic syscall restarting on
release builds and an assertion due to `errorno` being `EINTR` on debug
builds.

This fixes early wait returns which would lead to UB or cause crashes due to
hitting debug assertions in Skyline Emulator.

Signed-off-by: Mark Collins <mark@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18031>

radv: advertise VK_KHR_global_priority

Similar to VK_KHR_buffer_device_address which has also been promoted
from EXT.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Acked-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18061>

radv: remove radv_graphics_pipeline_info completely

RADV now uses vk_graphics_pipeline_state from Vulkan common code which
is more convenient for eg. graphics pipeline libraries.

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/18015>

radv: convert radv_vertex_input_info to vk_vertex_input_state

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/18015>

radv: convert radv_rendering_info to vk_render_pass_state

This also converts VkAttachmentSampleCountInfoAMD which is part of
vk_render_pass_state.

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/18015>

radv: convert radv_color_blend_info to vk_color_blend_state

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/18015>

radv: convert radv_multisample_info to vk_multisample_state

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/18015>

radv: convert radv_input_assembly_info to vk_input_assembly_state

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/18015>

radv: convert radv_fragment_shading_rate_info to vk_fragment_shading_rate_state

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/18015>

radv: convert radv_depth_stencil_info to vk_depth_stencil_state

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/18015>

radv: convert radv_rasterization_info to vk_rasterization_state

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/18015>

radv: convert radv_discard_rectangle_info to vk_discard_rectangles_state

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/18015>

radv: convert radv_viewport_info to vk_viewpoert_state

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/18015>

radv: convert radv_tessellation_info to vk_tessellation_state

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/18015>

radv: call vk_graphics_pipeline_state_fill() when initializing gfx pipelines

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/18015>

radv: slightly change the color write enable mask

For switching to the common graphics pipeline state.

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/18015>

radv: do not translate primitive topology during gfx info initialization

For switching to the common graphics pipeline state.

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/18015>

radv: do not translate blend op/factor during gfx info initialization

For switching to the common graphics pipeline state.

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/18015>

radv: do not translate logic op during gfx info initialization

For switching to the common graphics pipeline state.

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/18015>

radv: do not translate polygon mode during gfx info initialization

For switching to the common graphics pipeline state.

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/18015>

radv: simplify the depth clip enable/disable logic

For switching to the common graphics pipeline state.

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/18015>

radv: store the provoking vertex mode in the graphics pipeline info

For switching to the common graphics pipeline state.

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/18015>

vulkan: add attachment sample count info AMD support

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18015>

vulkan: fix initializing the primitive topology when it's dynamic

The PSO still contain the topology class and some Vulkan drivers like
RADV need to know this.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18015>

vulkan: initialize discardRectangleCount even if the state is dynamic

Only the array of rectangles should be dynamic.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18015>

intel/compiler: Use named NIR intrinsic const index accessors

In the early days of NIR, you had to prod at inst->const_index[]
directly, but a long while back, we added handy accessor functions
that let you use the actual name of the thing you want instead of
memorizing the exact order of parameters.

Also rewrite a comment I had a hard time parsing.

Reviewed-by: Ivan Briano <ivan.briano@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18067>

v3dv: vkCmdClearAttachments no longer generates its own RCL

We improved this a long time ago and now it emits a clear rect inside
the current subpass job instead of emitting its own job with its own
RCL, so we no longer need to handle this as an exception to the rule.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17854>

v3dv: add a heuristic for double-buffer mode

For this we add a scoring system that evaluates various aspects of
the draw calls in a job.

If the cost of the geometry side of the pipeline is too high, then
we may pay too high a price in double-buffer mode because with smaller
tile size may will probably have more vertex shader invocations in the
render and binning stages.

On the other hand, if rendering cost is not high enough, we may not
have enough rendering work to hide the latency of tile stores in
double-buffer mode.

Also, because we need to make a decision after we know all the draw
calls in a job, but the double-buffer enable bit comes in the
TILE_BINNING_MODE_CFG that needs to be emitted first in the binning
command list before the draw calls are recorded, if we decide to
enable it we need to rewrite that packet and we need to size the
tile state properly to account for the extra tiles. For this
purpose we delay tile state setup for render pass jobs until we are
finishing a job.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17854>

v3dv: tell job_compute_frame_tiling whether we want to use double-buffer mode

We want to have control over the double-buffer setting here so we can control
explicitly from the driver when we want to enable this mode.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17854>

v3dv: postpone tile state allocation for render pass jobs

These are jobs for which we may want to enable double-buffering,
which affects tile state allocation. Since the idea is that we
want to decide about double buffering late, we also want to
postpone allocation of the tile state until we are about to
emit the RCL for the job.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17854>

v3dv: add a v3dv_job_allocate_tile_state helper

If we enable double-buffer we are reducing the tile size, and thus,
we'll need more tiles and a larger tile state allocation, so we'll
need to call to this helper.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17854>

v3dv: fix tile state allocation

If we had decided that we only need one layer worth of tile state
we should actually only allocate one layer.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17854>

v3dv: track if a job is compatible with double-buffer mode

MSAA is not compatible with double-buffer mode. Also, jobs that emit
tile loads or that don't have any stores can't take advantage
of double-buffer mode.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17854>

v3dv: move check_needs_load/store helpers to unversioned code

These helpers don't depend on the V3D version and we are going
to need them outside v3dvx_cmd_buffer.c soon.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17854>

broadcom/compiler: track number of TMU operations in prog data

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17854>

docs/features: drop unsupported VK_KHR_display & VK_KHR_get_display_properties2 for lavapipe

Acked-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17796>

docs/features: start tracking extensions supported by powervr

Reviewed-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17796>

docs/features: start tracking extensions supported by panvk

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17796>

docs/features: start tracking extensions supported by dozen

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/6330
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17796>

docs/features: add missing extensions supported by venus

Reviewed-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17796>

docs/features: add missing extensions supported by turnip

Reviewed-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17796>

docs/features: add missing extensions supported by radv

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17796>

docs/features: add missing extensions supported by lavapipe

Acked-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17796>

docs/features: add missing extensions supported by anv

Reviewed-by: Ivan Briano <ivan.briano@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17796>

turnip: handle fence wrapping in autotune

fence values can be 0 after wrapping.  We should not treat 0 specially.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18040>

turnip: remove unused member in tu_submission_data

buffers_count is set but unused.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18040>

turnip: fix a use-after-free in autotune

When removing old histories, check against gpu fence.  Otherwise,
pending_results could have dangling pointers to the removed histories.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7055
Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18040>

zink: use a maybe more accurate wild guess for pcp-less gpl

this is only reachable with an env var, so being wrong is still okay,
but maybe be right slightly more often

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17955>

zink: (correctly) require extendedDynamicState2PatchControlPoints for GPL

this is otherwise broken

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17955>

zink: require EXT_non_seamless_cube_map for GPL support

it's impossible to precompile without this

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17955>

c11: Remove mtx_try from mtx enums as it's not a part of c11 standard

And we support for four types of mtx init enum values that define by C11 standard:
0 mtx_plain - a simple, non-recursive mutex is created.
2 mtx_timed - a non-recursive mutex, that supports timeout, is created.
1 mtx_plain | mtx_recursive - a recursive mutex is created.
3 mtx_timed | mtx_recursive - a recursive mutex, that supports timeout, is created.

musl library also use these enum combination

Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18036>

aco: Use v_cmpx pre GFX10.

Foz-DB Vega10:
Totals from 29508 (21.85% of 135041) affected shaders:
CodeSize: 184345656 -> 184345820 (+0.00%)
Instrs: 35906154 -> 35906195 (+0.00%)
Latency: 581696114 -> 581530021 (-0.03%); split: -0.03%, +0.00%
InvThroughput: 245625572 -> 245561351 (-0.03%); split: -0.03%, +0.00%
Copies: 3134925 -> 3278672 (+4.59%)

Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18049>

aco/assembler: Fix v_cmpx pre GFX10.

The second destination is implicit exec.

Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18049>

aco: Fix optimizing branching sequence with s_and_saveexec.

This optimization was broken for two reasons:
- s_and_saveexec has two operands, the copy value and exec
- s_and_saveexec has an exec read, so exec_write_used will always be true
  before we find branch_exec_val_idx

Foz-DB Navi21:
Totals from 31453 (23.31% of 134913) affected shaders:
CodeSize: 204831260 -> 204831156 (-0.00%)
Instrs: 38157117 -> 38157091 (-0.00%)
Latency: 533708882 -> 531211721 (-0.47%); split: -0.47%, +0.00%
InvThroughput: 107088408 -> 106719188 (-0.34%); split: -0.35%, +0.00%
Copies: 2326179 -> 2502490 (+7.58%)

Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18049>

aco/assembler: Fix v_cmpx with SDWA.

We need to use the implicit destination.

Fixes: baab6f18c91 ("aco: Optimize branching sequence during SSA elimination.")
Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18049>

aco: Check that we don't override exec_val operands during branching sequence optimization.

Fixes: baab6f18c91 ("aco: Optimize branching sequence during SSA elimination.")
Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18049>

aco: Check for instructions that inhibit the branching sequence optimization.

Fixes: baab6f18c91 ("aco: Optimize branching sequence during SSA elimination.")
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18049>

aco: Fix invalidated reference in branching sequence optimization.

Inserting in the instructions vector may invalidate the exec_val reference,
so do that last.

Fixes: baab6f18c91 ("aco: Optimize branching sequence during SSA elimination.")
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18049>

ci: do not install i386 sub-architecture in VK images

Not required by apitrace.

debian/x86_test-vk image size:
previous: 1.8G
after:    1.0G

v2: added back winehq-stable (missing symlinks)

Acked-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17834>

ci: use apt-get in x86_test-vk instead of apt

apt shouldn't be used in CI, see:
https://unix.stackexchange.com/questions/590699/should-i-use-apt-or-apt-get-in-shell-scripting

Acked-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17834>

ci: Report failures to GitLab for LAVA jobs

We were uploading the junit.xml to the artifacts, but weren't telling
GitLab about it.

Reviewed-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18028>

iris/ci: Set FDO_CI_CONCURRENT for all Chromebook jobs

For some reason we were missing these and that was causing some CPUs to
remain idle, and some boards to have too high load and some tests timing
out occasionally.

Set FDO_CI_CONCURRENT for the number of cores, plus one to account for
GPU and I/O waits.

Reviewed-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18028>

iris/ci: Test on Jasper Lake Chromebooks

8 such boards have been added to Collabora's lab, let's run all tests
from the selected set.

Reviewed-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18028>

nir/lower_gs_intrinsics: fix primitive count for points

When primitive is points, EndPrimitive can't be used to count
primitive. Need to use vertex count instead. And it's also not
needed to do vertex per primitive count and overwrite incomplete
primitive work for points.

Fixes: 2be99012e95 ("nir: Add ability to count emitted GS primitives.")
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17805>

zink: fix program cache comparisons

I missed this when pipe shader types were merged with mesa types

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18051>

zink: delete zink_shader_stage()

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18051>

zink: remove all pipe_shader_type usage

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18051>

zink: PIPE_SHADER_TYPES -> MESA_SHADER_STAGES

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18051>

zink: explicitly define ZINK_GFX_SHADER_COUNT

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18051>

zink: ZINK_SHADER_COUNT -> ZINK_GFX_SHADER_COUNT

more descriptive

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18051>

zink: use statically sized array for descriptor allocation

this will never be more than 100

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18051>

zink: rename _lazy descriptor stuff

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18051>

zink: reorder some descriptor structs

should yield slightly better packing

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18051>

zink: merge batch descriptor data onto batch state

also clean up some related descriptor code

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18051>

zink: merge program descriptor data onto program struct

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18051>

zink: merge context descriptor data onto context struct

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18051>

zink: merge batch descriptor data structs

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18051>

zink: merge context descriptor data structs

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18051>

zink: remove descriptor dummy_set

while well-intentioned, this didn't actually do anything

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18051>

zink: delete some unused descriptor struct members

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18051>

zink: move lazy descriptor data types into zink_types.h

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18051>

zink: move all types/enums into zink_types.h

this significantly untangles all the header includes and fixes an issue
where zink_descriptors.h couldn't be included by certain files due to
type conflicts

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18051>

zink: delete unused struct types

so old

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18051>