zink: require optimal_keys for GPL

this is pointless otherwise

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18911>

zink: set VK_PIPELINE_CREATE_RETAIN_LINK_TIME_OPTIMIZATION_INFO_BIT_EXT

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18911>

zink: further compress zink_gfx_output_key lookups

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18911>

zink: compress zink_gfx_output_key members

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18911>

zink: don't access prog->shaders array during pipeline compile

this may or may not have valid data

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18911>

zink: add tgl ci skip

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18921>

venus: Update mutable descriptor struct type

Signed-off-by: Dawn Han <dawnhan@google.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18766>

venus: support the promoted VK_EXT_mutable_descriptor_type

Signed-off-by: Dawn Han <dawnhan@google.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18766>

gallivm: remove dead cube map lod calculation code

Since commit d413fd02190d ("gallivm: Always take the per-pixel LOD path
for cubemaps.") the special cube rho code was not used anymore.

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18892>

vulkan/wsi/x11: Fix double free of modifiers.

Breaks radv when modifiers are used.

Fixes: 8775f08bf11 ("vulkan/wsi/x11: Configure images via params passed to wsi_swapchain_init()")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7386
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7398
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-By: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18919>

zink: use more dynamic state3 when available

this is an all-or-nothing type of thing, where either all the blend/ms
states are available or they aren't

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18890>

zink: use dynamic state3 rasterization states when available

this massively reduces the number of pipelines required

also temporarily disable GPL until a few more patches land since something's
weird and broken

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18890>

lavapipe: improve blend handling on pipeline binds

* use renderpass info for independent blend
* only reset blend state if no color attachments

Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18913>

aco: fix assembler.gfx11.vinterp test

This was missed. I guess CI doesn't have a recent enough LLVM for these
tests.

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/17710>

aco: limit hard clauses to 63 instructions

See https://reviews.llvm.org/D127391

fossil-db (gfx1100):
Totals from 4 (0.00% of 161689) affected shaders:
Latency: 24545 -> 24539 (-0.02%)
InvThroughput: 102867 -> 102835 (-0.03%)

fossil-db (navi10):
Totals from 4 (0.00% of 161220) affected shaders:
Latency: 25969 -> 25959 (-0.04%)
InvThroughput: 112917 -> 112869 (-0.04%)

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/17710>

aco/gfx11: update form_hard_clauses

See https://reviews.llvm.org/D127391

fossil-db (gfx1100):
Totals from 116 (0.07% of 161689) affected shaders:
Instrs: 124719 -> 124664 (-0.04%); split: -0.06%, +0.02%
CodeSize: 731660 -> 731440 (-0.03%); split: -0.04%, +0.01%
Latency: 2771695 -> 2771671 (-0.00%); split: -0.00%, +0.00%
InvThroughput: 1050309 -> 1050312 (+0.00%)
VClause: 3731 -> 3779 (+1.29%)

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/17710>

aco/gfx11: deallocate VGPRs at the end of the shader

fossil-db (gfx1100):
Totals from 65987 (40.81% of 161689) affected shaders:
Instrs: 57123207 -> 57199947 (+0.13%)
CodeSize: 308402500 -> 308709460 (+0.10%)
Latency: 680527139 -> 680527160 (+0.00%)
InvThroughput: 131620026 -> 131620045 (+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/17710>

aco: update sendmsg enum from LLVM

Add GFX11 enums and some new ones that apparently existed before.

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/17710>

aco/gfx11: fix s_waitcnt printing

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/17710>

aco: improve wait_imm unpack

Add GFX11 support and use wait_imm::unset_counter. Looping in the waitcnt
pass was probably broken on GFX11 because of this.

fossil-db (gfx1100):
Totals from 899 (0.56% of 161689) affected shaders:
Instrs: 1319368 -> 1319179 (-0.01%)
CodeSize: 7124640 -> 7123884 (-0.01%)
Latency: 26554304 -> 26404606 (-0.56%)
InvThroughput: 9032485 -> 8978773 (-0.59%); split: -0.59%, +0.00%

No navi10 fossil-db changes.

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/17710>

aco/gfx11: swap ds_cmpst_* data operands

According to an LLVM comment, these are swapped in GFX11.

No fossil-db changes.

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/17710>

zink: enable loop unrolling for loops that have soft fp64

The info that a loop contains fp64 ops is lost after full
software lowering.
Therefore, specifically unroll these loops in zink instead of
in vulkan, where the blown up body is no longer unrolled.
This helps a lot in KHR-Single-GL46.arrays_of_arrays_gl.SubroutineFunctionCalls2.

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18863>

nir: allow to fine tune unrolling for loops with soft fp64 ops

Lowered fp64 ops can blow up the loop bodies while still being
suitable for unrolling.
Allow for using different parameters to unroll loops with
soft fp64.

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18863>

nir: track whether a loop contains soft fp64 ops

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18863>

zink: fix spirv temp func variable class emission

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18863>

zink: fix nir_op_unpack_64_2x32 emission

Cc: mesa-stable
Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18863>

zink: lower 64bit pack when float64 not supported

SPIRV ops for 64bit pack/unpack require float64 cap.

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18863>

zink: optimise lower_64bit_vars for when int64 is supported

There are vulkan drivers that support Int64 but not Float64.
Optimise the 64bit lowering by converting doubles to uint64,
but keep matrix lowerig to float as glsl int64 matrices do
not exist.

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18863>

zink: lower 64bit vars for drivers without float64

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18863>

zink: call flush_resource when presenting garbage

this ensures the resource is queued for presentation

fixes #7359

Fixes: babf9474c43 ("zink: rework flush_frontbuffer to always flush")

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18899>

ci: Remove Wno-error=stringop-truncation from debian-arm64-asan

All the warnings are now fixed.

Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Acked-by: David Heidelberg <david.heidelberg@collabora.com>
Signed-off-by: Corentin Noël <corentin.noel@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18613>

gallium/hud: use snprintf(..., "%s", ...) instead of strncpy

This fixes a warning with stringop-truncation:
```
error: 'strncpy' output may be truncated copying 127 bytes from a string of length 255 [-Werror=stringop-truncation]
 1443 |             strncpy(graph->name, s, sizeof(graph->name)-1);
      |             ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
```

Signed-off-by: Corentin Noël <corentin.noel@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18613>

mesa/main: do not copy the exact size of the string

This fixes a compiler warning with stringop-truncation:
```
error: 'strncpy' specified bound 4096 equals destination size [-Werror=stringop-truncation]
  286 |    strncpy(s, msg, MAX_DEBUG_MESSAGE_LENGTH);
      |    ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
```

Signed-off-by: Corentin Noël <corentin.noel@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18613>

aco: add VMEMtoScalarWriteHazard tests

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/18270>

aco: improve printing of s_waitcnt_depctr

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/18270>

aco: improve printing of sgpr_null

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/18270>

aco: use some helpers in GFX10 hazard workarounds

No fossil-db changes.

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/18270>

aco: improve VMEMtoScalarWriteHazard s_waitcnt mitigation

fossil-db (navi10):
Totals from 171 (0.13% of 134563) affected shaders:
Instrs: 69348 -> 69144 (-0.29%)
CodeSize: 381112 -> 380296 (-0.21%)
Latency: 451534 -> 451387 (-0.03%)
InvThroughput: 88692 -> 88686 (-0.01%)

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/18270>

aco: fix VMEMtoScalarWriteHazard s_waitcnt mitigation

It doesn't make sense for a "s_waitcnt vmcnt(0)" to affect a store or DS
instruction.

LLVM checks for "s_waitcnt vmcnt(0) lgkmcnt(0) expcnt(0)" but ignores
s_waitcnt_vscnt (which I assume is a bug).

No fossil-db changes.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Fixes: bcf94bb933e ("aco: properly recognize that s_waitcnt mitigates VMEMtoScalarWriteHazard")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18270>

nir/opt_algebraic: Optimize more (a cmp b ? a : b) to min/max.

Foz-DB Navi21:
Totals from 112 (0.08% of 134913) affected shaders:
CodeSize: 1618384 -> 1618172 (-0.01%); split: -0.06%, +0.04%
Instrs: 307695 -> 307535 (-0.05%); split: -0.05%, +0.00%
Latency: 3590228 -> 3589658 (-0.02%); split: -0.02%, +0.00%
InvThroughput: 563692 -> 563447 (-0.04%); split: -0.05%, +0.01%
Copies: 24541 -> 24519 (-0.09%); split: -0.10%, +0.01%
Branches: 13480 -> 13468 (-0.09%)

Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18548>

pvr: Implement vkGetEventStatus 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/18864>

pvr: Add assert to check for non-NULL binding_desc.

Signed-off-by: Rajnesh Kanwal <rajnesh.kanwal@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Reviewed-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18866>

pvr: Call VALGRIND_FREELIKE_BLOCK before unmapping to avoid use after free.

Fixes: CID 1503261.

Signed-off-by: Rajnesh Kanwal <rajnesh.kanwal@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Reviewed-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18866>

pvr: Fix Unintentional integer overflow (OVERFLOW_BEFORE_WIDEN) errors.

Fixes: CID 1503255, 1503258, 1503265, 1503268, 1504078, 1507509.

Signed-off-by: Rajnesh Kanwal <rajnesh.kanwal@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Reviewed-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18866>

pvr: Zero init pbe_cs_words to avoid passing on garbage.

In case if the surface->resolve_type is not PVR_RESOLVE_TYPE_PBE
then we do not setup pbe_cs_words in which case we might end up
using garbage values.

Fixes: CID 1515466.

Signed-off-by: Rajnesh Kanwal <rajnesh.kanwal@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Reviewed-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18866>

pvr: Remove dead code from pvr_perform_start_of_render_clears.

Logically the hw_render would never be NULL so the if block is
never used and can be removed.

Fixes: CID1503270.

Signed-off-by: Rajnesh Kanwal <rajnesh.kanwal@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Reviewed-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18866>

pvr: Remove framebuffer NULL check from pvr_CmdBeginRenderPass2.

Framebuffer can not be NULL in pvr_CmdBeginRenderPass2.

Fixes: CID 1503259.

Signed-off-by: Rajnesh Kanwal <rajnesh.kanwal@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Reviewed-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18866>

pvr: Remove redundant assignment error.instruction.

Fixes: CID 1503256.

Signed-off-by: Rajnesh Kanwal <rajnesh.kanwal@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Reviewed-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18866>

pvr: Remove redundant instruction buffer check.

Fixes: CID 1503273.

Signed-off-by: Rajnesh Kanwal <rajnesh.kanwal@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Reviewed-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18866>

pvr: Convert attachment indexes to unsigned.

This commit converts attachment indexes used in renderpass to unsigned
and uses VK_ATTACHMENT_UNUSED for unused attachments instead of -1.

Also fixes some signed to unsigned comparisons.

Fixed: CID 1515587, 1515590, 1515597, 1515604.

Signed-off-by: Rajnesh Kanwal <rajnesh.kanwal@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Reviewed-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18866>

anv: limit to default fast clear color when image used for transfer

v2: use vk_image_layout_to_usage_flags for detecting
    transfer usage (Nanley)

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7189
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18836>

pvr: Handle indirect buffer address in pvr_setup_descriptor_mappings.

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/18872>

pvr: implement vkcmddispatchindirect 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/18872>

radv: enable the unified heap on APUs for Red Dead Redemption 2

This seems to resolve the memory explosion without hurting performance.
This workaround is only applied for native Vulkan.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18884>

radv: add a new drirc option to enable a unified heap on APUs

This option (radv_enable_unified_heap_on_apu) allows to force the
driver to expose only one heap of VRAM because some games seem to
perform better with that.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18884>

zink: enable glthread by default

huge perf gains.

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18858>

pipe-loader: fix zink driinfo header path

this feels like a compiler bug, but somehow just specifying the
shorter path ends up with a broken/partial header include

Fixes: d760a9151b7 ("gallium: Learn about kopper")

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18858>

zink: export PIPE_CAP_MAP_UNSYNCHRONIZED_THREAD_SAFE

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18858>

zink: export PIPE_CAP_ALLOW_MAPPED_BUFFERS_DURING_EXECUTION

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18858>

zink: fix/improve handling for multi-component bitfield ops

the original improvement for this correctly handled cases where
the offset/count values were swizzled with .xxxx, but it was broken
for any other swizzling

this adds a nir pass to more easily manipulate the swizzles and rewrite
the multi-component ops as multiple ops with the swizzled params

fixes #6697

Fixes: 8e97f51c676 ("zink: handle swizzled offset/count values for shader bitfield ops")

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18706>

llvmpipe/fs: convert outputs handling to explicit api

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Mihai Preda <mhpreda@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18876>

llvmpipe/fs: cleanup some remaining mask handling and reuse types

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Mihai Preda <mhpreda@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18876>

llvmpipe/fs: handle explicit types around blending and c/zs bufs calcs

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Mihai Preda <mhpreda@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18876>

llvmpipe/fs: use explicit api in viewport code.

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Mihai Preda <mhpreda@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18876>

llvmpipe/fs: pass mask type to alpha to coverage handler

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Mihai Preda <mhpreda@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18876>

llvmpipe/fs: add sample position type to the interp interface

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Mihai Preda <mhpreda@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18876>

llvmpipe/fs: port depth code to opaque pointer api

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Mihai Preda <mhpreda@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18876>

llvmpipe/fs: pass explicit mask_type into interp code.

This is so it is known for explicit pointer support

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Mihai Preda <mhpreda@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18876>

llvmpipe/fs: add mask_type for mask_store accesses.

This adds and passes around the type for mask stores.

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Mihai Preda <mhpreda@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18876>

gallivm/struct: add opaque ptr friendly pointer accessors.

These just add explicit types.

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Mihai Preda <mhpreda@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18876>

Revert "intel/compiler: Vectorize gl_TessLevelInner/Outer[] writes"

This reverts commit abba55382f5588e3b9dfbc517d5fe940b2917c7c.
The assertions I added late in the process broke shader-db, and my
quick fix broke CI, so let's just revert it for now and I'll resubmit
this later when it's working better.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7385
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18895>

freedreno/ir3: Validate our scheduling DAGs after construction.

This gives us some better explanation of a stack overflow in ir3_postsched
with shader-db:

IR3_SHADER_DEBUG=nouboopt ./run shaders/nexuiz/46.shader_test

DAG validation failed at:
  0x55f6570e8460: 0079:0107:002:  _meta:collect r1.w (wrmask=0xff), r1.w, r2.x, r2.y, r2.z, r2.w, r3.x, r3.y, r3.z, false-deps:_[0098:0126:000:  mov.u32u32], _[0112:0143:000:  mov.u32u32], _[0087:0113:000:  mov.u32u32], _[0113:0144:000:  mov.u32u32], _[0099:0127:000:  mov.u32u32], _[0088:0114:000:  mov.u32u32]

Nodes in stack:
  0x55f657102050: 0079:0103:009:  mov.u32u32 r1.w, r0.x, right=_[0080:0104:009:  mov.u32u32]

  0x55f6570e8460: 0079:0107:002:  _meta:collect r1.w (wrmask=0xff), r1.w, r2.x, r2.y, r2.z, r2.w, r3.x, r3.y, r3.z, false-deps:_[0098:0126:000:  mov.u32u32], _[0112:0143:000:  mov.u32u32], _[0087:0113:000:  mov.u32u32], _[0113:0144:000:  mov.u32u32], _[0099:0127:000:  mov.u32u32], _[0088:0114:000:  mov.u32u32]

  0x55f657075f80: 0083:0108:007:  samgq (f32)(xyz)r0.z (wrmask=0x7), r1.w (wrmask=0xff), s#3, t#3

  0x55f657051b60: 0104:0134:008:  ldc.offset0 r3.x (wrmask=0xf), imm[0.000000,0,0x0], r9.w

  0x55f657103040: 0112:0143:000:  mov.u32u32 r9.w, r0.x, right=_[0113:0144:000:  mov.u32u32]

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6656>

util/dag: Add a validation function.

I experienced a circular dag in freedreno, and wanted a way to see what
was going wrong.

Reviewed-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6656>

util/dag: Move the callback function pointer to the state.

No sense passing it as a recursive argument when we have this storage
right there.

Reviewed-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6656>

glthread: handle DeleteBuffers(n=-1) gracefully

this is a valid error condition

cc: mesa-stable

fixes:
GTF-GL46.gtf21.GL2FixedTests.buffer_objects.buffer_objects

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18856>

glthread: use signed vertex buffer offsets when available, don't require them

this is a great memory-saving optimization for drivers that support it,
but for drivesr that can't, it ends up taking slower paths instead
of just consuming more memory

but the year is 2022. we have all the memory. so use more memory if signed
offsets aren't supported

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18856>

d3d12: Style variable fix in d3d12_video_decoder_prepare_dxva_slices_control_av1

Reviewed-by: Giancarlo Devich <gdevich@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18875>

d3d12: Fix aliased usage of m_SliceControlBuffer in d3d12_video_decoder_prepare_dxva_slices_control_XXXX

Reviewed-by: Giancarlo Devich <gdevich@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18875>

d3d12: Support more video formats

Reviewed-by: Giancarlo Devich <gdevich@microsoft.com>
Adds support for P010 Encoding and Decoding
Also adds support for RGBA/RGBX UNORM/UINT in the video processor

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18875>

d3d12: Add support for AV1 Video Decode

Reviewed-by: Giancarlo Devich <gdevich@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18875>

frontends/va: Fix AV1 decode qm_* parameters reading from VADecPictureParameterBufferAV1

Reviewed-by: Ruijing Dong <ruijing.dong@amd.com>
Those are being read with | 0xF, but instead should be & 0xF to clear the bits above 4 (the field is 4 bits in size)

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18875>

frontends/va: Use is_video_format_supported in vlVaVidEngineBlit instead of format list

Reviewed-by: Ruijing Dong <ruijing.dong@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18875>

frontends/va: Support AV1 Decode with multiple tiles (num_elements > 1) in a single VASliceParameterBufferAV1 buffer

Reviewed-by: Ruijing Dong <ruijing.dong@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18875>

frontends/va: Add more params from VADecPictureParameterBufferAV1 and VASliceParameterBufferAV1 to pipe_av1_picture_desc

Reviewed-by: Ruijing Dong <ruijing.dong@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18875>

intel/compiler: fix loading of draw_id from task & mesh payload

Previously both destination and source were floats, so no casting was
performed, but with 7664c85b1da source register was reinterpreted as
unsigned integer, so MOV started casting that integer to float.

Fixes: 7664c85b1da ("intel/compiler: Create and use struct for TASK and MESH thread payloads")
Reviewed-by: Caio Oliveira <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18886>

ci/virgl: Enable virgl-iris-traces

All virgl-iris-traces tests are passing now after the crosvm revup that
fixed crosvm crash and after updating test checksums. Enable auto-testing
for the virgl-iris-traces pipeline stage.

Reviewed-by: Corentin Noël <corentin.noel@collabora.com>
Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18693>

ci: Update virgl-iris-traces checksums

Few virgl-iris-traces tests fail because resultant images now have few
pixels that changed. The actual reason of the change is unknown, we
assume it happened due to a new optimizations. Update the virgl-iris-traces
checksums.

Reviewed-by: Corentin Noël <corentin.noel@collabora.com>
Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18693>

ci: Bump crosvm version

Recent crosvm contains a fix for the unmapping memory on a wrong thread
that crashes crosvm. Bump the crosvm version.

Reviewed-by: Yiwei Zhang <zzyiwei@chromium.org>
Reviewed-by: Corentin Noël <corentin.noel@collabora.com>
Link: https://gitlab.freedesktop.org/virgl/virglrenderer/-/issues/282
Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18693>

venus/ci: remove redundant crosvm patch

VirglRenderer::export_blob doesn't need to handle opaque fd. That will
let rutabaga's export_blob to fail and fallback directly to the external
mapping path desired.

Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18693>

radv: ignore shader stages that don't need to be imported with GPL

Fixes an assertion with vkoverhead.

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/18869>

radv: fix initializing the blend state for NULL render pass with GPL

If a library with only the vertex input interface is created, this
would crash.

This fixes segfault with vkoverhead.

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/18869>

radv: fix initializing the attribute mask for VS input state and GPL

attribute_mask was unused when I first introduced this but now it's
used again.

This fixes a bunch of GPL regressions.

Fixes: 0feab7b9cfa ("radv: prepare the VS input state for prologs created with GPL")
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/18862>

zink: replace mixed_zs with zs feedback loops

this is the more technically correct way to handle such things

SoroushIMG <soroush.kashani@imgtec.com>

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18855>

zink: add a draw barrier when unbinding a sampler-bound fb surface

this is guaranteed to need some kind of layout change, so ensure the
barrier actually happens

cc: mesa-stable

SoroushIMG <soroush.kashani@imgtec.com>

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18855>

virgl: Add nir_lower_tex pass

In GLES hosts that do not support EXT_texture_shadow_lod we need to lower
the shadow texture calls with offset.

Signed-off-by: Corentin Noël <corentin.noel@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18296>

virgl: Add TEXTURE_SHADOW_LOD capability support

Bind PIPE_CAP_TEXTURE_SHADOW_LOD with its value so that we are not using
non-existing GLSL on host side.

Rename VIRGL_CAP_FAKE_FP64 into VIRGL_CAP_HOST_IS_GLES

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/18296>

radv: remove NULL checks when binding a graphics pipeline

The driver no longer bind NULL graphics pipelines, so these checks
are useless.

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/18873>

tu: Fix setting RB_DEPTH_CNTL::Z_CLAMP_ENABLE

I missed this when enabling pipeline libraries, and we were also setting
this to the wrong thing. Previously we were using rasterization state
when parsing depth/stencil indirectly via builder->depth_clip_disable,
which is not allowed with pipeline libraries. Fixing this is a bit
painful because now RB_DEPTH_CNTL can depend on state from both the
fragment shader library and the pre-rasterization library, in addition
to being disabled via output interface state.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18861>

freedreno/a6xx, tu: GRAS_CL_CNTL::UNK5 is Z_CLAMP_ENABLE

This changes the behavior for freedreno but it should ultimately be the
same for GL/GLES, given what mesa/st does.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18861>

Revert "freedreno,ir3: rename Z_CLAMP_ENABLE to Z_CLIP_DISABLE"

This reverts commit 6cb41c51884c2488194f74e2ce637b950c835620. It was
incorrect and the issue it was trying to fix was actually a zink bug.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18861>