intel/compiler: mask GS URB handles at thread payload construction

Reviewed-by: Caio Oliveira <caio.oliveira@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25195>

intel/compiler/mesh: implement IO for xe2

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25195>

intel/compiler/mesh: fix position of output URB handle for xe2

Reviewed-by: Caio Oliveira <caio.oliveira@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25195>

intel/fs: Delete manual 'inst->mlen' calculations from all uses of logical URB reads.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25195>

intel/fs: Delete manual 'inst->mlen' calculations from all uses of logical URB writes.

Rework:
 * Marcin: update emit_urb_indirect_vec4_write

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25195>

intel/fs: Specify number of data components of logical URB writes via control immediate.

This is what most logical SEND messages do when they take a variable
number of components.  'inst->mlen' is expected to be zero for logical
SEND opcodes, which are expected to behave like plain arithmetic
operations, so certain automated transformations (like SIMD lowering)
can manipulate them without opcode-specific special-casing.

Guessing the number of components from 'inst->mlen' has other
disadvantages, because it requires duplicating the logic that infers
the message payload size in every use of the instruction -- Instead we
can just do the computation once during logical send lowering.  In
addition on LNL platform this causes the 'inst->mlen' field of URB
writes to have units inconsistent with every other SEND instruction,
which is likely to lead to confusion and bugs down the road.

Rework:
 * Marcin: update emit_urb_indirect_vec4_write

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25195>

intel/fs/xe2+: Fix URB writes with 0 data components.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25195>

intel/compiler/xe2: Update TCS ICP handle code to support SIMD16

Rework:
 * Use ffs(grf_size_bytes) (s-b Ken)

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25195>

intel/compiler/xe2: Fix URB writes in TCS

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25195>

intel/compiler/xe2: URB fence uses LSC now

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25195>

intel/compiler/xe2: Update fs_visitor::emit_urb_writes to not assume SIMD8

v2: Account for 512b physical registers which causes the URB handle to be in FIXED_GFR 2 instead of 1.

XXX - Use fs_builder::vgrf() instead of open-coded dispatch_width calculations.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25195>

intel/fs: Fix Xe2 URB read/lowering with per-slot offsets

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25195>

intel/compiler: Use enum xe2_lsc_cache_load on xe2

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25195>

intel/compiler: Add enum xe2_lsc_cache_load

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25195>

intel/compiler: Use enum xe2_lsc_cache_store on xe2

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25195>

intel/compiler: Add enum xe2_lsc_cache_store

Rework:
 * Rohan: Fix enum value for L1WB_L3WB
 * Fix write-through comments (Ken)

Ref: bspec 71167
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25195>

intel/compiler: add initial support for URB_LOGICAL_SRC_CHANNEL_MASK to lower_urb_write_logical_send_xe2

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25195>

intel/compiler: add lsc_msg_desc_wcmask

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25195>

intel/compiler/xe2: Handle new URB write messages

Rework:
 * idr v1: Fix compilation error.
 * idr v2: Add support for per-channel offsets.
 * idr v3: get_lowered_simd_width is 16 on Xe2+.
 * idr v4: Add disassembly support.  Add validation support.
 * Sqaushed in changes Marcin Ślusarz's patches:
   * "intel/compiler: skip adding 0 to payload address"
   * "intel/compiler/xe2: drop masking off top 8 bits of URB handle"

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25195>

intel/compiler/xe2: Handle new URB read messages

Rework:
 * Sqaushed in changes Marcin Ślusarz's patches:
   * "intel/compiler: skip adding 0 to payload address"
   * "intel/compiler/xe2: drop masking off top 8 bits of URB handle"

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25195>

vulkan: Remove vk_get_physical_device_core_1_*_feature_ext

It's unused.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24773>

hasvk: Use the common GetPhysicalDeviceFeatures2 implementation

Reviewed-by: Julia Tatz <tatz.j@northeastern.edu>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24773>

nir/algebraic: optimize u2u32(a >> 32)

fossil-db (navi21):
Totals from 352 (0.44% of 79330) affected shaders:
Instrs: 271816 -> 271240 (-0.21%); split: -0.28%, +0.07%
CodeSize: 1546520 -> 1544448 (-0.13%); split: -0.23%, +0.09%
SpillVGPRs: 832 -> 827 (-0.60%); split: -1.08%, +0.48%
Latency: 4037120 -> 4021748 (-0.38%); split: -0.41%, +0.03%
InvThroughput: 1369540 -> 1362066 (-0.55%); split: -0.59%, +0.04%
VClause: 6476 -> 6471 (-0.08%); split: -0.12%, +0.05%
SClause: 6798 -> 6794 (-0.06%)
Copies: 44828 -> 44630 (-0.44%); split: -0.89%, +0.45%
Branches: 8845 -> 8844 (-0.01%); split: -0.05%, +0.03%
PreSGPRs: 14684 -> 14659 (-0.17%)

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25409>

nir/lower_int64: fix find_lsb(0)

If the high 32 bits were zero, this would be umin(find_lsb(lo), 31). This
evaluates to 31 if lo is also zero, instead of -1.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Fixes: 9293d8e64bc7 ("nir: Add find_lsb lowering to nir_lower_int64.")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25409>

tu/kgsl: Fix bitfield of DITHER_MODE_MRT6

The enum `adreno_rb_dither_mode` needs 2 bits, change the `high` to 13 to make sure 2 bits are used for `DITHER_MODE_MRT6`

Fixes: e03259974e2f2c8e8c0295f8dab56d88e7bd896c ("freedreno: Generate headers from xml files")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25439>

anv: Print warning that Xe2 is not supported rather than failing

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25411>

anv: Build for Xe2

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25411>

anv: Disable Ray Tracing on xe2 until our compiler supports Xe2 RT

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25411>

anv/blorp: Use anv_genX to set device->blorp.exec

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25411>

iris: Build for Xe2

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25411>

iris: add missing workaround for 3DSTATE_LINE_STIPPLE

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: mesa-stable
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25425>

anv: add missing workaround for 3DSTATE_LINE_STIPPLE

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: mesa-stable
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25425>

anv/iris: widen Wa_14015946265 to Gfx11+

We missed out that ICL+ added a programming requiring a CS_STALL.

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/25426>

tu: Expose VK_KHR_maintenance5

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25001>

freedreno/ci: Skip dEQP-VK.info.device_extensions

Copied from anv and radv.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25001>

tu: Add maintenance5 properties

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25001>

tu: Check for DEVICE_LOST in vkGetEventStatus()

Required by maintenance5.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25001>

tu: Support VkPipelineCreateFlags2CreateInfoKHR

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25001>

tu: Use new buffer usage flags

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25001>

tu: Implement vkGetRenderingAreaGranularityKHR

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25001>

tu: Implement vkGetImageSubresourceLayout2KHR and vkGetDeviceImageSubresourceLayoutKHR

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25001>

tu: Implement vkCmdBindIndexBuffer2KHR

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25001>

tu: Allow VK_WHOLE_SIZE in tu_CmdBindVertexBuffers2EXT pSizes

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25001>

tu/clear_blit: Allow VK_REMAINING_ARRAY_LAYERS as layerCount

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25001>

tu/clear_blit: Fix staging image view layer count

This seems to be a mistake that didn't affect anything.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25001>

freedreno/fdl: Support PIPE_FORMAT_R5G5B5A1_UNORM on a6xx

This is just a different swap of the same format.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25001>

tu: Support clearing A8_UNORM

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25001>

freedreno/fdl: Use A8_UNORM HW format for sampling

We've already been using this for 3d blits on turnip and it works fine.
There's no need to emulate it, and it won't work on turnip where we
can't swizzle the border colors.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25001>

zink: Rework color clamping and conversion

Before this commit, zink_format_clamp_channel_color() ignored the format
swizzle, so it was assuming that for emulated formats like alpha,
alpha-luminance etc. that the color had already been swizzled to match
the internal format rather than the emulated one. It's somewhat confusing
that passing in e.g. A8_UNORM actually means R8_UNORM, and led to a bug
when using VK_FORMAT_A8_UNORM for texture border colors because we
didn't swizzle it back. It also wouldn't have worked for media formats
like R10X6G10X6 due to the void channel in the middle.

In order to fix this, we need to untangle the mess in its users.

For convert_color() used when clearing, this means we now need to clamp
and then swizzle instead of swizzle and then clamp, and we can drop the
hack for A8_UNORM.

For texture border colors, the state tracker duplicates colors for the
emulated formats to help drivers, which zink was previously relying on,
but fixing zink_format_clamp_channel_color() breaks this because it
assumes that those duplicated colors are useless and clamps them.
However, because we know the format we can just swizzle the border color
ourself, which convert_color() was already doing. So, we pull that out
into a common zink_convert_color() function that handles both clamping
and format emulation, and have both clearing and border color handling
use it.

This fixes A8_UNORM in turnip+zink once we enable it.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25001>

v3dv: allow headless device without display device

When trying to create a headless Vulkan instance without a display
device available, device enumeration would fail.
Part of the physical device creation code already accounts for a missing
display device, allow device enumeration to continue in that case too so
headless instances can be created.

Signed-off-by: Erico Nunes <nunes.erico@gmail.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25313>

pvr: Force compile error on GNU void pointer arithmetic

This GNU extension caused a bug where a void pointer was modified then
interpreted as uint32_t* (instead of the other way round). Force a
compile error to (hopefully) prevent this from happening again.

Signed-off-by: Matt Coster <matt.coster@imgtec.com>
Reviewed-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25327>

pvr: Don't rely on GNU void pointer arithmetic

Besides being not standard C, one instance (in pvr_cmd_buffer.c) was a
bug caused by adding-then-casting, which would likely have been caught
if void pointer arithmetic were not allowed.

All instances detected by -Wpointer-arith have been fixed here.

Signed-off-by: Matt Coster <matt.coster@imgtec.com>
Reviewed-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25327>

aco: implement 64bit div find_lsb

This can be selected for divergent subgroupBallotFindLSB.

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25407>

aco: fix p_extract with v1 dst and s1 operand

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Fixes: f14023666ca ("aco: Allow p_extract to have different definition and operand sizes.")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25403>

aco/waitcnt: add print helpers

These may be useful in the future.

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/25373>

aco/waitcnt: replace wait_cnt::*_cnt with booleans

Previously, a loop could be revisited until a counter reaches it's
maximum:
loop {
   store()
}
Each visit of that loop would increase vs_cnt until it reaches max.

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/25373>

zink: lower vec8/16

Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25330>

nir: add nir_lower_alu_vec8_16_srcs pass

This pass is useful for vector based backends as we might end up with alu
instructions referencing vec8/vec16 values even though being vec4 or
smaller themselves.

This new pass intents to clean up any use of vec8/vec16 sources other
passes won't.

Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25330>

vc4: Fix mask RGBA validation at YUV blit

Solves regression on video players using GPU for
video decoding that just displays the video in green.

Fixes: d13da7782cd80 ("vc4: call blit paths in chain")
Reviewed-by: Juan A. Suarez <jasuarez@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25412>

ci/docs: drop extra overwritten rules

Every job re-defines its own rules, so there is no point in including these here.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25307>

iris: use intel_needs_workaround for Wa_14014414195 part 2

Commit 3ec953ed755 left couple of extra things, these changes are
based on a patch from Francisco Jerez.

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25306>

intel/compiler: Don't store stage name and abbrev

Those are used in the failure paths and are easily retriavable from the
stage itself, so no need to store them.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Sagar Ghuge <sagar.ghuge@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25367>

llvmpipe: reset viewport_index_slot in fb bind

I hit a problem running a set of lvp caselists with
dEQP-VK.transform_feedback.simple.draw_indirect_endqueryindexed_streamid_0_16,Crash

This was crashing due to
 assert(setup->viewport_index_slot < 0);
in try_update_scene_state

This was because a previous draw had set viewport index slot to 2,
but a clear then draw sequence never resets it, so the clear gets
a scene, and when the subsequent flush happens for that scene,
the viewport_index_slot is never updated. It only gets updated on
draws.

This just resets it as lp_setup_update_state will always pick up
the correct one for the next draw.

Cc: mesa-stable
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25387>

rusticl/mesa/context: fix clear_sampler_views

Even though drivers are required by documentation to handle the samplers
being NULL, st/mesa changed enough so it always sets the pointer.

As being similiar to st/mesa is more important than following docs, we
simply mimic st/mesa here and meet the expectations of drivers.

In the future we want to track the set state similiar to st/mesa, so
`clear_sampler_views` will probably go away and we'll just update what
we'll need.

Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: @LingMan <18294-LingMan@users.noreply.gitlab.freedesktop.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25363>

svga: sync resource content from backing resource before image upload

When a backing resource is created for a render target view when the
same resource is currently bound to a shader resource view, the content
update back to the original resource happens when the associated render
target view is unbound. But state update only happens at clear or draw
time. So if TexSubImage happens after BindFrameBuffer and before Draw,
the original texture resource that is mapped to for subimage update
would not have been updated. As a matter of fact at the subsequent state
update at the next draw, the render target views will be updated, the
content from the previous backing resource will be propogated to the
original resource, hence overwriting the changes from the last TexSubImage.

To fix the problem, this patch validates the texture resource, updates
any pending changes from the backing resource before transfer map upload
occurs.

Fixes the rendering issue demonstrated from the fbo_texsubimage_update trace

Reviewed-by: Martin Krastev <krastevm@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25386>

panfrost: advertise YUV formats for valhall

Signed-off-by: Italo Nicola <italonicola@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24479>

panfrost: use centered YUV chroma siting

This is likely a more reasonable default than co-sited samples, and will be
needed for valhall.

Signed-off-by: Italo Nicola <italonicola@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24479>

panfrost: prepare v9+ to support YUV sampling

Signed-off-by: Italo Nicola <italonicola@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24479>

panfrost: rename _needs_multiplanar_descriptor to _is_yuv

This is only really used to condition on YUV formats, even on v7, but
moreso now on v9+.

Signed-off-by: Italo Nicola <italonicola@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24479>

pan/genxml: add Width/Height fields to v9+ Plane descriptor

Signed-off-by: Italo Nicola <italonicola@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24479>

llvmpipe/fs: fix regression in sample mask handling from tgsi removal.

This got the bits the wrong way,
dEQP-VK.rasterization.frag_side_effects.color_at_beginning.sample_mask_after

Fixes: a63c2daf7ad7 ("llvmpipe/fs: start using nir info in some places.")
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25396>

freedreno: Add Adreno 643

Just another funny speedbin of A635

Signed-off-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25408>

freedreno: Include speedbin fallback in 690 chipid to fix probing

The kernel exposes CHIP_ID, which consists of:

[63:48] const 0 / reserved
[47:32] speedbin (default 0xffff)
[31:0 ] actual gpu chip ID

Fix the 690 chip id to make it probe correctly.
Keep the existing entry for the downstream KGSL driver.

This is essentially the same as Commit 6067aba9e6db ("freedreno: Include
speedbin fallback in 740 chipid to fix probing"), except for the older
core.

Fixes: 210c6c11cc61 ("freedreno+tu: Add a690 support")
Signed-off-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25408>

docs/meson: drop mention that our meson is ready

It's been ready for a very long time, and scons & autotools are long
gone with only meson left, so this note doesn't make much sense anymore.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25323>

ci: unify container and build jobs rules

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25294>

intel/pci_ids: Consistently use lowercase

The pci ids used a mixture of uppercase and lowercase, even in the same
file. The inconsistency introduces unnecessary complexity in regular
expressions.

Diff made in vim with `%s/\v(CHIPSET\()([^,]+)/\1\L\2\E/`.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>`
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25257>

meson/rusticl: add sha1_h

This ensures this file is generated.

Fixes: 20c90fed5a0 ("rusticl: added")
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25400>

radv/ci/vkcts-navi10: catch all the line-related flakes

Signed-off-by: Martin Roukala (né Peres) <martin.roukala@mupuf.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25395>

radv/ci/vkcts-navi21: document more flakes

Signed-off-by: Martin Roukala (né Peres) <martin.roukala@mupuf.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25395>

radv/ci/vkcts-navi21: catch all the line_stipple_(enable|params) flakes

There are so many of these tests, so let's just write a regular
expression to catch them all.

Signed-off-by: Martin Roukala (né Peres) <martin.roukala@mupuf.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25395>

gallium/auxiliary/vl: Only map the shader constants buffer in render

Don't map the buffer in vl_compositor_set_csc_matrix.
This avoids mapping the buffer twice with compute shaders.

Acked-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Thong Thai <thong.thai@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25258>

gallium/auxiliary/vl: Fix YUV to RGB bob compute shader deinterlacing

This reverts commit ef0d92459c3f and instead removes the y coordinate
divide by two in weave shaders. The scale ratio now works correctly
with interlaced input.

Also remove the YUV to YUV bob shader and reuse the progressive shader
for bob deinterlacing, same as in YUV to RGB case (video_buffer shader).

Acked-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Thong Thai <thong.thai@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25258>

gallium/auxiliary/vl: Use chroma offset in YUV to RGB weave compute shader

Acked-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Thong Thai <thong.thai@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25258>

gallium/auxiliary/vl: Add RGB to YUV compute shader

Compared to fragment shader that was being used before, it supports
chroma location and fixes wrong color at right/bottom edge when scaling.

Acked-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Thong Thai <thong.thai@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25258>

mesa: Ensure that the baselevel will never exceed the maximal supported number

GL_TEXTURE_BASE_LEVEL is only supported up to MAX_TEXTURE_LEVELS

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25332>

radv: declare shader_query_state for mesh/task shaders

This will control whether mesh/task queries is enabled.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25331>

radv: enable lowering of mesh/task shader queries when enabled

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25331>

radv: adjust lowering of intrinsic queries for mesh/task shaders

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25331>

radv: add GDS counters offset for mesh/task queries

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25331>

ac/nir: add lowering for task shader queries

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25331>

ac/nir: add lowering for mesh shader queries

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25331>

nir: rename atomic_add_gs_invocation_count_amd to make it more generic

It will be re-used to implement mesh/tash shader invocations queries.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25331>

mesa: Fix zeroing of new ParameterValues array entries when growing

On non-Windows OSes, align_realloc is the os_realloc_aligned() from
src/util/os_memory_aligned.h, which doesn't use realloc internally.
Instead, it uses os_malloc_aligned() and memcpy's over the old data,
which is why it needs an "old size" (unlike normal realloc).

In _mesa_reserve_parameter_storage, the call to align_realloc above
passes (oldValNum * sizeof(gl_constant_value)) as the old size, which
is all the actual data.  The actual allocation size of the array may
be larger (in fact, we allocate 16 extra components), which is tracked
in SizeValues.  After realloc, we memset to zero starting at the old
allocation size, to the new allocation size.

This would work if it were a real realloc.  However, because we actually
malloc + memcpy and only copy the previous /data/, not the allocated
size, and then memset from the old /allocated size/, our new copy will
have the spaces between the old data and the old allocation size neither
copied nor memset, leaving them as uninitialized garbage memory.

These values then get written to the shader cache, meaning that if you
compile the same shader multiple times, you may get different shader
cache entries.  This is bad for reproducible, deterministic compiles.

While at it, we also memset to zero in _mesa_add_parameter, as this
looks like another place where memset-to-zero is missing.

To reproduce this error, one can run shader-db:

$ MESA_SHADER_CACHE_DIR=a ./run -b shaders/godot3.4/49-28.shader_test
$ MESA_SHADER_CACHE_DIR=b ./run -b shaders/godot3.4/49-28.shader_test

and see an occasional difference in the end of the ParameterValues
array, where there's a padding gap between the last two elements that
was never zero-initialized.

Thanks to Mark Janes for discovering this and tracking it down together!

Cc: mesa-stable
Reviewed-by: Mark Janes <markjanes@swizzler.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25316>

amd/llvm: fix build with LLVM 18

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25388>

gallivm: fix build with LLVM 18

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25388>

nvk: Don't store the descriptor pool BO in the set

Instead, store an address, size, and mapped pointer.  This is more
consistent with how things like images work.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25357>

nvk/drm: Split exec as needed for large command buffers

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25357>

nvk/drm: Restructure nvk_queue_submit_drm_nouveau()

Now that we don't need the lock, we can return directly.  Also, now that
we don't have the old UAPI, we can clean things up and make the whole
function make a bit more sense.  Also, drop some pointless braces while
we're just moving code around.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25357>

nvk: Get rid of the tiled memory allocation helpers

These existed entirely to support shadow memory for VkImage cases where
we needed tiling.  Now that we have VM_BIND, these are no longer used so
we can drop the wrappers and just implement VkAllocate/FreeMemory again.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25357>

nvk: Drop the device-level mutex

This existed to let us lock the memory_objects list and for handling
BO-based vk_sync waits.  We don't have either of these things anymore so
there's no need for a device-level lock.  We already have fine-grained
locks around the data structures that need them.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25357>