zink: add RADV to list of broken drivers for EXT_color_write_enable

ref #6185

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15509>

panfrost: Process scissor state earlier

Otherwise, if batch->scissor_culls_everything is set for a single draw,
every draw after it in the batch will be skipped because the new
scissor/viewport state will never be processed. Process scissor state
early in draw_vbo to fix this interaction.

We do need to be careful: setting something on the batch can only happen when
we've decided on a batch. If we have to select a fresh batch due to too many
draws, that must happen first. This is pretty clear in the code but worth noting
for the diff.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reported-by: Icecream95 <ixn@disroot.org>
Reviewed-by: Icecream95 <ixn@disroot.org>
Fixes: 79356b2e ("panfrost: Skip rasterizer discard draws without side effects")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/5839
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/6136
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15365>

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

Cc: mesa-stable
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15424>

virgl/ci: Add jobs for running trace tests on LAVA

Provide new jobs virgl-lava-traces and virgl-lava-traces-performance to
run piglit trace tests on Intel based LAVA runners.

Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Reviewed-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15208>

ci: Dynamically adjust LIBGL_ALWAYS_SOFTWARE for crosvm

For an increased flexibility in operation, do not set
'CROSVM_LIBGL_ALWAYS_SOFTWARE=true' when *not* using llvmpipe
Gallium driver.

Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Reviewed-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15208>

ci: Allow specifying any shell command via HWCI_TEST_SCRIPT

Interpret the value of HWCI_TEST_SCRIPT environment variable as a shell
command. This allows, for example, to provide additional environment
variables: HWCI_TEST_SCRIPT="VAR1=VAL1 VAR2=VAL2 /path/to/script"

Additionally, add the missing execute permission flags to
gtest-runner.sh script.

Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Reviewed-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15208>

ci: Provide consistent results location in LAVA

There is an out-of-sync approach regarding the location of the results
folder: some scripts refer to it via $CI_PROJECT_DIR/results, while
others just assume it is located in the current working directory.

Usually $PWD points to $CI_PROJECT_DIR, but in some cases this is not
the case, hence let's ensure the 'results' folder can always be found
in the current working directory.

Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Reviewed-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15208>

ci: Remove obsolete CROSVM_TEST_SCRIPT env var

This was used in the past for passing the path to a script to be
executed inside a crosvm instance. Currently, setting this variable
has no effect, hence remove it from generate-env.sh.

Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Reviewed-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15208>

ci: Add PIGLIT_REPLAY_LOOP_TIMES to generate-env.sh

PIGLIT_REPLAY_LOOP_TIMES is currently used to override the default
configuration when running virgl trace performance tests in LAVA.

Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Reviewed-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15208>

ci: Use script relative paths in crosvm-runner

For an increased portability, do not rely on 'CI_PROJECT_DIR' to
reference script relative resources and, instead, compute their
paths based on the crosvm-runner.sh invocation file path as
indicated by $0.

Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Reviewed-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15208>

ci: Make kernel image available in LAVA for KVM use cases

In order to run a VM (e.g. crosvm) through HWCI_TEST_SCRIPT on a LAVA
target, it's necessary to download a kernel image on the target device.

When HWCI_KVM is set to 'true', we can safely assume HWCI_TEST_SCRIPT
contains a command or the path to a script which expects the kernel
image to be available under /lava-files/${KERNEL_IMAGE_NAME}.

Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Reviewed-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15208>

ci: Load KVM kernel module for LAVA runners

If 'HWCI_KVM' enviroment variable is set, load the KVM kernel module
specific to the detected CPU virtualisation extensions: vmx for Intel
VT and svm for AMD-V.

As an additional optimization, handle HWCI_KERNEL_MODULES probing in the
main shell process instead of creating an unnecessary subshell.

Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Reviewed-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15208>

ci: Enable KVM_AMD and KVM_INTEL kernel modules

Build and deploy KVM kernel modules in rootfs image to be used for
running crossvm in LAVA environment.

Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Reviewed-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15208>

ci: Add crosvm runtime dependencies for LAVA

Provide the required packages in the rootfs image in order to allow
running crosvm inside LAVA environment.

Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Reviewed-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15208>

ci: Build crosvm for LAVA runners

This is the first step to add support for running crosvm inside LAVA.

Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Reviewed-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15208>

ci: Set CI_JOB_JWT_FILE to a fixed path outside /tmp

Having CI_JOB_JWT_FILE pointing to path on /tmp makes it difficult to be
managed in VM contexts (e.g. crosvm) because the /tmp mountpoint usually
refers to a local filesystem rather than the host one.

Additionally, there is another restriction for 'piglit-traces-test' job
to have the file available on the root filesystem.

To avoid amending all the jobs that might be affected, let's just set
the variable to a fixed path '/minio_jwt'. Note we also need to do this
in the 'variables:' section instead of 'before_script:' in order to be
able to reference it in CI job variables, e.g. PIGLIT_REPLAY_EXTRA_ARGS

Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Reviewed-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15208>

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

This shell script will be used in environments (e.g. LAVA) where bash
is not available, hence let's make sure it is POSIX compliant in order
to be able to execute on any modern shell interpreter.

Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Reviewed-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15208>

ci: Avoid altering EXTRA_CARGO_ARGS environment variable

Use a dedicated DEQP_RUNNER_CARGO_ARGS variable instead of
EXTRA_CARGO_ARGS in build-deqp-runner.sh to pass custom arguments when
invoking 'cargo install'.

This is to avoid modifications of EXTRA_CARGO_ARGS which might have
negative side-effects in the scripts which rely on this variable and
import build-deqp-runner.sh instead of executing it in a subshell.

Fixes: 8729c6e9818 ("ci: Support building and installing deqp-runner from source")
Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Reviewed-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15208>

radv,aco: lower image descriptor loads in NIR

fossil-db (Sienna Cichlid):
Totals from 2926 (1.80% of 162293) affected shaders:
Instrs: 2315110 -> 2306644 (-0.37%); split: -0.37%, +0.00%
CodeSize: 12581592 -> 12546588 (-0.28%); split: -0.28%, +0.00%
VGPRs: 130216 -> 130208 (-0.01%)
SpillSGPRs: 477 -> 474 (-0.63%); split: -5.03%, +4.40%
Latency: 29686188 -> 29678804 (-0.02%); split: -0.05%, +0.02%
InvThroughput: 6926545 -> 6926286 (-0.00%); split: -0.02%, +0.02%
SClause: 73761 -> 72996 (-1.04%); split: -1.16%, +0.12%
Copies: 144068 -> 137279 (-4.71%); split: -4.78%, +0.07%
Branches: 47466 -> 47483 (+0.04%); split: -0.01%, +0.04%
PreSGPRs: 118042 -> 117377 (-0.56%); split: -1.34%, +0.77%

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12773>

radv,aco: lower texture descriptor loads in NIR

fossil-db (Sienna Cichlid):
Totals from 39445 (24.30% of 162293) affected shaders:
MaxWaves: 875988 -> 875972 (-0.00%)
Instrs: 35372561 -> 35234909 (-0.39%); split: -0.41%, +0.03%
CodeSize: 190237480 -> 189379240 (-0.45%); split: -0.47%, +0.02%
VGPRs: 1889856 -> 1889928 (+0.00%); split: -0.00%, +0.01%
SpillSGPRs: 10764 -> 10857 (+0.86%); split: -2.04%, +2.91%
SpillVGPRs: 1891 -> 1907 (+0.85%); split: -0.32%, +1.16%
Scratch: 260096 -> 261120 (+0.39%)
Latency: 477701150 -> 477578466 (-0.03%); split: -0.06%, +0.03%
InvThroughput: 87819847 -> 87830346 (+0.01%); split: -0.03%, +0.04%
VClause: 673353 -> 673829 (+0.07%); split: -0.04%, +0.11%
SClause: 1385396 -> 1366478 (-1.37%); split: -1.65%, +0.29%
Copies: 2327965 -> 2229134 (-4.25%); split: -4.58%, +0.34%
Branches: 906707 -> 906434 (-0.03%); split: -0.13%, +0.10%
PreSGPRs: 1874153 -> 1862698 (-0.61%); split: -1.34%, +0.73%
PreVGPRs: 1691382 -> 1691383 (+0.00%); split: -0.00%, +0.00%

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12773>

radv,aco: lower buffer descriptor loads in NIR

fossil-db (Sienna Cichlid):
Totals from 75420 (46.47% of 162293) affected shaders:
MaxWaves: 1878200 -> 1879228 (+0.05%); split: +0.06%, -0.00%
Instrs: 54021103 -> 54141370 (+0.22%); split: -0.04%, +0.26%
CodeSize: 287813520 -> 288293352 (+0.17%); split: -0.04%, +0.21%
VGPRs: 3267576 -> 3266296 (-0.04%); split: -0.04%, +0.00%
SpillSGPRs: 10445 -> 10904 (+4.39%); split: -0.31%, +4.70%
SpillVGPRs: 1818 -> 1811 (-0.39%); split: -1.05%, +0.66%
Scratch: 955392 -> 954368 (-0.11%)
Latency: 563477854 -> 562131282 (-0.24%); split: -0.31%, +0.08%
InvThroughput: 111860104 -> 111553968 (-0.27%); split: -0.30%, +0.02%
VClause: 958432 -> 961415 (+0.31%); split: -0.34%, +0.65%
SClause: 1917415 -> 1926952 (+0.50%); split: -0.69%, +1.19%
Copies: 3812945 -> 3916758 (+2.72%); split: -0.27%, +2.99%
Branches: 1611235 -> 1612022 (+0.05%); split: -0.04%, +0.08%
PreSGPRs: 3095505 -> 3126580 (+1.00%); split: -0.06%, +1.07%
PreVGPRs: 2773011 -> 2773013 (+0.00%)

Most regressions seem to be because ACO's convert_pointer_to_64_bit()
can't be CSE'd with radv_nir_apply_pipeline_layout()'s
convert_pointer_to_64_bit(). This should be improved by later commits.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12773>

radv,aco: lower vulkan_resource_index in NIR

fossil-db (Sienna Cichlid):
Totals from 31338 (19.31% of 162293) affected shaders:
MaxWaves: 758634 -> 758616 (-0.00%)
Instrs: 26398289 -> 26378282 (-0.08%); split: -0.09%, +0.01%
CodeSize: 141048208 -> 140971060 (-0.05%); split: -0.07%, +0.01%
VGPRs: 1373656 -> 1373736 (+0.01%)
SpillSGPRs: 9944 -> 9924 (-0.20%); split: -0.24%, +0.04%
SpillVGPRs: 1892 -> 1898 (+0.32%); split: -0.95%, +1.27%
Latency: 308570144 -> 308528462 (-0.01%); split: -0.03%, +0.02%
InvThroughput: 57698072 -> 57684901 (-0.02%); split: -0.07%, +0.04%
VClause: 440357 -> 440602 (+0.06%); split: -0.02%, +0.08%
SClause: 974724 -> 973315 (-0.14%); split: -0.18%, +0.04%
Copies: 1944925 -> 1945103 (+0.01%); split: -0.11%, +0.12%
Branches: 799444 -> 799461 (+0.00%); split: -0.00%, +0.00%
PreSGPRs: 1619860 -> 1619233 (-0.04%); split: -0.05%, +0.02%
PreVGPRs: 1252813 -> 1252863 (+0.00%); split: -0.00%, +0.00%

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12773>

ac/llvm: implement nir_tex_src_{texture,sampler}_handle

nir_tex_src_{texture,sampler}_handle is either the actual descriptor as a
vec4/vec8 or a pointer passed to load_sampler_desc.

The sample index isn't adjusted using FMASK when these are used.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12773>

ac/llvm: remove deref chasing for tg4 integer workaround

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12773>

ac/llvm: implement nir_intrinsic_bindless_image_sparse_load

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12773>

ac/llvm: remove deref requirement for image fmask loads

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12773>

nir: allow bindless image/texture/sampler handles to be vectors

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12773>

radv: move radv_declare_shader_args() out of shader_variant_compile()

Declaring them earlier will allow us to access them in NIR.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12773>

ac/llvm: implement implement load_{scalar,vector}_arg_amd and load_smem_amd

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12773>

aco: implement load_{scalar,vector}_arg_amd and load_smem_amd

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12773>

nir: add load_{scalar,vector}_arg_amd and load_smem_amd intrinsics

load_smem_gcn is similar to load_global/load_global_constant, but it's
guaranteed to use SMEM and it's much easier to utilize the format's 32-bit
offset source.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12773>

pvr: Add a Vulkan driver for Imagination Technologies PowerVR Rogue GPUs

Co-authored-by: Rajnesh Kanwal <rajnesh.kanwal@imgtec.com>
Co-authored-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Co-authored-by: Simon Perretta <simon.perretta@imgtec.com>
Co-authored-by: Alexander Wasey <Alexander.Wasey@imgtec.com>
Signed-off-by: Frank Binns <frank.binns@imgtec.com>
Signed-off-by: Rajnesh Kanwal <rajnesh.kanwal@imgtec.com>
Signed-off-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Signed-off-by: Simon Perretta <simon.perretta@imgtec.com>
Signed-off-by: Alexander Wasey <Alexander.Wasey@imgtec.com>
Acked-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15243>

turnip: Disallow non-linear tiling when casting R8G8 to other fmts

R8G8 have a different block width/height and height alignment from other
formats that would normally be compatible (like R16), and so if we are
trying to, for example, sample R16 as R8G8 we need to demote to linear.

Follows the fix in Freedreno: b97e3bb2e1ffef2dbdb843411c222a7654c0c7a0

Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15465>

turnip: Force linear mode for non-ubwc R8G8 formats

Non-UBWC tiled R8G8 is probably buggy since media formats are always
either linear or UBWC. There is no simple test to reproduce the bug.
However it was observed in the wild leading to an unrecoverable hang
on a650/a660.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/5926

Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15465>

frontends/va: fix coredump as creating surface with VAConfigAttrib

As creating surface with VAConfigAttrib, checking if modifier from attrib list is null

Signed-off-by: shanshengwang <shansheng.wang@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15483>

zink: use the current compute shader, not the base one

this is the current variant and the one that should be used

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15501>

zink: create compute pipeline after updating shader variants

it turns out shader variants don't work if you generate them after you
determine which variant to use

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15501>

radeonsi: NIR equivalent of si_create_clear_buffer_rmw_cs()

Replaced the existing internal TGSI compute shader, which clears
a read-modify-write buffer, with its NIR equivalent. The disassembly
shader generated by the new NIR variant is identical to the previous
implementation. These changes remove the additional conversion step
from TGSI to NIR for the shader at runtime. Tested on a Navi 23 card.

Reviewed-by: Mihai Preda <mhpreda@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15356>

amd/ac_gpu_info: fix warning on fread unused result

fixes this warning:
ignoring return value of 'fread' declared with attribute 'warn_unused_result' [-Wunused-result]

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15502>

intel: Add INTEL_DEBUG=noccs alias for INTEL_DEBUG=norbc

When CCS compression first came out on Skylake, we referred to it as
"renderbuffer compression", or RBC for short.  However, that name has
long since fallen out of favor, and we refer to it as CCS nearly
everywhere.

This patch renames DEBUG_NO_RBC to DEBUG_NO_CCS inside the codebase
for clarity, and adds INTEL_DEBUG=noccs.  The legacy INTEL_DEBUG=norbc
name continues to work, because it's one line of code and having both
names makes our lives easier in the interim.

Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15447>

nir: Print divergence status of SSA values if analysis was ever run.

After running divergence analysis, we include "div" or "con" for each
SSA def's divergence/convergence status:

vec1 32 div ssa_35 = fddy ssa_34
vec1 32 con ssa_36 = fddy ssa_6.x

We omit this before the first time divergence analysis has been run.

Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15445>

zink: ci updates

these should be properly disabled now

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15500>

zink: ci updates

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15392>

zink: use the right query type for primitives generated

this should've always been clipping invocations, but I got scared because
then tests with rasterization_discard=1 fail and I didn't handle that instead

Acked-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15392>

zink: use EXT_color_write_enable to mask out primgen+rasterizer_discard output

by disabling color and depth write, the side effects of force-disabling discard can
be mitigated

fixes:
KHR-GL46.tessellation_shader.single.isolines_tessellation
KHR-GL46.tessellation_shader.tessellation_control_to_tessellation_evaluation.data_pass_through
KHR-GL46.tessellation_shader.tessellation_invariance.invariance_rule3
KHR-GL46.tessellation_shader.tessellation_shader_point_mode.points_verification
KHR-GL46.tessellation_shader.tessellation_shader_quads_tessellation.degenerate_case
KHR-GL46.tessellation_shader.tessellation_shader_quads_tessellation.inner_tessellation_level_rounding
KHR-GL46.tessellation_shader.tessellation_shader_tessellation.gl_InvocationID_PatchVerticesIn_PrimitiveID
KHR-GL46.tessellation_shader.vertex.vertex_spacing

Acked-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15392>

zink: add an alternate path for EXT_color_write_enable usage

for drivers where this is broken/missing, the same effect can be achieved
by feeding the renderpass a framebuffer with null/dummy attachments

Acked-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15392>

zink: use EXT_color_write_enable when possible

this should only verify that drivers aren't completely broken by
enabling it and have no other changes

Acked-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15392>

zink: disable color_write_enable on ANV

this breaks the driver

Acked-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15392>

zink: start a unified driver workarounds struct

eventually anything that checks driverID should be moved here for ease of
reference/updating

Acked-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15392>

zink: force disable rasterization discard if primgen query is active

this query requires rasterization to pass the clipping invocations stage,
which means discard is impossible

Acked-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15392>

zink: hook up EXT_color_write_enable

no functional changes

Acked-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15392>

nir/linker: set varying from uniform as flat

Flat varying can save some rasterization compute cost
and register needed by shader.

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15341>

lima: enable nir lower_varying_from_uniform

Mali GPU pass varying by memory, so enable this optimization.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15341>

nir/linker: disable varying from uniform lowering by default

This fixes performance regression for Specviewperf/Energy
on AMD GPU. Other GPUs passing varying by memory may choose
to re-enable it as need.

Fixes: 26046250437 ("nir/linker: support uniform when optimizing varying")
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15341>

asahi: Port driver to macOS 12.x ABI

There's lots of reshuffling required. Nothing "interesting", though.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15482>

asahi: Don't clobber clear colours

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15482>

asahi: Handle flushes of depth-only rendering

Not totally right but this should be a start.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15482>

asahi: Wire in u_transfer_helper

We need it for emulating packed depth/stencil as separate depth/stencil
resources, populating separate_stencil for us as required.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15482>

asahi: Generate IOGPU attachments dynamically

Take a pipe_framebuffer_state and go from there. We need some care to
handle separate stencil, but the logic is largely routine.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15482>

asahi: Add separate_stencil, internal_format fields

Currently unused, but will be set when u_transfer_helper is integrated
for emulating packed depth/stencil.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15482>

asahi: Add size field to slices

Needed to size attachments for the kernel... for some reason. We already
compute this; just save it.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15482>

asahi: Identify IOGPU_MISC data structure

This will be elaborated upon soon.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15482>

asahi: Add stencil buffer attachment type

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15482>

asahi: Identify IOGPU Internal Pipelines structure

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15482>

asahi: Identify aux framebuffer data structure

Total guess at the name.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15482>

asahi: Identify IOGPU Clear Z/S structure

Not sure on the details yet but identify and dump the data structure to start.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15482>

util: trim trailing space for files src/util/**/*

Using the following bash script doing that
```
cd src/util
find . -type f -print0 | xargs -0 -n1 sed -i 's/[ \t]*$//'
```

Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15093>

ci/anv: Changes from enabling 8/16bpp CCS more

- Fixes in dEQP-VK.dynamic_rendering.suballocation.multisample_resolve.*
- Fails in dEQP-VK.drm_format_modifiers.export_import.*

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15420>

Revert "anv: Disable CCS_E for some 8/16bpp copies on TGL+"

This reverts commit d68b2db89ca38f3f6c0d3ebd65945c471d864d58.

With this change, no regressions have been observed with the
dEQP-VK.synchronization* test group. There are regressions with
dEQP-VK.drm_format_modifiers.export_import.*, but those have been
root-caused to be test issues (see 3575).

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/6125
Fixes: 57445adc891 ("anv: Re-enable CCS_E on TGL+")
Reviewed-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/15420>

gallium/opencl: set OCL_ICD_FILENAMES with devenv

So that `meson devenv` also sets up OpenCL.

Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15471>

r300: set PVS_LAST_VTX_SRC_INST properly to last input read

From docs:
The PVS Instruction which uses the Input Vertex Memory for the last
time. This value is used to free up the Input Vertex Slots ASAP.
This field must be set to a valid instruction.

Right now it is set to the last instruction. When the last read is
inside a loop, set it on the outhermost ENDLOOP. This could in theory
help performance, but none of my usual benchmarks including GLmark,
Unigine Sanctuary or Lightsmark show any measurable performance difference.

Suggested in: https://gitlab.freedesktop.org/mesa/mesa/-/issues/6045

Signed-off-by: Pavel Ondračka <pavel.ondracka@gmail.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15252>

nir: fix nir_sweep for printf

I hit a memory corruption trying to implement printf for Rusticl

Signed-off-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15474>

ci: enable intel-clc on some platforms

We'll have to figure out the cross compiling strategy, in particular
for Android. But as it stands we can't have the target & host llvm
packages installed at the same time so we can't really compile it.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Acked-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13171>

ci: enable llvm on debian-release build

Needed for intel-clc.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Acked-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13171>

ci: add clang/spirv-tools/llvm-spirv packages to fedora container

Needed for intel-clc.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Acked-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13171>

intel/clc: disable tool prior to Gfx12.5 platforms

This tool is currently only aimed at Gfx version 12.5+ with
COMPUTE_WALKER. We could make it work on earlier platforms but they
require pushing gl_SubgroupInvocation and the CLC code is missing the
back-end compiler set-up bits for that.

v2: Commit description by Jason

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13171>

intel/clc: specify supported extensions

Having everything ever known to man is confusing our SPIRV parser.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13171>

intel/clc: allow producing SPIRV files

Useful to debug the parser.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13171>

intel/clc: allow multiple CL files to be compiled together

v2: use util_dynarray_append() (Jason)
    identation fixes (Jason)

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13171>

intel: Add a little OpenCL C compiler binary

v2: Fix up indentation (Marcin)
    s/gen/gfx/ (Marcin)
    Deal with fd closing in success/fail cases (Marin)

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13171>

meson: try to find clang-cpp before going through each module

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Acked-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13171>

intel/fs: handle inline data on OpenCL style kernels

This is for Gfx12.5 with the COMPUTE_WALKER::Inline Data payload. We
do this in a similar way to the compute kernels.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13171>

intel/kernel: Implement some Intel built-in functions

v2: Document mangled function names (Marcin)
    Fixup progress & metadata (Marcin)

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13171>

intel/compiler: Add code for compiling CL-style SPIR-V kernels

v2: simplify INTEL_DEBUG expressions (Marcin)

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13171>

intel/debug: Dump KERNEL source when INTEL_DEBUG=cs

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13171>

intel/nir: Add optimizations to help OpenCL-style kernels

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13171>

intel/nir/fs: replace COMPUTE || KERNEL by gl_shader_stage_is_compute()

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Caio Oliveira <caio.oliveira@intel.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13171>

nir/print: Add support for generic pointers

The way they're handled is that deref->modes is treated as a bitfield of
possible modes.  Variables are required to have a specific mode and
derefs with deref_type_var are as well.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13171>

glx/egl: improve dri null screen related error messages.

Convert from `failed to create dri screen` to more exact error message
for easier debugging

Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15480>

crocus: replace opencoded slab_zalloc

Acked-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15277>

d3d12: replace opencoded slab_zalloc

Acked-by: Emma Anholt <emma@anholt.net>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15277>

etnaviv: replace opencoded slab_zalloc

Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Acked-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15277>

freedreno: replace opencoded slab_zalloc

Reviewed-by: Rob Clark <robdclark@chromium.org>
Acked-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15277>

lima: replace opencoded slab_zalloc

Acked-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15277>

iris: replace opencoded slab_zalloc

Acked-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15277>

v3d: replace opencoded slab_zalloc

Reviewed-by: Juan A. Suarez <jasuarez@igalia.com>
Acked-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15277>

vc4: replace opencoded slab_zalloc

Reviewed-by: Juan A. Suarez <jasuarez@igalia.com>
Acked-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15277>

virgl: replace opencoded slab_zalloc

Acked-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15277>

zink: replace opencoded slab_zalloc

Acked-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Acked-by: Emma Anholt <emma@anholt.net>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15277>

r600: replace opencoded slab_zalloc

Acked-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15277>