AlexIndustrial/mesa
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
166,782 Commits 1 Branch 0 Tags
6747d301557154584cccd25fecdb8c8686f5932f
Commit Graph

166782 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
Rob Clark 6747d30155 freedreno: Add seqno helper 
It is a pretty common pattern to allocate a non-zero sequence # for
lightweight checking if an object is the same, changed, for use in cache
keys, etc.  (And also pretty common to forget to handle the rollover
zero case.)  Add a helper for this.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21274>
 2023-02-16 19:57:13 +00:00
Rob Clark 8f2b22ba66 freedreno: Drop batch lock 
Now that we are not tracking cross-context batch dependencies, there is
no scenario where one context could trigger flushing another context's
batch.  So we can drop the batch lock intended to protect against this.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21274>
 2023-02-16 19:57:13 +00:00
Rob Clark 9a6de00e98 freedreno/batch: Stop tracking cross-context deps 
The app is expected to provide suitable cross-context synchronization
(fences, etc), so don't try to do it's job for them.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21274>
 2023-02-16 19:57:13 +00:00
Rob Clark a4b949fe61 freedreno: Avoid taking screen lock 
Avoid taking screen unlock for batch unref.  Instead just split the
destroy fxn into locked and unlocked variants.  That way we only end
up taking the screen lock on final unref but avoid it in the common
case.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21274>
 2023-02-16 19:57:13 +00:00
Rob Clark 35fc1595b3 freedreno/a6xx: Pre-compute PROG related LRZ state 
PROG state mostly just disables various LRZ related flags, which can
be handled as a simple mask.  The exception is ztest mode, which is
either overriden by PROG state, or we use the all 1's value (which
isn't valid from hw standpoint) to signal that it needs to be computed
at draw time, which fortunately fits in with the bitmask approach.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21274>
 2023-02-16 19:57:13 +00:00
Rob Clark c938101bb5 freedreno: Move FD_MESA_DEBUG cases out of draw_vbo 
If the debug options are enabled, just plug in a debug version of
draw_vbo with the additional checks.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21274>
 2023-02-16 19:57:13 +00:00
Rob Clark 8942f4b734 freedreno: Move blend out of dirty-rsc tracking 
This was not doing any actual resource tracking, just updating
gmem_reason.  And furthermore, a6xx+ doesn't care about the bits
it was setting.  So move this to per-gen backend for the gens that
need it, and avoid setting FD_DIRTY_RESOURCE when FD_DIRTY_BLEND
is set.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21274>
 2023-02-16 19:57:13 +00:00
Rob Clark 67d4bc7be4 freedreno/a6xx: Remove tex-state refcnting 
Now that we use a flag to trigger the tex state invalidation coming from
other contexts, we can drop the refcnt'ing.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21274>
 2023-02-16 19:57:13 +00:00
Rob Clark cfd4721ee0 freedreno/drm: Make rb refcnt non-atomic 
Now that the one special case where multiple threads could race to
ref/unref, we can go back to using non-atomic refcnts.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21274>
 2023-02-16 19:57:13 +00:00
Rob Clark f91bcd2455 freedreno/a6xx: Do tex-state invalidates in same ctx 
If a resource invalidate is triggered by a different ctx (potentially on
a different thread) simply flag that the tex state needs invalidation,
but defer handling it to the ctx that owns the tex state.

This will let us remove atomic refcnt'ing on the tex state, and more
importantly atomic refcnt'ing on the fd_ringbuffer (as this was the one
special case where rb's could be accessed from multiple threads).

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21274>
 2023-02-16 19:57:13 +00:00
Rob Clark e7993d68e2 freedreno/a6xx: Multi-draw support 
Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21274>
 2023-02-16 19:57:13 +00:00
Rob Clark cc31997f1b freedreno/a6xx: Split out flush_streamout() helper 
Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21274>
 2023-02-16 19:57:13 +00:00
Rob Clark 911d67bdad freedreno/a6xx: Drop unused return 
Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21274>
 2023-02-16 19:57:13 +00:00
Rob Clark c4e2e821a2 freedreno: Push num_draws down to backend 
Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21274>
 2023-02-16 19:57:13 +00:00
Rob Clark 6bfee9e669 freedreno: Account for multi-draw in num_draws 
Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21274>
 2023-02-16 19:57:13 +00:00
Daniel Schürmann f6251b21f9 radv/rt: don't hash maxPipelineRayRecursionDepth 
The stack size has no effect on the generated shader anymore.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21159>
 2023-02-16 19:37:25 +00:00
Daniel Schürmann 8e718c5b63 radv/rt: use dynamic_callable_stack_base also for static stack_sizes 
This patch also removes rt_pipeline->dynamic_stack_size and replaces
it by checking for rt_pipeline->stack_size == -1u.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21159>
 2023-02-16 19:37:25 +00:00
Daniel Schürmann 2649a1f272 radv/rt: introduce and set rt_pipeline->stack_size 
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21159>
 2023-02-16 19:37:25 +00:00
Daniel Schürmann b338d59047 radv: unconditionally enable scratch for RT shaders 
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21159>
 2023-02-16 19:37:25 +00:00
Daniel Schürmann aa362b4b6f radv: rename shader_info->cs.uses_sbt -> shader_info->cs.is_rt_shader 
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21159>
 2023-02-16 19:37:25 +00:00
Konstantin Seurer 72d9604db0 radv: Clean up dynamic RT stack allocation 
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21159>
 2023-02-16 19:37:25 +00:00
Sidney Just fc84c63e17 zink: Add missing features to the profile file 
Fixes: 2ea481b2f0 ("Zink: add Zink profiles file")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20920>
 2023-02-16 19:11:57 +00:00
Sidney Just 60e0322092 zink: add check for samplerMirrorClampToEdge Vulkan 1.2 feature 
This adds a check to advertise PIPE_CAP_TEXTURE_MIRROR_CLAMP_TO_EDGE when either the extension is present or the Vulkan 1.2 feature is enabled.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20920>
 2023-02-16 19:11:57 +00:00
Emma Anholt ed62eec58b hasvk: Fix SPIR-V warning about TF unsupported on gen7. 
It's supported now.

Fixes: d82826ad44 ("anv: Implement VK_EXT_transform_feedback on Gen7")
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21228>
 2023-02-16 18:11:44 +00:00
Emma Anholt 98455470ea hasvk: Silence conformance warning in CI. 
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21228>
 2023-02-16 18:11:44 +00:00
Emma Anholt 570acf5655 ci: Add a manual full and 1/10th hasvk CTS runs. 
These are manual since they're on a runner in my basement that sometimes
can go down, but it'll be nice to have this for throwing the rare hasvk MR
at.

Acked-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21228>
 2023-02-16 18:11:44 +00:00
Danylo Piliaiev be976e0aa6 ci/tu: Add 1/200 pass to test for stale reg usage 
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21226>
 2023-02-16 17:43:10 +00:00
Danylo Piliaiev 86f82d4224 docs/freedreno: Add info about stale reg stomper dbg option 
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21226>
 2023-02-16 17:43:10 +00:00
Danylo Piliaiev a66d9c815d turnip: Add debug option to find usage of stale reg values 
MESA_VK_ABORT_ON_DEVICE_LOSS=1 \
TU_DEBUG_STALE_REGS_RANGE=0x00000c00,0x0000be01 \
TU_DEBUG_STALE_REGS_FLAGS=cmdbuf,renderpass \
./app

To pinpoint the reg causing a failure reducing regs range could be
used for bisection. Some failures may be caused by multi-reg combination,
in such case set 'inverse' flag which would change the meaning of reg
range to "do not stomp these regs".

Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21226>
 2023-02-16 17:43:10 +00:00
Timur Kristóf 084d10a702 aco: Remove MTBUF zero operand. 
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21363>
 2023-02-16 17:16:34 +00:00
Timur Kristóf afdacf4dcc aco: Don't set scalar offset on buffer load instructions when it's zero. 
This helps generate slightly more optimal instructions.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21363>
 2023-02-16 17:16:34 +00:00
José Roberto de Souza e050a00b9f intel/common: Move i915 files to i915 folder 
Following the organization done in intel/dev and intel/vulkan.

Probably due to some rebase issue we had a duplicated copyright header
in intel_gem_i915.h that is being removed in here too.

Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21256>
 2023-02-16 16:24:36 +00:00
Mike Blumenkrantz 41286f100e vl/dri3: avoid deadlocking when polling deleted windows for events 
upcoming xserver releases will emit PresentConfigureNotify with this
flag set when a window is destroyed, ensuring drivers
don't poll infinitely and deadlock

Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>

Acked-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21339>
 2023-02-16 15:55:47 +00:00
Mike Blumenkrantz 819cbf329a vulkan/wsi: avoid deadlocking dri3 when polling deleted windows for events 
upcoming xserver releases will emit PresentConfigureNotify with this
flag set when a window is destroyed, ensuring drivers
don't poll infinitely and deadlock

fixes #6685

cc: mesa-stable

Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Acked-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21339>
 2023-02-16 15:55:47 +00:00
Mike Blumenkrantz 91de576a7f dri3: avoid deadlocking when polling deleted windows for events 
upcoming xserver releases will emit PresentConfigureNotify with this
flag set when a window is destroyed, ensuring drivers
don't poll infinitely and deadlock

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/116

cc: mesa-stable

Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>

Acked-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21339>
 2023-02-16 15:55:47 +00:00
Timur Kristóf 4621ffdec1 aco: Get rid of redundant load_vmem_mubuf function. 
Call emit_load directly from visit_load_buffer instead.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Acked-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21358>
 2023-02-16 15:29:37 +00:00
Timur Kristóf 74f1b77046 radv: Move VS input lowering to new file: radv_nir_lower_vs_inputs. 
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Acked-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21358>
 2023-02-16 15:29:37 +00:00
Timur Kristóf 450e173de0 ac/llvm: Change ac_build_tbuffer_load to take format and channel type. 
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Qiang Yu <yuq825@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21358>
 2023-02-16 15:29:37 +00:00
Timur Kristóf 0ae778ca59 ac/llvm: Fix ac_build_buffer_load to work with more than 4 channels. 
LLVM is unable to select instructions for num_channels > 4, so we
workaround that by manually splitting larger buffer loads.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Qiang Yu <yuq825@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21358>
 2023-02-16 15:29:37 +00:00
Timur Kristóf a2755fc203 ac/llvm: Fix buffer_load_amd with larger than 32-bit channel sizes. 
LLVM is unable to select instructions for larger than 32-bit channel types.
Workaround by using i32 and casting to the correct type later.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Qiang Yu <yuq825@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21358>
 2023-02-16 15:29:37 +00:00
Timur Kristóf b5b0ded4c1 ac/llvm: Remove "structurized" argument and instead check vindex. 
Change ac_build_buffer_load_common and ac_build_tbuffer_load so
the use structurized load when the vindex argument is not NULL.
Adjust callers to match the new behaviour.

This fixes the load_buffer_amd intrinsic with index source.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Qiang Yu <yuq825@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21358>
 2023-02-16 15:29:37 +00:00
Timur Kristóf 881c52ba19 ac: Port ACO's get_fetch_format to ac_get_safe_fetch_size. 
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Acked-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21358>
 2023-02-16 15:29:36 +00:00
Timur Kristóf 2e9f5aadd0 nir: Clarify comment above load_buffer_amd. 
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21358>
 2023-02-16 15:29:36 +00:00
Tapani Pälli f815daa285 mesa/st: refactor st_destroy_texcompress_compute condition 
What happened is that we did not call init but still called dtor, this
makes us do some unnecessary work.

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21309>
 2023-02-16 15:00:13 +00:00
Iago Toral Quiroga 93952916c6 v3dv: fix stencil view aspect selection of depth/stencil image 
Instead of fixing the swizzle to X001, we should compose this
swizzle with the image view's own swizzle.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21354>
 2023-02-16 14:47:39 +00:00
Guilherme Gallo eba566c854 ci/lava: Fix LAVA logs issues for Collabora jobs 
Since the Collabora LAVA update related to the downtime from
https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21119, the
LAVA logs from Collabora continued to use the hack for older versions
which digested some control characters, such as carriage returns acting
as newlines, which made it necessary to recover from split lines to make
Gitlab sections work in job logs as expected.

Collabora's LAVA instance now gives a more raw log output. It is
necessary to pay attention to newlines at the end of each log message,
which may cause double newlines when printed with Python built-in
`print` function. I decided to remove the repeating `\n` from the
received log messages to make them transparent to LogFollower users.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/8242

Signed-off-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21325>
 2023-02-16 13:08:41 +00:00
Guilherme Gallo c85f3fbbb7 ci/lava: Add LavaFarm class to find LAVA farm from runner tag 
LavaFarm is a class created to handle the different types of LAVA farms
and their tags in Mesa CI. Since specific jobs may require different
types of LAVA farms to run on, it is essential to determine which farm
the runner is running on to configure the job correctly.

LavaFarm provides an easy-to-use interface for checking the runner tag
and returning the corresponding LAVA farm, making it simple for Mesa CI
to configure jobs appropriately. By adding tests for LavaFarm, the team
can ensure that this class is functioning as expected, allowing for the
smooth execution of Mesa CI jobs on the correct LAVA farm.

The tests ensure that get_lava_farm returns the correct LavaFarm value
when given invalid or valid tags and that it returns LavaFarm.UNKNOWN
when no tag is provided. The tests use Hypothesis strategies to generate
various labels and farms for testing.

Example of use:
```
from lava.utils.lava_farm import LavaFarm, get_lava_farm

lava_farm = get_lava_farm()
if lava_farm == LavaFarm.DUMMY:
    # Configure the job for the DUMMY farm
    ...
elif lava_farm == LavaFarm.COLLABORA:
    # Configure the job for the COLLABORA farm
    ...
elif lava_farm == LavaFarm.KERNELCI:
    # Configure the job for the KERNELCI farm
    ...
else:
    # Handle the case where the LAVA farm is unknown
    ...
```

Signed-off-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21325>
 2023-02-16 13:08:41 +00:00
Guilherme Gallo ca110658b7 ci/lava: Move LAVA dependencies to pip 
Use requirements.txt and requirements-test.txt to organize better Python
dependencies related to LAVA.

Now LAVA tooling can use recent and fixed library versions.
And test-related libs will not trigger container rebuilding anymore.

Signed-off-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21325>
 2023-02-16 13:08:41 +00:00
Konstantin Seurer cf9e1b953c radv: Hash VK_PIPELINE_CREATE_RAY_TRACING_NO_NULL_* flags 
Found by inspection.

Fixes: 687a82d ("radv/rt: Handle no-null shader flags")
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21320>
 2023-02-16 11:42:25 +00:00
Michel Dänzer 3e9c131482 frontend/dri: Initialize callbacks in dri_swrast_kms_init_screen 
This was missed in the commit below.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/8232
Fixes: 7d5b1cd02c ("frontend/dri: move callbacks from the VTable into dri_screen, dri_drawable")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21333>
 2023-02-16 10:21:44 +00:00