After the transition to vk_meta, all clear tests seem to pass.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29451>
Variable copy lowering might generate new IOs, so let's make sure
those are lowered before the IOs.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29451>
There's no real blocker here, so let's just allow it.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29451>
Compressed formats were disabled because of the different layout for
u_interleaved(compressed), where the 16x16 texel tile model for
non-compressed is replaced by a 4x4 block tile. This prevents us from
creating RGBA views of compressed images, which break copies.
The simple option is to keep compressed images linear. The more complex
one would be to patch coordinates/frag-coords on the shader side when
we're dealing with a compressed texture/color-attachment, but given
we're using vk_meta for copies, it's not something we can easily do,
so let's go for the first option and see if we want to revisit it
later.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29451>
When both the depth and stencil attachments are enabled, and the
underlying image is AFBC, it must be an combined Z24S8 format.
Make sure this is the case when initializing the fbinfo object.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29451>
Use vk_meta_resolve_rendering() to handle render-pass resolves.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29451>
Replace our custom panvk_meta stuff by something based on vk_meta.
That's less gen-specific code to maintain which will help with v10
support.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29451>
We move the modifier selection after the panvk_image initialization
so we have less VkXxx objects to retrieve, and split the 3 cases
(explicit modifier, explicit modifier list ordered by preference and
driver auto-selection based on the image attributes) to make things
clearer.
We also check formats against the pan_best_modifiers array which is
sorted by GPU order of preference and serve as an is_valid_modifier()
too.
Finally, we add the following new constraints on u-tiled/AFBC:
- don't allow u-tiled on images with compressed formats that might have
non-compressed views
- don't allow u-tiled or AFBC on 1D images
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29451>
Those helpers will be used to save/restore the command buffer
state when calling vk_meta helpers.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29451>
The command buffer logic is not supposed to modify these objects.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29451>
We are about to allow buffer memory allocation from the cmd_buffer
memory pool. In order to allow that, we need to tweak the NULL checks
done in panvk_buffer_{gpu_ptr,range}() to use the device address
instead of the bo field.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29451>
Needed image copies in Vulkan, where we create RGBA views of
compressed images to do the copy.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29451>
We do free the device object, but never call vk_device_finish() on
it.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29451>
It's usually clearer to describe what the error unwinding entrypoint
does than which type of error it's supposed to handle, because some
entrypoints can be called from different error paths.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29451>
Specifically:
- `alpine-build-testing`
- `fedora-release`
- `.meson-arm` (inherited by all the arm build jobs)
- `debian-arm64` (because it overrides the inherited list)
- `debian-vulkan`
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30658>
The Vulkan loader tries each driver until it finds one that's accepting
the device. Don't complain about panvk being non-conformant until we're
sure the device is for us.
Reported-by: Rob Clark <robclark@freedesktop.org>
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Acked-by: Rob Clark <robclark@freedesktop.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30756>
These errors are nicer to read than the ones from live variable analysis.
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/30440>
Otherwise, there is no definition of this temporary. Fixes fail in future
validation.
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/30440>
The CI-wide default is 20min, so if we set a 20min job timeout here, we
can't get the results of our jobs when they timeout.
Instead of setting the test timeout to 15min, which would be too short
for some jobs, leave it at 20min (but be explicit about it and protect
against a future change of that default), and bump the job timeout by
5min to allow for results to be uploaded.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30694>
and we would prevoisly incorrectly apply MergeSplits in this case
That is,
a, b = split(x)
y = merge(b, a)
could be incorrectly optimized to
y = x
before this commit
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24984>
We were currently emitting logical atomic instructions with a packed
destination region for sub-dword LSC atomics, along the lines of:
> untyped_atomic_logical(32) dst<1>:HF, ...
However, these instructions use an LSC data size D16U32, which means
that the 16b data on the return payload is expanded to 32b by the LSC
shared function, so we were lying to the compiler about the location
of the individual channels on the return payload, its execution
masking, etc. This is why the hacks that manually set the
'inst->size_written' of the instruction were required.
In some cases this worked, but any non-trivial manipulation of the
instruction destination by lowering or optimization passes could have
led to corruption, as has been reproduced in deqp-vk during
lower_simd_width() for shaders that use 16-bit atomics in SIMD32
dispatch mode.
Note that LSC sub-dword reads aren't affected by this because they use
raw UD destinations and specify the actual bit size of the operation
datatype as the immediate SURFACE_LOGICAL_SRC_IMM_ARG, which doesn't
work for atomic operations since that immediate specifies the atomic
opcode.
Instead, have the logical operation implement the behavior of 16-bit
destinations correctly instead of silently replacing the 16-bit region
with an inconsistent 32-bit region -- This is done by emitting the MOV
instructions used to pack the data from the UD temporary into the
packed destination from the lower_logical_sends() pass instead of from
the NIR translation pass.
Fixes: 43169dbbe5 ("intel/compiler: Support 16 bit float ops")
Reviewed-by: Sagar Ghuge <sagar.ghuge@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30683>
This improves drivers in the following ways:
* iris_hiz_exec() and crocus_hiz_exec() gets rid of the narrowly-used
update_clear_depth parameters.
* iris avoids fast-clearing if the aux state is CLEAR. crocus avoids
this too, but didn't actually need it in the first place.
* iris updates the value once per fast_clear_depth() call instead of
doing an update for each layer being cleared.
* anv now updates the clear value when transitioning from an undefined
layout instead of doing so on every fast-clear. This should be safer
because we don't perform state cache invalidates when changing the
clear value. So, existing surface states won't have any stale values.
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30520>
The next patch will be update the clear value. Move the stalling flush
to the top of fast_clear_depth() so that there are no users of the clear
value when it is replaced.
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30520>
We need to invalidate the state cache when updating the value in the
indirect clear color so that existing surface states can pick up the new
value.
Cc: mesa-stable
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30520>
this pbo shader works by iterating over the framebuffer size
and storing a value to an offset for each source pixel. if the
number of pixels being written out does not correspond to fragcoord
to the extent that certain source pixels are not written at all, however,
then this method should not be used in order to avoid giving broken results
cc: mesa-stable
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30689>
