Since b3133e250e ("gallium: add pipe_context::resource_release to
eliminate buffer refcounting") we need to take a reference for every bound
buffer object.
As we create image views on buffers, and kinda take partly reference
already just do it properly for now so we don't end up with
use-after-frees in drivers.
Fixes: dee9600a ("zink: eliminate buffer refcounting to improve performance")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37350>
The value read was incorrect.
This change was tested on palm, barts and cayman, it fixes all the variants
of the advanced-sso-simple test:
khr-gl4[56]/es_31_compatibility/shader_image_load_store/advanced-sso-simple: fail pass
khr-gles31/core/shader_image_load_store/advanced-sso-simple: fail pass
Cc: mesa-stable
Signed-off-by: Patrick Lerda <patrick9876@free.fr>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37225>
We are about to allow any type that is not FLOAT on v9+ at the
pan_afbc_format() level, but this regresses
dEQP-GLES31.functional.copy_image.non_compressed.viewclass_16_bits.*_snorm*
tests because of the clamping that's done on values that fall outside
the [-2^(b-1)+1, 2^(b-1)-1] range (b being the number of bits in the
SNORM component).
In order to fix that, we would have to use a _UNORM type when copying,
but:
1. There are many places where internal copies can happen and it's hard
to identify all of them
2. If we do it at the panfrost_blit_no_afbc_legalization() level, we
might do format re-interpretation that's not wanted by the gallium
layer
Given AFBC(SNORM) has not been supported so far, let's just go for the
simple solution and filter it out explicitly in panfrost_should_afbc().
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Christoph Pillmayer <christoph.pillmayer@arm.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37158>
We are about to extend pan_afbc_format() to support AFBC(Z32),
but the the AFBC(Z32) + S8 combination doesn't work on v7-, so
we need to filter AFBC(Z32) out in panfrost_should_afbc() on
v7-.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Christoph Pillmayer <christoph.pillmayer@arm.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37158>
Stencil-only attachments should go in the stencil view, not the
depth/stencil one.
Cc: mesa-stable
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37158>
Now it should return explicite modifier and plane number.
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31658>
To query if a DRI_IMAGE has separate plane buffer handle.
radeonsi's DCC compressed dri image has multiple planes
(compressed data and metadata). But they share the same
buffer with different offsets, so we only need one buffer
handle for them.
This attribute is used to distinguish single buffer multi
plane image from multi buffer multi plane image.
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31658>
The CML job runs into leaks in various tests, and the only way to
re-enable the job is to disable leak detection for now.
Signed-off-by: Valentine Burley <valentine.burley@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37330>
This fixes font rendering artifacts when Chromium/Firefox use Office365
PowerPoint set into fullscreen presentation mode.
Fixes: 5b951bcdd7 ("v3d: Enable Early-Z with discards when depth updates are disabled")
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37349>
EGL is going to really export each plane's offset/stride, panfrost
need to fix for passing the piglit test:
spec@ext_image_dma_buf_import@ext_image_dma_buf_import-export
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/13853
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Eric R. Smith <eric.smith@collabora.com>
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37247>
Reference the original code, each plane resource have
the same image plane array.
Fixes: 53e5e07c4b ("pan: Add the concept of modifier handler")
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Eric R. Smith <eric.smith@collabora.com>
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37247>
this sucks, but if the gallium expectation is that changing vertex state
always requires a vb update, vbuf can no longer deduplicate repeated
calls
Fixes: 1638d486ff ("gallium/u_threaded,st/mesa: add a merged set_vertex_elements_and_buffers call")
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37277>
many times (especially in viewperf), vertex buffers and draw buffers
have no prior access, which means incurring a heavy call to the barrier
function which will just return a no-op anyway
instead, have an inline helper to no-op these cases and just add the
expected access
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37277>
even doing this mechanism inside the barrier function doesn't yield
anywhere near the same gains (~15-20% in viewperf catia), mostly for
the draw buffers, so I'd assume it's just added register pressure
from the huge draw function
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37277>
Following the example of dri_sw_winsys, in order to support
importing client-allocated (u)dmabufs into GL renderers as well
as KMS.
This crucially allows Wayland compositors running on vkms + llvmpipe
to behave much more like on HW GPUs/display engines, making CI testing
more robust and intuitive.
Signed-off-by: Robert Mader <robert.mader@collabora.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37305>
The main intention is to allow to pass dmabuf handles, however as its
type differs depending on platform, just pass the whole whandle.
Also make use of it in llvmpipe_resource_from_handle(), in preparation
for the next commit.
Signed-off-by: Robert Mader <robert.mader@collabora.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37305>
