In Vulkan we configure our integer RTs to clamp automatically, so with logic
operations we need to be careful and avoid overflows by discarding any bits
that won't fit in the RT component size.
Fixes remaining CTS test failures in:
dEQP-VK.pipeline.logic_op.*
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10801>
This avoids debug builds to assert crash. Components that don't exist
won't be used and will be eventually DCEd, so simply lower them to 0.
Fixes CTS tests like these in debug builds:
dEQP-VK.pipeline.logic_op.r8_uint.clear
dEQP-VK.pipeline.logic_op.r8_uint.and
dEQP-VK.pipeline.logic_op.r8_uint.and_reverse
dEQP-VK.pipeline.logic_op.r8_uint.copy
dEQP-VK.pipeline.logic_op.r8_uint.and_inverted
dEQP-VK.pipeline.logic_op.r8_uint.no_op
dEQP-VK.pipeline.logic_op.r8_uint.xor
dEQP-VK.pipeline.logic_op.r8_uint.or
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10801>
There is a regression that made it impossible to export gem
handles with write access.
That is, a client may export gem handles of each buffer plane, then
export dmabuf fds using these handles, and mmap these dmabuf in
a different process (this is what Chromium does).
After https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4861,
it became impossible as mmap resulted in EACCESS error as slightly
different approach was taken for exporting these gem handles.
This CL fixes exporting gem handles (which are exported from dmabuf
fds) by adding the DRM_RDWR flag.
Cc: mesa-stable
Fixes#3119
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10850>
There is a regression that made it impossible to export gem
handles with write access.
That is, a client may export gem handles of each buffer plane, then
export dmabuf fds using these handles, and mmap these dmabuf in
a different process (this is what Chromium does).
After https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4861,
it became impossible as mmap resulted in EACCESS error as slightly
different approach was taken for exporting these gem handles.
This CL fixes exporting gem handles (which are exported from dmabuf
fds) by adding the DRM_RDWR flag.
Cc: mesa-stable
Fixes#3119
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10851>
It's perfectly legal to declare multiple SSBOs that point to the same
binding/descriptor_set with different access mask. Currently, it will
always get the first one in the list that matches binding/desc_set
regardless of the access mask, but other variables might have different
access mask.
Fix this by being conservative if another variable uses the same
binding/desc_set because we can't get it reliably without adding
a new field to vulkan_resource_index.
This fixes rendering issues in Resident Evil Village with vkd3d-proton.
This bug has been uncovered by ("spirv: Don't remove variables used by
resource indexing intrinsics") because variables are no longer removed
No fossils-db changes.
Cc: 21.1 mesa-stable
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10692>
Would allow earlier faults instead of having corrupted cmdstream.
This was already done to Freedreno long ago in:
04aff7e4 "freedreno: make cmdstream bo's read-only to GPU"
Since private memory should be GPU writable it is now allocated
separately, instead of suballocation from now read-only cmdstream.
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10807>
Flush batches when a new batch accessing the same set of BOs comes in
instead of delaying that operation. This greatly simplifies the
dependency tracking logic and shouldn't hurt the perfs (it might even
improve the latency since jobs are now flushed earlier).
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10842>
The PAN_BATCH_DEBUG section was too wide, potentially leading to
double insertion of the same batch in the readers array. Let's get rid
of this already_accessed parameter and make sure
panfrost_batch_update_bo_access() is only called if the RW flags have
changed.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10842>
We don't want blit batches to interfere with subsequent draw calls, but
we want to keep them in the list of pending batches (so they gets flushed
when glFlush() is called), so calling panfrost_freeze_batch() is not an
option. Reset the ctx->batch pointer manually.
Note that there's no risk to have those batches re-used to queue
unrelated draws because we created a unique pipe_surface object per
blit call, and the batch logic use the surface pointers to calculate
the batch key hash.
Fixes: e1649e916d67 ("panfrost: Use pan_blit() when PAN_MESA_DEBUG=panblit")
Reported-by: Icecream95 <ixn@disroot.org>
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Tested-by: Icecream95 <ixn@disroot.org>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10842>
BOs in the cache are chronological, so we try oldest BOs first. That
means if we find the oldest BO is busy, likely every BO is busy, and we
should bail early. This dramatically reduces the useless cycles spent in
bo_wait.
I studied the BO cache of the following drivers, all of which handle
this correctly: iris, lima, etnaviv, freedreno, vc4, v3d, v3dv.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10794>
The atomic can be costly, so skip it. This should be safe because
we never have pipe_transfer as the only live reference of a buffer.
Drivers can't do this yet because something needs to hold the resource
reference while a transfer object is being passed via u_threaded_context.
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10662>
We don't want to have to deal with vector phis in freedreno, because
vectors are always split/unsplit around vectorized instructions anyways,
and the stated reason for not scalarising them (it hurting coalescing)
won't apply to us because we won't be using nir_from_ssa. Add this
option so that we don't have to do the equivalent thing while
translating from NIR.
Reviewed-by: Rob Clark <robdclark@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10809>
