When allocating descriptor sets, ensure we correctly calculate the new
size if the last binding is an inline uniform block with a variable
count.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Olivia Lee <olivia.lee@collabora.com>
Reviewed-by: Christoph Pillmayer <christoph.pillmayer@arm.com>
Fixes: 5fe5e317 ("panvk: advertise descriptor indexing on valhall")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35545>
maxInlineUniformTotalSize is derived from
MAX_INLINE_UNIFORM_BLOCK_DESCRIPTORS, but due to missing parenthesis
around the expression, the value calculated is incorrect.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Olivia Lee <olivia.lee@collabora.com>
Reviewed-by: Christoph Pillmayer <christoph.pillmayer@arm.com>
Fixes: 0440d1dd03 ("panvk: Advertise VK_EXT_inline_uniform_block")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35545>
The copy job was missing a dependency on the indirect job causing issues
on various "dEQP-VK.synchronization.*indirect*" and
"dEQP-VK.synchronization2.*indirect*" tests.
Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com>
Fixes: 8b4e515ca2 ("panvk: Implement indirect dispatch on JM")
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35639>
for various math helpers.
the clzll version isn't available since long long is 128-bit for CL.
Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35639>
We were assigning the scratch size of the whole lib... that was causing
very big TLS usage on draw indirect.
With this TLS usage is way lower now at the cost of running more pass
for every variants.
Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35639>
At some point CL C shader switched to a format hash approach and we
forgot to properly migrate here.
This should fix all printf again.
Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35639>
These copy the stats with any padding zero'd.
The layout is the same as the C struct, so it's still usable as one, if
the alignment is correct.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35583>
The radeonsi code has been improved to handle reference lists
better, port the code to radv and vulkan API.
This has been updated to for Vulkan
Reviewed-by: David Rosca <david.rosca@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35628>
This seems to canonicalize NaNs when the source format is a 16-bit
floating point format, likely because it implicitly converts to 32b.
Fixes OpenCL-CTS:
./test_basic imagearraycopy
./test_basic imagearraycopy3d
Signed-off-by: Rob Clark <rob.clark@oss.qualcomm.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35589>
If a driver is (or is contemplating) providing it's own fs_override then
it needs to know whether the blitter is going to use txf or not. So
split this logic out into a helper.
Signed-off-by: Rob Clark <rob.clark@oss.qualcomm.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35589>
Extract out a helper to check for f16 formats from turnip so it can be
used elsewhere.
Signed-off-by: Rob Clark <rob.clark@oss.qualcomm.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35589>
The GL frontend is perfectly good at lowering it like we do. Cuts out a
bunch of duplicate code.
We still have ucp_enables for the FS due to lowering of CLIPDIST to
discards in the FS.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8953>
More driver boilerplate we don't need. We do still have to go through the
transformation in st_atom_clip.c because the clip state can get used in
draw fallbacks. This revealed that lima had a dirty bit being set that
nobody was reading.
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8953>
The GL frontend handles it for us since c1442030ec ("vc4: lower clip
planes in st/mesa"). Now we should just be dropping the VS re-recompiles
we were doing, along with dead code.
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8953>
Decouple the .venus-rules definition from lavapipe to prepare for adding
hardware driver coverage for Venus.
Signed-off-by: Valentine Burley <valentine.burley@collabora.com>
Co-authored-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Co-authored-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35522>
The GitLab runner tags are case sensitive, and Flip-hatch's tag was
incorrectly lowercase. This prevented jobs from being picked up
by the runner.
Also remove an outdated note, as the device now appears to be stable.
Signed-off-by: Valentine Burley <valentine.burley@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35522>
Current LAVA jobs use dEQP instead of Android CTS, so there's no need
to include Android CTS in the rootfs. Removing both Cuttlefish and
Android CTS reduces the rootfs size from ~2GB to ~500MB.
Signed-off-by: Valentine Burley <valentine.burley@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35522>
Instead of including Cuttlefish in the core LAVA rootfs, make it an
optional LAVA overlay. This reduces the size of the core rootfs and
keeps Cuttlefish cached when the container is rebuilt, leading to
faster deploys.
Signed-off-by: Valentine Burley <valentine.burley@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35522>
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes: 252cac1e5c ("anv: avoid memory type changes with INTEL_DEBUG=noccs")
Tested-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35615>
This should help vk-icd-loader skip libraries of the wrong bit width.
Reviewed-by: Eric Engestrom <eric@igalia.com>
Acked-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Acked-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Acked-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35567>
It's ill-defined at best since it doesn't even initialize the
vk_object_base and its only use was NVK and that use is now gone.
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35624>
This struct gathers up all the sampler state from VkSamplerCreateInfo
and its pNext chain into a single struct. The struct is has no pointers
and has uses -Wpadded to ensure no holes. This means it's hashable and
mem-comparable. We also make give vk_ycbcr_conversion_state -Wpadded
because vk_sampler_state has a copy of vk_ycbcr_conversion_state
embedded in it.
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35624>
This removes a very old hack which will also allow us to enable DCC
for multiplanar formats eventually and to reduce the combined
image+sampler descriptor size from 96 to 48 on RDNA3+.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35457>
