Commit f64a7c1506 ("pan/layout: Check the wsi_layout consistency in
wsi_row_pitch_to_row_stride()") introduced stricter alignments on pre-v7
for linear/tiled imports. It turns out alignment constraints on pre-v7 are
weaker than they are on v7+, and this got lost in the refactoring.
Let's fix linear_or_tiled_row_align_req() helper so we can import images
that don't have their base address/row_pitch aligned on a cacheline, and
only cacheline alignment for images we create ourselves.
Also add unittests to make sure we don't regress that in the future.
With this being tested, we can also use
pan_linear_or_tiled_row_align_req() in offset_align_for_mod()
(test-layout.cpp) instead of open-coding it.
Fixes: f64a7c1506 ("pan/layout: Check the wsi_layout consistency in wsi_row_pitch_to_row_stride()")
Fixes: 916f75a2a6 ("pan/layout: Test WSI import behavior on all supported format/mods")
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Reviewed-by: Eric R. Smith <eric.smith@collabora.com>
Tested-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35555>
The blocksize of for R8G8_R8G8_xxx formats is 4 bytes, not 2 bytes.
Fixes: 4d9a4e8228 ("pan/image: Teach pan_image/layout about planar images")
Fixes: 916f75a2a6 ("pan/layout: Test WSI import behavior on all supported format/mods")
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Reviewed-by: Eric R. Smith <eric.smith@collabora.com>
Tested-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35555>
Right now the headers are not counted when we calculate the total slice
size of an AFBC(3D) image. Fix that by special-casing size_B
initialization for AFBC.
I couldn't get back to the original commit introducing this mistakes,
so I'm flagging for backport instead of adding a proper Fixes tag.
Backport-to: 25.1
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Reviewed-by: Eric R. Smith <eric.smith@collabora.com>
Tested-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35555>
Fail the image creation if the image layout initialization fails instead
of silently ignoring the problem.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35555>
Maximum texture dimension is 2^16, but we're limited by the 32-bit
fields that are used to pass strides/sizes in various descriptors.
Assuming RGBA32_FLOAT is the biggest format we support, that gives us a
16k-1 image size for 2D and cube map, and 512 for 3D.
Change our GetPhysicalDeviceImageFormatProperties2() implementation so
that smaller formats can still advertise bigger image sizes.
Fixes: d5ed77800e ("panvk: Fix GetPhysicalDeviceProperties2() to report accurate info")
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35555>
AFBC headers point to their tile with a 32-bit offset, meaning the last
header has to cross the entire body to reach its tile, so we can't
have a body size that's bigger than UINT32_MAX.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Reviewed-by: Eric R. Smith <eric.smith@collabora.com>
Tested-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35555>
We will soon make pan_image_layout_init() return false even for
non-explicit layout case, which will allow panfrost_can_create_resource()
to properly report when the resource size doesn't fix some HW limits.
But before we can do that, we need to let panfrost_resource_setup()
return an error instead of asserting() that pan_image_layout_init()
always return true. Add a new panfrost_resource_try_setup() helper
and define panfrost_resource_setup() as wrapper around it with the
existing assert(valid), and make panfrost_can_create_resource() use the
new helper.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Reviewed-by: Eric R. Smith <eric.smith@collabora.com>
Tested-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35555>
AFBC headers point to their tile with a 32-bit offset, meaning the last
header has to cross the entire body to reach its tile, so we can't
have a body size that's bigger than UINT32_MAX.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Reviewed-by: Eric R. Smith <eric.smith@collabora.com>
Tested-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35555>
If we store the temporary value in a [u]int32_t, the result might be
silently truncated, making the overflow check in the u_pack helpers
useless.
Make sure we use 64-bit fields as soon as the size is greater than 31
bits to prevent that.
It forces us to fix some types in the xml and helper arguments (sint
wrongly defined as uint) and add explicit casts on u32 subtraction
whose result is stored in an s32.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35555>
In most cases this has little to no effect, but AV1 decode on VCN5
is up to 40% slower when the context buffer is in GTT.
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35684>
We were pointing the new_node.next at the previous node but trying to
iterate over the nodes starting at the first one.
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35687>
The linked list will be used to store occlusion queries to signal after a
renderpass.
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35687>
The Vulkan spec is pretty clear and says:
"VUID-VkSamplerCreateInfo-None-04012
The maximum number of samplers with custom border colors which can be
simultaneously created on a device is implementation-dependent and
specified by the maxCustomBorderColorSamplers member of the
VkPhysicalDeviceCustomBorderColorPropertiesEXT structure"
So this check can be removed because it would be an application bug
otherwise.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35708>
Like Van Gogh, Mendocino is RDNA 2. This new job is intended to replace
the existing radeonsi-vangogh-glcts job, which couldn't run the full
test suite pre-merge.
Signed-off-by: Valentine Burley <valentine.burley@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35682>
The full job previously ran in the nightly pipeline, but its runtime is
within the recommended 10-minute limit reported by deqp-runner. This
allows us to run it pre-merge and remove the separate nightly job.
Signed-off-by: Valentine Burley <valentine.burley@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35682>
We can't actually enable MSAA for images with sample count 1, and
without MSAA active, the sample location machinery does not get used.
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35504>
In panvk we often need to allocate and free address ranges in the
address space represented by panvk_device::as::heap. In most
situations, panvk_device::as::lock has to be held while performing
an operation on panvk_device::as::heap, so there are a lot of
instances of exactly the same 3 locs locking the as::lock,
allocating or freeing a range in as::heap, and unlocking the lock.
These helpers are the less verbose replacements for these 3 locs
that should be used going forward.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35692>
Make the queue entry points forward to the implementations
appropriate for the queue's type.
Rename panvk_queue to panvk_gpu_queue.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35692>
We will soon have a panvk_bind_queue object to expose bind queues,
so let's modify the code to make this reasonably clean by letting
the CSF/JM backing keep their queue object opaque.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-By: Caterina Shablia <caterina.shablia@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35692>
It was absent when initialising a panfrost_resource from a winsys handle.
Signed-off-by: Adrián Larumbe <adrian.larumbe@collabora.com>
Fixes: 7da251fc72 ("panfrost: Check in sources for command stream")
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34224>
Extend the panfrost resource screen interface with a function that lets
the pipe resource the BO belongs to share a label pointer for debugging
purposes.
Then append this resource label to the existing driver tag and label the
backing kernel BO accordingly.
A panfrost_resource::user_label field was added to store the argument
user label. Because in the case of an AFBC-packed resource, user tagging
can happen before the packed BO is created, we need to store it
separately, so that all the necessary label fields are available before
a complete label can be given to the kernel driver.
Signed-off-by: Adrián Larumbe <adrian.larumbe@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34224>
At the moment, the string passed to glObjectLabel() stays only within the
confines of GL objects. It would be nice if that textual description could
be made known to the underlying drivers somehow.
Expand the pipe screen function interface to allow passing GL object labels
down to the UM driver that handles a particular pipe screen.
Not all GL objects have an associated GL resources, so this behaviour has
been implemented for those which are associated with at least one.
Signed-off-by: Adrián Larumbe <adrian.larumbe@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34224>
When creating a BO that backs a panfrost resource, append additional
information. To this end, add a new helper that composes it from
a base resource label and some parameters of interest.
To avoid dealing with situations in which a BO is referenced by multiple
resources, we only allow the BO creator to claim ownership at resource
creation time. There are, however, code paths in which a panfrost
resource's BO is assigned from an existing resource. In those cases, we
transfer ownership between resources.
In those cases, we assign the old BO a label in static memory without
passing it down to the kernel, because in a majority of times that BO
will immediately reach a refcnt of 0, leading to its destruction. That
way we avoid a useless round-trip to the kernel through the BO labelling
ioctl().
Signed-off-by: Adrián Larumbe <adrian.larumbe@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34224>
In preparation of a future commit in which UM will be able to pass
down a label from the GL layer and append it to the existing BO's driver
label, we want to disallow resource handling of imported BO labels.
To this end, add a boolean flag to the panfrost_resource definition so
that future commits can take this into account when managing resource
labels.
Signed-off-by: Adrián Larumbe <adrian.larumbe@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34224>
To this end, provide a helper that passes the label to the kmod layer
upon demand. Helper returns the old label, in preparation for future
commits in which the label might not be in static RO memory, but
dynamically alocated instead, and its lifetime managed by the resource
layer.
Signed-off-by: Adrián Larumbe <adrian.larumbe@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34224>
When panfrost_resource_init_afbc_headers() fails, freeing the newly
created resource is not enough, because we need to unreference its BOs.
This will also take care of freeing its resource label.
Also replace instances of FREE() in error-handling paths with
panfrost_resource_destroy(), as it is capable of handling partially
initialised resources.
Signed-off-by: Adrián Larumbe <adrian.larumbe@collabora.com>
Fixes: e3f2bc7963 ("panfrost: handle mmap failures")
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34224>
