The existing refcounting code is correct, unless si_texture_from_winsys_buffer
fails in which case we get a refcount error.
The error code path will use si_texture_reference(&tex, NULL), which
will drop a reference to the memobj buffer, but none was taken yet.
Reviewed-by: Ganesh Belgur Ramachandra <ganesh.belgurramachandra@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36228>
The functionality is provided by isl_surf_supports_ccs().
Also, move the protected content restriction to
iris_resource_configure_aux(). I'm not aware of any reason protected
content wouldn't support CCS. However, to keep this series simple,
enabling that combination is left for another time.
Reviewed-by: Jianxun Zhang <jianxun.zhang@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32120>
Allow CCS for non-display linear surfaces in isl_surf_supports_ccs().
We're going to rely more on the helper to determine CCS-enabling for Xe2
on iris.
Reviewed-by: Jianxun Zhang <jianxun.zhang@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32120>
The compression format is only used during rendering.
This prevents drivers from hitting an unreachable when we start enabling
CCS on linear surfaces which may have non-renderable and non-pow2 formats.
For now, continue to use the surface format instead of the view format
to look up the CMF. This strategy should return the optimal CMF for
compressed surfaces that undergo redescription during copies.
Reviewed-by: Jianxun Zhang <jianxun.zhang@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32120>
On XeKMD, BOs need to be created with a vm_id of zero in order to get
prime handles. That only occurs if the image was created with
PIPE_BIND_SHARED/BO_ALLOC_SHARED. Ensure that shareable images have this
flag in iris_flush_resource().
Fixes the dmabufshare demo on BMG with INTEL_DEBUG=noccs and mesa hacked
to disable suballocation.
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/13511
Reviewed-by: Jianxun Zhang <jianxun.zhang@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32120>
The AHB type is no longer needed there as the assert check is no longer
used on the export path. One less DETECT_OS_ANDROID.
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36252>
Per spec VUID-VkMemoryAllocateInfo-allocationSize-07897
> If the parameters do not define an import or export operation,
allocationSize must be greater than 0.
So early return there is simply messing up with external memory.
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36252>
It's helping buggy applications that may use the returned
value without checking for an error first.
For instance, viewperf20/maya uses a sequence like this:
glGenBuffers(n = 1, buffers = &3458)
...
glGetNamedBufferParameteriv(3458, GL_BUFFER_SIZE, &p)
Since 3458 isn't associated to an object yet, _mesa_lookup_bufferobj_err
will return NULL, leaving 'p' with its original value.
It seems to randomly trigger an exit from the benchmark,
presumably because the content in 'p' is random (for instance
-1879044180).
Cc: mesa-stable
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36190>
Every PAN_FAULT_INJECTION_RATE allocation from a mempool, we generate an
invalid GPU address which will cause a GPU fault.
This is useful to debug the GPU reset / context recover path.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31868>
Panthor can report the group state, so let's provide a
get_device_reset_status() querying this state and turning it into a
pipe_reset_status status.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31868>
We are going to need proper variants for the vertex stage to emulate
geometry/tesselation stages.
This switch our codebase to use something similar as Honeykrisp.
Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com>
Reviewed-by: Olivia Lee <olivia.lee@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36131>
We were missing parenthesis on pan_cast_and_pack macros causing various
issues.
Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com>
Fixes: 410e5a36ec ("pan/genxml: Rework gen_pack.py to support OpenCL")
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36261>
Use the existing DRIVER_NAME mechanism to pick up common skips.
This is less error prone than manually adding the skips.
A redundant freedreno-a618-skips.txt is also dropped, as it's already
included via GPU_VERSION.
Signed-off-by: Valentine Burley <valentine.burley@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36215>
The cheza runners were decommissioned.
Rename the restricted trace results to a618 (same GPU generation) to keep
the history.
Signed-off-by: Valentine Burley <valentine.burley@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36215>
The KIND_BPP stuff is actually Blackwell B+ and has a few naming
differences from what we landed originally. This fixes all that.
Cc: 25.2
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36251>
Newer generation copy headers (including Turing) have a few enums with
more than one enumerant having the same value. Skip these in the
printer and in the Rust bindings.
Cc: 25.2
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36251>
Otherwise, the uniform_bool could point to a temporary which might be
DCE'd, since it's not used by the s_cselect.
fossil-db (navi21):
Totals from 1 (0.00% of 79825) affected shaders:
Instrs: 1267 -> 1269 (+0.16%)
Latency: 91071 -> 91103 (+0.04%)
SALU: 283 -> 285 (+0.71%)
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36177>
The optimizer shouldn't consider a lanemask to be a uniform boolean unless
it's either 0 or -1. Optimizations involving s_not/s_xor might not work
properly otherwise.
No fossil-db changes (navi21).
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36177>
NULL was forgotten.
This issue was detected by the gcc sanitizer.
==30873==ERROR: AddressSanitizer: global-buffer-overflow on address 0x7f94f9ad14b0 at pc 0x7f94f04905bc bp 0x7ffe3b1ebcf0 sp 0x7ffe3b1ebce8
READ of size 8 at 0x7f94f9ad14b0 thread T0
#0 0x7f94f04905bb in loader_bind_extensions ../src/loader/loader.c:783
#1 0x7f94ee656232 in setupLoaderExtensions ../src/gallium/frontends/dri/dri_util.c:88
#2 0x7f94ee656232 in driCreateNewScreen3 ../src/gallium/frontends/dri/dri_util.c:111
#3 0x7f94f98d9a14 in dri2_create_screen ../src/egl/drivers/dri2/egl_dri2.c:774
#4 0x7f94f98dfbea in surfaceless_probe_device ../src/egl/drivers/dri2/platform_surfaceless.c:298
#5 0x7f94f98dfbea in dri2_initialize_surfaceless ../src/egl/drivers/dri2/platform_surfaceless.c:390
#6 0x7f94f98dbc4f in dri2_initialize ../src/egl/drivers/dri2/egl_dri2.c:861
#7 0x7f94f98dbc4f in dri2_initialize ../src/egl/drivers/dri2/egl_dri2.c:832
#8 0x7f94f9892843 in eglInitialize ../src/egl/main/eglapi.c:696
Fixes: 56ed5cf1bb ("dri: Get rid of __DRIbackgroundCallableExtension")
Signed-off-by: Patrick Lerda <patrick9876@free.fr>
Reviewed-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36255>
* use tex builder (what I came for)
* use nir_create_variable_with_location
* drop redundant internal = true's, the shaders are already initialized that way
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36050>
this is already set when building the gs copy shader since it is a nir_builder
simple shader
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36050>
* use the tex builder (what I came for)
* use fmul_imm and fix weird wrapping
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36050>
for intrinsics, we have these really nice builders using designated initializers
+ macros to specify optional indices. texture instrs have even more craziness
involved, but we can do the same trick. this commit takes the existing "fixed
form" deref-centric tex builders and generalizes them to work with non-deref
textures, making it useful also for GL and late VK passes, while providing an
API that strives to be ergonomic and consistent.
this series only implements a subset of possible texture operations for now, but
more generalizing could be added as people have need.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36050>
