Since dbbf566588 ("aco,ac/llvm,radeonsi: lower f2f16 to f2f16_rtz in nir")
radeonsi behavior changed and some of the core fp16 ops broke as a result.
We should explicitly specify the rounding mode until we add an gallium API
for drivers to advertize what they prefer.
Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/26655>
(cherry picked from commit d136583c82)
Needed by Davinci Resolve.
There are two parts to this fix:
1. flush dependencies also on flush, not just finish
2. move the dependency checking logic into Queue::flush as otherwise we
miss required implicit flushes.
Fixes: 8616c0a52c ("rusticl/event: flush queues from dependencies")
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: @LingMan <18294-LingMan@users.noreply.gitlab.freedesktop.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/26053>
(cherry picked from commit 8cbb84dc42)
This saves us a few CPU cycles and makes properly fixing implicit flushes
less expensive.
Fixes: 8616c0a52c ("rusticl/event: flush queues from dependencies")
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: @LingMan <18294-LingMan@users.noreply.gitlab.freedesktop.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/26053>
(cherry picked from commit 52e41d4c97)
If application requests to map surface that was most recently used
as decoder or postproc target and also doesn't explicitly set the
map flags (vaMapBuffer2) it's very likely the intent is to read from
this surface, so we need to map it as such.
This fixes regression on radeonsi where mapping NV12 surfaces for
reading would fail with applications using vaDeriveImage. The reason
for this is that the VA frontend doesn't allow vaDeriveImage for
interlaced surfaces so the applications would use vaGetImage fallback,
but radeonsi doesn't allocate NV12 surfaces as interlaced anymore.
This also fixes mapping other formats surfaces (P010, RGBx, ...)
for reading, which never worked before.
Cc: mesa-stable
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/9935
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/10048
Reviewed-by: Leo Liu <leo.liu@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/26008>
(cherry picked from commit c638e61ef5)
bindgen 0.69.0 broke the `--version` switch, resulting in misleading errors about requiring at
least bindgen 0.62 or about unexpected arguments.
Ideally the build system would fetch the correct bindgen version automatically like cargo does.
Until then, provide a hopefully more helpful error message to the user.
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/26046>
(cherry picked from commit 1f1ec1c6bc)
Host pointer allocations are all linear laid out, so just tell the drivers
in case they don't assume this implicitly.
Fixes: 71a9af4910 ("rusticl/mem: support read/write/copy ops for images")
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25937>
(cherry picked from commit 7f08036abc)
This resolves a memory leak when the application drops its last reference
to the queue, but never waits explicitly.
The problem was, that the queue was refed by QueueState::last and that ref
only gets dropped on a blocking wait. This is problematic as non user
Event objects also hold a ref on the Queue they are created on, therefore
causing a cyclic ref relation.
In order to resolve it, just use a weak reference. A failure of upgrading
the Weak ref is not an issue as in this case we'd only wait on an already
destroyed or processed event. The worker thread already makes sure
everything stays in sync.
Fixes: 5b3ff7e3f3 ("rusticl/queue: overhaul of the queue+event handling")
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: @LingMan <18294-LingMan@users.noreply.gitlab.freedesktop.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25926>
(cherry picked from commit 9a3af6e1d8)
It's kinda pointless to have it too big, it also causes weird shaders to
be generated and causes stack overflows in `nir_opt_gcm`.
Nothing needs big values here anyway.
Cc: mesa-stable
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25837>
(cherry picked from commit 694001eef7)
Don't try to handle emulation bytes and only parse first NAL unit
if the packed header has no emulation bytes.
This fixes parsing packed headers from gstreamer.
Reviewed-by: Ruijing Dong <ruijing.dong@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25565>
This is used for parsing VA packed headers and those can be without
emulation prevention bytes.
Add emulation_bytes argument to vl_rbsp_init and skip all emulation
bytes handling when set.
Reviewed-by: Ruijing Dong <ruijing.dong@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25565>
The assert in vlVaSetSurfaceContext would very rarely fail because
the mutex was already unlocked when calling this function from
vlVaBeginPicture.
Keep the lock until returning from vlVaBeginPicture, as that's what
other functions are already doing.
Reviewed-by: Leo Liu <leo.liu@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25823>
This was an ill-advised extension. While we advertised SWAP_COPY support,
we might implement it with a back-copy from the front buffer. And we
never advertised EXCHANGE because we couldn't guarantee it. So, if you
actually used this extension to try to reduce app redraws of the back
buffer, you might actually increase the bandwidth you used. Whoops.
Instead, GLX_EXT_buffer_age and the similar EGL extension give you
feedback on what's left in your back buffer, letting you do minimum
redraws.
This reduces our GLX visual+fbconfig count from 1410 to 940 on an llvmpipe
X server. Reducing visual counts will improve test runtime for
visual-iterating tests like piglit's glx-visuals-*.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25650>
With this change we are able to query the render node fd of a
render-only device compatible with a given KMS-only device (at the
egl/dri2 level).
It uses pipe_loader_get_compatible_render_capable_device_fd(), which was
added in commit "pipe-loader: add
pipe_loader_get_compatible_render_capable_device_fd()".
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24825>
We have exclusive access in Drop, so we can use `get_mut` instead of having to `lock`.
Since that borrows `self` mutably but `call` also needs to borrow `self`, we `take` the Vec with
callbacks out of `self` so the mutable borrow can end before running `call`.
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25669>
These details are supposed to be inaccessbile and were only made pub during the initial
implementation phase.
Add an `#[allow(dead_code)]` marker since the CreateContextCB isn't hooked up yet which would cause
a warning about unused fields otherwise.
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25669>
Needed because some events may consume their inputs. E.g. it will shortly be needed for the SVMFreeCb.
SVMMemFill will also soon require mutable access.
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25669>
