Descriptor buffers were a neat attempt at organizing descriptors.
Simple, robust, reliable.
Unfortunately, driver support never caught on, and neither did validation
layer support.
Now they're being replaced by descriptor heaps, which promises to be
the future. We'll see how it goes.
They require no special video decode extension. The push descriptor
extension was a placeholder for when the check was performed inside
the init function.
The issue is that .flush gets called asynchronously, and modifies the
video session state while its being used for decoding. This did not
result in issues since all known vendors do not keep important state
there, but its not compliant with the specs.
Its not necessary to flush the decoder at all when seeking,
so simply don't.
Fixes#20487
The decoder will reinit the hwaccel upon pixfmt/dimension changes,
so we can remove the f->use32bit and is_rgb variants of all shaders.
This speeds up init time.
We have no use for 14-bit pixel formats for now, so remove support for gray14,
which was broken due to the LSB padding issue.
Similarly YUVA at 10/12 bit was broken for the same reason.
This commit also makes it possible for the encoder to choose a different
quantization table on a per-slice basis, as well as adding this capability
to the decoder.
Also, this commit fully fixes decoding of context=1 encoded files.
This reduces the intermediate VRAM used for RGB decoding by a
factor of 100x for 6k video.
This also speeds the decoder up by 16% for 4k RGB24 and 31% for 6k video.
This is equivalent to what the software decoder does, but with less pointers.
This patch adds a fully-featured level 3 and 4 decoder for FFv1,
supporting Golomb and all Range coding variants, all pixel formats,
and all features, except for the newly added floating-point formats.
On a 6000 Ada, for 3840x2160 bgr0 content at 50Mbps (standard desktop
recording), it is able to do 400fps.
An Alder Lake with 24 threads can barely do 100fps.
