The v3 spec reserves CTU size 256. Currently, we use an uint8_t* table to hold cb_width and cb_height.
If a CTU size of 256 is not split, cb_width and cb_height will overflow to 0.
To avoid switching to uint16_t, rejecting CTU size 256 provides a simple solution.
If the current stage reports an error, some variables may not be correctly initialized.
Scheduling the next stage could lead to the use of uninitialized variables.
According to section 6.3.3 "Spatial or component-wise partitionings,"
CTUs should fully cover slices with no overlaps, gaps, or additions.
No overlaps are ensured by task_init_parse.
No gaps and no additions are ensured by this patch.
Co-authored-by: Frank Plowman <post@frankplowman.com>
According to section 6.3.3, 'Spatial or component-wise partitionings,'
Subpictures must cover the entire picture.
Therefore, the total number of subpicture slices should equal the number of picture slices
Co-authored-by: Frank Plowman <post@frankplowman.com>
Previously the delta phase was fixed point fractional with 2^32 fractions,
which caused inaccuracies in the output frequency, unless the input
frequency*2^32 was divisable by the sample rate.
This patch improves frequency accuracy by tracking subfractions of the delta
phase fractions. For this we are using a denominator which is a multiple of the
sample rate, making sure that integer frequencies are always accurately
represented.
Signed-off-by: Marton Balint <cus@passwd.hu>
The filter currently uses inaccurate frequencies, this is in preparation for
fixing that, by using numbers that will map to the equivalent value in the
future code.
Signed-off-by: Marton Balint <cus@passwd.hu>
alignas is portable than compiler's specific __attribute__. It do
have a limitation, that alignas don't support specify aligment on
the declarations of struct (it works for specify alignment on struct
fields), which only used by avcodec/cavs, and is removed now.
Signed-off-by: Zhao Zhili <zhilizhao@tencent.com>
There are another three cavs_vector variables, ff_cavs_dir_mv,
ff_cavs_intra_mv and un_mv. They don't need align to 8.
Signed-off-by: Zhao Zhili <zhilizhao@tencent.com>
SAN FRME objects specify width/height as well as offsets from top/left.
These offsets need to be taken into account for the diff buffers
as well. Fixes playback of all SAN videos of "Shadows of the Empire",
which have a constant top offset of 60 (640x272 video on a 640x480 window)
and show tons of ghosting and block artifacts.
Signed-off-by: Manuel Lauss <manuel.lauss@gmail.com>
Signed-off-by: Anton Khirnov <anton@khirnov.net>
The codec47 header provides colors to use to clear the
2 difference buffers. This fixes artifacts (black hair, faces) in
Curse of Monkey Island "CURSERNG.SAN" video.
Signed-off-by: Manuel Lauss <manuel.lauss@gmail.com>
Signed-off-by: Anton Khirnov <anton@khirnov.net>
Outlaws' RAE.SAN file did not properly fade out in certain scenes,
but simply turned to black as soon as the first XPAL chunk with
the apply command was encountered.
Signed-off-by: Manuel Lauss <manuel.lauss@gmail.com>
Signed-off-by: Anton Khirnov <anton@khirnov.net>
Some codec47 frames come with an interpolation table:
Combining 2 adjacent pixels into a 16bit value, this value can
then be used as index into the interpolation table to get a new
pixel value. This is used for subcodec 1, which encodes a key-
frame at half width and height, and makes these frame less blocky
appearing.
Signed-off-by: Manuel Lauss <manuel.lauss@gmail.com>
Signed-off-by: Anton Khirnov <anton@khirnov.net>
When RPU is removed EL should also be removed. This only applies to
HEVC as AV1 based Profile 10 does not support EL at all.
Signed-off-by: Gnattu OC <gnattuoc@me.com>
Signed-off-by: Anton Khirnov <anton@khirnov.net>
The changes to output frame props in query_frame overlapped since
578ac59887. Move the copy frame props before the changes.
Signed-off-by: Fei Wang <fei.w.wang@intel.com>
According to the GL_EXT_buffer_reference spec alignment
"must be a power of two and be greater than or equal to the largest scalar/component type in the block."
This means by using u32vec2 we can drop the requirement alignment from 8 bytes to 4 bytes
and save a pack64 call in reverse8 (though I assume in most ISAs that compiles to nothing)
Allows the vc2 vulkan encoder to function without setting PB_UNALIGNED
If caller wrote a divisible by eight number of bits it would write an extra byte.
Also increment by to_write instead of BUF_BYTES which overly pads the bitstream.
The code used to use atomic, but over time, this got broken.
This commit also remmoves the is-the-last-submission-ready
shortcut, which rarely did anything.
There's also value in relying on the fact that contexts
always carry their frames in a strictly incremental order
with no gaps.
The function is quite important to ensure that the output
is always going to be sufficient, and it can change version to
version, so exposing it makes sense.
This allows the encoder to fully saturate all queues the GPU
has, giving a good 10% in certain cases and resolutions.
This also improves error resilience if an allocation fails,
and properly cleans up after itself if it does.
This function dates back a long time ago, before vkfmt_from_pixfmt2.
When it was converted over, the thought was that this was far too
restrictive to demand storage images for each format.
With the new clever function, it makes sure to check that the compatible
subformats a format can be used as support storage capabilities.
This gets rid of fake support for RGB48/RGB96 which some implementations
offer but don't support using as storage images.
