The entire set of 32 coefficients and corresponding past 32 samples can
fit in a single vector (with LMUL=8) exactly, but... since widening
double the needed vector sizes, we still end up too short with 128-bit
vectors. This adds a very simple version for future 256+-bit hardware,
and for pred_orders values up to 16, and a bit more involved loop for
for 128-bit hardware with pred_orders between 17 and 32.
With 128-bit hardware, the benchmarks look like this:
flac_lpc_32_13_c: 30152.0
flac_lpc_32_13_rvv_i32: 10244.7
flac_lpc_32_16_c: 37314.2
flac_lpc_32_16_rvv_i32: 10126.2
flac_lpc_32_29_c: 61910.0
flac_lpc_32_29_rvv_i32: 14495.2
flac_lpc_32_32_c: 68204.0
flac_lpc_32_32_rvv_i32: 13273.7
decorrelate_ls, _rs and _ms are decorrelate[1], [2] and [3] respectively.
The code ended up testing indep ([0]) as twice, ms never, and misnaming
the other two.
Segmented loads are slow, so here we use unit-strided load and narrowing shifts.
c910:
fcmul_add_c: 2179
fcmul_add_rvv_f64: 1652
c908:
fcmul_add_c: 4891.2
fcmul_add_rvv_f64: 2399.5
Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>
This commit adds support for VP8 bitstream read methods to the cbs
codec. This enables the trace_headers bitstream filter to support VP8,
in addition to AV1, H.264, H.265, and VP9. This can be useful for
debugging VP8 stream issues.
The CBS VP8 implements a simple VP8 boolean decoder using GetBitContext
to read the bitstream.
Only the read methods `read_unit` and `split_fragment` are implemented.
The write methods `write_unit` and `assemble_fragment` return the error
code AVERROR_PATCHWELCOME. This is because CBS VP8 write is unlikely to
be used by any applications at the moment. The write methods can be
added later if there is a real need for them.
TESTS: ffmpeg -i fate-suite/vp8/frame_size_change.webm -vcodec copy
-bsf:v trace_headers -f null -
Signed-off-by: Jianhui Dai <jianhui.j.dai@intel.com>
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
This commit exports the `vp8_token_update_probs` variable to internal
library scope to facilitate its reuse within the library.
Signed-off-by: Jianhui Dai <jianhui.j.dai@intel.com>
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
Better performance can probably be achieved with a more intricate
unrolled loop, but this is a start:
add_hfyu_left_pred_bgr32_c: 15084.0
add_hfyu_left_pred_bgr32_rvv_i32: 10280.2
This would actually be cleaner with the RISC-V P extension, but that is
not ratified yet (I think?) and usually not supported if V is supported.
Current code tracks min/max pts for each stream separately; then when
the file ends it combines them with last frame's duration to compute the
total duration of each stream; finally it selects the longest of those
durations as the file duration.
This is incorrect - the total file duration is the largest timestamp
difference between any frames, regardless of the stream.
Also change the way the last frame information is reported from decoders
to the muxer - previously it would be just the last frame's duration,
now the end timestamp is sent, which is simpler.
Changes the result of the fate-ffmpeg-streamloop-transcode-av test,
where the timestamps are shifted slightly forward. Note that the
matroska demuxer does not return the first audio packet after seeking
(due to buggy interaction betwen the generic code and the demuxer), so
there is a gap in audio.
This ensures that tq_receive() will always return EOF after all streams
were receive-finished, even though the sending side might not have
closed them yet. This may allow the receiver to avoid manually tracking
which streams it has already closed.
It is common for subtitle streams to have large gaps between packets.
When the caller is interleaving packets from multiple files, it can
easily happen that two successive subtitle packets trigger this limit,
even though no excessive buffering is happening.
Should fix#7064
With explicit unrolling, we can skip half of the sign bit flips, and
the compiler is then better able to optimise the scalar loop:
predictor_c: 31376.0 (before)
predictor_c: 23703.0 (after)
This is restricted to 128-bit vectors as larger vector sizes could read
past the end of the noise array. Support for future hardware with larger
vector sizes is left for some other time.
hf_apply_noise_0_c: 2319.7
hf_apply_noise_0_rvv_f32: 1229.0
hf_apply_noise_1_c: 2539.0
hf_apply_noise_1_rvv_f32: 1244.7
hf_apply_noise_2_c: 2319.7
hf_apply_noise_2_rvv_f32: 1232.7
hf_apply_noise_3_c: 2541.2
hf_apply_noise_3_rvv_f32: 1244.2
The tested functions treat s_m[i] == 0 as a special case. Other than
that, the functions are slightly complicated vector additions.
This actually makes the zero case happen pseudorandomly.
In my personal opinion, we should not need to support unaligned YUY2
pixel maps. They should always be aligned to at least 32 bits, and the
current code assumes just 16 bits. However checkasm does test for
unaligned input bitmaps. QEMU accepts it, but real hardware dose not.
In this particular case, we can at the same time improve performance and
handle unaligned inputs, so do just that.
uyvytoyuv422_c: 104379.0
uyvytoyuv422_c: 104060.0
uyvytoyuv422_rvv_i32: 25284.0 (before)
uyvytoyuv422_rvv_i32: 19303.2 (after)
This saves three scratch registers and three instructions per line. The
performance gains are mostly negligible. The main point is to free up
registers for further rework.
Fixes OOM when caller keeps adding frames into filtergraph
that reached EOF by other means, for example EOF is signalled
by other filter in filtergraph or by buffersink.
This commit fixes issue with missing SPS/PPS headers in video
encoded by AMF AV1 encoder.
Missing headers leads to broken seek in MPV video player.
Default value for property AV1_HEADER_INSERTION_MODE shouldn't be setup
to NONE (no headers insertion). We need to skip definition of this property,
because default value depends on USAGE property.
Signed-off-by: Dmitrii Ovchinnikov <ovchinnikov.dmitrii@gmail.com>
With 5 accumulator vectors and 6 inputs, this can only use LMUL=2.
Also the number of vector loop iterations is small, just 5 on 128-bit
vector hardware.
The vector loop is somewhat unusual in that it processes data in
descending memory order, in order to save on vector slides:
in descending order, we can extract elements to carry over to the next
iteration from the bottom of the vectors directly. With ascending order
(see in the Opus postfilter function), there are no ways to get the top
elements directly. On the downside, this requires the use of separate
shift and sub (the would-be SH3SUB instruction does not exist), with
a small pipeline stall on the vector load address.
The edge cases in scalar are done in scalar as this saves on loads
and remains significantly faster than C.
autocorrelate_c: 669.2
autocorrelate_rvv_f32: 421.0
Given the size of the data set, strided memory accesses cannot be avoided.
We can still do better than the current code.
ps_hybrid_synthesis_deint_c: 12065.5
ps_hybrid_synthesis_deint_rvv_i32: 13650.2 (before)
ps_hybrid_synthesis_deint_rvv_i64: 8181.0 (after)
Segmented loads may be slower than not. So this advantageously uses a
unit-strided load and narrowing shifts instead.
Before:
ps_add_squares_c: 60757.7
ps_add_squares_rvv_f32: 22242.5
After:
ps_add_squares_c: 60516.0
ps_add_squares_rvv_i64: 17067.7
Rémi Denis-Courmont
Jun Zhao
Diederik de Haas via ffmpeg-devel
Paul B Mahol
sunyuechi
James Almer
Dai, Jianhui J
Zhao Zhili
Anton Khirnov
Gyan Doshi
Evgeny Pavlov
Sebastian Ramacher