The fixed-point AAC decoder is the only user of the fixed-point sinewin
tables from sinewin; and it only uses a few of them (about 10% when
counting by size). This means that guarding initializing these tables by
an AVOnce (as done in 3719122065) is
unnecessary for them. Furthermore the array of pointers to the
individual arrays is also unneeded.
Therefore this commit moves these tables directly into aacdec_fixed.c;
this is done by ridding the original sinewin.h and sinewin_tablegen.h
headers completely of any fixed-point code at the cost of a bit of
duplicated code (the alternative is an ugly ifdef-mess).
This saves about 58KB from the binary when using hardcoded tables (as
these tables are hardcoded in this scenario); when not using hardcoded
tables, most of these savings only affect the .bss segment, but the rest
(< 1KB) contains relocations (i.e. savings in .data.rel.ro).
Reviewed-by: Lynne <dev@lynne.ee>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
Neither module should depend on the other.
Move shared functions to its own file for this purpose, and ensure
source files are compiled only when the required modules are enabled.
Signed-off-by: James Almer <jamrial@gmail.com>
The only call to ff_intel_h263_decode_picture_header() is already behind
"if (CONFIG_H263I_DECODER)".
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
2ef2496cd1 used ff_vorbis_channel_layouts
in flac.c, but added a dependency to the FLAC decoder only; lateron
aba0278e9f added the dependency of the
FLAC parser and encoder on vorbis_data.o. Yet when the original commit
was reverted in aba0278e9f, the two other
dependencies were not removed. This commit fixes this.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
No longer used by anything.
Unfortunately the old FFT_FLOAT/FFT_FIXED_32 is left as-is. It's
simply too much work for code meant to be all removed anyway.
The AC3 encoder used to be a separate library called "Aften", which
got merged into libavcodec (literally, SVN commits and all).
The merge preserved as much features from the library as possible.
The code had two versions - a fixed point version and a floating
point version. FFmpeg had floating point DSP code used by other
codecs, the AC3 decoder including, so the floating-point DSP was
simply replaced with FFmpeg's own functions.
However, FFmpeg had no fixed-point audio code at that point. So
the encoder brought along its own fixed-point DSP functions,
including a fixed-point MDCT.
The fixed-point MDCT itself is trivially just a float MDCT with a
different type and each multiply being a fixed-point multiply.
So over time, it got refactored, and the FFT used for all other codecs
was templated.
Due to design decisions at the time, the fixed-point version of the
encoder operates at 16-bits of precision. Although convenient, this,
even at the time, was inadequate and inefficient. The encoder is noisy,
does not produce output comparable to the float encoder, and even
rings at higher frequencies due to the badly approximated winow function.
Enter MIPS (owned by Imagination Technologies at the time). They wanted
quick fixed-point decoding on their FPUless cores. So they contributed
patches to template the AC3 decoder so it had both a fixed-point
and a floating-point version. They also did the same for the AAC decoder.
They however, used 32-bit samples. Not 16-bits. And we did not have
32-bit fixed-point DSP functions, including an MDCT. But instead of
templating our MDCT to output 3 versions (float, 32-bit fixed and 16-bit fixed),
they simply copy-pasted their own MDCT into ours, and completely
ifdeffed our own MDCT code out if a 32-bit fixed point MDCT was selected.
This is also the status quo nowadays - 2 separate MDCTs, one which
produces floating point and 16-bit fixed point versions, and one
sort-of integrated which produces 32-bit MDCT.
MIPS weren't all that interested in encoding, so they left the encoder
as-is, and they didn't care much about the ifdeffery, mess or quality - it's
not their problem.
So the MDCT/FFT code has always been a thorn in anyone looking to clean up
code's eye.
Backstory over. Internally AC3 operates on 25-bit fixed-point coefficients.
So for the floating point version, the encoder simply runs the float MDCT,
and converts the resulting coefficients to 25-bit fixed-point, as AC3 is inherently
a fixed-point codec. For the fixed-point version, the input is 16-bit samples,
so to maximize precision the frame samples are analyzed and the highest set
bit is detected via ac3_max_msb_abs_int16(), and the coefficients are then
scaled up via ac3_lshift_int16(), so the input for the FFT is always at least 14 bits,
computed in normalize_samples(). After FFT, the coefficients are scaled up to 25 bits.
This patch simply changes the encoder to accept 32-bit samples, reusing
the already well-optimized 32-bit MDCT code, allowing us to clean up and drop
a large part of a very messy code of ours, as well as prepare for the future lavu/tx
conversion. The coefficients are simply scaled down to 25 bits during windowing,
skipping 2 separate scalings, as the hacks to extend precision are simply no longer
necessary. There's no point in running the MDCT always at 32 bits when you're
going to drop 6 bits off anyway, the headroom is plenty, and the MDCT rounds
properly.
This also makes the encoder even slightly more accurate over the float version,
as there's no coefficient conversion step necessary.
SIZE SAVINGS:
ARM32:
HARDCODED TABLES:
BASE - 10709590
DROP DSP - 10702872 - diff: -6.56KiB
DROP MDCT - 10667932 - diff: -34.12KiB - both: -40.68KiB
DROP FFT - 10336652 - diff: -323.52KiB - all: -364.20KiB
SOFTCODED TABLES:
BASE - 9685096
DROP DSP - 9678378 - diff: -6.56KiB
DROP MDCT - 9643466 - diff: -34.09KiB - both: -40.65KiB
DROP FFT - 9573918 - diff: -67.92KiB - all: -108.57KiB
ARM64:
HARDCODED TABLES:
BASE - 14641112
DROP DSP - 14633806 - diff: -7.13KiB
DROP MDCT - 14604812 - diff: -28.31KiB - both: -35.45KiB
DROP FFT - 14286826 - diff: -310.53KiB - all: -345.98KiB
SOFTCODED TABLES:
BASE - 13636238
DROP DSP - 13628932 - diff: -7.13KiB
DROP MDCT - 13599866 - diff: -28.38KiB - both: -35.52KiB
DROP FFT - 13542080 - diff: -56.43KiB - all: -91.95KiB
x86:
HARDCODED TABLES:
BASE - 12367336
DROP DSP - 12354698 - diff: -12.34KiB
DROP MDCT - 12331024 - diff: -23.12KiB - both: -35.46KiB
DROP FFT - 12029788 - diff: -294.18KiB - all: -329.64KiB
SOFTCODED TABLES:
BASE - 11358094
DROP DSP - 11345456 - diff: -12.34KiB
DROP MDCT - 11321742 - diff: -23.16KiB - both: -35.50KiB
DROP FFT - 11276946 - diff: -43.75KiB - all: -79.25KiB
PERFORMANCE (10min random s32le):
ARM32 - before - 39.9x - 0m15.046s
ARM32 - after - 28.2x - 0m21.525s
Speed: -30%
ARM64 - before - 36.1x - 0m16.637s
ARM64 - after - 36.0x - 0m16.727s
Speed: -0.5%
x86 - before - 184x - 0m3.277s
x86 - after - 190x - 0m3.187s
Speed: +3%
Up until now, the SpeedHQ encoder called a wrong function for init:
void ff_init_uni_ac_vlc(const uint8_t huff_size_ac[256],
uint8_t *uni_ac_vlc_len);
Yet the first argument actually used is of type RLTable; the size of
said struct is less than 256 if the size of a pointer is four, leading
to an access beyond the end of the RLTable.
This commit fixes this by calling the actually intended function:
init_uni_ac_vlc() from mpeg12enc.c. It was intended to use this
function [1], yet doing so was forgotten when the patch was actually
applied.
[1]: https://ffmpeg.org/pipermail/ffmpeg-devel/2020-July/266187.html
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
SMVJPEG stores frames as slices of a big JPEG image. The decoder is
implemented as a wrapper that instantiates a full internal MJPEG
decoder, then forwards the decoded frames with offset data pointers.
This is unnecessarily complex and fragile, not supporting useful decoder
capabilities like direct rendering.
Re-implement the decoder inside the MJPEG decoder, which is accomplished
by returning each decoded frame multiple times, setting cropping
information appropriately on each instance.
One peculiar aspect of the previous design is that since
- the smvjpeg decoder returns one frame per input packet
- there are multiple frames in each packets (the aformentioned slices)
the demuxer needs to return each packet multiple times.
This is now also eliminated - the demuxer now returns each packet
exactly once, with the duration set to the number of frames it decodes
to.
This also removes one of the last remaining internal uses of the old
video decoding API.
Both the fixed as well as the floating point mpegaudio decoders use
LUTs of type int8_t and uint32_t with 32K entries each; these tables
are completely the same, yet they are not shared. This commit makes
them shared. When both fixed as well as floating point decoders are
enabled, this saves 160KiB from the bss segment for a normal build
(translating into 160KiB less memory usage if both a shared as well as
a floating point decoder have actually been used) and 160KiB from the
binary for a build with hardcoded tables.
It also means that the code to create said LUTs is no longer duplicated
(for a normal build).
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
The manual states "there is virtually no reason to use that encoder.".
It supports less sample formats than the native encoder, is less efficient
than the native encoder and is also slower and pretty much remains untested.
libwavpack also isn't being fuzzed, which given that we plug the parameters
without any sanitizing them looks concerning.
This AV1 decoder is currently only used for hardware accelerated decoding.
It can be extended into a native decoder in the future, so set its name to
"av1" and temporarily give it the lowest priority in the codec list.
Signed-off-by: Fei Wang <fei.w.wang@intel.com>
Signed-off-by: James Almer <jamrial@gmail.com>
Paul B Mahol
Andreas Rheinhardt
James Almer
Nicolas Gaullier
Xu Guangxin
Mark Thompson
Lynne
Anton Khirnov
Peter Ross
Mohammad Izadi
Jean-Baptiste Kempf
Hendrik Leppkes
Timo Rothenpieler
Fei Wang
Zane van Iperen
hwren
Clément Bœsch