virtualenv/FFmpeg
1
0
Fork 0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2024-12-12 19:18:44 +02:00
Code Issues Releases Activity
cab876f5f4
FFmpeg/libavcodec/opusenc.c

754 lines
27 KiB
C
Raw Normal View History

opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
/*
* Opus encoder
* Copyright (c) 2017 Rostislav Pehlivanov <atomnuker@gmail.com>
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
avcodec/opus: Move defines to better places Move ROUND_MUL* macros to their only users and the Celt macros to opus_celt.h. Also improve the other headers a bit while at it. Reviewed-by: Lynne <dev@lynne.ee> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2022-10-03 22:47:58 +02:00
#include <float.h>
avcodec/encode: Always use intermediate buffer in ff_alloc_packet2() Up until now, ff_alloc_packet2() has a min_size parameter: It is supposed to be a lower bound on the final size of the packet to allocate. If it is not too far from the upper bound (namely, if it is at least half the upper bound), then ff_alloc_packet2() already allocates the final, already refcounted packet; if it is not, then the packet is not refcounted and its data only points to a buffer owned by the AVCodecContext (in this case, the packet will be made refcounted in encode_simple_internal() in libavcodec/encode.c). The goal of this was to avoid data copies and intermediate buffers if one has a precise lower bound. Yet those encoders for which precise lower bounds exist have recently been switched to ff_get_encode_buffer() (which automatically allocates final buffers), leaving only two encoders to actually set the min_size to something else than zero (namely aliaspixenc and hapenc). Both of these encoders use a very low lower bound that is not helpful in any nontrivial case. This commit therefore removes the min_size parameter as well as the codepath in ff_alloc_packet2() for the allocation of final buffers. Furthermore, the function has been renamed to ff_alloc_packet() and moved to encode.h alongside ff_get_encode_buffer(). Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2021-05-11 15:17:13 +02:00
#include "encode.h"
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
#include "opusenc.h"
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
#include "opus_pvq.h"
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
#include "opusenc_psy.h"
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
#include "opustab.h"
avcodec/avcodec: Stop including channel_layout.h in avcodec.h Also include channel_layout.h directly wherever used. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2021-06-12 01:10:58 +02:00
#include "libavutil/channel_layout.h"
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
#include "libavutil/float_dsp.h"
lavu/mem: move the DECLARE_ALIGNED macro family to mem_internal on next+1 bump They are not properly namespaced and not intended for public use.
2020-05-27 14:54:38 +02:00
#include "libavutil/mem_internal.h"
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
#include "libavutil/opt.h"
#include "bytestream.h"
#include "audio_frame_queue.h"
avcodec/internal: Move FF_CODEC_CAP_* to a new header codec_internal.h Also move FF_CODEC_TAGS_END as well as struct AVCodecDefault. This reduces the amount of files that have to include internal.h (which comes with quite a lot of indirect inclusions), as e.g. most encoders don't need it. It is furthemore in preparation for moving the private part of AVCodec out of the public codec.h. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2022-03-16 19:18:28 +02:00
#include "codec_internal.h"
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
typedef struct OpusEncContext {
AVClass *av_class;
OpusEncOptions options;
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
OpusPsyContext psyctx;
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
AVCodecContext *avctx;
AudioFrameQueue afq;
AVFloatDSPContext *dsp;
opus: convert encoder and decoder to lavu/tx This commit changes both the encoder and decoder to use the new lavu/tx code, which has faster C transforms and more assembly optimizations.
2022-09-24 00:46:44 +02:00
AVTXContext *tx[CELT_BLOCK_NB];
av_tx_fn tx_fn[CELT_BLOCK_NB];
opus_pvq: port to allow for SIMD functions Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-05-10 07:47:44 +02:00
CeltPVQ *pvq;
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
struct FFBufQueue bufqueue;
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
uint8_t enc_id[64];
int enc_id_bits;
OpusPacketInfo packet;
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
int channels;
CeltFrame *frame;
OpusRangeCoder *rc;
/* Actual energy the decoder will have */
float last_quantized_energy[OPUS_MAX_CHANNELS][CELT_MAX_BANDS];
DECLARE_ALIGNED(32, float, scratch)[2048];
} OpusEncContext;
static void opus_write_extradata(AVCodecContext *avctx)
{
uint8_t *bs = avctx->extradata;
bytestream_put_buffer(&bs, "OpusHead", 8);
bytestream_put_byte (&bs, 0x1);
opus: convert to new channel layout API Signed-off-by: James Almer <jamrial@gmail.com>
2019-05-28 14:38:10 +02:00
bytestream_put_byte (&bs, avctx->ch_layout.nb_channels);
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
bytestream_put_le16 (&bs, avctx->initial_padding);
bytestream_put_le32 (&bs, avctx->sample_rate);
bytestream_put_le16 (&bs, 0x0);
bytestream_put_byte (&bs, 0x0); /* Default layout */
}
static int opus_gen_toc(OpusEncContext *s, uint8_t *toc, int *size, int *fsize_needed)
{
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
int tmp = 0x0, extended_toc = 0;
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
static const int toc_cfg[][OPUS_MODE_NB][OPUS_BANDWITH_NB] = {
/* Silk Hybrid Celt Layer */
/* NB MB WB SWB FB NB MB WB SWB FB NB MB WB SWB FB Bandwidth */
{ { 0, 0, 0, 0, 0 }, { 0, 0, 0, 0, 0 }, { 17, 0, 21, 25, 29 } }, /* 2.5 ms */
{ { 0, 0, 0, 0, 0 }, { 0, 0, 0, 0, 0 }, { 18, 0, 22, 26, 30 } }, /* 5 ms */
{ { 1, 5, 9, 0, 0 }, { 0, 0, 0, 13, 15 }, { 19, 0, 23, 27, 31 } }, /* 10 ms */
{ { 2, 6, 10, 0, 0 }, { 0, 0, 0, 14, 16 }, { 20, 0, 24, 28, 32 } }, /* 20 ms */
{ { 3, 7, 11, 0, 0 }, { 0, 0, 0, 0, 0 }, { 0, 0, 0, 0, 0 } }, /* 40 ms */
{ { 4, 8, 12, 0, 0 }, { 0, 0, 0, 0, 0 }, { 0, 0, 0, 0, 0 } }, /* 60 ms */
};
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
int cfg = toc_cfg[s->packet.framesize][s->packet.mode][s->packet.bandwidth];
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
*fsize_needed = 0;
if (!cfg)
return 1;
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
if (s->packet.frames == 2) { /* 2 packets */
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
if (s->frame[0].framebits == s->frame[1].framebits) { /* same size */
tmp = 0x1;
} else { /* different size */
tmp = 0x2;
*fsize_needed = 1; /* put frame sizes in the packet */
}
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
} else if (s->packet.frames > 2) {
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
tmp = 0x3;
extended_toc = 1;
}
tmp |= (s->channels > 1) << 2; /* Stereo or mono */
tmp |= (cfg - 1) << 3; /* codec configuration */
*toc++ = tmp;
if (extended_toc) {
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int i = 0; i < (s->packet.frames - 1); i++)
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
*fsize_needed |= (s->frame[i].framebits != s->frame[i + 1].framebits);
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
tmp = (*fsize_needed) << 7; /* vbr flag */
tmp |= (0) << 6; /* padding flag */
tmp |= s->packet.frames;
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
*toc++ = tmp;
}
*size = 1 + extended_toc;
return 0;
}
static void celt_frame_setup_input(OpusEncContext *s, CeltFrame *f)
{
AVFrame *cur = NULL;
const int subframesize = s->avctx->frame_size;
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
int subframes = OPUS_BLOCK_SIZE(s->packet.framesize) / subframesize;
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
cur = ff_bufqueue_get(&s->bufqueue);
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int ch = 0; ch < f->channels; ch++) {
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
CeltBlock *b = &f->block[ch];
const void *input = cur->extended_data[ch];
size_t bps = av_get_bytes_per_sample(cur->format);
memcpy(b->overlap, input, bps*cur->nb_samples);
}
av_frame_free(&cur);
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int sf = 0; sf < subframes; sf++) {
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
if (sf != (subframes - 1))
cur = ff_bufqueue_get(&s->bufqueue);
else
cur = ff_bufqueue_peek(&s->bufqueue, 0);
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int ch = 0; ch < f->channels; ch++) {
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
CeltBlock *b = &f->block[ch];
const void *input = cur->extended_data[ch];
const size_t bps = av_get_bytes_per_sample(cur->format);
const size_t left = (subframesize - cur->nb_samples)*bps;
const size_t len = FFMIN(subframesize, cur->nb_samples)*bps;
memcpy(&b->samples[sf*subframesize], input, len);
memset(&b->samples[cur->nb_samples], 0, left);
}
/* Last frame isn't popped off and freed yet - we need it for overlap */
if (sf != (subframes - 1))
av_frame_free(&cur);
}
}
/* Apply the pre emphasis filter */
static void celt_apply_preemph_filter(OpusEncContext *s, CeltFrame *f)
{
const int subframesize = s->avctx->frame_size;
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
const int subframes = OPUS_BLOCK_SIZE(s->packet.framesize) / subframesize;
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
/* Filter overlap */
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int ch = 0; ch < f->channels; ch++) {
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
CeltBlock *b = &f->block[ch];
float m = b->emph_coeff;
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int i = 0; i < CELT_OVERLAP; i++) {
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
float sample = b->overlap[i];
b->overlap[i] = sample - m;
m = sample * CELT_EMPH_COEFF;
}
b->emph_coeff = m;
}
/* Filter the samples but do not update the last subframe's coeff - overlap ^^^ */
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int sf = 0; sf < subframes; sf++) {
for (int ch = 0; ch < f->channels; ch++) {
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
CeltBlock *b = &f->block[ch];
float m = b->emph_coeff;
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int i = 0; i < subframesize; i++) {
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
float sample = b->samples[sf*subframesize + i];
b->samples[sf*subframesize + i] = sample - m;
m = sample * CELT_EMPH_COEFF;
}
if (sf != (subframes - 1))
b->emph_coeff = m;
}
}
}
/* Create the window and do the mdct */
static void celt_frame_mdct(OpusEncContext *s, CeltFrame *f)
{
opusenc: use float_dsp for non-transient windowing Also fixes transient windowing Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-07-14 05:50:07 +02:00
float *win = s->scratch, *temp = s->scratch + 1920;
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
if (f->transient) {
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int ch = 0; ch < f->channels; ch++) {
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
CeltBlock *b = &f->block[ch];
float *src1 = b->overlap;
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int t = 0; t < f->blocks; t++) {
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
float *src2 = &b->samples[CELT_OVERLAP*t];
opusenc: use float_dsp for non-transient windowing Also fixes transient windowing Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-07-14 05:50:07 +02:00
s->dsp->vector_fmul(win, src1, ff_celt_window, 128);
opusenc: use float_dsp for transient mdcts vector_fmul_reverse requires padding the window at the front Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-07-12 05:49:21 +02:00
s->dsp->vector_fmul_reverse(&win[CELT_OVERLAP], src2,
avcodec/opustab: Avoid indirection to access ff_celt_window Currently, it is accessed via a pointer (ff_celt_window) exported from opustab.h which points inside a static array (ff_celt_window_padded) in opustab.h. Instead export ff_celt_window_padded directly and make opustab.h a static const pointer pointing inside ff_celt_window_padded. Also mark all the declarations in opustab.h as hidden, so that the compiler knows that ff_celt_window has a fixed offset from the code even when compiling position-independent code. Reviewed-by: Lynne <dev@lynne.ee> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2022-10-03 17:21:53 +02:00
ff_celt_window_padded, 128);
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
src1 = src2;
opus: convert encoder and decoder to lavu/tx This commit changes both the encoder and decoder to use the new lavu/tx code, which has faster C transforms and more assembly optimizations.
2022-09-24 00:46:44 +02:00
s->tx_fn[0](s->tx[0], b->coeffs + t, win, sizeof(float)*f->blocks);
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
}
}
} else {
int blk_len = OPUS_BLOCK_SIZE(f->size), wlen = OPUS_BLOCK_SIZE(f->size + 1);
int rwin = blk_len - CELT_OVERLAP, lap_dst = (wlen - blk_len - CELT_OVERLAP) >> 1;
opusenc: use float_dsp for non-transient windowing Also fixes transient windowing Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-07-14 05:50:07 +02:00
memset(win, 0, wlen*sizeof(float));
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int ch = 0; ch < f->channels; ch++) {
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
CeltBlock *b = &f->block[ch];
opusenc: use float_dsp for non-transient windowing Also fixes transient windowing Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-07-14 05:50:07 +02:00
/* Overlap */
s->dsp->vector_fmul(temp, b->overlap, ff_celt_window, 128);
memcpy(win + lap_dst, temp, CELT_OVERLAP*sizeof(float));
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
opusenc: use float_dsp for non-transient windowing Also fixes transient windowing Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-07-14 05:50:07 +02:00
/* Samples, flat top window */
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
memcpy(&win[lap_dst + CELT_OVERLAP], b->samples, rwin*sizeof(float));
opusenc: use float_dsp for non-transient windowing Also fixes transient windowing Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-07-14 05:50:07 +02:00
/* Samples, windowed */
s->dsp->vector_fmul_reverse(temp, b->samples + rwin,
avcodec/opustab: Avoid indirection to access ff_celt_window Currently, it is accessed via a pointer (ff_celt_window) exported from opustab.h which points inside a static array (ff_celt_window_padded) in opustab.h. Instead export ff_celt_window_padded directly and make opustab.h a static const pointer pointing inside ff_celt_window_padded. Also mark all the declarations in opustab.h as hidden, so that the compiler knows that ff_celt_window has a fixed offset from the code even when compiling position-independent code. Reviewed-by: Lynne <dev@lynne.ee> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2022-10-03 17:21:53 +02:00
ff_celt_window_padded, 128);
opusenc: use float_dsp for non-transient windowing Also fixes transient windowing Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-07-14 05:50:07 +02:00
memcpy(win + lap_dst + blk_len, temp, CELT_OVERLAP*sizeof(float));
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
opus: convert encoder and decoder to lavu/tx This commit changes both the encoder and decoder to use the new lavu/tx code, which has faster C transforms and more assembly optimizations.
2022-09-24 00:46:44 +02:00
s->tx_fn[f->size](s->tx[f->size], b->coeffs, win, sizeof(float));
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
}
}
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int ch = 0; ch < f->channels; ch++) {
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
CeltBlock *block = &f->block[ch];
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int i = 0; i < CELT_MAX_BANDS; i++) {
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
float ener = 0.0f;
opusenc: remove unused header entries and simplify normalization Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-04-03 14:13:14 +02:00
int band_offset = ff_celt_freq_bands[i] << f->size;
int band_size = ff_celt_freq_range[i] << f->size;
float *coeffs = &block->coeffs[band_offset];
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int j = 0; j < band_size; j++)
opusenc: remove unused header entries and simplify normalization Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-04-03 14:13:14 +02:00
ener += coeffs[j]*coeffs[j];
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
block->lin_energy[i] = sqrtf(ener) + FLT_EPSILON;
ener = 1.0f/block->lin_energy[i];
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int j = 0; j < band_size; j++)
opusenc: remove unused header entries and simplify normalization Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-04-03 14:13:14 +02:00
coeffs[j] *= ener;
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
block->energy[i] = log2f(block->lin_energy[i]) - ff_celt_mean_energy[i];
/* CELT_ENERGY_SILENCE is what the decoder uses and its not -infinity */
block->energy[i] = FFMAX(block->energy[i], CELT_ENERGY_SILENCE);
}
}
}
opus_celt: deduplicate band quantization/dequantization function No point in having the same code twice to do exactly the same thing. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-12-04 10:55:45 +02:00
static void celt_enc_tf(CeltFrame *f, OpusRangeCoder *rc)
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
{
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
int tf_select = 0, diff = 0, tf_changed = 0, tf_select_needed;
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
int bits = f->transient ? 2 : 4;
tf_select_needed = ((f->size && (opus_rc_tell(rc) + bits + 1) <= f->framebits));
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int i = f->start_band; i < f->end_band; i++) {
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
if ((opus_rc_tell(rc) + bits + tf_select_needed) <= f->framebits) {
const int tbit = (diff ^ 1) == f->tf_change[i];
ff_opus_rc_enc_log(rc, tbit, bits);
diff ^= tbit;
tf_changed |= diff;
}
bits = f->transient ? 4 : 5;
}
if (tf_select_needed && ff_celt_tf_select[f->size][f->transient][0][tf_changed] !=
ff_celt_tf_select[f->size][f->transient][1][tf_changed]) {
ff_opus_rc_enc_log(rc, f->tf_select, 1);
tf_select = f->tf_select;
}
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int i = f->start_band; i < f->end_band; i++)
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
f->tf_change[i] = ff_celt_tf_select[f->size][f->transient][tf_select][f->tf_change[i]];
}
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
static void celt_enc_quant_pfilter(OpusRangeCoder *rc, CeltFrame *f)
{
float gain = f->pf_gain;
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
int txval, octave = f->pf_octave, period = f->pf_period, tapset = f->pf_tapset;
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
ff_opus_rc_enc_log(rc, f->pfilter, 1);
if (!f->pfilter)
return;
/* Octave */
txval = FFMIN(octave, 6);
ff_opus_rc_enc_uint(rc, txval, 6);
octave = txval;
/* Period */
txval = av_clip(period - (16 << octave) + 1, 0, (1 << (4 + octave)) - 1);
ff_opus_rc_put_raw(rc, period, 4 + octave);
period = txval + (16 << octave) - 1;
/* Gain */
txval = FFMIN(((int)(gain / 0.09375f)) - 1, 7);
ff_opus_rc_put_raw(rc, txval, 3);
gain = 0.09375f * (txval + 1);
/* Tapset */
if ((opus_rc_tell(rc) + 2) <= f->framebits)
ff_opus_rc_enc_cdf(rc, tapset, ff_celt_model_tapset);
else
tapset = 0;
/* Finally create the coeffs */
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int i = 0; i < 2; i++) {
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
CeltBlock *block = &f->block[i];
block->pf_period_new = FFMAX(period, CELT_POSTFILTER_MINPERIOD);
block->pf_gains_new[0] = gain * ff_celt_postfilter_taps[tapset][0];
block->pf_gains_new[1] = gain * ff_celt_postfilter_taps[tapset][1];
block->pf_gains_new[2] = gain * ff_celt_postfilter_taps[tapset][2];
}
}
opusenc: switch between intra/inter mode for coarse energy Saves around 5kbps. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-04-08 01:53:31 +02:00
static void exp_quant_coarse(OpusRangeCoder *rc, CeltFrame *f,
float last_energy[][CELT_MAX_BANDS], int intra)
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
{
float alpha, beta, prev[2] = { 0, 0 };
opusenc: switch between intra/inter mode for coarse energy Saves around 5kbps. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-04-08 01:53:31 +02:00
const uint8_t *pmod = ff_celt_coarse_energy_dist[f->size][intra];
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
/* Inter is really just differential coding */
if (opus_rc_tell(rc) + 3 <= f->framebits)
opusenc: switch between intra/inter mode for coarse energy Saves around 5kbps. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-04-08 01:53:31 +02:00
ff_opus_rc_enc_log(rc, intra, 3);
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
else
opusenc: switch between intra/inter mode for coarse energy Saves around 5kbps. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-04-08 01:53:31 +02:00
intra = 0;
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
opusenc: switch between intra/inter mode for coarse energy Saves around 5kbps. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-04-08 01:53:31 +02:00
if (intra) {
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
alpha = 0.0f;
opus: simplify coarse energy beta coefficients Just put the subtraction in the table. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-07-18 21:47:30 +02:00
beta = 1.0f - (4915.0f/32768.0f);
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
} else {
alpha = ff_celt_alpha_coef[f->size];
opus: simplify coarse energy beta coefficients Just put the subtraction in the table. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-07-18 21:47:30 +02:00
beta = ff_celt_beta_coef[f->size];
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
}
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int i = f->start_band; i < f->end_band; i++) {
for (int ch = 0; ch < f->channels; ch++) {
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
CeltBlock *block = &f->block[ch];
const int left = f->framebits - opus_rc_tell(rc);
opusenc: switch between intra/inter mode for coarse energy Saves around 5kbps. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-04-08 01:53:31 +02:00
const float last = FFMAX(-9.0f, last_energy[ch][i]);
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
float diff = block->energy[i] - prev[ch] - last*alpha;
int q_en = lrintf(diff);
if (left >= 15) {
ff_opus_rc_enc_laplace(rc, &q_en, pmod[i << 1] << 7, pmod[(i << 1) + 1] << 6);
} else if (left >= 2) {
q_en = av_clip(q_en, -1, 1);
opusenc: fix coarse energy quantization with 2 bits left Fixes CID1400584 Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-16 00:51:37 +02:00
ff_opus_rc_enc_cdf(rc, 2*q_en + 3*(q_en < 0), ff_celt_model_energy_small);
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
} else if (left >= 1) {
q_en = av_clip(q_en, -1, 0);
ff_opus_rc_enc_log(rc, (q_en & 1), 1);
} else q_en = -1;
block->error_energy[i] = q_en - diff;
prev[ch] += beta * q_en;
}
}
}
opus_celt: deduplicate band quantization/dequantization function No point in having the same code twice to do exactly the same thing. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-12-04 10:55:45 +02:00
static void celt_quant_coarse(CeltFrame *f, OpusRangeCoder *rc,
opusenc: switch between intra/inter mode for coarse energy Saves around 5kbps. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-04-08 01:53:31 +02:00
float last_energy[][CELT_MAX_BANDS])
{
uint32_t inter, intra;
OPUS_RC_CHECKPOINT_SPAWN(rc);
exp_quant_coarse(rc, f, last_energy, 1);
intra = OPUS_RC_CHECKPOINT_BITS(rc);
OPUS_RC_CHECKPOINT_ROLLBACK(rc);
exp_quant_coarse(rc, f, last_energy, 0);
inter = OPUS_RC_CHECKPOINT_BITS(rc);
if (inter > intra) { /* Unlikely */
OPUS_RC_CHECKPOINT_ROLLBACK(rc);
exp_quant_coarse(rc, f, last_energy, 1);
}
}
opus_celt: deduplicate band quantization/dequantization function No point in having the same code twice to do exactly the same thing. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-12-04 10:55:45 +02:00
static void celt_quant_fine(CeltFrame *f, OpusRangeCoder *rc)
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
{
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int i = f->start_band; i < f->end_band; i++) {
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
if (!f->fine_bits[i])
continue;
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int ch = 0; ch < f->channels; ch++) {
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
CeltBlock *block = &f->block[ch];
int quant, lim = (1 << f->fine_bits[i]);
float offset, diff = 0.5f - block->error_energy[i];
quant = av_clip(floor(diff*lim), 0, lim - 1);
ff_opus_rc_put_raw(rc, quant, f->fine_bits[i]);
offset = 0.5f - ((quant + 0.5f) * (1 << (14 - f->fine_bits[i])) / 16384.0f);
block->error_energy[i] -= offset;
}
}
}
static void celt_quant_final(OpusEncContext *s, OpusRangeCoder *rc, CeltFrame *f)
{
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int priority = 0; priority < 2; priority++) {
for (int i = f->start_band; i < f->end_band && (f->framebits - opus_rc_tell(rc)) >= f->channels; i++) {
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
if (f->fine_priority[i] != priority || f->fine_bits[i] >= CELT_MAX_FINE_BITS)
continue;
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int ch = 0; ch < f->channels; ch++) {
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
CeltBlock *block = &f->block[ch];
const float err = block->error_energy[i];
const float offset = 0.5f * (1 << (14 - f->fine_bits[i] - 1)) / 16384.0f;
const int sign = FFABS(err + offset) < FFABS(err - offset);
ff_opus_rc_put_raw(rc, sign, 1);
block->error_energy[i] -= offset*(1 - 2*sign);
}
}
}
}
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
static void celt_encode_frame(OpusEncContext *s, OpusRangeCoder *rc,
CeltFrame *f, int index)
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
{
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
ff_opus_rc_enc_init(rc);
ff_opus_psy_celt_frame_init(&s->psyctx, f, index);
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
celt_frame_setup_input(s, f);
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
if (f->silence) {
if (f->framebits >= 16)
ff_opus_rc_enc_log(rc, 1, 15); /* Silence (if using explicit singalling) */
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int ch = 0; ch < s->channels; ch++)
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
memset(s->last_quantized_energy[ch], 0.0f, sizeof(float)*CELT_MAX_BANDS);
return;
}
/* Filters */
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
celt_apply_preemph_filter(s, f);
if (f->pfilter) {
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
ff_opus_rc_enc_log(rc, 0, 15);
celt_enc_quant_pfilter(rc, f);
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
}
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
/* Transform */
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
celt_frame_mdct(s, f);
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
/* Need to handle transient/non-transient switches at any point during analysis */
while (ff_opus_psy_celt_frame_process(&s->psyctx, f, index))
celt_frame_mdct(s, f);
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
ff_opus_rc_enc_init(rc);
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
/* Silence */
ff_opus_rc_enc_log(rc, 0, 15);
/* Pitch filter */
if (!f->start_band && opus_rc_tell(rc) + 16 <= f->framebits)
celt_enc_quant_pfilter(rc, f);
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
/* Transient flag */
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
if (f->size && opus_rc_tell(rc) + 3 <= f->framebits)
ff_opus_rc_enc_log(rc, f->transient, 3);
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
/* Main encoding */
opus: merge encoder and decoder bitallocation functions into one There's no difference apart from which entropy coding functions get called. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-12-30 19:02:54 +02:00
celt_quant_coarse (f, rc, s->last_quantized_energy);
celt_enc_tf (f, rc);
ff_celt_bitalloc (f, rc, 1);
celt_quant_fine (f, rc);
ff_celt_quant_bands(f, rc);
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
/* Anticollapse bit */
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
if (f->anticollapse_needed)
ff_opus_rc_put_raw(rc, f->anticollapse, 1);
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
/* Final per-band energy adjustments from leftover bits */
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
celt_quant_final(s, rc, f);
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int ch = 0; ch < f->channels; ch++) {
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
CeltBlock *block = &f->block[ch];
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int i = 0; i < CELT_MAX_BANDS; i++)
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
s->last_quantized_energy[ch][i] = block->energy[i] + block->error_energy[i];
}
}
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
static inline int write_opuslacing(uint8_t *dst, int v)
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
{
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
dst[0] = FFMIN(v - FFALIGN(v - 255, 4), v);
dst[1] = v - dst[0] >> 2;
return 1 + (v >= 252);
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
}
static void opus_packet_assembler(OpusEncContext *s, AVPacket *avpkt)
{
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
int offset, fsize_needed;
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
/* Write toc */
opus_gen_toc(s, avpkt->data, &offset, &fsize_needed);
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
/* Frame sizes if needed */
if (fsize_needed) {
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int i = 0; i < s->packet.frames - 1; i++) {
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
offset += write_opuslacing(avpkt->data + offset,
s->frame[i].framebits >> 3);
}
}
/* Packets */
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int i = 0; i < s->packet.frames; i++) {
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
ff_opus_rc_enc_end(&s->rc[i], avpkt->data + offset,
s->frame[i].framebits >> 3);
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
offset += s->frame[i].framebits >> 3;
}
avpkt->size = offset;
}
/* Used as overlap for the first frame and padding for the last encoded packet */
static AVFrame *spawn_empty_frame(OpusEncContext *s)
{
AVFrame *f = av_frame_alloc();
opus: convert to new channel layout API Signed-off-by: James Almer <jamrial@gmail.com>
2019-05-28 14:38:10 +02:00
int ret;
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
if (!f)
return NULL;
f->format = s->avctx->sample_fmt;
f->nb_samples = s->avctx->frame_size;
opus: convert to new channel layout API Signed-off-by: James Almer <jamrial@gmail.com>
2019-05-28 14:38:10 +02:00
ret = av_channel_layout_copy(&f->ch_layout, &s->avctx->ch_layout);
if (ret < 0) {
av_frame_free(&f);
return NULL;
}
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
if (av_frame_get_buffer(f, 4)) {
av_frame_free(&f);
return NULL;
}
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int i = 0; i < s->channels; i++) {
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
size_t bps = av_get_bytes_per_sample(f->format);
memset(f->extended_data[i], 0, bps*f->nb_samples);
}
return f;
}
static int opus_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
const AVFrame *frame, int *got_packet_ptr)
{
OpusEncContext *s = avctx->priv_data;
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
int ret, frame_size, alloc_size = 0;
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
if (frame) { /* Add new frame to queue */
if ((ret = ff_af_queue_add(&s->afq, frame)) < 0)
return ret;
ff_bufqueue_add(avctx, &s->bufqueue, av_frame_clone(frame));
} else {
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
ff_opus_psy_signal_eof(&s->psyctx);
opusenc: fix infinite loop if flushing encoder upon init The issue is that the afq still has samples as on init it counts the overlap used as a delay to adjust the PTS it generates, hence we can't rely on it right after init. So just check to see if any frames have been encoded. frame_number can't be anything but 0 right after init and can only be set by lavc. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-12-12 22:05:33 +02:00
if (!s->afq.remaining_samples || !avctx->frame_number)
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
return 0; /* We've been flushed and there's nothing left to encode */
}
/* Run the psychoacoustic system */
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
if (ff_opus_psy_process(&s->psyctx, &s->packet))
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
return 0;
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
frame_size = OPUS_BLOCK_SIZE(s->packet.framesize);
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
if (!frame) {
/* This can go negative, that's not a problem, we only pad if positive */
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
int pad_empty = s->packet.frames*(frame_size/s->avctx->frame_size) - s->bufqueue.available + 1;
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
/* Pad with empty 2.5 ms frames to whatever framesize was decided,
* this should only happen at the very last flush frame. The frames
* allocated here will be freed (because they have no other references)
* after they get used by celt_frame_setup_input() */
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int i = 0; i < pad_empty; i++) {
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
AVFrame *empty = spawn_empty_frame(s);
if (!empty)
return AVERROR(ENOMEM);
ff_bufqueue_add(avctx, &s->bufqueue, empty);
}
}
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int i = 0; i < s->packet.frames; i++) {
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
celt_encode_frame(s, &s->rc[i], &s->frame[i], i);
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
alloc_size += s->frame[i].framebits >> 3;
}
/* Worst case toc + the frame lengths if needed */
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
alloc_size += 2 + s->packet.frames*2;
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
avcodec/encode: Always use intermediate buffer in ff_alloc_packet2() Up until now, ff_alloc_packet2() has a min_size parameter: It is supposed to be a lower bound on the final size of the packet to allocate. If it is not too far from the upper bound (namely, if it is at least half the upper bound), then ff_alloc_packet2() already allocates the final, already refcounted packet; if it is not, then the packet is not refcounted and its data only points to a buffer owned by the AVCodecContext (in this case, the packet will be made refcounted in encode_simple_internal() in libavcodec/encode.c). The goal of this was to avoid data copies and intermediate buffers if one has a precise lower bound. Yet those encoders for which precise lower bounds exist have recently been switched to ff_get_encode_buffer() (which automatically allocates final buffers), leaving only two encoders to actually set the min_size to something else than zero (namely aliaspixenc and hapenc). Both of these encoders use a very low lower bound that is not helpful in any nontrivial case. This commit therefore removes the min_size parameter as well as the codepath in ff_alloc_packet2() for the allocation of final buffers. Furthermore, the function has been renamed to ff_alloc_packet() and moved to encode.h alongside ff_get_encode_buffer(). Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2021-05-11 15:17:13 +02:00
if ((ret = ff_alloc_packet(avctx, avpkt, alloc_size)) < 0)
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
return ret;
/* Assemble packet */
opus_packet_assembler(s, avpkt);
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
/* Update the psychoacoustic system */
avcodec/opusenc_psy: Remove unused function parameter Reviewed-by: Lynne <dev@lynne.ee> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2022-10-07 16:11:45 +02:00
ff_opus_psy_postencode_update(&s->psyctx, s->frame);
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
/* Remove samples from queue and skip if needed */
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
ff_af_queue_remove(&s->afq, s->packet.frames*frame_size, &avpkt->pts, &avpkt->duration);
if (s->packet.frames*frame_size > avpkt->duration) {
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
uint8_t *side = av_packet_new_side_data(avpkt, AV_PKT_DATA_SKIP_SAMPLES, 10);
if (!side)
return AVERROR(ENOMEM);
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
AV_WL32(&side[4], s->packet.frames*frame_size - avpkt->duration + 120);
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
}
*got_packet_ptr = 1;
return 0;
}
static av_cold int opus_encode_end(AVCodecContext *avctx)
{
OpusEncContext *s = avctx->priv_data;
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int i = 0; i < CELT_BLOCK_NB; i++)
opus: convert encoder and decoder to lavu/tx This commit changes both the encoder and decoder to use the new lavu/tx code, which has faster C transforms and more assembly optimizations.
2022-09-24 00:46:44 +02:00
av_tx_uninit(&s->tx[i]);
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
opus_pvq: port to allow for SIMD functions Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-05-10 07:47:44 +02:00
ff_celt_pvq_uninit(&s->pvq);
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
av_freep(&s->dsp);
av_freep(&s->frame);
av_freep(&s->rc);
ff_af_queue_close(&s->afq);
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
ff_opus_psy_end(&s->psyctx);
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
ff_bufqueue_discard_all(&s->bufqueue);
return 0;
}
static av_cold int opus_encode_init(AVCodecContext *avctx)
{
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
int ret, max_frames;
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
OpusEncContext *s = avctx->priv_data;
s->avctx = avctx;
opus: convert to new channel layout API Signed-off-by: James Almer <jamrial@gmail.com>
2019-05-28 14:38:10 +02:00
s->channels = avctx->ch_layout.nb_channels;
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
/* Opus allows us to change the framesize on each packet (and each packet may
* have multiple frames in it) but we can't change the codec's frame size on
* runtime, so fix it to the lowest possible number of samples and use a queue
* to accumulate AVFrames until we have enough to encode whatever the encoder
* decides is the best */
avctx->frame_size = 120;
/* Initial padding will change if SILK is ever supported */
avctx->initial_padding = 120;
if (!avctx->bit_rate) {
int coupled = ff_opus_default_coupled_streams[s->channels - 1];
avctx->bit_rate = coupled*(96000) + (s->channels - coupled*2)*(48000);
} else if (avctx->bit_rate < 6000 || avctx->bit_rate > 255000 * s->channels) {
int64_t clipped_rate = av_clip(avctx->bit_rate, 6000, 255000 * s->channels);
libavcodec/opusenc: use correct format specifiers Squelches the following compiler warnings: libavcodec/opusenc.c:1051:16: warning: format specifies type 'long' but the argument has type 'long long' [-Wformat] avctx->bit_rate/1000, clipped_rate/1000); ^~~~~~~~~~~~~~~~~~~~ libavcodec/opusenc.c:1051:38: warning: format specifies type 'long' but the argument has type 'long long' [-Wformat] avctx->bit_rate/1000, clipped_rate/1000); ^~~~~~~~~~~~~~~~~
2017-03-26 20:48:28 +02:00
av_log(avctx, AV_LOG_ERROR, "Unsupported bitrate %"PRId64" kbps, clipping to %"PRId64" kbps\n",
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
avctx->bit_rate/1000, clipped_rate/1000);
avctx->bit_rate = clipped_rate;
}
/* Extradata */
avctx->extradata_size = 19;
avctx->extradata = av_malloc(avctx->extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
if (!avctx->extradata)
return AVERROR(ENOMEM);
opus_write_extradata(avctx);
ff_af_queue_init(avctx, &s->afq);
opus_celt: deduplicate band quantization/dequantization function No point in having the same code twice to do exactly the same thing. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-12-04 10:55:45 +02:00
if ((ret = ff_celt_pvq_init(&s->pvq, 1)) < 0)
opus_pvq: port to allow for SIMD functions Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-05-10 07:47:44 +02:00
return ret;
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
if (!(s->dsp = avpriv_float_dsp_alloc(avctx->flags & AV_CODEC_FLAG_BITEXACT)))
return AVERROR(ENOMEM);
/* I have no idea why a base scaling factor of 68 works, could be the twiddles */
opus: convert encoder and decoder to lavu/tx This commit changes both the encoder and decoder to use the new lavu/tx code, which has faster C transforms and more assembly optimizations.
2022-09-24 00:46:44 +02:00
for (int i = 0; i < CELT_BLOCK_NB; i++) {
const float scale = 68 << (CELT_BLOCK_NB - 1 - i);
if ((ret = av_tx_init(&s->tx[i], &s->tx_fn[i], AV_TX_FLOAT_MDCT, 0, 15 << (i + 3), &scale, 0)))
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
return AVERROR(ENOMEM);
opus: convert encoder and decoder to lavu/tx This commit changes both the encoder and decoder to use the new lavu/tx code, which has faster C transforms and more assembly optimizations.
2022-09-24 00:46:44 +02:00
}
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
/* Zero out previous energy (matters for inter first frame) */
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int ch = 0; ch < s->channels; ch++)
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
memset(s->last_quantized_energy[ch], 0.0f, sizeof(float)*CELT_MAX_BANDS);
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
/* Allocate an empty frame to use as overlap for the first frame of audio */
ff_bufqueue_add(avctx, &s->bufqueue, spawn_empty_frame(s));
if (!ff_bufqueue_peek(&s->bufqueue, 0))
return AVERROR(ENOMEM);
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
if ((ret = ff_opus_psy_init(&s->psyctx, s->avctx, &s->bufqueue, &s->options)))
return ret;
/* Frame structs and range coder buffers */
max_frames = ceilf(FFMIN(s->options.max_delay_ms, 120.0f)/2.5f);
s->frame = av_malloc(max_frames*sizeof(CeltFrame));
if (!s->frame)
return AVERROR(ENOMEM);
s->rc = av_malloc(max_frames*sizeof(OpusRangeCoder));
if (!s->rc)
return AVERROR(ENOMEM);
opusenc: use for loops with declarations Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2018-05-18 03:38:08 +02:00
for (int i = 0; i < max_frames; i++) {
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
s->frame[i].dsp = s->dsp;
s->frame[i].avctx = s->avctx;
s->frame[i].seed = 0;
s->frame[i].pvq = s->pvq;
opusenc: add apply_phase_inv option By popular request. Does the same as in libopusenc.
2020-05-25 15:41:33 +02:00
s->frame[i].apply_phase_inv = s->options.apply_phase_inv;
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
s->frame[i].block[0].emph_coeff = s->frame[i].block[1].emph_coeff = 0.0f;
}
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
return 0;
}
#define OPUSENC_FLAGS AV_OPT_FLAG_ENCODING_PARAM | AV_OPT_FLAG_AUDIO_PARAM
static const AVOption opusenc_options[] = {
opusenc: implement a psychoacoustic system This commit implements a psychoacoustic system for the native Opus encoder. Its unlike any other psychoacoustic system known since its capable of using a lookahead to make better choices on how to treat the current frame and how many bits to allocate for it (and future frames). Also, whilst the main bulk of the analysis function has to run in a single thread, the per-frame anaylsis functions does not modify the main psychoacoustic context, so in the future it will be fairly trivial to run those as slice threads. Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-09-23 01:38:37 +02:00
{ "opus_delay", "Maximum delay in milliseconds", offsetof(OpusEncContext, options.max_delay_ms), AV_OPT_TYPE_FLOAT, { .dbl = OPUS_MAX_LOOKAHEAD }, 2.5f, OPUS_MAX_LOOKAHEAD, OPUSENC_FLAGS, "max_delay_ms" },
opusenc: add apply_phase_inv option By popular request. Does the same as in libopusenc.
2020-05-25 15:41:33 +02:00
{ "apply_phase_inv", "Apply intensity stereo phase inversion", offsetof(OpusEncContext, options.apply_phase_inv), AV_OPT_TYPE_BOOL, { .i64 = 1 }, 0, 1, OPUSENC_FLAGS, "apply_phase_inv" },
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
{ NULL },
};
static const AVClass opusenc_class = {
.class_name = "Opus encoder",
.item_name = av_default_item_name,
.option = opusenc_options,
.version = LIBAVUTIL_VERSION_INT,
};
avcodec/codec_internal: Rename AVCodecDefault->FFCodecDefault This structure is no longer declared in a public header, so using an FF-prefix is more appropriate. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2022-03-16 22:26:11 +02:00
static const FFCodecDefault opusenc_defaults[] = {
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
{ "b", "0" },
{ "compression_level", "10" },
{ NULL },
};
avcodec/codec_internal: Add FFCodec, hide internal part of AVCodec Up until now, codec.h contains both public and private parts of AVCodec. This exposes the internals of AVCodec to users and leads them into the temptation of actually using them and forces us to forward-declare structures and types that users can't use at all. This commit changes this by adding a new structure FFCodec to codec_internal.h that extends AVCodec, i.e. contains the public AVCodec as first member; the private fields of AVCodec are moved to this structure, leaving codec.h clean. Reviewed-by: Anton Khirnov <anton@khirnov.net> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2022-03-16 22:09:54 +02:00
const FFCodec ff_opus_encoder = {
.p.name = "opus",
avcodec/codec_internal: Add macro to set AVCodec.long_name It reduces typing: Before this patch, there were 105 codecs whose long_name-definition exceeded the 80 char line length limit. Now there are only nine of them. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2022-08-29 13:38:02 +02:00
CODEC_LONG_NAME("Opus"),
avcodec/codec_internal: Add FFCodec, hide internal part of AVCodec Up until now, codec.h contains both public and private parts of AVCodec. This exposes the internals of AVCodec to users and leads them into the temptation of actually using them and forces us to forward-declare structures and types that users can't use at all. This commit changes this by adding a new structure FFCodec to codec_internal.h that extends AVCodec, i.e. contains the public AVCodec as first member; the private fields of AVCodec are moved to this structure, leaving codec.h clean. Reviewed-by: Anton Khirnov <anton@khirnov.net> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2022-03-16 22:09:54 +02:00
.p.type = AVMEDIA_TYPE_AUDIO,
.p.id = AV_CODEC_ID_OPUS,
avcodec: Make ff_alloc_packet() based encoders accept user buffers Up until now, these encoders received non-refcounted packets (whose data was owned by the corresponding AVCodecContext) from ff_alloc_packet(); these packets were made refcounted lateron by av_packet_make_refcounted() generically. This commit makes these encoders accept user-supplied buffers by replacing av_packet_make_refcounted() with an equivalent function that is based upon get_encode_buffer(). (I am pretty certain that one can also set the flag for mpegvideo- based encoders, but I want to double-check this later. What is certain is that it reallocates the buffer owned by the AVCodecContext which should maybe be moved to encode.c, so that proresenc_kostya.c and ttaenc.c can make use of it, too.) Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2021-05-11 20:52:13 +02:00
.p.capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_DELAY |
AV_CODEC_CAP_SMALL_LAST_FRAME | AV_CODEC_CAP_EXPERIMENTAL,
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
.defaults = opusenc_defaults,
avcodec/codec_internal: Add FFCodec, hide internal part of AVCodec Up until now, codec.h contains both public and private parts of AVCodec. This exposes the internals of AVCodec to users and leads them into the temptation of actually using them and forces us to forward-declare structures and types that users can't use at all. This commit changes this by adding a new structure FFCodec to codec_internal.h that extends AVCodec, i.e. contains the public AVCodec as first member; the private fields of AVCodec are moved to this structure, leaving codec.h clean. Reviewed-by: Anton Khirnov <anton@khirnov.net> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2022-03-16 22:09:54 +02:00
.p.priv_class = &opusenc_class,
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
.priv_data_size = sizeof(OpusEncContext),
.init = opus_encode_init,
avcodec/codec_internal: Use union for FFCodec decode/encode callbacks This is possible, because every given FFCodec has to implement exactly one of these. Doing so decreases sizeof(FFCodec) and therefore decreases the size of the binary. Notice that in case of position-independent code the decrease is in .data.rel.ro, so that this translates to decreased memory consumption. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2022-03-30 23:28:24 +02:00
FF_CODEC_ENCODE_CB(opus_encode_frame),
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
.close = opus_encode_end,
avcodec: Make init-threadsafety the default and remove FF_CODEC_CAP_INIT_THREADSAFE All our native codecs are already init-threadsafe (only wrappers for external libraries and hwaccels are typically not marked as init-threadsafe yet), so it is only natural for this to also be the default state. Reviewed-by: Anton Khirnov <anton@khirnov.net> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2022-07-10 00:05:45 +02:00
.caps_internal = FF_CODEC_CAP_INIT_CLEANUP,
avcodec/codec_internal: Add FFCodec, hide internal part of AVCodec Up until now, codec.h contains both public and private parts of AVCodec. This exposes the internals of AVCodec to users and leads them into the temptation of actually using them and forces us to forward-declare structures and types that users can't use at all. This commit changes this by adding a new structure FFCodec to codec_internal.h that extends AVCodec, i.e. contains the public AVCodec as first member; the private fields of AVCodec are moved to this structure, leaving codec.h clean. Reviewed-by: Anton Khirnov <anton@khirnov.net> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2022-03-16 22:09:54 +02:00
.p.supported_samplerates = (const int []){ 48000, 0 },
avcodec/codec_internal: Avoid deprecation warnings for channel_layouts AVCodec.channel_layouts is deprecated and Clang (unlike GCC) warns when setting this field in a codec definition. Fortunately, Clang (unlike GCC) allows to use FF_DISABLE_DEPRECATION_WARNINGS inside a definition (of an FFCodec), so that one can create simple macros to set AVCodec.channel_layouts that also suppress deprecation warnings for Clang. (Notice that some of the codec definitions were already inside FF_DISABLE/ENABLE_DEPRECATION_WARNINGS (that were not guarded by FF_API_OLD_CHANNEL_LAYOUT); these have been removed. Also notice that setting AVCodec.channel_layouts was not guarded by FF_API_OLD_CHANNEL_LAYOUT either, so testing disabling it it without removing all the codeblocks would not have worked.) Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2022-09-26 01:10:51 +02:00
CODEC_OLD_CHANNEL_LAYOUTS(AV_CH_LAYOUT_MONO, AV_CH_LAYOUT_STEREO)
avcodec/codec_internal: Add FFCodec, hide internal part of AVCodec Up until now, codec.h contains both public and private parts of AVCodec. This exposes the internals of AVCodec to users and leads them into the temptation of actually using them and forces us to forward-declare structures and types that users can't use at all. This commit changes this by adding a new structure FFCodec to codec_internal.h that extends AVCodec, i.e. contains the public AVCodec as first member; the private fields of AVCodec are moved to this structure, leaving codec.h clean. Reviewed-by: Anton Khirnov <anton@khirnov.net> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2022-03-16 22:09:54 +02:00
.p.ch_layouts = (const AVChannelLayout []){ AV_CHANNEL_LAYOUT_MONO,
opus: convert to new channel layout API Signed-off-by: James Almer <jamrial@gmail.com>
2019-05-28 14:38:10 +02:00
AV_CHANNEL_LAYOUT_STEREO, { 0 } },
avcodec/codec_internal: Add FFCodec, hide internal part of AVCodec Up until now, codec.h contains both public and private parts of AVCodec. This exposes the internals of AVCodec to users and leads them into the temptation of actually using them and forces us to forward-declare structures and types that users can't use at all. This commit changes this by adding a new structure FFCodec to codec_internal.h that extends AVCodec, i.e. contains the public AVCodec as first member; the private fields of AVCodec are moved to this structure, leaving codec.h clean. Reviewed-by: Anton Khirnov <anton@khirnov.net> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2022-03-16 22:09:54 +02:00
.p.sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_FLTP,
opus: add a native Opus encoder This marks the first time anyone has written an Opus encoder without using any libopus code. The aim of the encoder is to prove how far the format can go by writing the craziest encoder for it. Right now the encoder's basic, it only supports CBR encoding, however internally every single feature the CELT layer has is implemented (except the pitch pre-filter which needs to work well with the rest of whatever gets implemented). Psychoacoustic and rate control systems are under development. The encoder takes in frames of 120 samples and depending on the value of opus_delay the plan is to use the extra buffered frames as lookahead. Right now the encoder will pick the nearest largest legal frame size and won't use the lookahead, but that'll change once there's a psychoacoustic system. Even though its a pretty basic encoder its already outperforming any other native encoder FFmpeg has by a huge amount. The PVQ search algorithm is faster and more accurate than libopus's algorithm so the encoder's performance is close to that of libopus at zero complexity (libopus has more SIMD). The algorithm might be ported to libopus or other codecs using PVQ in the future. The encoder still has a few minor bugs, like desyncs at ultra low bitrates (below 9kbps with 20ms frames). Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-02-11 02:25:08 +02:00
AV_SAMPLE_FMT_NONE },
};
Reference in New Issue Copy Permalink