From a5789be297faa409c00eb62c30afd0965f4d70ea Mon Sep 17 00:00:00 2001 From: Paul B Mahol Date: Sat, 23 Sep 2023 16:49:25 +0200 Subject: [PATCH 1/2] avcodec: add QOA decoder and parser Signed-off-by: Paul B Mahol --- libavcodec/Makefile | 2 + libavcodec/allcodecs.c | 1 + libavcodec/codec_desc.c | 7 ++ libavcodec/codec_id.h | 1 + libavcodec/parsers.c | 1 + libavcodec/qoa_parser.c | 89 ++++++++++++++++++++ libavcodec/qoadec.c | 175 ++++++++++++++++++++++++++++++++++++++++ 7 files changed, 276 insertions(+) create mode 100644 libavcodec/qoa_parser.c create mode 100644 libavcodec/qoadec.c diff --git a/libavcodec/Makefile b/libavcodec/Makefile index a1eb83f92d..c70d27a5c8 100644 --- a/libavcodec/Makefile +++ b/libavcodec/Makefile @@ -623,6 +623,7 @@ OBJS-$(CONFIG_QCELP_DECODER) += qcelpdec.o \ OBJS-$(CONFIG_QDM2_DECODER) += qdm2.o OBJS-$(CONFIG_QDMC_DECODER) += qdmc.o OBJS-$(CONFIG_QDRAW_DECODER) += qdrw.o +OBJS-$(CONFIG_QOA_DECODER) += qoadec.o OBJS-$(CONFIG_QOI_DECODER) += qoidec.o OBJS-$(CONFIG_QOI_ENCODER) += qoienc.o OBJS-$(CONFIG_QPEG_DECODER) += qpeg.o @@ -1205,6 +1206,7 @@ OBJS-$(CONFIG_OPUS_PARSER) += opus_parser.o opus_parse.o \ vorbis_data.o OBJS-$(CONFIG_PNG_PARSER) += png_parser.o OBJS-$(CONFIG_PNM_PARSER) += pnm_parser.o pnm.o +OBJS-$(CONFIG_QOA_PARSER) += qoa_parser.o OBJS-$(CONFIG_QOI_PARSER) += qoi_parser.o OBJS-$(CONFIG_RV34_PARSER) += rv34_parser.o OBJS-$(CONFIG_SBC_PARSER) += sbc_parser.o diff --git a/libavcodec/allcodecs.c b/libavcodec/allcodecs.c index 4ea6515c6c..86d0972742 100644 --- a/libavcodec/allcodecs.c +++ b/libavcodec/allcodecs.c @@ -523,6 +523,7 @@ extern const FFCodec ff_paf_audio_decoder; extern const FFCodec ff_qcelp_decoder; extern const FFCodec ff_qdm2_decoder; extern const FFCodec ff_qdmc_decoder; +extern const FFCodec ff_qoa_decoder; extern const FFCodec ff_ra_144_encoder; extern const FFCodec ff_ra_144_decoder; extern const FFCodec ff_ra_288_decoder; diff --git a/libavcodec/codec_desc.c b/libavcodec/codec_desc.c index 3ddd2fd3ee..fe838b49cb 100644 --- a/libavcodec/codec_desc.c +++ b/libavcodec/codec_desc.c @@ -3434,6 +3434,13 @@ static const AVCodecDescriptor codec_descriptors[] = { .long_name = NULL_IF_CONFIG_SMALL("Sonarc (Speech Compression)"), .props = AV_CODEC_PROP_INTRA_ONLY | AV_CODEC_PROP_LOSSLESS, }, + { + .id = AV_CODEC_ID_QOA, + .type = AVMEDIA_TYPE_AUDIO, + .name = "qoa", + .long_name = NULL_IF_CONFIG_SMALL("QOA (Quite OK Audio)"), + .props = AV_CODEC_PROP_INTRA_ONLY | AV_CODEC_PROP_LOSSY, + }, /* subtitle codecs */ { diff --git a/libavcodec/codec_id.h b/libavcodec/codec_id.h index 73352f5461..c313536f7b 100644 --- a/libavcodec/codec_id.h +++ b/libavcodec/codec_id.h @@ -546,6 +546,7 @@ enum AVCodecID { AV_CODEC_ID_AC4, AV_CODEC_ID_OSQ, AV_CODEC_ID_SONARC, + AV_CODEC_ID_QOA, /* subtitle codecs */ AV_CODEC_ID_FIRST_SUBTITLE = 0x17000, ///< A dummy ID pointing at the start of subtitle codecs. diff --git a/libavcodec/parsers.c b/libavcodec/parsers.c index 5128009cd4..2ed96120f1 100644 --- a/libavcodec/parsers.c +++ b/libavcodec/parsers.c @@ -65,6 +65,7 @@ extern const AVCodecParser ff_mpegvideo_parser; extern const AVCodecParser ff_opus_parser; extern const AVCodecParser ff_png_parser; extern const AVCodecParser ff_pnm_parser; +extern const AVCodecParser ff_qoa_parser; extern const AVCodecParser ff_qoi_parser; extern const AVCodecParser ff_rv34_parser; extern const AVCodecParser ff_sbc_parser; diff --git a/libavcodec/qoa_parser.c b/libavcodec/qoa_parser.c new file mode 100644 index 0000000000..0e5537a326 --- /dev/null +++ b/libavcodec/qoa_parser.c @@ -0,0 +1,89 @@ +/* + * QOA parser + * + * 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 + */ + +#include "parser.h" + +typedef struct QOAParseContext { + ParseContext pc; + int frame_size; + int remaining; + int duration; +} QOAParseContext; + +static int qoa_parse(AVCodecParserContext *s, AVCodecContext *avctx, + const uint8_t **poutbuf, int *poutbuf_size, + const uint8_t *buf, int buf_size) +{ + QOAParseContext *qoa = s->priv_data; + uint64_t state = qoa->pc.state64; + int next = END_NOT_FOUND, i = 0; + + *poutbuf_size = 0; + *poutbuf = NULL; + + if (s->flags & PARSER_FLAG_COMPLETE_FRAMES) { + next = buf_size; + } else { + if (!qoa->frame_size) { + for (; i < buf_size; i++) { + state = (state << 8) | buf[i]; + if (((state & 0xFFFF) > 0 && (state >> 56))) { + qoa->frame_size = state & 0xFFFF; + qoa->duration = (state >> 16) & 0xFFFF; + break; + } + } + } + + if (qoa->frame_size) { + if (!qoa->remaining) + qoa->remaining = qoa->frame_size; + if (qoa->remaining <= buf_size) { + next = qoa->remaining; + qoa->remaining = 0; + qoa->frame_size = 0; + state = 0; + } else { + qoa->remaining -= buf_size; + } + } + + qoa->pc.state64 = state; + if (ff_combine_frame(&qoa->pc, next, &buf, &buf_size) < 0) { + *poutbuf = NULL; + *poutbuf_size = 0; + return buf_size; + } + + s->duration = qoa->duration; + } + + *poutbuf = buf; + *poutbuf_size = buf_size; + + return next; +} + +const AVCodecParser ff_qoa_parser = { + .codec_ids = { AV_CODEC_ID_QOA }, + .priv_data_size = sizeof(QOAParseContext), + .parser_parse = qoa_parse, + .parser_close = ff_parse_close, +}; diff --git a/libavcodec/qoadec.c b/libavcodec/qoadec.c new file mode 100644 index 0000000000..9b2abae833 --- /dev/null +++ b/libavcodec/qoadec.c @@ -0,0 +1,175 @@ +/* + * QOA decoder + * + * 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 + */ + +#include "avcodec.h" +#include "codec_internal.h" +#include "decode.h" +#include "get_bits.h" +#include "bytestream.h" +#include "mathops.h" + +#define QOA_SLICE_LEN 20 +#define QOA_LMS_LEN 4 + +typedef struct QOAChannel { + int history[QOA_LMS_LEN]; + int weights[QOA_LMS_LEN]; +} QOAChannel; + +typedef struct QOAContext { + QOAChannel *ch; +} QOAContext; + +static const int16_t qoa_dequant_tab[16][8] = { + { 1, -1, 3, -3, 5, -5, 7, -7}, + { 5, -5, 18, -18, 32, -32, 49, -49}, + { 16, -16, 53, -53, 95, -95, 147, -147}, + { 34, -34, 113, -113, 203, -203, 315, -315}, + { 63, -63, 210, -210, 378, -378, 588, -588}, + { 104, -104, 345, -345, 621, -621, 966, -966}, + { 158, -158, 528, -528, 950, -950, 1477, -1477}, + { 228, -228, 760, -760, 1368, -1368, 2128, -2128}, + { 316, -316, 1053, -1053, 1895, -1895, 2947, -2947}, + { 422, -422, 1405, -1405, 2529, -2529, 3934, -3934}, + { 548, -548, 1828, -1828, 3290, -3290, 5117, -5117}, + { 696, -696, 2320, -2320, 4176, -4176, 6496, -6496}, + { 868, -868, 2893, -2893, 5207, -5207, 8099, -8099}, + {1064, -1064, 3548, -3548, 6386, -6386, 9933, -9933}, + {1286, -1286, 4288, -4288, 7718, -7718, 12005, -12005}, + {1536, -1536, 5120, -5120, 9216, -9216, 14336, -14336}, +}; + +static av_cold int qoa_decode_init(AVCodecContext *avctx) +{ + QOAContext *s = avctx->priv_data; + + avctx->sample_fmt = AV_SAMPLE_FMT_S16; + + s->ch = av_calloc(avctx->ch_layout.nb_channels, sizeof(*s->ch)); + if (!s->ch) + return AVERROR(ENOMEM); + + return 0; +} + +static int qoa_lms_predict(QOAChannel *lms) +{ + int prediction = 0; + for (int i = 0; i < QOA_LMS_LEN; i++) + prediction += lms->weights[i] * lms->history[i]; + return prediction >> 13; +} + +static void qoa_lms_update(QOAChannel *lms, int sample, int residual) +{ + int delta = residual >> 4; + for (int i = 0; i < QOA_LMS_LEN; i++) + lms->weights[i] += lms->history[i] < 0 ? -delta : delta; + for (int i = 0; i < QOA_LMS_LEN-1; i++) + lms->history[i] = lms->history[i+1]; + lms->history[QOA_LMS_LEN-1] = sample; +} + +static int qoa_decode_frame(AVCodecContext *avctx, AVFrame *frame, + int *got_frame_ptr, AVPacket *avpkt) +{ + QOAContext *s = avctx->priv_data; + int ret, frame_size, nb_channels; + GetByteContext gb; + int16_t *samples; + + bytestream2_init(&gb, avpkt->data, avpkt->size); + + nb_channels = bytestream2_get_byte(&gb); + if (avctx->ch_layout.nb_channels != nb_channels) + return AVERROR_INVALIDDATA; + + avctx->sample_rate = bytestream2_get_be24(&gb); + frame->nb_samples = bytestream2_get_be16(&gb); + frame_size = bytestream2_get_be16(&gb); + if (frame_size > avpkt->size) + return AVERROR_INVALIDDATA; + + if (frame_size < 8 + QOA_LMS_LEN * 4 * nb_channels + + 8LL * frame->nb_samples * nb_channels / QOA_SLICE_LEN) + return AVERROR_INVALIDDATA; + + if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) + return ret; + samples = (int16_t *)frame->data[0]; + + for (int ch = 0; ch < nb_channels; ch++) { + QOAChannel *qch = &s->ch[ch]; + + for (int n = 0; n < QOA_LMS_LEN; n++) + qch->history[n] = sign_extend(bytestream2_get_be16u(&gb), 16); + for (int n = 0; n < QOA_LMS_LEN; n++) + qch->weights[n] = sign_extend(bytestream2_get_be16u(&gb), 16); + } + + for (int sample_index = 0; sample_index < frame->nb_samples * nb_channels; + sample_index += QOA_SLICE_LEN) { + for (int ch = 0; ch < nb_channels; ch++) { + QOAChannel *lms = &s->ch[ch]; + uint64_t slice = bytestream2_get_be64u(&gb); + int scalefactor = (slice >> 60) & 0xf; + int slice_start = sample_index * nb_channels + ch; + int slice_end = av_clip(sample_index + QOA_SLICE_LEN, 0, frame->nb_samples) * nb_channels + ch; + + for (int si = slice_start; si < slice_end; si += nb_channels) { + int predicted = qoa_lms_predict(lms); + int quantized = (slice >> 57) & 0x7; + int dequantized = qoa_dequant_tab[scalefactor][quantized]; + int reconstructed = av_clip_int16(predicted + dequantized); + + samples[si] = reconstructed; + slice <<= 3; + + qoa_lms_update(lms, reconstructed, dequantized); + } + } + } + + *got_frame_ptr = 1; + + return avpkt->size; +} + +static av_cold int qoa_decode_end(AVCodecContext *avctx) +{ + QOAContext *s = avctx->priv_data; + av_freep(&s->ch); + return 0; +} + +const FFCodec ff_qoa_decoder = { + .p.name = "qoa", + CODEC_LONG_NAME("QOA (Quite OK Audio)"), + .p.type = AVMEDIA_TYPE_AUDIO, + .p.id = AV_CODEC_ID_QOA, + .priv_data_size = sizeof(QOAContext), + .init = qoa_decode_init, + FF_CODEC_DECODE_CB(qoa_decode_frame), + .close = qoa_decode_end, + .p.capabilities = AV_CODEC_CAP_CHANNEL_CONF | + AV_CODEC_CAP_DR1, + .p.sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_S16, + AV_SAMPLE_FMT_NONE }, +}; -- 2.42.0