diff options
author | Michael Niedermayer <michaelni@gmx.at> | 2011-10-24 01:01:21 +0200 |
---|---|---|
committer | Michael Niedermayer <michaelni@gmx.at> | 2011-10-24 01:01:21 +0200 |
commit | 2b0cdb7364fbe1e8a7e97d36ba36dc67fb24095e (patch) | |
tree | 696306b4a66d7598f2795f9fbae4557135ff9104 /libavcodec | |
parent | 42d44ec306d0656ace5cea63867f70fa686141cb (diff) | |
parent | 16ad77b357ebbe74a7bc9568904c328a2722651e (diff) | |
download | ffmpeg-2b0cdb7364fbe1e8a7e97d36ba36dc67fb24095e.tar.gz |
Merge remote-tracking branch 'qatar/master'
* qatar/master:
Move id3v2 tag writing to a separate file.
swscale: add missing colons to x86 assembly yuv2planeX.
g722: split decoder and encoder into separate files
cosmetics: remove extra spaces before end-of-statement semi-colons
vorbisdec: check output buffer size before writing output
wavpack: calculate bpp using av_get_bytes_per_sample()
ac3enc: Set max value for mode options correctly
lavc: move get_b_cbp() from h263.h to mpeg4videoenc.c
mpeg12: move closed_gop from MpegEncContext to Mpeg1Context
mpeg12: move full_pel from MpegEncContext to Mpeg1Context
mpeg12: move Mpeg1Context from mpeg12.c to mpeg12.h
mpegvideo: remove some unused variables from MpegEncContext.
Conflicts:
libavcodec/mpeg12.c
libavformat/mp3enc.c
Merged-by: Michael Niedermayer <michaelni@gmx.at>
Diffstat (limited to 'libavcodec')
-rw-r--r-- | libavcodec/Makefile | 4 | ||||
-rw-r--r-- | libavcodec/ac3enc_opts_template.c | 6 | ||||
-rw-r--r-- | libavcodec/arm/asm-offsets.h | 12 | ||||
-rw-r--r-- | libavcodec/g722.c | 427 | ||||
-rw-r--r-- | libavcodec/g722.h | 74 | ||||
-rw-r--r-- | libavcodec/g722dec.c | 147 | ||||
-rw-r--r-- | libavcodec/g722enc.c | 311 | ||||
-rw-r--r-- | libavcodec/h263.h | 42 | ||||
-rw-r--r-- | libavcodec/h264.c | 1 | ||||
-rw-r--r-- | libavcodec/mpeg12.c | 35 | ||||
-rw-r--r-- | libavcodec/mpeg12.h | 16 | ||||
-rw-r--r-- | libavcodec/mpeg4videoenc.c | 41 | ||||
-rw-r--r-- | libavcodec/mpegvideo.c | 1 | ||||
-rw-r--r-- | libavcodec/mpegvideo.h | 7 | ||||
-rw-r--r-- | libavcodec/vdpau.c | 7 | ||||
-rw-r--r-- | libavcodec/vdpau_internal.h | 3 | ||||
-rw-r--r-- | libavcodec/vorbisdec.c | 24 | ||||
-rw-r--r-- | libavcodec/wavpack.c | 4 |
18 files changed, 641 insertions, 521 deletions
diff --git a/libavcodec/Makefile b/libavcodec/Makefile index 4941b6f0b2..d90bcacb8a 100644 --- a/libavcodec/Makefile +++ b/libavcodec/Makefile @@ -519,8 +519,8 @@ OBJS-$(CONFIG_ADPCM_EA_R1_DECODER) += adpcm.o OBJS-$(CONFIG_ADPCM_EA_R2_DECODER) += adpcm.o OBJS-$(CONFIG_ADPCM_EA_R3_DECODER) += adpcm.o OBJS-$(CONFIG_ADPCM_EA_XAS_DECODER) += adpcm.o -OBJS-$(CONFIG_ADPCM_G722_DECODER) += g722.o -OBJS-$(CONFIG_ADPCM_G722_ENCODER) += g722.o +OBJS-$(CONFIG_ADPCM_G722_DECODER) += g722.o g722dec.o +OBJS-$(CONFIG_ADPCM_G722_ENCODER) += g722.o g722enc.o OBJS-$(CONFIG_ADPCM_G726_DECODER) += g726.o OBJS-$(CONFIG_ADPCM_G726_ENCODER) += g726.o OBJS-$(CONFIG_ADPCM_IMA_AMV_DECODER) += adpcm.o adpcm_data.o diff --git a/libavcodec/ac3enc_opts_template.c b/libavcodec/ac3enc_opts_template.c index acb4ab477a..a5f848e200 100644 --- a/libavcodec/ac3enc_opts_template.c +++ b/libavcodec/ac3enc_opts_template.c @@ -45,7 +45,7 @@ static const AVOption eac3_options[] = { /* other metadata options */ {"copyright", "Copyright Bit", OFFSET(copyright), AV_OPT_TYPE_INT, {.dbl = AC3ENC_OPT_NONE }, AC3ENC_OPT_NONE, 1, AC3ENC_PARAM}, {"dialnorm", "Dialogue Level (dB)", OFFSET(dialogue_level), AV_OPT_TYPE_INT, {.dbl = -31 }, -31, -1, AC3ENC_PARAM}, -{"dsur_mode", "Dolby Surround Mode", OFFSET(dolby_surround_mode), AV_OPT_TYPE_INT, {.dbl = AC3ENC_OPT_NONE }, AC3ENC_OPT_NONE, AC3ENC_OPT_MODE_OFF, AC3ENC_PARAM, "dsur_mode"}, +{"dsur_mode", "Dolby Surround Mode", OFFSET(dolby_surround_mode), AV_OPT_TYPE_INT, {.dbl = AC3ENC_OPT_NONE }, AC3ENC_OPT_NONE, AC3ENC_OPT_MODE_ON, AC3ENC_PARAM, "dsur_mode"}, {"notindicated", "Not Indicated (default)", 0, AV_OPT_TYPE_CONST, {.dbl = AC3ENC_OPT_NOT_INDICATED }, INT_MIN, INT_MAX, AC3ENC_PARAM, "dsur_mode"}, {"on", "Dolby Surround Encoded", 0, AV_OPT_TYPE_CONST, {.dbl = AC3ENC_OPT_MODE_ON }, INT_MIN, INT_MAX, AC3ENC_PARAM, "dsur_mode"}, {"off", "Not Dolby Surround Encoded", 0, AV_OPT_TYPE_CONST, {.dbl = AC3ENC_OPT_MODE_OFF }, INT_MIN, INT_MAX, AC3ENC_PARAM, "dsur_mode"}, @@ -59,11 +59,11 @@ static const AVOption eac3_options[] = { {"ltrt_surmixlev", "Lt/Rt Surround Mix Level", OFFSET(ltrt_surround_mix_level), AV_OPT_TYPE_FLOAT, {.dbl = -1.0 }, -1.0, 2.0, AC3ENC_PARAM}, {"loro_cmixlev", "Lo/Ro Center Mix Level", OFFSET(loro_center_mix_level), AV_OPT_TYPE_FLOAT, {.dbl = -1.0 }, -1.0, 2.0, AC3ENC_PARAM}, {"loro_surmixlev", "Lo/Ro Surround Mix Level", OFFSET(loro_surround_mix_level), AV_OPT_TYPE_FLOAT, {.dbl = -1.0 }, -1.0, 2.0, AC3ENC_PARAM}, -{"dsurex_mode", "Dolby Surround EX Mode", OFFSET(dolby_surround_ex_mode), AV_OPT_TYPE_INT, {.dbl = AC3ENC_OPT_NONE }, AC3ENC_OPT_NONE, AC3ENC_OPT_MODE_OFF, AC3ENC_PARAM, "dsurex_mode"}, +{"dsurex_mode", "Dolby Surround EX Mode", OFFSET(dolby_surround_ex_mode), AV_OPT_TYPE_INT, {.dbl = AC3ENC_OPT_NONE }, AC3ENC_OPT_NONE, AC3ENC_OPT_MODE_ON, AC3ENC_PARAM, "dsurex_mode"}, {"notindicated", "Not Indicated (default)", 0, AV_OPT_TYPE_CONST, {.dbl = AC3ENC_OPT_NOT_INDICATED }, INT_MIN, INT_MAX, AC3ENC_PARAM, "dsurex_mode"}, {"on", "Dolby Surround EX Encoded", 0, AV_OPT_TYPE_CONST, {.dbl = AC3ENC_OPT_MODE_ON }, INT_MIN, INT_MAX, AC3ENC_PARAM, "dsurex_mode"}, {"off", "Not Dolby Surround EX Encoded", 0, AV_OPT_TYPE_CONST, {.dbl = AC3ENC_OPT_MODE_OFF }, INT_MIN, INT_MAX, AC3ENC_PARAM, "dsurex_mode"}, -{"dheadphone_mode", "Dolby Headphone Mode", OFFSET(dolby_headphone_mode), AV_OPT_TYPE_INT, {.dbl = AC3ENC_OPT_NONE }, AC3ENC_OPT_NONE, AC3ENC_OPT_MODE_OFF, AC3ENC_PARAM, "dheadphone_mode"}, +{"dheadphone_mode", "Dolby Headphone Mode", OFFSET(dolby_headphone_mode), AV_OPT_TYPE_INT, {.dbl = AC3ENC_OPT_NONE }, AC3ENC_OPT_NONE, AC3ENC_OPT_MODE_ON, AC3ENC_PARAM, "dheadphone_mode"}, {"notindicated", "Not Indicated (default)", 0, AV_OPT_TYPE_CONST, {.dbl = AC3ENC_OPT_NOT_INDICATED }, INT_MIN, INT_MAX, AC3ENC_PARAM, "dheadphone_mode"}, {"on", "Dolby Headphone Encoded", 0, AV_OPT_TYPE_CONST, {.dbl = AC3ENC_OPT_MODE_ON }, INT_MIN, INT_MAX, AC3ENC_PARAM, "dheadphone_mode"}, {"off", "Not Dolby Headphone Encoded", 0, AV_OPT_TYPE_CONST, {.dbl = AC3ENC_OPT_MODE_OFF }, INT_MIN, INT_MAX, AC3ENC_PARAM, "dheadphone_mode"}, diff --git a/libavcodec/arm/asm-offsets.h b/libavcodec/arm/asm-offsets.h index 17a540031b..5cfc5cb10c 100644 --- a/libavcodec/arm/asm-offsets.h +++ b/libavcodec/arm/asm-offsets.h @@ -29,11 +29,11 @@ #endif /* MpegEncContext */ -#define Y_DC_SCALE 0xac -#define C_DC_SCALE 0xb0 -#define AC_PRED 0xb4 -#define BLOCK_LAST_INDEX 0xb8 -#define H263_AIC 0xe8 -#define INTER_SCANTAB_RASTER_END 0x130 +#define Y_DC_SCALE 0xa8 +#define C_DC_SCALE 0xac +#define AC_PRED 0xb0 +#define BLOCK_LAST_INDEX 0xb4 +#define H263_AIC 0xe4 +#define INTER_SCANTAB_RASTER_END 0x12c #endif /* AVCODEC_ARM_ASM_OFFSETS_H */ diff --git a/libavcodec/g722.c b/libavcodec/g722.c index 30c6f5313c..e8e74242b8 100644 --- a/libavcodec/g722.c +++ b/libavcodec/g722.c @@ -36,45 +36,8 @@ * respectively of each byte are ignored. */ -#include "avcodec.h" #include "mathops.h" -#include "get_bits.h" - -#define PREV_SAMPLES_BUF_SIZE 1024 - -#define FREEZE_INTERVAL 128 - -typedef struct { - int16_t prev_samples[PREV_SAMPLES_BUF_SIZE]; ///< memory of past decoded samples - int prev_samples_pos; ///< the number of values in prev_samples - - /** - * The band[0] and band[1] correspond respectively to the lower band and higher band. - */ - struct G722Band { - int16_t s_predictor; ///< predictor output value - int32_t s_zero; ///< previous output signal from zero predictor - int8_t part_reconst_mem[2]; ///< signs of previous partially reconstructed signals - int16_t prev_qtzd_reconst; ///< previous quantized reconstructed signal (internal value, using low_inv_quant4) - int16_t pole_mem[2]; ///< second-order pole section coefficient buffer - int32_t diff_mem[6]; ///< quantizer difference signal memory - int16_t zero_mem[6]; ///< Seventh-order zero section coefficient buffer - int16_t log_factor; ///< delayed 2-logarithmic quantizer factor - int16_t scale_factor; ///< delayed quantizer scale factor - } band[2]; - - struct TrellisNode { - struct G722Band state; - uint32_t ssd; - int path; - } *node_buf[2], **nodep_buf[2]; - - struct TrellisPath { - int value; - int prev; - } *paths[2]; -} G722Context; - +#include "g722.h" static const int8_t sign_lookup[2] = { -1, 1 }; @@ -85,7 +48,7 @@ static const int16_t inv_log2_table[32] = { 3444, 3520, 3597, 3676, 3756, 3838, 3922, 4008 }; static const int16_t high_log_factor_step[2] = { 798, -214 }; -static const int16_t high_inv_quant[4] = { -926, -202, 926, 202 }; +const int16_t ff_g722_high_inv_quant[4] = { -926, -202, 926, 202 }; /** * low_log_factor_step[index] == wl[rl42[index]] */ @@ -93,11 +56,11 @@ static const int16_t low_log_factor_step[16] = { -60, 3042, 1198, 538, 334, 172, 58, -30, 3042, 1198, 538, 334, 172, 58, -30, -60 }; -static const int16_t low_inv_quant4[16] = { +const int16_t ff_g722_low_inv_quant4[16] = { 0, -2557, -1612, -1121, -786, -530, -323, -150, 2557, 1612, 1121, 786, 530, 323, 150, 0 }; -static const int16_t low_inv_quant6[64] = { +const int16_t ff_g722_low_inv_quant6[64] = { -17, -17, -17, -17, -3101, -2738, -2376, -2088, -1873, -1689, -1535, -1399, -1279, -1170, -1072, -982, -899, -822, -750, -682, -618, -558, -501, -447, @@ -173,10 +136,10 @@ static int inline linear_scale_factor(const int log_factor) return shift < 0 ? wd1 >> -shift : wd1 << shift; } -static void update_low_predictor(struct G722Band *band, const int ilow) +void ff_g722_update_low_predictor(struct G722Band *band, const int ilow) { do_adaptive_prediction(band, - band->scale_factor * low_inv_quant4[ilow] >> 10); + band->scale_factor * ff_g722_low_inv_quant4[ilow] >> 10); // quantizer adaptation band->log_factor = av_clip((band->log_factor * 127 >> 7) + @@ -184,7 +147,7 @@ static void update_low_predictor(struct G722Band *band, const int ilow) band->scale_factor = linear_scale_factor(band->log_factor - (8 << 11)); } -static void update_high_predictor(struct G722Band *band, const int dhigh, +void ff_g722_update_high_predictor(struct G722Band *band, const int dhigh, const int ihigh) { do_adaptive_prediction(band, dhigh); @@ -195,7 +158,7 @@ static void update_high_predictor(struct G722Band *band, const int dhigh, band->scale_factor = linear_scale_factor(band->log_factor - (10 << 11)); } -static void apply_qmf(const int16_t *prev_samples, int *xout1, int *xout2) +void ff_g722_apply_qmf(const int16_t *prev_samples, int *xout1, int *xout2) { int i; @@ -206,377 +169,3 @@ static void apply_qmf(const int16_t *prev_samples, int *xout1, int *xout2) MAC16(*xout1, prev_samples[2*i+1], qmf_coeffs[11-i]); } } - -static av_cold int g722_init(AVCodecContext * avctx) -{ - G722Context *c = avctx->priv_data; - - if (avctx->channels != 1) { - av_log(avctx, AV_LOG_ERROR, "Only mono tracks are allowed.\n"); - return AVERROR_INVALIDDATA; - } - avctx->sample_fmt = AV_SAMPLE_FMT_S16; - - switch (avctx->bits_per_coded_sample) { - case 8: - case 7: - case 6: - break; - default: - av_log(avctx, AV_LOG_WARNING, "Unsupported bits_per_coded_sample [%d], " - "assuming 8\n", - avctx->bits_per_coded_sample); - case 0: - avctx->bits_per_coded_sample = 8; - break; - } - - c->band[0].scale_factor = 8; - c->band[1].scale_factor = 2; - c->prev_samples_pos = 22; - - if (avctx->lowres) - avctx->sample_rate /= 2; - - if (avctx->trellis) { - int frontier = 1 << avctx->trellis; - int max_paths = frontier * FREEZE_INTERVAL; - int i; - for (i = 0; i < 2; i++) { - c->paths[i] = av_mallocz(max_paths * sizeof(**c->paths)); - c->node_buf[i] = av_mallocz(2 * frontier * sizeof(**c->node_buf)); - c->nodep_buf[i] = av_mallocz(2 * frontier * sizeof(**c->nodep_buf)); - } - } - - return 0; -} - -static av_cold int g722_close(AVCodecContext *avctx) -{ - G722Context *c = avctx->priv_data; - int i; - for (i = 0; i < 2; i++) { - av_freep(&c->paths[i]); - av_freep(&c->node_buf[i]); - av_freep(&c->nodep_buf[i]); - } - return 0; -} - -#if CONFIG_ADPCM_G722_DECODER -static const int16_t low_inv_quant5[32] = { - -35, -35, -2919, -2195, -1765, -1458, -1219, -1023, - -858, -714, -587, -473, -370, -276, -190, -110, - 2919, 2195, 1765, 1458, 1219, 1023, 858, 714, - 587, 473, 370, 276, 190, 110, 35, -35 -}; - -static const int16_t *low_inv_quants[3] = { low_inv_quant6, low_inv_quant5, - low_inv_quant4 }; - -static int g722_decode_frame(AVCodecContext *avctx, void *data, - int *data_size, AVPacket *avpkt) -{ - G722Context *c = avctx->priv_data; - int16_t *out_buf = data; - int j, out_len = 0; - const int skip = 8 - avctx->bits_per_coded_sample; - const int16_t *quantizer_table = low_inv_quants[skip]; - GetBitContext gb; - - init_get_bits(&gb, avpkt->data, avpkt->size * 8); - - for (j = 0; j < avpkt->size; j++) { - int ilow, ihigh, rlow; - - ihigh = get_bits(&gb, 2); - ilow = get_bits(&gb, 6 - skip); - skip_bits(&gb, skip); - - rlow = av_clip((c->band[0].scale_factor * quantizer_table[ilow] >> 10) - + c->band[0].s_predictor, -16384, 16383); - - update_low_predictor(&c->band[0], ilow >> (2 - skip)); - - if (!avctx->lowres) { - const int dhigh = c->band[1].scale_factor * - high_inv_quant[ihigh] >> 10; - const int rhigh = av_clip(dhigh + c->band[1].s_predictor, - -16384, 16383); - int xout1, xout2; - - update_high_predictor(&c->band[1], dhigh, ihigh); - - c->prev_samples[c->prev_samples_pos++] = rlow + rhigh; - c->prev_samples[c->prev_samples_pos++] = rlow - rhigh; - apply_qmf(c->prev_samples + c->prev_samples_pos - 24, - &xout1, &xout2); - out_buf[out_len++] = av_clip_int16(xout1 >> 12); - out_buf[out_len++] = av_clip_int16(xout2 >> 12); - if (c->prev_samples_pos >= PREV_SAMPLES_BUF_SIZE) { - memmove(c->prev_samples, - c->prev_samples + c->prev_samples_pos - 22, - 22 * sizeof(c->prev_samples[0])); - c->prev_samples_pos = 22; - } - } else - out_buf[out_len++] = rlow; - } - *data_size = out_len << 1; - return avpkt->size; -} - -AVCodec ff_adpcm_g722_decoder = { - .name = "g722", - .type = AVMEDIA_TYPE_AUDIO, - .id = CODEC_ID_ADPCM_G722, - .priv_data_size = sizeof(G722Context), - .init = g722_init, - .decode = g722_decode_frame, - .long_name = NULL_IF_CONFIG_SMALL("G.722 ADPCM"), - .max_lowres = 1, -}; -#endif - -#if CONFIG_ADPCM_G722_ENCODER -static const int16_t low_quant[33] = { - 35, 72, 110, 150, 190, 233, 276, 323, - 370, 422, 473, 530, 587, 650, 714, 786, - 858, 940, 1023, 1121, 1219, 1339, 1458, 1612, - 1765, 1980, 2195, 2557, 2919 -}; - -static inline void filter_samples(G722Context *c, const int16_t *samples, - int *xlow, int *xhigh) -{ - int xout1, xout2; - c->prev_samples[c->prev_samples_pos++] = samples[0]; - c->prev_samples[c->prev_samples_pos++] = samples[1]; - apply_qmf(c->prev_samples + c->prev_samples_pos - 24, &xout1, &xout2); - *xlow = xout1 + xout2 >> 13; - *xhigh = xout1 - xout2 >> 13; - if (c->prev_samples_pos >= PREV_SAMPLES_BUF_SIZE) { - memmove(c->prev_samples, - c->prev_samples + c->prev_samples_pos - 22, - 22 * sizeof(c->prev_samples[0])); - c->prev_samples_pos = 22; - } -} - -static inline int encode_high(const struct G722Band *state, int xhigh) -{ - int diff = av_clip_int16(xhigh - state->s_predictor); - int pred = 141 * state->scale_factor >> 8; - /* = diff >= 0 ? (diff < pred) + 2 : diff >= -pred */ - return ((diff ^ (diff >> (sizeof(diff)*8-1))) < pred) + 2*(diff >= 0); -} - -static inline int encode_low(const struct G722Band* state, int xlow) -{ - int diff = av_clip_int16(xlow - state->s_predictor); - /* = diff >= 0 ? diff : -(diff + 1) */ - int limit = diff ^ (diff >> (sizeof(diff)*8-1)); - int i = 0; - limit = limit + 1 << 10; - if (limit > low_quant[8] * state->scale_factor) - i = 9; - while (i < 29 && limit > low_quant[i] * state->scale_factor) - i++; - return (diff < 0 ? (i < 2 ? 63 : 33) : 61) - i; -} - -static int g722_encode_trellis(AVCodecContext *avctx, - uint8_t *dst, int buf_size, void *data) -{ - G722Context *c = avctx->priv_data; - const int16_t *samples = data; - int i, j, k; - int frontier = 1 << avctx->trellis; - struct TrellisNode **nodes[2]; - struct TrellisNode **nodes_next[2]; - int pathn[2] = {0, 0}, froze = -1; - struct TrellisPath *p[2]; - - for (i = 0; i < 2; i++) { - nodes[i] = c->nodep_buf[i]; - nodes_next[i] = c->nodep_buf[i] + frontier; - memset(c->nodep_buf[i], 0, 2 * frontier * sizeof(*c->nodep_buf)); - nodes[i][0] = c->node_buf[i] + frontier; - nodes[i][0]->ssd = 0; - nodes[i][0]->path = 0; - nodes[i][0]->state = c->band[i]; - } - - for (i = 0; i < buf_size >> 1; i++) { - int xlow, xhigh; - struct TrellisNode *next[2]; - int heap_pos[2] = {0, 0}; - - for (j = 0; j < 2; j++) { - next[j] = c->node_buf[j] + frontier*(i & 1); - memset(nodes_next[j], 0, frontier * sizeof(**nodes_next)); - } - - filter_samples(c, &samples[2*i], &xlow, &xhigh); - - for (j = 0; j < frontier && nodes[0][j]; j++) { - /* Only k >> 2 affects the future adaptive state, therefore testing - * small steps that don't change k >> 2 is useless, the orignal - * value from encode_low is better than them. Since we step k - * in steps of 4, make sure range is a multiple of 4, so that - * we don't miss the original value from encode_low. */ - int range = j < frontier/2 ? 4 : 0; - struct TrellisNode *cur_node = nodes[0][j]; - - int ilow = encode_low(&cur_node->state, xlow); - - for (k = ilow - range; k <= ilow + range && k <= 63; k += 4) { - int decoded, dec_diff, pos; - uint32_t ssd; - struct TrellisNode* node; - - if (k < 0) - continue; - - decoded = av_clip((cur_node->state.scale_factor * - low_inv_quant6[k] >> 10) - + cur_node->state.s_predictor, -16384, 16383); - dec_diff = xlow - decoded; - -#define STORE_NODE(index, UPDATE, VALUE)\ - ssd = cur_node->ssd + dec_diff*dec_diff;\ - /* Check for wraparound. Using 64 bit ssd counters would \ - * be simpler, but is slower on x86 32 bit. */\ - if (ssd < cur_node->ssd)\ - continue;\ - if (heap_pos[index] < frontier) {\ - pos = heap_pos[index]++;\ - assert(pathn[index] < FREEZE_INTERVAL * frontier);\ - node = nodes_next[index][pos] = next[index]++;\ - node->path = pathn[index]++;\ - } else {\ - /* Try to replace one of the leaf nodes with the new \ - * one, but not always testing the same leaf position */\ - pos = (frontier>>1) + (heap_pos[index] & ((frontier>>1) - 1));\ - if (ssd >= nodes_next[index][pos]->ssd)\ - continue;\ - heap_pos[index]++;\ - node = nodes_next[index][pos];\ - }\ - node->ssd = ssd;\ - node->state = cur_node->state;\ - UPDATE;\ - c->paths[index][node->path].value = VALUE;\ - c->paths[index][node->path].prev = cur_node->path;\ - /* Sift the newly inserted node up in the heap to restore \ - * the heap property */\ - while (pos > 0) {\ - int parent = (pos - 1) >> 1;\ - if (nodes_next[index][parent]->ssd <= ssd)\ - break;\ - FFSWAP(struct TrellisNode*, nodes_next[index][parent],\ - nodes_next[index][pos]);\ - pos = parent;\ - } - STORE_NODE(0, update_low_predictor(&node->state, k >> 2), k); - } - } - - for (j = 0; j < frontier && nodes[1][j]; j++) { - int ihigh; - struct TrellisNode *cur_node = nodes[1][j]; - - /* We don't try to get any initial guess for ihigh via - * encode_high - since there's only 4 possible values, test - * them all. Testing all of these gives a much, much larger - * gain than testing a larger range around ilow. */ - for (ihigh = 0; ihigh < 4; ihigh++) { - int dhigh, decoded, dec_diff, pos; - uint32_t ssd; - struct TrellisNode* node; - - dhigh = cur_node->state.scale_factor * - high_inv_quant[ihigh] >> 10; - decoded = av_clip(dhigh + cur_node->state.s_predictor, - -16384, 16383); - dec_diff = xhigh - decoded; - - STORE_NODE(1, update_high_predictor(&node->state, dhigh, ihigh), ihigh); - } - } - - for (j = 0; j < 2; j++) { - FFSWAP(struct TrellisNode**, nodes[j], nodes_next[j]); - - if (nodes[j][0]->ssd > (1 << 16)) { - for (k = 1; k < frontier && nodes[j][k]; k++) - nodes[j][k]->ssd -= nodes[j][0]->ssd; - nodes[j][0]->ssd = 0; - } - } - - if (i == froze + FREEZE_INTERVAL) { - p[0] = &c->paths[0][nodes[0][0]->path]; - p[1] = &c->paths[1][nodes[1][0]->path]; - for (j = i; j > froze; j--) { - dst[j] = p[1]->value << 6 | p[0]->value; - p[0] = &c->paths[0][p[0]->prev]; - p[1] = &c->paths[1][p[1]->prev]; - } - froze = i; - pathn[0] = pathn[1] = 0; - memset(nodes[0] + 1, 0, (frontier - 1)*sizeof(**nodes)); - memset(nodes[1] + 1, 0, (frontier - 1)*sizeof(**nodes)); - } - } - - p[0] = &c->paths[0][nodes[0][0]->path]; - p[1] = &c->paths[1][nodes[1][0]->path]; - for (j = i; j > froze; j--) { - dst[j] = p[1]->value << 6 | p[0]->value; - p[0] = &c->paths[0][p[0]->prev]; - p[1] = &c->paths[1][p[1]->prev]; - } - c->band[0] = nodes[0][0]->state; - c->band[1] = nodes[1][0]->state; - - return i; -} - -static int g722_encode_frame(AVCodecContext *avctx, - uint8_t *dst, int buf_size, void *data) -{ - G722Context *c = avctx->priv_data; - const int16_t *samples = data; - int i; - - if (avctx->trellis) - return g722_encode_trellis(avctx, dst, buf_size, data); - - for (i = 0; i < buf_size >> 1; i++) { - int xlow, xhigh, ihigh, ilow; - filter_samples(c, &samples[2*i], &xlow, &xhigh); - ihigh = encode_high(&c->band[1], xhigh); - ilow = encode_low(&c->band[0], xlow); - update_high_predictor(&c->band[1], c->band[1].scale_factor * - high_inv_quant[ihigh] >> 10, ihigh); - update_low_predictor(&c->band[0], ilow >> 2); - *dst++ = ihigh << 6 | ilow; - } - return i; -} - -AVCodec ff_adpcm_g722_encoder = { - .name = "g722", - .type = AVMEDIA_TYPE_AUDIO, - .id = CODEC_ID_ADPCM_G722, - .priv_data_size = sizeof(G722Context), - .init = g722_init, - .close = g722_close, - .encode = g722_encode_frame, - .long_name = NULL_IF_CONFIG_SMALL("G.722 ADPCM"), - .sample_fmts = (const enum AVSampleFormat[]){AV_SAMPLE_FMT_S16,AV_SAMPLE_FMT_NONE}, -}; -#endif - diff --git a/libavcodec/g722.h b/libavcodec/g722.h new file mode 100644 index 0000000000..5edb6c8119 --- /dev/null +++ b/libavcodec/g722.h @@ -0,0 +1,74 @@ +/* + * Copyright (c) CMU 1993 Computer Science, Speech Group + * Chengxiang Lu and Alex Hauptmann + * Copyright (c) 2005 Steve Underwood <steveu at coppice.org> + * Copyright (c) 2009 Kenan Gillet + * Copyright (c) 2010 Martin Storsjo + * + * This file is part of Libav. + * + * Libav 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. + * + * Libav 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 Libav; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#ifndef AVCODEC_G722_H +#define AVCODEC_G722_H + +#include <stdint.h> + +#define PREV_SAMPLES_BUF_SIZE 1024 + +typedef struct { + int16_t prev_samples[PREV_SAMPLES_BUF_SIZE]; ///< memory of past decoded samples + int prev_samples_pos; ///< the number of values in prev_samples + + /** + * The band[0] and band[1] correspond respectively to the lower band and higher band. + */ + struct G722Band { + int16_t s_predictor; ///< predictor output value + int32_t s_zero; ///< previous output signal from zero predictor + int8_t part_reconst_mem[2]; ///< signs of previous partially reconstructed signals + int16_t prev_qtzd_reconst; ///< previous quantized reconstructed signal (internal value, using low_inv_quant4) + int16_t pole_mem[2]; ///< second-order pole section coefficient buffer + int32_t diff_mem[6]; ///< quantizer difference signal memory + int16_t zero_mem[6]; ///< Seventh-order zero section coefficient buffer + int16_t log_factor; ///< delayed 2-logarithmic quantizer factor + int16_t scale_factor; ///< delayed quantizer scale factor + } band[2]; + + struct TrellisNode { + struct G722Band state; + uint32_t ssd; + int path; + } *node_buf[2], **nodep_buf[2]; + + struct TrellisPath { + int value; + int prev; + } *paths[2]; +} G722Context; + +extern const int16_t ff_g722_high_inv_quant[4]; +extern const int16_t ff_g722_low_inv_quant4[16]; +extern const int16_t ff_g722_low_inv_quant6[64]; + +void ff_g722_update_low_predictor(struct G722Band *band, const int ilow); + +void ff_g722_update_high_predictor(struct G722Band *band, const int dhigh, + const int ihigh); + +void ff_g722_apply_qmf(const int16_t *prev_samples, int *xout1, int *xout2); + +#endif /* AVCODEC_G722_H */ diff --git a/libavcodec/g722dec.c b/libavcodec/g722dec.c new file mode 100644 index 0000000000..51f87c9d3c --- /dev/null +++ b/libavcodec/g722dec.c @@ -0,0 +1,147 @@ +/* + * Copyright (c) CMU 1993 Computer Science, Speech Group + * Chengxiang Lu and Alex Hauptmann + * Copyright (c) 2005 Steve Underwood <steveu at coppice.org> + * Copyright (c) 2009 Kenan Gillet + * Copyright (c) 2010 Martin Storsjo + * + * This file is part of Libav. + * + * Libav 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. + * + * Libav 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 Libav; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/** + * @file + * G.722 ADPCM audio decoder + * + * This G.722 decoder is a bit-exact implementation of the ITU G.722 + * specification for all three specified bitrates - 64000bps, 56000bps + * and 48000bps. It passes the ITU tests. + * + * @note For the 56000bps and 48000bps bitrates, the lowest 1 or 2 bits + * respectively of each byte are ignored. + */ + +#include "avcodec.h" +#include "get_bits.h" +#include "g722.h" + +static av_cold int g722_decode_init(AVCodecContext * avctx) +{ + G722Context *c = avctx->priv_data; + + if (avctx->channels != 1) { + av_log(avctx, AV_LOG_ERROR, "Only mono tracks are allowed.\n"); + return AVERROR_INVALIDDATA; + } + avctx->sample_fmt = AV_SAMPLE_FMT_S16; + + switch (avctx->bits_per_coded_sample) { + case 8: + case 7: + case 6: + break; + default: + av_log(avctx, AV_LOG_WARNING, "Unsupported bits_per_coded_sample [%d], " + "assuming 8\n", + avctx->bits_per_coded_sample); + case 0: + avctx->bits_per_coded_sample = 8; + break; + } + + c->band[0].scale_factor = 8; + c->band[1].scale_factor = 2; + c->prev_samples_pos = 22; + + if (avctx->lowres) + avctx->sample_rate /= 2; + + return 0; +} + +static const int16_t low_inv_quant5[32] = { + -35, -35, -2919, -2195, -1765, -1458, -1219, -1023, + -858, -714, -587, -473, -370, -276, -190, -110, + 2919, 2195, 1765, 1458, 1219, 1023, 858, 714, + 587, 473, 370, 276, 190, 110, 35, -35 +}; + +static const int16_t *low_inv_quants[3] = { ff_g722_low_inv_quant6, + low_inv_quant5, + ff_g722_low_inv_quant4 }; + +static int g722_decode_frame(AVCodecContext *avctx, void *data, + int *data_size, AVPacket *avpkt) +{ + G722Context *c = avctx->priv_data; + int16_t *out_buf = data; + int j, out_len = 0; + const int skip = 8 - avctx->bits_per_coded_sample; + const int16_t *quantizer_table = low_inv_quants[skip]; + GetBitContext gb; + + init_get_bits(&gb, avpkt->data, avpkt->size * 8); + + for (j = 0; j < avpkt->size; j++) { + int ilow, ihigh, rlow; + + ihigh = get_bits(&gb, 2); + ilow = get_bits(&gb, 6 - skip); + skip_bits(&gb, skip); + + rlow = av_clip((c->band[0].scale_factor * quantizer_table[ilow] >> 10) + + c->band[0].s_predictor, -16384, 16383); + + ff_g722_update_low_predictor(&c->band[0], ilow >> (2 - skip)); + + if (!avctx->lowres) { + const int dhigh = c->band[1].scale_factor * + ff_g722_high_inv_quant[ihigh] >> 10; + const int rhigh = av_clip(dhigh + c->band[1].s_predictor, + -16384, 16383); + int xout1, xout2; + + ff_g722_update_high_predictor(&c->band[1], dhigh, ihigh); + + c->prev_samples[c->prev_samples_pos++] = rlow + rhigh; + c->prev_samples[c->prev_samples_pos++] = rlow - rhigh; + ff_g722_apply_qmf(c->prev_samples + c->prev_samples_pos - 24, + &xout1, &xout2); + out_buf[out_len++] = av_clip_int16(xout1 >> 12); + out_buf[out_len++] = av_clip_int16(xout2 >> 12); + if (c->prev_samples_pos >= PREV_SAMPLES_BUF_SIZE) { + memmove(c->prev_samples, + c->prev_samples + c->prev_samples_pos - 22, + 22 * sizeof(c->prev_samples[0])); + c->prev_samples_pos = 22; + } + } else + out_buf[out_len++] = rlow; + } + *data_size = out_len << 1; + return avpkt->size; +} + +AVCodec ff_adpcm_g722_decoder = { + .name = "g722", + .type = AVMEDIA_TYPE_AUDIO, + .id = CODEC_ID_ADPCM_G722, + .priv_data_size = sizeof(G722Context), + .init = g722_decode_init, + .decode = g722_decode_frame, + .long_name = NULL_IF_CONFIG_SMALL("G.722 ADPCM"), + .max_lowres = 1, +}; diff --git a/libavcodec/g722enc.c b/libavcodec/g722enc.c new file mode 100644 index 0000000000..bdc30d5d07 --- /dev/null +++ b/libavcodec/g722enc.c @@ -0,0 +1,311 @@ +/* + * Copyright (c) CMU 1993 Computer Science, Speech Group + * Chengxiang Lu and Alex Hauptmann + * Copyright (c) 2005 Steve Underwood <steveu at coppice.org> + * Copyright (c) 2009 Kenan Gillet + * Copyright (c) 2010 Martin Storsjo + * + * This file is part of Libav. + * + * Libav 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. + * + * Libav 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 Libav; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/** + * @file + * G.722 ADPCM audio encoder + */ + +#include "avcodec.h" +#include "g722.h" + +#define FREEZE_INTERVAL 128 + +static av_cold int g722_encode_init(AVCodecContext * avctx) +{ + G722Context *c = avctx->priv_data; + + if (avctx->channels != 1) { + av_log(avctx, AV_LOG_ERROR, "Only mono tracks are allowed.\n"); + return AVERROR_INVALIDDATA; + } + + c->band[0].scale_factor = 8; + c->band[1].scale_factor = 2; + c->prev_samples_pos = 22; + + if (avctx->trellis) { + int frontier = 1 << avctx->trellis; + int max_paths = frontier * FREEZE_INTERVAL; + int i; + for (i = 0; i < 2; i++) { + c->paths[i] = av_mallocz(max_paths * sizeof(**c->paths)); + c->node_buf[i] = av_mallocz(2 * frontier * sizeof(**c->node_buf)); + c->nodep_buf[i] = av_mallocz(2 * frontier * sizeof(**c->nodep_buf)); + } + } + + return 0; +} + +static av_cold int g722_encode_close(AVCodecContext *avctx) +{ + G722Context *c = avctx->priv_data; + int i; + for (i = 0; i < 2; i++) { + av_freep(&c->paths[i]); + av_freep(&c->node_buf[i]); + av_freep(&c->nodep_buf[i]); + } + return 0; +} + +static const int16_t low_quant[33] = { + 35, 72, 110, 150, 190, 233, 276, 323, + 370, 422, 473, 530, 587, 650, 714, 786, + 858, 940, 1023, 1121, 1219, 1339, 1458, 1612, + 1765, 1980, 2195, 2557, 2919 +}; + +static inline void filter_samples(G722Context *c, const int16_t *samples, + int *xlow, int *xhigh) +{ + int xout1, xout2; + c->prev_samples[c->prev_samples_pos++] = samples[0]; + c->prev_samples[c->prev_samples_pos++] = samples[1]; + ff_g722_apply_qmf(c->prev_samples + c->prev_samples_pos - 24, &xout1, &xout2); + *xlow = xout1 + xout2 >> 13; + *xhigh = xout1 - xout2 >> 13; + if (c->prev_samples_pos >= PREV_SAMPLES_BUF_SIZE) { + memmove(c->prev_samples, + c->prev_samples + c->prev_samples_pos - 22, + 22 * sizeof(c->prev_samples[0])); + c->prev_samples_pos = 22; + } +} + +static inline int encode_high(const struct G722Band *state, int xhigh) +{ + int diff = av_clip_int16(xhigh - state->s_predictor); + int pred = 141 * state->scale_factor >> 8; + /* = diff >= 0 ? (diff < pred) + 2 : diff >= -pred */ + return ((diff ^ (diff >> (sizeof(diff)*8-1))) < pred) + 2*(diff >= 0); +} + +static inline int encode_low(const struct G722Band* state, int xlow) +{ + int diff = av_clip_int16(xlow - state->s_predictor); + /* = diff >= 0 ? diff : -(diff + 1) */ + int limit = diff ^ (diff >> (sizeof(diff)*8-1)); + int i = 0; + limit = limit + 1 << 10; + if (limit > low_quant[8] * state->scale_factor) + i = 9; + while (i < 29 && limit > low_quant[i] * state->scale_factor) + i++; + return (diff < 0 ? (i < 2 ? 63 : 33) : 61) - i; +} + +static int g722_encode_trellis(AVCodecContext *avctx, + uint8_t *dst, int buf_size, void *data) +{ + G722Context *c = avctx->priv_data; + const int16_t *samples = data; + int i, j, k; + int frontier = 1 << avctx->trellis; + struct TrellisNode **nodes[2]; + struct TrellisNode **nodes_next[2]; + int pathn[2] = {0, 0}, froze = -1; + struct TrellisPath *p[2]; + + for (i = 0; i < 2; i++) { + nodes[i] = c->nodep_buf[i]; + nodes_next[i] = c->nodep_buf[i] + frontier; + memset(c->nodep_buf[i], 0, 2 * frontier * sizeof(*c->nodep_buf)); + nodes[i][0] = c->node_buf[i] + frontier; + nodes[i][0]->ssd = 0; + nodes[i][0]->path = 0; + nodes[i][0]->state = c->band[i]; + } + + for (i = 0; i < buf_size >> 1; i++) { + int xlow, xhigh; + struct TrellisNode *next[2]; + int heap_pos[2] = {0, 0}; + + for (j = 0; j < 2; j++) { + next[j] = c->node_buf[j] + frontier*(i & 1); + memset(nodes_next[j], 0, frontier * sizeof(**nodes_next)); + } + + filter_samples(c, &samples[2*i], &xlow, &xhigh); + + for (j = 0; j < frontier && nodes[0][j]; j++) { + /* Only k >> 2 affects the future adaptive state, therefore testing + * small steps that don't change k >> 2 is useless, the orignal + * value from encode_low is better than them. Since we step k + * in steps of 4, make sure range is a multiple of 4, so that + * we don't miss the original value from encode_low. */ + int range = j < frontier/2 ? 4 : 0; + struct TrellisNode *cur_node = nodes[0][j]; + + int ilow = encode_low(&cur_node->state, xlow); + + for (k = ilow - range; k <= ilow + range && k <= 63; k += 4) { + int decoded, dec_diff, pos; + uint32_t ssd; + struct TrellisNode* node; + + if (k < 0) + continue; + + decoded = av_clip((cur_node->state.scale_factor * + ff_g722_low_inv_quant6[k] >> 10) + + cur_node->state.s_predictor, -16384, 16383); + dec_diff = xlow - decoded; + +#define STORE_NODE(index, UPDATE, VALUE)\ + ssd = cur_node->ssd + dec_diff*dec_diff;\ + /* Check for wraparound. Using 64 bit ssd counters would \ + * be simpler, but is slower on x86 32 bit. */\ + if (ssd < cur_node->ssd)\ + continue;\ + if (heap_pos[index] < frontier) {\ + pos = heap_pos[index]++;\ + assert(pathn[index] < FREEZE_INTERVAL * frontier);\ + node = nodes_next[index][pos] = next[index]++;\ + node->path = pathn[index]++;\ + } else {\ + /* Try to replace one of the leaf nodes with the new \ + * one, but not always testing the same leaf position */\ + pos = (frontier>>1) + (heap_pos[index] & ((frontier>>1) - 1));\ + if (ssd >= nodes_next[index][pos]->ssd)\ + continue;\ + heap_pos[index]++;\ + node = nodes_next[index][pos];\ + }\ + node->ssd = ssd;\ + node->state = cur_node->state;\ + UPDATE;\ + c->paths[index][node->path].value = VALUE;\ + c->paths[index][node->path].prev = cur_node->path;\ + /* Sift the newly inserted node up in the heap to restore \ + * the heap property */\ + while (pos > 0) {\ + int parent = (pos - 1) >> 1;\ + if (nodes_next[index][parent]->ssd <= ssd)\ + break;\ + FFSWAP(struct TrellisNode*, nodes_next[index][parent],\ + nodes_next[index][pos]);\ + pos = parent;\ + } + STORE_NODE(0, ff_g722_update_low_predictor(&node->state, k >> 2), k); + } + } + + for (j = 0; j < frontier && nodes[1][j]; j++) { + int ihigh; + struct TrellisNode *cur_node = nodes[1][j]; + + /* We don't try to get any initial guess for ihigh via + * encode_high - since there's only 4 possible values, test + * them all. Testing all of these gives a much, much larger + * gain than testing a larger range around ilow. */ + for (ihigh = 0; ihigh < 4; ihigh++) { + int dhigh, decoded, dec_diff, pos; + uint32_t ssd; + struct TrellisNode* node; + + dhigh = cur_node->state.scale_factor * + ff_g722_high_inv_quant[ihigh] >> 10; + decoded = av_clip(dhigh + cur_node->state.s_predictor, + -16384, 16383); + dec_diff = xhigh - decoded; + + STORE_NODE(1, ff_g722_update_high_predictor(&node->state, dhigh, ihigh), ihigh); + } + } + + for (j = 0; j < 2; j++) { + FFSWAP(struct TrellisNode**, nodes[j], nodes_next[j]); + + if (nodes[j][0]->ssd > (1 << 16)) { + for (k = 1; k < frontier && nodes[j][k]; k++) + nodes[j][k]->ssd -= nodes[j][0]->ssd; + nodes[j][0]->ssd = 0; + } + } + + if (i == froze + FREEZE_INTERVAL) { + p[0] = &c->paths[0][nodes[0][0]->path]; + p[1] = &c->paths[1][nodes[1][0]->path]; + for (j = i; j > froze; j--) { + dst[j] = p[1]->value << 6 | p[0]->value; + p[0] = &c->paths[0][p[0]->prev]; + p[1] = &c->paths[1][p[1]->prev]; + } + froze = i; + pathn[0] = pathn[1] = 0; + memset(nodes[0] + 1, 0, (frontier - 1)*sizeof(**nodes)); + memset(nodes[1] + 1, 0, (frontier - 1)*sizeof(**nodes)); + } + } + + p[0] = &c->paths[0][nodes[0][0]->path]; + p[1] = &c->paths[1][nodes[1][0]->path]; + for (j = i; j > froze; j--) { + dst[j] = p[1]->value << 6 | p[0]->value; + p[0] = &c->paths[0][p[0]->prev]; + p[1] = &c->paths[1][p[1]->prev]; + } + c->band[0] = nodes[0][0]->state; + c->band[1] = nodes[1][0]->state; + + return i; +} + +static int g722_encode_frame(AVCodecContext *avctx, + uint8_t *dst, int buf_size, void *data) +{ + G722Context *c = avctx->priv_data; + const int16_t *samples = data; + int i; + + if (avctx->trellis) + return g722_encode_trellis(avctx, dst, buf_size, data); + + for (i = 0; i < buf_size >> 1; i++) { + int xlow, xhigh, ihigh, ilow; + filter_samples(c, &samples[2*i], &xlow, &xhigh); + ihigh = encode_high(&c->band[1], xhigh); + ilow = encode_low(&c->band[0], xlow); + ff_g722_update_high_predictor(&c->band[1], c->band[1].scale_factor * + ff_g722_high_inv_quant[ihigh] >> 10, ihigh); + ff_g722_update_low_predictor(&c->band[0], ilow >> 2); + *dst++ = ihigh << 6 | ilow; + } + return i; +} + +AVCodec ff_adpcm_g722_encoder = { + .name = "g722", + .type = AVMEDIA_TYPE_AUDIO, + .id = CODEC_ID_ADPCM_G722, + .priv_data_size = sizeof(G722Context), + .init = g722_encode_init, + .close = g722_encode_close, + .encode = g722_encode_frame, + .long_name = NULL_IF_CONFIG_SMALL("G.722 ADPCM"), + .sample_fmts = (const enum AVSampleFormat[]){AV_SAMPLE_FMT_S16,AV_SAMPLE_FMT_NONE}, +}; diff --git a/libavcodec/h263.h b/libavcodec/h263.h index b2b6613536..1a3c85e22e 100644 --- a/libavcodec/h263.h +++ b/libavcodec/h263.h @@ -200,48 +200,6 @@ static inline int get_p_cbp(MpegEncContext * s, return cbp; } -static inline int get_b_cbp(MpegEncContext * s, DCTELEM block[6][64], - int motion_x, int motion_y, int mb_type){ - int cbp=0, i; - - if(s->flags & CODEC_FLAG_CBP_RD){ - int score=0; - const int lambda= s->lambda2 >> (FF_LAMBDA_SHIFT - 6); - - for(i=0; i<6; i++){ - if(s->coded_score[i] < 0){ - score += s->coded_score[i]; - cbp |= 1 << (5 - i); - } - } - - if(cbp){ - int zero_score= -6; - if ((motion_x | motion_y | s->dquant | mb_type) == 0){ - zero_score-= 4; //2*MV + mb_type + cbp bit - } - - zero_score*= lambda; - if(zero_score <= score){ - cbp=0; - } - } - - for (i = 0; i < 6; i++) { - if (s->block_last_index[i] >= 0 && ((cbp >> (5 - i))&1)==0 ){ - s->block_last_index[i]= -1; - s->dsp.clear_block(s->block[i]); - } - } - }else{ - for (i = 0; i < 6; i++) { - if (s->block_last_index[i] >= 0) - cbp |= 1 << (5 - i); - } - } - return cbp; -} - static inline void memsetw(short *tab, int val, int n) { int i; diff --git a/libavcodec/h264.c b/libavcodec/h264.c index 5f8b7925a4..6ec77f5b44 100644 --- a/libavcodec/h264.c +++ b/libavcodec/h264.c @@ -1060,7 +1060,6 @@ static av_cold void common_init(H264Context *h){ h->dequant_coeff_pps= -1; s->unrestricted_mv=1; - s->decode=1; //FIXME s->dsp.dct_bits = 16; dsputil_init(&s->dsp, s->avctx); // needed so that idct permutation is known early diff --git a/libavcodec/mpeg12.c b/libavcodec/mpeg12.c index 7d469530d0..b5dfc42cc7 100644 --- a/libavcodec/mpeg12.c +++ b/libavcodec/mpeg12.c @@ -735,8 +735,9 @@ static void exchange_uv(MpegEncContext *s) #define MT_16X8 2 #define MT_DMV 3 -static int mpeg_decode_mb(MpegEncContext *s, DCTELEM block[12][64]) +static int mpeg_decode_mb(Mpeg1Context *s1, DCTELEM block[12][64]) { + MpegEncContext *s = &s1->mpeg_enc_ctx; int i, j, k, cbp, val, mb_type, motion_type; const int mb_block_count = 4 + (1 << s->chroma_format); @@ -910,7 +911,7 @@ static int mpeg_decode_mb(MpegEncContext *s, DCTELEM block[12][64]) s->mv[i][0][1]= s->last_mv[i][0][1]= s->last_mv[i][1][1] = mpeg_decode_motion(s, s->mpeg_f_code[i][1], s->last_mv[i][0][1]); /* full_pel: only for MPEG-1 */ - if (s->full_pel[i]) { + if (s1->full_pel[i]) { s->mv[i][0][0] <<= 1; s->mv[i][0][1] <<= 1; } @@ -1112,20 +1113,6 @@ static int mpeg_decode_mb(MpegEncContext *s, DCTELEM block[12][64]) return 0; } -typedef struct Mpeg1Context { - MpegEncContext mpeg_enc_ctx; - int mpeg_enc_ctx_allocated; /* true if decoding context allocated */ - int repeat_field; /* true if we must repeat the field */ - AVPanScan pan_scan; /**< some temporary storage for the panscan */ - int slice_count; - int swap_uv;//indicate VCR2 - int save_aspect_info; - int save_width, save_height, save_progressive_seq; - AVRational frame_rate_ext; ///< MPEG-2 specific framerate modificator - int sync; ///< Did we reach a sync point like a GOP/SEQ/KEYFrame? - int tmpgexs; -} Mpeg1Context; - static av_cold int mpeg_decode_init(AVCodecContext *avctx) { Mpeg1Context *s = avctx->priv_data; @@ -1376,7 +1363,7 @@ static int mpeg1_decode_picture(AVCodecContext *avctx, vbv_delay = get_bits(&s->gb, 16); if (s->pict_type == AV_PICTURE_TYPE_P || s->pict_type == AV_PICTURE_TYPE_B) { - s->full_pel[0] = get_bits1(&s->gb); + s1->full_pel[0] = get_bits1(&s->gb); f_code = get_bits(&s->gb, 3); if (f_code == 0 && (avctx->err_recognition & AV_EF_BITSTREAM)) return -1; @@ -1384,7 +1371,7 @@ static int mpeg1_decode_picture(AVCodecContext *avctx, s->mpeg_f_code[0][1] = f_code; } if (s->pict_type == AV_PICTURE_TYPE_B) { - s->full_pel[1] = get_bits1(&s->gb); + s1->full_pel[1] = get_bits1(&s->gb); f_code = get_bits(&s->gb, 3); if (f_code == 0 && (avctx->err_recognition & AV_EF_BITSTREAM)) return -1; @@ -1532,7 +1519,7 @@ static void mpeg_decode_picture_coding_extension(Mpeg1Context *s1) { MpegEncContext *s = &s1->mpeg_enc_ctx; - s->full_pel[0] = s->full_pel[1] = 0; + s1->full_pel[0] = s1->full_pel[1] = 0; s->mpeg_f_code[0][0] = get_bits(&s->gb, 4); s->mpeg_f_code[0][1] = get_bits(&s->gb, 4); s->mpeg_f_code[1][0] = get_bits(&s->gb, 4); @@ -1763,7 +1750,7 @@ static int mpeg_decode_slice(Mpeg1Context *s1, int mb_y, if (CONFIG_MPEG_XVMC_DECODER && s->avctx->xvmc_acceleration > 1) ff_xvmc_init_block(s); // set s->block - if (mpeg_decode_mb(s, s->block) < 0) + if (mpeg_decode_mb(s1, s->block) < 0) return -1; if (s->current_picture.f.motion_val[0] && !s->encoding) { // note motion_val is normally NULL unless we want to extract the MVs @@ -2171,7 +2158,7 @@ static void mpeg_decode_gop(AVCodecContext *avctx, time_code_seconds = get_bits(&s->gb, 6); time_code_pictures = get_bits(&s->gb, 6); - s->closed_gop = get_bits1(&s->gb); + s1->closed_gop = get_bits1(&s->gb); /*broken_link indicate that after editing the reference frames of the first B-Frames after GOP I-Frame are missing (open gop)*/ @@ -2321,7 +2308,8 @@ static int decode_chunks(AVCodecContext *avctx, } if (CONFIG_VDPAU && uses_vdpau(avctx)) - ff_vdpau_mpeg_picture_complete(s2, buf, buf_size, s->slice_count); + ff_vdpau_mpeg_picture_complete(s, buf, buf_size, s->slice_count); + if (slice_end(avctx, picture)) { if (s2->last_picture_ptr || s2->low_delay) //FIXME merge with the stuff in mpeg_decode_slice @@ -2450,7 +2438,7 @@ static int decode_chunks(AVCodecContext *avctx, if (s2->last_picture_ptr == NULL) { /* Skip B-frames if we do not have reference frames and gop is not closed */ if (s2->pict_type == AV_PICTURE_TYPE_B) { - if (!s2->closed_gop) + if (!s->closed_gop) break; } } @@ -2535,6 +2523,7 @@ static void flush(AVCodecContext *avctx) Mpeg1Context *s = avctx->priv_data; s->sync=0; + s->closed_gop = 0; ff_mpeg_flush(avctx); } diff --git a/libavcodec/mpeg12.h b/libavcodec/mpeg12.h index 541b66bd94..2f75d39454 100644 --- a/libavcodec/mpeg12.h +++ b/libavcodec/mpeg12.h @@ -30,6 +30,22 @@ extern VLC ff_dc_lum_vlc; extern VLC ff_dc_chroma_vlc; +typedef struct Mpeg1Context { + MpegEncContext mpeg_enc_ctx; + int mpeg_enc_ctx_allocated; /* true if decoding context allocated */ + int repeat_field; /* true if we must repeat the field */ + AVPanScan pan_scan; /**< some temporary storage for the panscan */ + int slice_count; + int swap_uv;//indicate VCR2 + int save_aspect_info; + int save_width, save_height, save_progressive_seq; + AVRational frame_rate_ext; ///< MPEG-2 specific framerate modificator + int sync; ///< Did we reach a sync point like a GOP/SEQ/KEYFrame? + int full_pel[2]; + int closed_gop; ///< GOP is closed + int tmpgexs; +} Mpeg1Context; + extern uint8_t ff_mpeg12_static_rl_table_store[2][2][2*MAX_RUN + MAX_LEVEL + 3]; void ff_mpeg12_common_init(MpegEncContext *s); diff --git a/libavcodec/mpeg4videoenc.c b/libavcodec/mpeg4videoenc.c index dac9a9c885..0c5c5f2280 100644 --- a/libavcodec/mpeg4videoenc.c +++ b/libavcodec/mpeg4videoenc.c @@ -425,6 +425,46 @@ static inline void mpeg4_encode_blocks(MpegEncContext * s, DCTELEM block[6][64], } } +static inline int get_b_cbp(MpegEncContext * s, DCTELEM block[6][64], + int motion_x, int motion_y, int mb_type) +{ + int cbp = 0, i; + + if (s->flags & CODEC_FLAG_CBP_RD) { + int score = 0; + const int lambda = s->lambda2 >> (FF_LAMBDA_SHIFT - 6); + + for (i = 0; i < 6; i++) + if (s->coded_score[i] < 0) { + score += s->coded_score[i]; + cbp |= 1 << (5 - i); + } + + if (cbp) { + int zero_score = -6; + if ((motion_x | motion_y | s->dquant | mb_type) == 0) + zero_score -= 4; //2*MV + mb_type + cbp bit + + zero_score *= lambda; + if (zero_score <= score) + cbp = 0; + } + + for (i = 0; i < 6; i++) { + if (s->block_last_index[i] >= 0 && ((cbp >> (5 - i)) & 1) == 0) { + s->block_last_index[i] = -1; + s->dsp.clear_block(s->block[i]); + } + } + } else { + for (i = 0; i < 6; i++) { + if (s->block_last_index[i] >= 0) + cbp |= 1 << (5 - i); + } + } + return cbp; +} + //FIXME this is duplicated to h263.c static const int dquant_code[5]= {1,0,9,2,3}; @@ -1213,7 +1253,6 @@ static av_cold int encode_init(AVCodecContext *avctx) s->inter_ac_vlc_length = uni_mpeg4_inter_rl_len; s->inter_ac_vlc_last_length= uni_mpeg4_inter_rl_len + 128*64; s->luma_dc_vlc_length= uni_DCtab_lum_len; - s->chroma_dc_vlc_length= uni_DCtab_chrom_len; s->ac_esc_length= 7+2+1+6+1+12+1; s->y_dc_scale_table= ff_mpeg4_y_dc_scale_table; s->c_dc_scale_table= ff_mpeg4_c_dc_scale_table; diff --git a/libavcodec/mpegvideo.c b/libavcodec/mpegvideo.c index fffa36a456..69dec5f11d 100644 --- a/libavcodec/mpegvideo.c +++ b/libavcodec/mpegvideo.c @@ -2425,7 +2425,6 @@ void ff_mpeg_flush(AVCodecContext *avctx){ s->current_picture_ptr = s->last_picture_ptr = s->next_picture_ptr = NULL; s->mb_x= s->mb_y= 0; - s->closed_gop= 0; s->parse_context.state= -1; s->parse_context.frame_start_found= 0; diff --git a/libavcodec/mpegvideo.h b/libavcodec/mpegvideo.h index d97212c358..93839e1f89 100644 --- a/libavcodec/mpegvideo.h +++ b/libavcodec/mpegvideo.h @@ -236,7 +236,6 @@ typedef struct MpegEncContext { int coded_picture_number; ///< used to set pic->coded_picture_number, should not be used for/by anything else int picture_number; //FIXME remove, unclear definition int picture_in_gop_number; ///< 0-> first pic in gop, ... - int b_frames_since_non_b; ///< used for encoding, relative to not yet reordered input int mb_width, mb_height; ///< number of MBs horizontally & vertically int mb_stride; ///< mb_width+1 used for some arrays to allow simple addressing of left & top MBs without sig11 int b8_stride; ///< 2*mb_width+1 used for some 8x8 block arrays to allow simple addressing @@ -305,7 +304,6 @@ typedef struct MpegEncContext { int last_dc[3]; ///< last DC values for MPEG1 int16_t *dc_val_base; int16_t *dc_val[3]; ///< used for mpeg4 DC prediction, all 3 arrays must be continuous - int16_t dc_cache[4*5]; const uint8_t *y_dc_scale_table; ///< qscale -> y_dc_scale table const uint8_t *c_dc_scale_table; ///< qscale -> c_dc_scale table const uint8_t *chroma_qscale_table; ///< qscale -> chroma_qscale (h263) @@ -333,7 +331,6 @@ typedef struct MpegEncContext { int *lambda_table; int adaptive_quant; ///< use adaptive quantization int dquant; ///< qscale difference to prev qscale - int closed_gop; ///< MPEG1/2 GOP is closed int pict_type; ///< AV_PICTURE_TYPE_I, AV_PICTURE_TYPE_P, AV_PICTURE_TYPE_B, ... int last_pict_type; //FIXME removes int last_non_b_pict_type; ///< used for mpeg4 gmc b-frames & ratecontrol @@ -345,7 +342,6 @@ typedef struct MpegEncContext { /* motion compensation */ int unrestricted_mv; ///< mv can point outside of the coded picture int h263_long_vectors; ///< use horrible h263v1 long vector mode - int decode; ///< if 0 then decoding will be skipped (for encoding b frames for example) DSPContext dsp; ///< pointers for accelerated dsp functions int f_code; ///< forward MV resolution @@ -440,7 +436,6 @@ typedef struct MpegEncContext { uint8_t *inter_ac_vlc_length; uint8_t *inter_ac_vlc_last_length; uint8_t *luma_dc_vlc_length; - uint8_t *chroma_dc_vlc_length; #define UNI_AC_ENC_INDEX(run,level) ((run)*128 + (level)) int coded_score[8]; @@ -462,7 +457,6 @@ typedef struct MpegEncContext { void *opaque; ///< private data for the user /* bit rate control */ - int64_t wanted_bits; int64_t total_bits; int frame_bits; ///< bits used for the current frame int next_lambda; ///< next lambda used for retrying to encode a frame @@ -643,7 +637,6 @@ typedef struct MpegEncContext { int chroma_y_shift; int progressive_frame; - int full_pel[2]; int interlaced_dct; int first_slice; int first_field; ///< is 1 for the first field of a field picture 0 otherwise diff --git a/libavcodec/vdpau.c b/libavcodec/vdpau.c index ce5103a6af..120c9dcf82 100644 --- a/libavcodec/vdpau.c +++ b/libavcodec/vdpau.c @@ -190,9 +190,10 @@ void ff_vdpau_h264_picture_complete(MpegEncContext *s) render->bitstream_buffers_used = 0; } -void ff_vdpau_mpeg_picture_complete(MpegEncContext *s, const uint8_t *buf, +void ff_vdpau_mpeg_picture_complete(Mpeg1Context *s1, const uint8_t *buf, int buf_size, int slice_count) { + MpegEncContext *s = &s1->mpeg_enc_ctx; struct vdpau_render_state *render, *last, *next; int i; @@ -211,8 +212,8 @@ void ff_vdpau_mpeg_picture_complete(MpegEncContext *s, const uint8_t *buf, render->info.mpeg.alternate_scan = s->alternate_scan; render->info.mpeg.q_scale_type = s->q_scale_type; render->info.mpeg.top_field_first = s->top_field_first; - render->info.mpeg.full_pel_forward_vector = s->full_pel[0]; // MPEG-1 only. Set 0 for MPEG-2 - render->info.mpeg.full_pel_backward_vector = s->full_pel[1]; // MPEG-1 only. Set 0 for MPEG-2 + render->info.mpeg.full_pel_forward_vector = s1->full_pel[0]; // MPEG-1 only. Set 0 for MPEG-2 + render->info.mpeg.full_pel_backward_vector = s1->full_pel[1]; // MPEG-1 only. Set 0 for MPEG-2 render->info.mpeg.f_code[0][0] = s->mpeg_f_code[0][0]; // For MPEG-1 fill both horiz. & vert. render->info.mpeg.f_code[0][1] = s->mpeg_f_code[0][1]; render->info.mpeg.f_code[1][0] = s->mpeg_f_code[1][0]; diff --git a/libavcodec/vdpau_internal.h b/libavcodec/vdpau_internal.h index 0a8d0b6b55..2d8a5d8e6f 100644 --- a/libavcodec/vdpau_internal.h +++ b/libavcodec/vdpau_internal.h @@ -26,11 +26,12 @@ #include <stdint.h> #include "mpegvideo.h" +#include "mpeg12.h" void ff_vdpau_add_data_chunk(MpegEncContext *s, const uint8_t *buf, int buf_size); -void ff_vdpau_mpeg_picture_complete(MpegEncContext *s, const uint8_t *buf, +void ff_vdpau_mpeg_picture_complete(Mpeg1Context *s1, const uint8_t *buf, int buf_size, int slice_count); void ff_vdpau_h264_picture_start(MpegEncContext *s); diff --git a/libavcodec/vorbisdec.c b/libavcodec/vorbisdec.c index 59d551e894..675de02d86 100644 --- a/libavcodec/vorbisdec.c +++ b/libavcodec/vorbisdec.c @@ -966,7 +966,7 @@ static int vorbis_parse_id_hdr(vorbis_context *vc) static av_cold int vorbis_decode_init(AVCodecContext *avccontext) { - vorbis_context *vc = avccontext->priv_data ; + vorbis_context *vc = avccontext->priv_data; uint8_t *headers = avccontext->extradata; int headers_len = avccontext->extradata_size; uint8_t *header_start[3]; @@ -1030,7 +1030,7 @@ static av_cold int vorbis_decode_init(AVCodecContext *avccontext) avccontext->sample_rate = vc->audio_samplerate; avccontext->frame_size = FFMIN(vc->blocksize[0], vc->blocksize[1]) >> 2; - return 0 ; + return 0; } // Decode audiopackets ------------------------------------------------- @@ -1608,10 +1608,10 @@ static int vorbis_decode_frame(AVCodecContext *avccontext, { const uint8_t *buf = avpkt->data; int buf_size = avpkt->size; - vorbis_context *vc = avccontext->priv_data ; + vorbis_context *vc = avccontext->priv_data; GetBitContext *gb = &(vc->gb); const float *channel_ptrs[255]; - int i, len; + int i, len, out_size; if (!buf_size) return 0; @@ -1630,12 +1630,19 @@ static int vorbis_decode_frame(AVCodecContext *avccontext, if (!vc->first_frame) { vc->first_frame = 1; *data_size = 0; - return buf_size ; + return buf_size; } av_dlog(NULL, "parsed %d bytes %d bits, returned %d samples (*ch*bits) \n", get_bits_count(gb) / 8, get_bits_count(gb) % 8, len); + out_size = len * vc->audio_channels * + av_get_bytes_per_sample(avccontext->sample_fmt); + if (*data_size < out_size) { + av_log(avccontext, AV_LOG_ERROR, "output buffer is too small\n"); + return AVERROR(EINVAL); + } + if (vc->audio_channels > 8) { for (i = 0; i < vc->audio_channels; i++) channel_ptrs[i] = vc->channel_floors + i * len; @@ -1651,10 +1658,9 @@ static int vorbis_decode_frame(AVCodecContext *avccontext, vc->fmt_conv.float_to_int16_interleave(data, channel_ptrs, len, vc->audio_channels); - *data_size = len * vc->audio_channels * - av_get_bytes_per_sample(avccontext->sample_fmt); + *data_size = out_size; - return buf_size ; + return buf_size; } // Close decoder @@ -1665,7 +1671,7 @@ static av_cold int vorbis_decode_close(AVCodecContext *avccontext) vorbis_free(vc); - return 0 ; + return 0; } AVCodec ff_vorbis_decoder = { diff --git a/libavcodec/wavpack.c b/libavcodec/wavpack.c index 0a0d8ab23e..5ff9795ef8 100644 --- a/libavcodec/wavpack.c +++ b/libavcodec/wavpack.c @@ -808,15 +808,13 @@ static int wavpack_decode_block(AVCodecContext *avctx, int block_no, } s->frame_flags = AV_RL32(buf); buf += 4; if(s->frame_flags&0x80){ - bpp = sizeof(float); avctx->sample_fmt = AV_SAMPLE_FMT_FLT; } else if((s->frame_flags&0x03) <= 1){ - bpp = 2; avctx->sample_fmt = AV_SAMPLE_FMT_S16; } else { - bpp = 4; avctx->sample_fmt = AV_SAMPLE_FMT_S32; } + bpp = av_get_bytes_per_sample(avctx->sample_fmt); samples = (uint8_t*)samples + bpp * wc->ch_offset; s->stereo = !(s->frame_flags & WV_MONO); |