yading@10: /*
yading@10:  * WMA compatible encoder
yading@10:  * Copyright (c) 2007 Michael Niedermayer
yading@10:  *
yading@10:  * This file is part of FFmpeg.
yading@10:  *
yading@10:  * FFmpeg is free software; you can redistribute it and/or
yading@10:  * modify it under the terms of the GNU Lesser General Public
yading@10:  * License as published by the Free Software Foundation; either
yading@10:  * version 2.1 of the License, or (at your option) any later version.
yading@10:  *
yading@10:  * FFmpeg is distributed in the hope that it will be useful,
yading@10:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
yading@10:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
yading@10:  * Lesser General Public License for more details.
yading@10:  *
yading@10:  * You should have received a copy of the GNU Lesser General Public
yading@10:  * License along with FFmpeg; if not, write to the Free Software
yading@10:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
yading@10:  */
yading@10: 
yading@10: #include "avcodec.h"
yading@10: #include "internal.h"
yading@10: #include "wma.h"
yading@10: #include "libavutil/avassert.h"
yading@10: 
yading@10: 
yading@10: static int encode_init(AVCodecContext * avctx){
yading@10:     WMACodecContext *s = avctx->priv_data;
yading@10:     int i, flags1, flags2, block_align;
yading@10:     uint8_t *extradata;
yading@10: 
yading@10:     s->avctx = avctx;
yading@10: 
yading@10:     if(avctx->channels > MAX_CHANNELS) {
yading@10:         av_log(avctx, AV_LOG_ERROR, "too many channels: got %i, need %i or fewer\n",
yading@10:                avctx->channels, MAX_CHANNELS);
yading@10:         return AVERROR(EINVAL);
yading@10:     }
yading@10: 
yading@10:     if (avctx->sample_rate > 48000) {
yading@10:         av_log(avctx, AV_LOG_ERROR, "sample rate is too high: %d > 48kHz\n",
yading@10:                avctx->sample_rate);
yading@10:         return AVERROR(EINVAL);
yading@10:     }
yading@10: 
yading@10:     if(avctx->bit_rate < 24*1000) {
yading@10:         av_log(avctx, AV_LOG_ERROR, "bitrate too low: got %i, need 24000 or higher\n",
yading@10:                avctx->bit_rate);
yading@10:         return AVERROR(EINVAL);
yading@10:     }
yading@10: 
yading@10:     /* extract flag infos */
yading@10:     flags1 = 0;
yading@10:     flags2 = 1;
yading@10:     if (avctx->codec->id == AV_CODEC_ID_WMAV1) {
yading@10:         extradata= av_malloc(4);
yading@10:         avctx->extradata_size= 4;
yading@10:         AV_WL16(extradata, flags1);
yading@10:         AV_WL16(extradata+2, flags2);
yading@10:     } else if (avctx->codec->id == AV_CODEC_ID_WMAV2) {
yading@10:         extradata= av_mallocz(10);
yading@10:         avctx->extradata_size= 10;
yading@10:         AV_WL32(extradata, flags1);
yading@10:         AV_WL16(extradata+4, flags2);
yading@10:     }else
yading@10:         av_assert0(0);
yading@10:     avctx->extradata= extradata;
yading@10:     s->use_exp_vlc = flags2 & 0x0001;
yading@10:     s->use_bit_reservoir = flags2 & 0x0002;
yading@10:     s->use_variable_block_len = flags2 & 0x0004;
yading@10:     if (avctx->channels == 2)
yading@10:         s->ms_stereo = 1;
yading@10: 
yading@10:     ff_wma_init(avctx, flags2);
yading@10: 
yading@10:     /* init MDCT */
yading@10:     for(i = 0; i < s->nb_block_sizes; i++)
yading@10:         ff_mdct_init(&s->mdct_ctx[i], s->frame_len_bits - i + 1, 0, 1.0);
yading@10: 
yading@10:     block_align        = avctx->bit_rate * (int64_t)s->frame_len /
yading@10:                          (avctx->sample_rate * 8);
yading@10:     block_align        = FFMIN(block_align, MAX_CODED_SUPERFRAME_SIZE);
yading@10:     avctx->block_align = block_align;
yading@10: 
yading@10:     avctx->frame_size = avctx->delay = s->frame_len;
yading@10: 
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: 
yading@10: static void apply_window_and_mdct(AVCodecContext * avctx, const AVFrame *frame)
yading@10: {
yading@10:     WMACodecContext *s = avctx->priv_data;
yading@10:     float **audio      = (float **)frame->extended_data;
yading@10:     int len            = frame->nb_samples;
yading@10:     int window_index= s->frame_len_bits - s->block_len_bits;
yading@10:     FFTContext *mdct = &s->mdct_ctx[window_index];
yading@10:     int ch;
yading@10:     const float * win = s->windows[window_index];
yading@10:     int window_len = 1 << s->block_len_bits;
yading@10:     float n = 2.0 * 32768.0 / window_len;
yading@10: 
yading@10:     for (ch = 0; ch < avctx->channels; ch++) {
yading@10:         memcpy(s->output, s->frame_out[ch], window_len * sizeof(*s->output));
yading@10:         s->fdsp.vector_fmul_scalar(s->frame_out[ch], audio[ch], n, len);
yading@10:         s->fdsp.vector_fmul_reverse(&s->output[window_len], s->frame_out[ch], win, len);
yading@10:         s->fdsp.vector_fmul(s->frame_out[ch], s->frame_out[ch], win, len);
yading@10:         mdct->mdct_calc(mdct, s->coefs[ch], s->output);
yading@10:     }
yading@10: }
yading@10: 
yading@10: //FIXME use for decoding too
yading@10: static void init_exp(WMACodecContext *s, int ch, const int *exp_param){
yading@10:     int n;
yading@10:     const uint16_t *ptr;
yading@10:     float v, *q, max_scale, *q_end;
yading@10: 
yading@10:     ptr = s->exponent_bands[s->frame_len_bits - s->block_len_bits];
yading@10:     q = s->exponents[ch];
yading@10:     q_end = q + s->block_len;
yading@10:     max_scale = 0;
yading@10:     while (q < q_end) {
yading@10:         /* XXX: use a table */
yading@10:         v = pow(10, *exp_param++ * (1.0 / 16.0));
yading@10:         max_scale= FFMAX(max_scale, v);
yading@10:         n = *ptr++;
yading@10:         do {
yading@10:             *q++ = v;
yading@10:         } while (--n);
yading@10:     }
yading@10:     s->max_exponent[ch] = max_scale;
yading@10: }
yading@10: 
yading@10: static void encode_exp_vlc(WMACodecContext *s, int ch, const int *exp_param){
yading@10:     int last_exp;
yading@10:     const uint16_t *ptr;
yading@10:     float *q, *q_end;
yading@10: 
yading@10:     ptr = s->exponent_bands[s->frame_len_bits - s->block_len_bits];
yading@10:     q = s->exponents[ch];
yading@10:     q_end = q + s->block_len;
yading@10:     if (s->version == 1) {
yading@10:         last_exp= *exp_param++;
yading@10:         av_assert0(last_exp-10 >= 0 && last_exp-10 < 32);
yading@10:         put_bits(&s->pb, 5, last_exp - 10);
yading@10:         q+= *ptr++;
yading@10:     }else
yading@10:         last_exp = 36;
yading@10:     while (q < q_end) {
yading@10:         int exp = *exp_param++;
yading@10:         int code = exp - last_exp + 60;
yading@10:         av_assert1(code >= 0 && code < 120);
yading@10:         put_bits(&s->pb, ff_aac_scalefactor_bits[code], ff_aac_scalefactor_code[code]);
yading@10:         /* XXX: use a table */
yading@10:         q+= *ptr++;
yading@10:         last_exp= exp;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static int encode_block(WMACodecContext *s, float (*src_coefs)[BLOCK_MAX_SIZE], int total_gain){
yading@10:     int v, bsize, ch, coef_nb_bits, parse_exponents;
yading@10:     float mdct_norm;
yading@10:     int nb_coefs[MAX_CHANNELS];
yading@10:     static const int fixed_exp[25]={20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20};
yading@10: 
yading@10:     //FIXME remove duplication relative to decoder
yading@10:     if (s->use_variable_block_len) {
yading@10:         av_assert0(0); //FIXME not implemented
yading@10:     }else{
yading@10:         /* fixed block len */
yading@10:         s->next_block_len_bits = s->frame_len_bits;
yading@10:         s->prev_block_len_bits = s->frame_len_bits;
yading@10:         s->block_len_bits = s->frame_len_bits;
yading@10:     }
yading@10: 
yading@10:     s->block_len = 1 << s->block_len_bits;
yading@10: //     assert((s->block_pos + s->block_len) <= s->frame_len);
yading@10:     bsize = s->frame_len_bits - s->block_len_bits;
yading@10: 
yading@10:     //FIXME factor
yading@10:     v = s->coefs_end[bsize] - s->coefs_start;
yading@10:     for (ch = 0; ch < s->avctx->channels; ch++)
yading@10:         nb_coefs[ch] = v;
yading@10:     {
yading@10:         int n4 = s->block_len / 2;
yading@10:         mdct_norm = 1.0 / (float)n4;
yading@10:         if (s->version == 1) {
yading@10:             mdct_norm *= sqrt(n4);
yading@10:         }
yading@10:     }
yading@10: 
yading@10:     if (s->avctx->channels == 2) {
yading@10:         put_bits(&s->pb, 1, !!s->ms_stereo);
yading@10:     }
yading@10: 
yading@10:     for (ch = 0; ch < s->avctx->channels; ch++) {
yading@10:         s->channel_coded[ch] = 1; //FIXME only set channel_coded when needed, instead of always
yading@10:         if (s->channel_coded[ch]) {
yading@10:             init_exp(s, ch, fixed_exp);
yading@10:         }
yading@10:     }
yading@10: 
yading@10:     for (ch = 0; ch < s->avctx->channels; ch++) {
yading@10:         if (s->channel_coded[ch]) {
yading@10:             WMACoef *coefs1;
yading@10:             float *coefs, *exponents, mult;
yading@10:             int i, n;
yading@10: 
yading@10:             coefs1 = s->coefs1[ch];
yading@10:             exponents = s->exponents[ch];
yading@10:             mult = pow(10, total_gain * 0.05) / s->max_exponent[ch];
yading@10:             mult *= mdct_norm;
yading@10:             coefs = src_coefs[ch];
yading@10:             if (s->use_noise_coding && 0) {
yading@10:                 av_assert0(0); //FIXME not implemented
yading@10:             } else {
yading@10:                 coefs += s->coefs_start;
yading@10:                 n = nb_coefs[ch];
yading@10:                 for(i = 0;i < n; i++){
yading@10:                     double t= *coefs++ / (exponents[i] * mult);
yading@10:                     if(t<-32768 || t>32767)
yading@10:                         return -1;
yading@10: 
yading@10:                     coefs1[i] = lrint(t);
yading@10:                 }
yading@10:             }
yading@10:         }
yading@10:     }
yading@10: 
yading@10:     v = 0;
yading@10:     for (ch = 0; ch < s->avctx->channels; ch++) {
yading@10:         int a = s->channel_coded[ch];
yading@10:         put_bits(&s->pb, 1, a);
yading@10:         v |= a;
yading@10:     }
yading@10: 
yading@10:     if (!v)
yading@10:         return 1;
yading@10: 
yading@10:     for(v= total_gain-1; v>=127; v-= 127)
yading@10:         put_bits(&s->pb, 7, 127);
yading@10:     put_bits(&s->pb, 7, v);
yading@10: 
yading@10:     coef_nb_bits= ff_wma_total_gain_to_bits(total_gain);
yading@10: 
yading@10:     if (s->use_noise_coding) {
yading@10:         for (ch = 0; ch < s->avctx->channels; ch++) {
yading@10:             if (s->channel_coded[ch]) {
yading@10:                 int i, n;
yading@10:                 n = s->exponent_high_sizes[bsize];
yading@10:                 for(i=0;i<n;i++) {
yading@10:                     put_bits(&s->pb, 1, s->high_band_coded[ch][i]= 0);
yading@10:                     if (0)
yading@10:                         nb_coefs[ch] -= s->exponent_high_bands[bsize][i];
yading@10:                 }
yading@10:             }
yading@10:         }
yading@10:     }
yading@10: 
yading@10:     parse_exponents = 1;
yading@10:     if (s->block_len_bits != s->frame_len_bits) {
yading@10:         put_bits(&s->pb, 1, parse_exponents);
yading@10:     }
yading@10: 
yading@10:     if (parse_exponents) {
yading@10:         for (ch = 0; ch < s->avctx->channels; ch++) {
yading@10:             if (s->channel_coded[ch]) {
yading@10:                 if (s->use_exp_vlc) {
yading@10:                     encode_exp_vlc(s, ch, fixed_exp);
yading@10:                 } else {
yading@10:                     av_assert0(0); //FIXME not implemented
yading@10: //                    encode_exp_lsp(s, ch);
yading@10:                 }
yading@10:             }
yading@10:         }
yading@10:     } else {
yading@10:         av_assert0(0); //FIXME not implemented
yading@10:     }
yading@10: 
yading@10:     for (ch = 0; ch < s->avctx->channels; ch++) {
yading@10:         if (s->channel_coded[ch]) {
yading@10:             int run, tindex;
yading@10:             WMACoef *ptr, *eptr;
yading@10:             tindex = (ch == 1 && s->ms_stereo);
yading@10:             ptr = &s->coefs1[ch][0];
yading@10:             eptr = ptr + nb_coefs[ch];
yading@10: 
yading@10:             run=0;
yading@10:             for(;ptr < eptr; ptr++){
yading@10:                 if(*ptr){
yading@10:                     int level= *ptr;
yading@10:                     int abs_level= FFABS(level);
yading@10:                     int code= 0;
yading@10:                     if(abs_level <= s->coef_vlcs[tindex]->max_level){
yading@10:                         if(run < s->coef_vlcs[tindex]->levels[abs_level-1])
yading@10:                             code= run + s->int_table[tindex][abs_level-1];
yading@10:                     }
yading@10: 
yading@10:                     av_assert2(code < s->coef_vlcs[tindex]->n);
yading@10:                     put_bits(&s->pb, s->coef_vlcs[tindex]->huffbits[code], s->coef_vlcs[tindex]->huffcodes[code]);
yading@10: 
yading@10:                     if(code == 0){
yading@10:                         if(1<<coef_nb_bits <= abs_level)
yading@10:                             return -1;
yading@10: 
yading@10:                         put_bits(&s->pb, coef_nb_bits, abs_level);
yading@10:                         put_bits(&s->pb, s->frame_len_bits, run);
yading@10:                     }
yading@10:                     put_bits(&s->pb, 1, level < 0); //FIXME the sign is fliped somewhere
yading@10:                     run=0;
yading@10:                 }else{
yading@10:                     run++;
yading@10:                 }
yading@10:             }
yading@10:             if(run)
yading@10:                 put_bits(&s->pb, s->coef_vlcs[tindex]->huffbits[1], s->coef_vlcs[tindex]->huffcodes[1]);
yading@10:         }
yading@10:         if (s->version == 1 && s->avctx->channels >= 2) {
yading@10:             avpriv_align_put_bits(&s->pb);
yading@10:         }
yading@10:     }
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: static int encode_frame(WMACodecContext *s, float (*src_coefs)[BLOCK_MAX_SIZE], uint8_t *buf, int buf_size, int total_gain){
yading@10:     init_put_bits(&s->pb, buf, buf_size);
yading@10: 
yading@10:     if (s->use_bit_reservoir) {
yading@10:         av_assert0(0);//FIXME not implemented
yading@10:     }else{
yading@10:         if(encode_block(s, src_coefs, total_gain) < 0)
yading@10:             return INT_MAX;
yading@10:     }
yading@10: 
yading@10:     avpriv_align_put_bits(&s->pb);
yading@10: 
yading@10:     return put_bits_count(&s->pb) / 8 - s->avctx->block_align;
yading@10: }
yading@10: 
yading@10: static int encode_superframe(AVCodecContext *avctx, AVPacket *avpkt,
yading@10:                              const AVFrame *frame, int *got_packet_ptr)
yading@10: {
yading@10:     WMACodecContext *s = avctx->priv_data;
yading@10:     int i, total_gain, ret, error;
yading@10: 
yading@10:     s->block_len_bits= s->frame_len_bits; //required by non variable block len
yading@10:     s->block_len = 1 << s->block_len_bits;
yading@10: 
yading@10:     apply_window_and_mdct(avctx, frame);
yading@10: 
yading@10:     if (s->ms_stereo) {
yading@10:         float a, b;
yading@10:         int i;
yading@10: 
yading@10:         for(i = 0; i < s->block_len; i++) {
yading@10:             a = s->coefs[0][i]*0.5;
yading@10:             b = s->coefs[1][i]*0.5;
yading@10:             s->coefs[0][i] = a + b;
yading@10:             s->coefs[1][i] = a - b;
yading@10:         }
yading@10:     }
yading@10: 
yading@10:     if ((ret = ff_alloc_packet2(avctx, avpkt, 2 * MAX_CODED_SUPERFRAME_SIZE)) < 0)
yading@10:         return ret;
yading@10: 
yading@10:     total_gain= 128;
yading@10:     for(i=64; i; i>>=1){
yading@10:         error = encode_frame(s, s->coefs, avpkt->data, avpkt->size,
yading@10:                                  total_gain - i);
yading@10:         if(error<=0)
yading@10:             total_gain-= i;
yading@10:     }
yading@10: 
yading@10:     while(total_gain <= 128 && error > 0)
yading@10:         error = encode_frame(s, s->coefs, avpkt->data, avpkt->size, total_gain++);
yading@10:     av_assert0((put_bits_count(&s->pb) & 7) == 0);
yading@10:     i= avctx->block_align - (put_bits_count(&s->pb)+7)/8;
yading@10:     av_assert0(i>=0);
yading@10:     while(i--)
yading@10:         put_bits(&s->pb, 8, 'N');
yading@10: 
yading@10:     flush_put_bits(&s->pb);
yading@10:     av_assert0(put_bits_ptr(&s->pb) - s->pb.buf == avctx->block_align);
yading@10: 
yading@10:     if (frame->pts != AV_NOPTS_VALUE)
yading@10:         avpkt->pts = frame->pts - ff_samples_to_time_base(avctx, avctx->delay);
yading@10: 
yading@10:     avpkt->size = avctx->block_align;
yading@10:     *got_packet_ptr = 1;
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: #if CONFIG_WMAV1_ENCODER
yading@10: AVCodec ff_wmav1_encoder = {
yading@10:     .name           = "wmav1",
yading@10:     .type           = AVMEDIA_TYPE_AUDIO,
yading@10:     .id             = AV_CODEC_ID_WMAV1,
yading@10:     .priv_data_size = sizeof(WMACodecContext),
yading@10:     .init           = encode_init,
yading@10:     .encode2        = encode_superframe,
yading@10:     .close          = ff_wma_end,
yading@10:     .sample_fmts    = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_FLTP,
yading@10:                                                      AV_SAMPLE_FMT_NONE },
yading@10:     .long_name      = NULL_IF_CONFIG_SMALL("Windows Media Audio 1"),
yading@10: };
yading@10: #endif
yading@10: #if CONFIG_WMAV2_ENCODER
yading@10: AVCodec ff_wmav2_encoder = {
yading@10:     .name           = "wmav2",
yading@10:     .type           = AVMEDIA_TYPE_AUDIO,
yading@10:     .id             = AV_CODEC_ID_WMAV2,
yading@10:     .priv_data_size = sizeof(WMACodecContext),
yading@10:     .init           = encode_init,
yading@10:     .encode2        = encode_superframe,
yading@10:     .close          = ff_wma_end,
yading@10:     .sample_fmts    = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_FLTP,
yading@10:                                                      AV_SAMPLE_FMT_NONE },
yading@10:     .long_name      = NULL_IF_CONFIG_SMALL("Windows Media Audio 2"),
yading@10: };
yading@10: #endif