00001
00002
00003
00004
00005
00006
00007
00008
00009
00010
00011
00012
00013
00014
00015
00016
00017
00018
00019
00020
00021
00022 #include "avcodec.h"
00023 #include "internal.h"
00024 #include "wma.h"
00025 #include "libavutil/avassert.h"
00026
00027 #undef NDEBUG
00028 #include <assert.h>
00029
00030
00031 static int encode_init(AVCodecContext * avctx){
00032 WMACodecContext *s = avctx->priv_data;
00033 int i, flags1, flags2;
00034 uint8_t *extradata;
00035
00036 s->avctx = avctx;
00037
00038 if(avctx->channels > MAX_CHANNELS) {
00039 av_log(avctx, AV_LOG_ERROR, "too many channels: got %i, need %i or fewer",
00040 avctx->channels, MAX_CHANNELS);
00041 return AVERROR(EINVAL);
00042 }
00043
00044 if (avctx->sample_rate > 48000) {
00045 av_log(avctx, AV_LOG_ERROR, "sample rate is too high: %d > 48kHz",
00046 avctx->sample_rate);
00047 return AVERROR(EINVAL);
00048 }
00049
00050 if(avctx->bit_rate < 24*1000) {
00051 av_log(avctx, AV_LOG_ERROR, "bitrate too low: got %i, need 24000 or higher\n",
00052 avctx->bit_rate);
00053 return AVERROR(EINVAL);
00054 }
00055
00056
00057 flags1 = 0;
00058 flags2 = 1;
00059 if (avctx->codec->id == CODEC_ID_WMAV1) {
00060 extradata= av_malloc(4);
00061 avctx->extradata_size= 4;
00062 AV_WL16(extradata, flags1);
00063 AV_WL16(extradata+2, flags2);
00064 } else if (avctx->codec->id == CODEC_ID_WMAV2) {
00065 extradata= av_mallocz(10);
00066 avctx->extradata_size= 10;
00067 AV_WL32(extradata, flags1);
00068 AV_WL16(extradata+4, flags2);
00069 }else
00070 assert(0);
00071 avctx->extradata= extradata;
00072 s->use_exp_vlc = flags2 & 0x0001;
00073 s->use_bit_reservoir = flags2 & 0x0002;
00074 s->use_variable_block_len = flags2 & 0x0004;
00075 if (avctx->channels == 2)
00076 s->ms_stereo = 1;
00077
00078 ff_wma_init(avctx, flags2);
00079
00080
00081 for(i = 0; i < s->nb_block_sizes; i++)
00082 ff_mdct_init(&s->mdct_ctx[i], s->frame_len_bits - i + 1, 0, 1.0);
00083
00084 s->block_align = avctx->bit_rate * (int64_t)s->frame_len /
00085 (avctx->sample_rate * 8);
00086 s->block_align = FFMIN(s->block_align, MAX_CODED_SUPERFRAME_SIZE);
00087 avctx->block_align = s->block_align;
00088 avctx->bit_rate = avctx->block_align * 8LL * avctx->sample_rate /
00089 s->frame_len;
00090
00091 avctx->frame_size = avctx->delay = s->frame_len;
00092
00093 #if FF_API_OLD_ENCODE_AUDIO
00094 avctx->coded_frame = &s->frame;
00095 avcodec_get_frame_defaults(avctx->coded_frame);
00096 #endif
00097
00098 return 0;
00099 }
00100
00101
00102 static void apply_window_and_mdct(AVCodecContext * avctx, const signed short * audio, int len) {
00103 WMACodecContext *s = avctx->priv_data;
00104 int window_index= s->frame_len_bits - s->block_len_bits;
00105 FFTContext *mdct = &s->mdct_ctx[window_index];
00106 int i, j, channel;
00107 const float * win = s->windows[window_index];
00108 int window_len = 1 << s->block_len_bits;
00109 float n = window_len/2;
00110
00111 for (channel = 0; channel < avctx->channels; channel++) {
00112 memcpy(s->output, s->frame_out[channel], sizeof(float)*window_len);
00113 j = channel;
00114 for (i = 0; i < len; i++, j += avctx->channels){
00115 s->output[i+window_len] = audio[j] / n * win[window_len - i - 1];
00116 s->frame_out[channel][i] = audio[j] / n * win[i];
00117 }
00118 mdct->mdct_calc(mdct, s->coefs[channel], s->output);
00119 }
00120 }
00121
00122
00123 static void init_exp(WMACodecContext *s, int ch, const int *exp_param){
00124 int n;
00125 const uint16_t *ptr;
00126 float v, *q, max_scale, *q_end;
00127
00128 ptr = s->exponent_bands[s->frame_len_bits - s->block_len_bits];
00129 q = s->exponents[ch];
00130 q_end = q + s->block_len;
00131 max_scale = 0;
00132 while (q < q_end) {
00133
00134 v = pow(10, *exp_param++ * (1.0 / 16.0));
00135 max_scale= FFMAX(max_scale, v);
00136 n = *ptr++;
00137 do {
00138 *q++ = v;
00139 } while (--n);
00140 }
00141 s->max_exponent[ch] = max_scale;
00142 }
00143
00144 static void encode_exp_vlc(WMACodecContext *s, int ch, const int *exp_param){
00145 int last_exp;
00146 const uint16_t *ptr;
00147 float *q, *q_end;
00148
00149 ptr = s->exponent_bands[s->frame_len_bits - s->block_len_bits];
00150 q = s->exponents[ch];
00151 q_end = q + s->block_len;
00152 if (s->version == 1) {
00153 last_exp= *exp_param++;
00154 assert(last_exp-10 >= 0 && last_exp-10 < 32);
00155 put_bits(&s->pb, 5, last_exp - 10);
00156 q+= *ptr++;
00157 }else
00158 last_exp = 36;
00159 while (q < q_end) {
00160 int exp = *exp_param++;
00161 int code = exp - last_exp + 60;
00162 assert(code >= 0 && code < 120);
00163 put_bits(&s->pb, ff_aac_scalefactor_bits[code], ff_aac_scalefactor_code[code]);
00164
00165 q+= *ptr++;
00166 last_exp= exp;
00167 }
00168 }
00169
00170 static int encode_block(WMACodecContext *s, float (*src_coefs)[BLOCK_MAX_SIZE], int total_gain){
00171 int v, bsize, ch, coef_nb_bits, parse_exponents;
00172 float mdct_norm;
00173 int nb_coefs[MAX_CHANNELS];
00174 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};
00175
00176
00177 if (s->use_variable_block_len) {
00178 assert(0);
00179 }else{
00180
00181 s->next_block_len_bits = s->frame_len_bits;
00182 s->prev_block_len_bits = s->frame_len_bits;
00183 s->block_len_bits = s->frame_len_bits;
00184 }
00185
00186 s->block_len = 1 << s->block_len_bits;
00187
00188 bsize = s->frame_len_bits - s->block_len_bits;
00189
00190
00191 v = s->coefs_end[bsize] - s->coefs_start;
00192 for(ch = 0; ch < s->nb_channels; ch++)
00193 nb_coefs[ch] = v;
00194 {
00195 int n4 = s->block_len / 2;
00196 mdct_norm = 1.0 / (float)n4;
00197 if (s->version == 1) {
00198 mdct_norm *= sqrt(n4);
00199 }
00200 }
00201
00202 if (s->nb_channels == 2) {
00203 put_bits(&s->pb, 1, !!s->ms_stereo);
00204 }
00205
00206 for(ch = 0; ch < s->nb_channels; ch++) {
00207 s->channel_coded[ch] = 1;
00208 if (s->channel_coded[ch]) {
00209 init_exp(s, ch, fixed_exp);
00210 }
00211 }
00212
00213 for(ch = 0; ch < s->nb_channels; ch++) {
00214 if (s->channel_coded[ch]) {
00215 WMACoef *coefs1;
00216 float *coefs, *exponents, mult;
00217 int i, n;
00218
00219 coefs1 = s->coefs1[ch];
00220 exponents = s->exponents[ch];
00221 mult = pow(10, total_gain * 0.05) / s->max_exponent[ch];
00222 mult *= mdct_norm;
00223 coefs = src_coefs[ch];
00224 if (s->use_noise_coding && 0) {
00225 assert(0);
00226 } else {
00227 coefs += s->coefs_start;
00228 n = nb_coefs[ch];
00229 for(i = 0;i < n; i++){
00230 double t= *coefs++ / (exponents[i] * mult);
00231 if(t<-32768 || t>32767)
00232 return -1;
00233
00234 coefs1[i] = lrint(t);
00235 }
00236 }
00237 }
00238 }
00239
00240 v = 0;
00241 for(ch = 0; ch < s->nb_channels; ch++) {
00242 int a = s->channel_coded[ch];
00243 put_bits(&s->pb, 1, a);
00244 v |= a;
00245 }
00246
00247 if (!v)
00248 return 1;
00249
00250 for(v= total_gain-1; v>=127; v-= 127)
00251 put_bits(&s->pb, 7, 127);
00252 put_bits(&s->pb, 7, v);
00253
00254 coef_nb_bits= ff_wma_total_gain_to_bits(total_gain);
00255
00256 if (s->use_noise_coding) {
00257 for(ch = 0; ch < s->nb_channels; ch++) {
00258 if (s->channel_coded[ch]) {
00259 int i, n;
00260 n = s->exponent_high_sizes[bsize];
00261 for(i=0;i<n;i++) {
00262 put_bits(&s->pb, 1, s->high_band_coded[ch][i]= 0);
00263 if (0)
00264 nb_coefs[ch] -= s->exponent_high_bands[bsize][i];
00265 }
00266 }
00267 }
00268 }
00269
00270 parse_exponents = 1;
00271 if (s->block_len_bits != s->frame_len_bits) {
00272 put_bits(&s->pb, 1, parse_exponents);
00273 }
00274
00275 if (parse_exponents) {
00276 for(ch = 0; ch < s->nb_channels; ch++) {
00277 if (s->channel_coded[ch]) {
00278 if (s->use_exp_vlc) {
00279 encode_exp_vlc(s, ch, fixed_exp);
00280 } else {
00281 assert(0);
00282
00283 }
00284 }
00285 }
00286 } else {
00287 assert(0);
00288 }
00289
00290 for(ch = 0; ch < s->nb_channels; ch++) {
00291 if (s->channel_coded[ch]) {
00292 int run, tindex;
00293 WMACoef *ptr, *eptr;
00294 tindex = (ch == 1 && s->ms_stereo);
00295 ptr = &s->coefs1[ch][0];
00296 eptr = ptr + nb_coefs[ch];
00297
00298 run=0;
00299 for(;ptr < eptr; ptr++){
00300 if(*ptr){
00301 int level= *ptr;
00302 int abs_level= FFABS(level);
00303 int code= 0;
00304 if(abs_level <= s->coef_vlcs[tindex]->max_level){
00305 if(run < s->coef_vlcs[tindex]->levels[abs_level-1])
00306 code= run + s->int_table[tindex][abs_level-1];
00307 }
00308
00309 assert(code < s->coef_vlcs[tindex]->n);
00310 put_bits(&s->pb, s->coef_vlcs[tindex]->huffbits[code], s->coef_vlcs[tindex]->huffcodes[code]);
00311
00312 if(code == 0){
00313 if(1<<coef_nb_bits <= abs_level)
00314 return -1;
00315
00316 put_bits(&s->pb, coef_nb_bits, abs_level);
00317 put_bits(&s->pb, s->frame_len_bits, run);
00318 }
00319 put_bits(&s->pb, 1, level < 0);
00320 run=0;
00321 }else{
00322 run++;
00323 }
00324 }
00325 if(run)
00326 put_bits(&s->pb, s->coef_vlcs[tindex]->huffbits[1], s->coef_vlcs[tindex]->huffcodes[1]);
00327 }
00328 if (s->version == 1 && s->nb_channels >= 2) {
00329 avpriv_align_put_bits(&s->pb);
00330 }
00331 }
00332 return 0;
00333 }
00334
00335 static int encode_frame(WMACodecContext *s, float (*src_coefs)[BLOCK_MAX_SIZE], uint8_t *buf, int buf_size, int total_gain){
00336 init_put_bits(&s->pb, buf, buf_size);
00337
00338 if (s->use_bit_reservoir) {
00339 assert(0);
00340 }else{
00341 if(encode_block(s, src_coefs, total_gain) < 0)
00342 return INT_MAX;
00343 }
00344
00345 avpriv_align_put_bits(&s->pb);
00346
00347 return put_bits_count(&s->pb)/8 - s->block_align;
00348 }
00349
00350 static int encode_superframe(AVCodecContext *avctx, AVPacket *avpkt,
00351 const AVFrame *frame, int *got_packet_ptr)
00352 {
00353 WMACodecContext *s = avctx->priv_data;
00354 const int16_t *samples = (const int16_t *)frame->data[0];
00355 int i, total_gain, ret;
00356
00357 s->block_len_bits= s->frame_len_bits;
00358 s->block_len = 1 << s->block_len_bits;
00359
00360 apply_window_and_mdct(avctx, samples, frame->nb_samples);
00361
00362 if (s->ms_stereo) {
00363 float a, b;
00364 int i;
00365
00366 for(i = 0; i < s->block_len; i++) {
00367 a = s->coefs[0][i]*0.5;
00368 b = s->coefs[1][i]*0.5;
00369 s->coefs[0][i] = a + b;
00370 s->coefs[1][i] = a - b;
00371 }
00372 }
00373
00374 if ((ret = ff_alloc_packet2(avctx, avpkt, 2 * MAX_CODED_SUPERFRAME_SIZE)))
00375 return ret;
00376
00377 #if 1
00378 total_gain= 128;
00379 for(i=64; i; i>>=1){
00380 int error = encode_frame(s, s->coefs, avpkt->data, avpkt->size,
00381 total_gain - i);
00382 if(error<0)
00383 total_gain-= i;
00384 }
00385 #else
00386 total_gain= 90;
00387 best = encode_frame(s, s->coefs, avpkt->data, avpkt->size, total_gain);
00388 for(i=32; i; i>>=1){
00389 int scoreL = encode_frame(s, s->coefs, avpkt->data, avpkt->size, total_gain - i);
00390 int scoreR = encode_frame(s, s->coefs, avpkt->data, avpkt->size, total_gain + i);
00391 av_log(NULL, AV_LOG_ERROR, "%d %d %d (%d)\n", scoreL, best, scoreR, total_gain);
00392 if(scoreL < FFMIN(best, scoreR)){
00393 best = scoreL;
00394 total_gain -= i;
00395 }else if(scoreR < best){
00396 best = scoreR;
00397 total_gain += i;
00398 }
00399 }
00400 #endif
00401
00402 encode_frame(s, s->coefs, avpkt->data, avpkt->size, total_gain);
00403 av_assert0((put_bits_count(&s->pb) & 7) == 0);
00404 i= s->block_align - (put_bits_count(&s->pb)+7)/8;
00405 av_assert0(i>=0);
00406 while(i--)
00407 put_bits(&s->pb, 8, 'N');
00408
00409 flush_put_bits(&s->pb);
00410 av_assert0(put_bits_ptr(&s->pb) - s->pb.buf == s->block_align);
00411
00412 if (frame->pts != AV_NOPTS_VALUE)
00413 avpkt->pts = frame->pts - ff_samples_to_time_base(avctx, avctx->delay);
00414
00415 avpkt->size = s->block_align;
00416 *got_packet_ptr = 1;
00417 return 0;
00418 }
00419
00420 AVCodec ff_wmav1_encoder = {
00421 .name = "wmav1",
00422 .type = AVMEDIA_TYPE_AUDIO,
00423 .id = CODEC_ID_WMAV1,
00424 .priv_data_size = sizeof(WMACodecContext),
00425 .init = encode_init,
00426 .encode2 = encode_superframe,
00427 .close = ff_wma_end,
00428 .sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_S16,
00429 AV_SAMPLE_FMT_NONE },
00430 .long_name = NULL_IF_CONFIG_SMALL("Windows Media Audio 1"),
00431 };
00432
00433 AVCodec ff_wmav2_encoder = {
00434 .name = "wmav2",
00435 .type = AVMEDIA_TYPE_AUDIO,
00436 .id = CODEC_ID_WMAV2,
00437 .priv_data_size = sizeof(WMACodecContext),
00438 .init = encode_init,
00439 .encode2 = encode_superframe,
00440 .close = ff_wma_end,
00441 .sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_S16,
00442 AV_SAMPLE_FMT_NONE },
00443 .long_name = NULL_IF_CONFIG_SMALL("Windows Media Audio 2"),
00444 };