January 2009 - FFmpeg-soc - lists.ffmpeg.org

[soc]: r4014 - wmapro/wma3dec.c
by faust3 01 Feb '09

01 Feb '09

Author: faust3 Date: Sat Jan 31 19:08:18 2009 New Revision: 4014 Log: use vector_fmul_window for the windowing Modified: wmapro/wma3dec.c Modified: wmapro/wma3dec.c ============================================================================== --- wmapro/wma3dec.c Tue Jan 27 21:57:50 2009 (r4013) +++ wmapro/wma3dec.c Sat Jan 31 19:08:18 2009 (r4014) @@ -1074,43 +1074,22 @@ static void wma_window(WMA3DecodeContext for(i=0;i<s->channels_for_cur_subframe;i++){ int c = s->channel_indexes_for_cur_subframe[i]; int j = s->channel[c].cur_subframe; - int x; - // FIXME: use dsp.vector_fmul_window float* start; - float* end; float* window; int prev_block_len = s->channel[c].prev_block_len; int block_len = s->channel[c].subframe_len[j]; - int len; - int winlen; + int winlen = prev_block_len; start = &s->channel[c].out[s->samples_per_frame/2 + s->channel[c].subframe_offset[j] - prev_block_len /2 ]; - end = &s->channel[c].out[s->samples_per_frame/2 + s->channel[c].subframe_offset[j] + block_len /2 - 1]; if(block_len <= prev_block_len){ start += (prev_block_len - block_len)/2; - len = block_len/2; winlen = block_len; - }else{ - end -= (block_len - prev_block_len)/2; - len = prev_block_len /2; - winlen = prev_block_len; } -// float* rs = &s->channel[c].out[s->samples_per_frame/2 + s->channel[c].subframe_offset[i]]; -// printf("Dstart %i %i end %i win %i prev %i\n",s->frame_num+1, start - rs,end -rs,winlen,prev_block_len); - - // FIXME: untangle the windowing so DSP functions can be used window = s->windows[av_log2(winlen)-BLOCK_MIN_BITS]; - for(x=0;x<len;x++){ - float ts = *start; - float te = *end; - float sin_v = window[x]; - float cos_v = window[winlen - x -1]; - *start = cos_v * ts - sin_v * te; - *end = cos_v * te + sin_v * ts; - ++start; - --end; - } + + s->dsp.vector_fmul_window(start, start, start + winlen/2, window, 0, winlen/2); + s->channel[c].prev_block_len = block_len; } }

3 2

[soc]: r4013 - libavfilter/diffs/02_ffmpeg_filters.diff
by vitor 27 Jan '09

27 Jan '09

Author: vitor Date: Tue Jan 27 21:57:50 2009 New Revision: 4013 Log: Fix build when not using --enable-avfilter. Patch by Cédric Schieli - cschieli <> gmail com Modified: libavfilter/diffs/02_ffmpeg_filters.diff Modified: libavfilter/diffs/02_ffmpeg_filters.diff ============================================================================== --- libavfilter/diffs/02_ffmpeg_filters.diff Tue Jan 27 06:34:33 2009 (r4012) +++ libavfilter/diffs/02_ffmpeg_filters.diff Tue Jan 27 21:57:50 2009 (r4013) @@ -250,15 +250,17 @@ Index: ffmpeg.c if (ost->video_pad) { av_picture_pad((AVPicture*)final_picture, (AVPicture *)padding_src, -@@ -1190,6 +1383,7 @@ +@@ -1190,6 +1383,9 @@ static short *samples= NULL; AVSubtitle subtitle, *subtitle_to_free; int got_subtitle; ++#if CONFIG_AVFILTER + int loop; ++#endif if(ist->next_pts == AV_NOPTS_VALUE) ist->next_pts= ist->pts; -@@ -1307,6 +1501,15 @@ +@@ -1307,6 +1503,15 @@ &buffer_to_free); } @@ -274,13 +276,15 @@ Index: ffmpeg.c // preprocess audio (volume) if (ist->st->codec->codec_type == CODEC_TYPE_AUDIO) { if (audio_volume != 256) { -@@ -1328,10 +1531,16 @@ +@@ -1328,10 +1533,18 @@ if (pts > now) usleep(pts - now); } - ++#if CONFIG_AVFILTER + loop = ist->st->codec->codec_type != CODEC_TYPE_VIDEO || + !ist->out_video_filter || avfilter_poll_frame(ist->out_video_filter->inputs[0]); ++#endif /* if output time reached then transcode raw format, encode packets and output them */ if (start_time == 0 || ist->pts >= start_time) @@ -302,12 +306,14 @@ Index: ffmpeg.c do_video_out(os, ost, ist, &picture, &frame_size); if (vstats_filename && frame_size) do_video_stats(os, ost, frame_size); -@@ -1411,7 +1623,15 @@ +@@ -1411,7 +1627,17 @@ av_free_packet(&opkt); } } ++#if CONFIG_AVFILTER + loop = (ist->st->codec->codec_type == CODEC_TYPE_VIDEO) && + ist->out_video_filter && avfilter_poll_frame(ist->out_video_filter->inputs[0]); ++#endif } + +#if CONFIG_AVFILTER

1 0

[soc]: r1195 - in jpeg2000: j2k.c j2k.h j2kdec.c j2kenc.c
by Stefan Walter Klein 27 Jan '09

27 Jan '09

Hi! I just read about your efforts, including jpeg2000 in ffmpeg. Sadly I could neither find a downloadable version, nor sth. like 'comming soon'..:-( Would you tell me, how far things are?! Is the development going on? Can I help in any way!? All the best and keep on rokkin'! Stefan Klein -- Beste Grüße!

2 1

[soc]: r4012 - amr/amrnbfloatdec.c
by kmalaussene 27 Jan '09

27 Jan '09

Author: kmalaussene Date: Tue Jan 27 06:34:33 2009 New Revision: 4012 Log: Fix decoding of pitch lag frac to match reference implementation. Some of its value were widely out of bound. Modified: amr/amrnbfloatdec.c Modified: amr/amrnbfloatdec.c ============================================================================== --- amr/amrnbfloatdec.c Mon Jan 26 22:44:11 2009 (r4011) +++ amr/amrnbfloatdec.c Tue Jan 27 06:34:33 2009 (r4012) @@ -501,8 +501,9 @@ static void decode_pitch_lag_3(AMRContex p->pitch_lag_frac = 0; }else if(pitch_index < 12) { // 1/3 fractional precision for [t1_temp-1 2/3, t1_temp+2/3] - p->pitch_lag_int = ( ((pitch_index - 5)*10923)>>15 ) + t1_temp - 1; + p->pitch_lag_int = ( ((pitch_index - 5)*10923)>>15 ) - 1; p->pitch_lag_frac = pitch_index - p->pitch_lag_int*3 - 9; + p->pitch_lag_int += t1_temp; }else { // integer only precision for [t1_temp+1, t1_temp+4] p->pitch_lag_int = pitch_index + t1_temp - 11;

1 0

[soc]: r4011 - in libavfilter/diffs: 01_ffplay_filters.diff 02_ffmpeg_filters.diff
by stefano 26 Jan '09

26 Jan '09

Author: stefano Date: Mon Jan 26 22:44:11 2009 New Revision: 4011 Log: Add checks for the validity of the parsed graph, fix crashes. Modified: libavfilter/diffs/01_ffplay_filters.diff libavfilter/diffs/02_ffmpeg_filters.diff Modified: libavfilter/diffs/01_ffplay_filters.diff ============================================================================== --- libavfilter/diffs/01_ffplay_filters.diff Mon Jan 26 22:28:54 2009 (r4010) +++ libavfilter/diffs/01_ffplay_filters.diff Mon Jan 26 22:44:11 2009 (r4011) @@ -211,7 +211,7 @@ Index: ffplay.c /* update the bitmap content */ SDL_UnlockYUVOverlay(vp->bmp); -@@ -1335,54 +1406,264 @@ +@@ -1335,54 +1406,266 @@ return queue_picture(is, src_frame, pts); } @@ -435,6 +435,8 @@ Index: ffplay.c + } + avfilter_graph_add_filter(graph, filt_src); + avfilter_graph_add_filter(graph, filt_out); ++ ++ if(avfilter_graph_check_validity(graph, NULL)) goto the_end; + if(avfilter_graph_config_formats(graph)) goto the_end; + if(avfilter_config_links(filt_out)) goto the_end; + @@ -507,7 +509,7 @@ Index: ffplay.c av_free(frame); return 0; } -@@ -2168,6 +2449,12 @@ +@@ -2168,6 +2451,12 @@ /* free all pictures */ for(i=0;i<VIDEO_PICTURE_QUEUE_SIZE; i++) { vp = &is->pictq[i]; @@ -520,7 +522,7 @@ Index: ffplay.c if (vp->bmp) { SDL_FreeYUVOverlay(vp->bmp); vp->bmp = NULL; -@@ -2505,6 +2792,9 @@ +@@ -2505,6 +2794,9 @@ { "ec", OPT_INT | HAS_ARG | OPT_EXPERT, {(void*)&error_concealment}, "set error concealment options", "bit_mask" }, { "sync", HAS_ARG | OPT_FUNC2 | OPT_EXPERT, {(void*)opt_sync}, "set audio-video sync. type (type=audio/video/ext)", "type" }, { "threads", HAS_ARG | OPT_FUNC2 | OPT_EXPERT, {(void*)opt_thread_count}, "thread count", "count" }, @@ -530,7 +532,7 @@ Index: ffplay.c { "default", OPT_FUNC2 | HAS_ARG | OPT_AUDIO | OPT_VIDEO | OPT_EXPERT, {(void*)opt_default}, "generic catch all option", "" }, { NULL, }, }; -@@ -2553,7 +2843,9 @@ +@@ -2553,7 +2845,9 @@ avctx_opts[i]= avcodec_alloc_context2(i); } avformat_opts = av_alloc_format_context(); Modified: libavfilter/diffs/02_ffmpeg_filters.diff ============================================================================== --- libavfilter/diffs/02_ffmpeg_filters.diff Mon Jan 26 22:28:54 2009 (r4010) +++ libavfilter/diffs/02_ffmpeg_filters.diff Mon Jan 26 22:44:11 2009 (r4011) @@ -40,7 +40,7 @@ Index: ffmpeg.c } AVInputStream; typedef struct AVInputFile { -@@ -301,6 +318,174 @@ +@@ -301,6 +318,176 @@ static struct termios oldtty; #endif @@ -200,6 +200,8 @@ Index: ffmpeg.c + } + + /* configure all the filter links */ ++ if(avfilter_graph_check_validity(filt_graph_all, NULL)) ++ return -1; + if(avfilter_graph_config_formats(filt_graph_all)) + return -1; + if(avfilter_config_links(ist->out_video_filter)) @@ -215,7 +217,7 @@ Index: ffmpeg.c static void term_exit(void) { #if HAVE_TERMIOS_H -@@ -886,6 +1071,9 @@ +@@ -886,6 +1073,9 @@ if (nb_frames <= 0) return; @@ -225,7 +227,7 @@ Index: ffmpeg.c if (ost->video_crop) { if (av_picture_crop((AVPicture *)&picture_crop_temp, (AVPicture *)in_picture, dec->pix_fmt, ost->topBand, ost->leftBand) < 0) { av_log(NULL, AV_LOG_ERROR, "error cropping picture\n"); -@@ -897,6 +1085,7 @@ +@@ -897,6 +1087,7 @@ } else { formatted_picture = in_picture; } @@ -233,7 +235,7 @@ Index: ffmpeg.c final_picture = formatted_picture; padding_src = formatted_picture; -@@ -914,12 +1103,14 @@ +@@ -914,12 +1105,14 @@ } } @@ -248,7 +250,7 @@ Index: ffmpeg.c if (ost->video_pad) { av_picture_pad((AVPicture*)final_picture, (AVPicture *)padding_src, -@@ -1190,6 +1381,7 @@ +@@ -1190,6 +1383,7 @@ static short *samples= NULL; AVSubtitle subtitle, *subtitle_to_free; int got_subtitle; @@ -256,7 +258,7 @@ Index: ffmpeg.c if(ist->next_pts == AV_NOPTS_VALUE) ist->next_pts= ist->pts; -@@ -1307,6 +1499,15 @@ +@@ -1307,6 +1501,15 @@ &buffer_to_free); } @@ -272,7 +274,7 @@ Index: ffmpeg.c // preprocess audio (volume) if (ist->st->codec->codec_type == CODEC_TYPE_AUDIO) { if (audio_volume != 256) { -@@ -1328,10 +1529,16 @@ +@@ -1328,10 +1531,16 @@ if (pts > now) usleep(pts - now); } @@ -290,7 +292,7 @@ Index: ffmpeg.c for(i=0;i<nb_ostreams;i++) { int frame_size; -@@ -1354,6 +1561,9 @@ +@@ -1354,6 +1563,9 @@ do_audio_out(os, ost, ist, data_buf, data_size); break; case CODEC_TYPE_VIDEO: @@ -300,7 +302,7 @@ Index: ffmpeg.c do_video_out(os, ost, ist, &picture, &frame_size); if (vstats_filename && frame_size) do_video_stats(os, ost, frame_size); -@@ -1411,7 +1621,15 @@ +@@ -1411,7 +1623,15 @@ av_free_packet(&opkt); } } @@ -316,7 +318,7 @@ Index: ffmpeg.c av_free(buffer_to_free); /* XXX: allocate the subtitles in the codec ? */ if (subtitle_to_free) { -@@ -1842,10 +2060,21 @@ +@@ -1842,10 +2062,21 @@ fprintf(stderr, "Cannot get resampling context\n"); av_exit(1); } @@ -338,7 +340,7 @@ Index: ffmpeg.c break; case CODEC_TYPE_SUBTITLE: ost->encoding_needed = 1; -@@ -3835,6 +4064,9 @@ +@@ -3835,6 +4066,9 @@ #if CONFIG_VHOOK { "vhook", HAS_ARG | OPT_EXPERT | OPT_VIDEO, {(void*)add_frame_hooker}, "insert video processing module", "module" }, #endif

1 0

[soc]: r4010 - in libavfilter: allfilters.c checkout.sh diffs/01_ffplay_filters.diff diffs/02_ffmpeg_filters.diff
by stefano 26 Jan '09

26 Jan '09

Author: stefano Date: Mon Jan 26 22:28:54 2009 New Revision: 4010 Log: Update to FFmpeg r16809. Modified: libavfilter/allfilters.c libavfilter/checkout.sh libavfilter/diffs/01_ffplay_filters.diff libavfilter/diffs/02_ffmpeg_filters.diff Modified: libavfilter/allfilters.c ============================================================================== --- libavfilter/allfilters.c Mon Jan 26 20:20:40 2009 (r4009) +++ libavfilter/allfilters.c Mon Jan 26 22:28:54 2009 (r4010) @@ -24,7 +24,7 @@ #define REGISTER_FILTER(X,x,y) { \ extern AVFilter avfilter_##y##_##x ; \ - if(ENABLE_##X##_FILTER ) avfilter_register(&avfilter_##y##_##x ); } + if(CONFIG_##X##_FILTER ) avfilter_register(&avfilter_##y##_##x ); } void avfilter_register_all(void) { Modified: libavfilter/checkout.sh ============================================================================== --- libavfilter/checkout.sh Mon Jan 26 20:20:40 2009 (r4009) +++ libavfilter/checkout.sh Mon Jan 26 22:28:54 2009 (r4010) @@ -1,7 +1,7 @@ #! /bin/sh echo "checking out pristine ffmpeg" -svn checkout svn://svn.ffmpeg.org/ffmpeg/trunk/ ffmpeg -r16562 +svn checkout svn://svn.ffmpeg.org/ffmpeg/trunk/ ffmpeg -r16809 echo "patching ffmpeg" for diff in $(ls $(pwd)/diffs/*.diff); do patch -d ffmpeg -p0 -i $diff; done Modified: libavfilter/diffs/01_ffplay_filters.diff ============================================================================== --- libavfilter/diffs/01_ffplay_filters.diff Mon Jan 26 20:20:40 2009 (r4009) +++ libavfilter/diffs/01_ffplay_filters.diff Mon Jan 26 22:28:54 2009 (r4010) @@ -1,12 +1,12 @@ Index: ffplay.c =================================================================== ---- ffplay.c (revision 16562) +--- ffplay.c (revision 16809) +++ ffplay.c (working copy) @@ -29,6 +29,12 @@ #include "libavcodec/audioconvert.h" #include "libavcodec/opt.h" -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER +# include "libavfilter/avfilter.h" +# include "libavfilter/avfiltergraph.h" +# include "libavfilter/graphparser.h" @@ -19,7 +19,7 @@ Index: ffplay.c /* NOTE: the size must be big enough to compensate the hardware audio buffersize size */ #define SAMPLE_ARRAY_SIZE (2*65536) -+#if !ENABLE_AVFILTER ++#if !CONFIG_AVFILTER static int sws_flags = SWS_BICUBIC; +#endif @@ -32,7 +32,7 @@ Index: ffplay.c + enum PixelFormat pix_fmt; int allocated; + -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER + AVFilterPicRef *picref; +#endif } VideoPicture; @@ -43,7 +43,7 @@ Index: ffplay.c char filename[1024]; int width, height, xleft, ytop; + -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER + AVFilterContext *out_video_filter; ///<the last filter in the video chain +#endif } VideoState; @@ -53,7 +53,7 @@ Index: ffplay.c static int error_recognition = FF_ER_CAREFUL; static int error_concealment = 3; static int decoder_reorder_pts= 0; -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER +static char *vfilters = NULL; +#endif @@ -63,7 +63,7 @@ Index: ffplay.c vp = &is->pictq[is->pictq_rindex]; if (vp->bmp) { -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER + if (vp->picref->pixel_aspect.num == 0) + aspect_ratio = 0; + else @@ -89,7 +89,7 @@ Index: ffplay.c } else if(!is_full_screen && screen_width){ w = screen_width; h = screen_height; -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER + }else if (is->out_video_filter && is->out_video_filter->inputs[0]){ + w = is->out_video_filter->inputs[0]->w; + h = is->out_video_filter->inputs[0]->h; @@ -105,7 +105,7 @@ Index: ffplay.c if (vp->bmp) SDL_FreeYUVOverlay(vp->bmp); -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER + if (vp->picref) + avfilter_unref_pic(vp->picref); + vp->picref = NULL; @@ -140,7 +140,7 @@ Index: ffplay.c VideoPicture *vp; int dst_pix_fmt; AVPicture pict; -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER + AVPicture pict_src; +#else static struct SwsContext *img_convert_ctx; @@ -153,7 +153,7 @@ Index: ffplay.c /* alloc or resize hardware picture buffer */ if (!vp->bmp || -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER + vp->width != is->out_video_filter->inputs[0]->w || + vp->height != is->out_video_filter->inputs[0]->h) { +#else @@ -167,7 +167,7 @@ Index: ffplay.c /* if the frame is not skipped, then display it */ if (vp->bmp) { -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER + if(vp->picref) + avfilter_unref_pic(vp->picref); + vp->picref = src_frame->opaque; @@ -181,7 +181,7 @@ Index: ffplay.c pict.linesize[1] = vp->bmp->pitches[2]; pict.linesize[2] = vp->bmp->pitches[1]; + -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER + pict_src.data[0] = src_frame->data[0]; + pict_src.data[1] = src_frame->data[1]; + pict_src.data[2] = src_frame->data[2]; @@ -253,7 +253,7 @@ Index: ffplay.c + return 0; +} + -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER +typedef struct { + VideoState *is; + AVFrame *frame; @@ -388,7 +388,7 @@ Index: ffplay.c + { .name = NULL }}, + .outputs = (AVFilterPad[]) {{ .name = NULL }}, +}; -+#endif /* ENABLE_AVFILTER */ ++#endif /* CONFIG_AVFILTER */ + static int video_thread(void *arg) { @@ -401,7 +401,7 @@ Index: ffplay.c + int ret; + -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER + AVFilterContext *filt_src = NULL, *filt_out = NULL; + AVFilterGraph *graph = av_mallocz(sizeof(AVFilterGraph)); + @@ -443,7 +443,7 @@ Index: ffplay.c + for(;;) { - while (is->paused && !is->videoq.abort_request) { -+#if !ENABLE_AVFILTER ++#if !CONFIG_AVFILTER + AVPacket pkt; +#endif + while (is->paused && !is->videoq.abort_request) @@ -451,7 +451,7 @@ Index: ffplay.c - } - if (packet_queue_get(&is->videoq, pkt, 1) < 0) - break; -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER + ret = get_filtered_video_frame(filt_out, frame, &pts_int); +#else + ret = get_video_frame(is, frame, &pts_int, &pkt); @@ -474,7 +474,7 @@ Index: ffplay.c + pts = pts_int; + pts /= AV_TIME_BASE; + ret = output_picture2(is, frame, pts); -+#if !ENABLE_AVFILTER ++#if !CONFIG_AVFILTER + av_free_packet(&pkt); +#endif + if (ret < 0) @@ -501,17 +501,17 @@ Index: ffplay.c stream_pause(cur_stream); } the_end: -+ #if ENABLE_AVFILTER ++ #if CONFIG_AVFILTER + avfilter_destroy_graph(graph); + #endif av_free(frame); return 0; } -@@ -2170,6 +2451,12 @@ +@@ -2168,6 +2449,12 @@ /* free all pictures */ for(i=0;i<VIDEO_PICTURE_QUEUE_SIZE; i++) { vp = &is->pictq[i]; -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER + if (vp->picref) { + avfilter_unref_pic(vp->picref); + vp->picref = NULL; @@ -520,21 +520,21 @@ Index: ffplay.c if (vp->bmp) { SDL_FreeYUVOverlay(vp->bmp); vp->bmp = NULL; -@@ -2507,6 +2794,9 @@ +@@ -2505,6 +2792,9 @@ { "ec", OPT_INT | HAS_ARG | OPT_EXPERT, {(void*)&error_concealment}, "set error concealment options", "bit_mask" }, { "sync", HAS_ARG | OPT_FUNC2 | OPT_EXPERT, {(void*)opt_sync}, "set audio-video sync. type (type=audio/video/ext)", "type" }, { "threads", HAS_ARG | OPT_FUNC2 | OPT_EXPERT, {(void*)opt_thread_count}, "thread count", "count" }, -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER + { "vfilters", OPT_STRING | HAS_ARG, {(void*)&vfilters}, "video filters", "filter list" }, +#endif { "default", OPT_FUNC2 | HAS_ARG | OPT_AUDIO | OPT_VIDEO | OPT_EXPERT, {(void*)opt_default}, "generic catch all option", "" }, { NULL, }, }; -@@ -2555,7 +2845,9 @@ +@@ -2553,7 +2843,9 @@ avctx_opts[i]= avcodec_alloc_context2(i); } avformat_opts = av_alloc_format_context(); -+#if !ENABLE_AVFILTER ++#if !CONFIG_AVFILTER sws_opts = sws_getContext(16,16,0, 16,16,0, sws_flags, NULL,NULL,NULL); +#endif Modified: libavfilter/diffs/02_ffmpeg_filters.diff ============================================================================== --- libavfilter/diffs/02_ffmpeg_filters.diff Mon Jan 26 20:20:40 2009 (r4009) +++ libavfilter/diffs/02_ffmpeg_filters.diff Mon Jan 26 22:28:54 2009 (r4010) @@ -1,26 +1,26 @@ Index: ffmpeg.c =================================================================== ---- ffmpeg.c (revision 16562) +--- ffmpeg.c (revision 16809) +++ ffmpeg.c (working copy) @@ -41,6 +41,13 @@ #include "libavutil/avstring.h" #include "libavformat/os_support.h" -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER +# include "libavfilter/avfilter.h" +# include "libavfilter/avfiltergraph.h" +# include "libavfilter/graphparser.h" +# include "libavfilter/vsrc_buffer.h" +#endif + - #ifdef HAVE_SYS_RESOURCE_H + #if HAVE_SYS_RESOURCE_H #include <sys/types.h> #include <sys/resource.h> @@ -148,6 +155,9 @@ static int loop_input = 0; static int loop_output = AVFMT_NOOUTPUTLOOP; static int qp_hist = 0; -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER +static char *vfilters = NULL; +#endif @@ -30,7 +30,7 @@ Index: ffmpeg.c is not defined */ int64_t pts; /* current pts */ int is_start; /* is 1 at the start and after a discontinuity */ -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER + AVFilterContext *out_video_filter; + AVFilterContext *input_video_filter; + AVFrame *filter_frame; @@ -44,7 +44,7 @@ Index: ffmpeg.c static struct termios oldtty; #endif -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER +typedef struct { + int pix_fmt; +} FilterOutPriv; @@ -210,16 +210,16 @@ Index: ffmpeg.c + + return 0; +} -+#endif /* ENABLE_AVFILTER */ ++#endif /* CONFIG_AVFILTER */ + static void term_exit(void) { - #ifdef HAVE_TERMIOS_H + #if HAVE_TERMIOS_H @@ -886,6 +1071,9 @@ if (nb_frames <= 0) return; -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER + formatted_picture = in_picture; +#else if (ost->video_crop) { @@ -237,7 +237,7 @@ Index: ffmpeg.c } } -+#if !ENABLE_AVFILTER ++#if !CONFIG_AVFILTER if (ost->video_resample) { padding_src = NULL; final_picture = &ost->pict_tmp; @@ -260,7 +260,7 @@ Index: ffmpeg.c &buffer_to_free); } -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER + if (ist->st->codec->codec_type == CODEC_TYPE_VIDEO && ist->input_video_filter) { + // add it to be filtered + av_vsrc_buffer_add_frame(ist->input_video_filter, &picture, @@ -282,7 +282,7 @@ Index: ffmpeg.c /* if output time reached then transcode raw format, encode packets and output them */ if (start_time == 0 || ist->pts >= start_time) -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER + while(loop) { + if (ist->st->codec->codec_type == CODEC_TYPE_VIDEO && ist->out_video_filter) + get_filtered_video_pic(ist->out_video_filter, &ist->picref, &picture, &ist->pts); @@ -294,7 +294,7 @@ Index: ffmpeg.c do_audio_out(os, ost, ist, data_buf, data_size); break; case CODEC_TYPE_VIDEO: -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER + ost->st->codec->sample_aspect_ratio = ist->picref->pixel_aspect; +#endif do_video_out(os, ost, ist, &picture, &frame_size); @@ -308,7 +308,7 @@ Index: ffmpeg.c + ist->out_video_filter && avfilter_poll_frame(ist->out_video_filter->inputs[0]); } + -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER + if(ist->picref) + avfilter_unref_pic(ist->picref); + } @@ -316,11 +316,11 @@ Index: ffmpeg.c av_free(buffer_to_free); /* XXX: allocate the subtitles in the codec ? */ if (subtitle_to_free) { -@@ -1840,10 +2058,21 @@ +@@ -1842,10 +2060,21 @@ fprintf(stderr, "Cannot get resampling context\n"); av_exit(1); } -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER + ost->resample_height = icodec->height; +#else ost->resample_height = icodec->height - (frame_topBand + frame_bottomBand); @@ -329,7 +329,7 @@ Index: ffmpeg.c ost->encoding_needed = 1; ist->decoding_needed = 1; + -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER + if (configure_filters(ist, ost)) { + fprintf(stderr, "Error opening filters!\n"); + exit(1); @@ -338,11 +338,11 @@ Index: ffmpeg.c break; case CODEC_TYPE_SUBTITLE: ost->encoding_needed = 1; -@@ -3831,6 +4060,9 @@ - #ifdef CONFIG_VHOOK +@@ -3835,6 +4064,9 @@ + #if CONFIG_VHOOK { "vhook", HAS_ARG | OPT_EXPERT | OPT_VIDEO, {(void*)add_frame_hooker}, "insert video processing module", "module" }, #endif -+#if ENABLE_AVFILTER ++#if CONFIG_AVFILTER + { "vfilters", OPT_STRING | HAS_ARG, {(void*)&vfilters}, "video filters", "filter list" }, +#endif { "intra_matrix", HAS_ARG | OPT_EXPERT | OPT_VIDEO, {(void*)opt_intra_matrix}, "specify intra matrix coeffs", "matrix" },

1 0

[soc]: r4009 - libavfilter/diffs/02_ffmpeg_filters.diff
by vitor 26 Jan '09

26 Jan '09

Author: vitor Date: Mon Jan 26 20:20:40 2009 New Revision: 4009 Log: Fix "-vcodec copy" when using a libavfilter patched ffmpeg. Patch by Cédric Schieli cschieli ! gmail ! com Modified: libavfilter/diffs/02_ffmpeg_filters.diff Modified: libavfilter/diffs/02_ffmpeg_filters.diff ============================================================================== --- libavfilter/diffs/02_ffmpeg_filters.diff Mon Jan 26 15:07:28 2009 (r4008) +++ libavfilter/diffs/02_ffmpeg_filters.diff Mon Jan 26 20:20:40 2009 (r4009) @@ -261,7 +261,7 @@ Index: ffmpeg.c } +#if ENABLE_AVFILTER -+ if (ist->st->codec->codec_type == CODEC_TYPE_VIDEO) { ++ if (ist->st->codec->codec_type == CODEC_TYPE_VIDEO && ist->input_video_filter) { + // add it to be filtered + av_vsrc_buffer_add_frame(ist->input_video_filter, &picture, + ist->pts, @@ -278,13 +278,13 @@ Index: ffmpeg.c } - + loop = ist->st->codec->codec_type != CODEC_TYPE_VIDEO || -+ avfilter_poll_frame(ist->out_video_filter->inputs[0]); ++ !ist->out_video_filter || avfilter_poll_frame(ist->out_video_filter->inputs[0]); /* if output time reached then transcode raw format, encode packets and output them */ if (start_time == 0 || ist->pts >= start_time) +#if ENABLE_AVFILTER + while(loop) { -+ if (ist->st->codec->codec_type == CODEC_TYPE_VIDEO) ++ if (ist->st->codec->codec_type == CODEC_TYPE_VIDEO && ist->out_video_filter) + get_filtered_video_pic(ist->out_video_filter, &ist->picref, &picture, &ist->pts); +#endif for(i=0;i<nb_ostreams;i++) { @@ -305,7 +305,7 @@ Index: ffmpeg.c } } + loop = (ist->st->codec->codec_type == CODEC_TYPE_VIDEO) && -+ avfilter_poll_frame(ist->out_video_filter->inputs[0]); ++ ist->out_video_filter && avfilter_poll_frame(ist->out_video_filter->inputs[0]); } + +#if ENABLE_AVFILTER

1 0

[soc]: r4008 - in aacenc: aac_enc.patch aaccoder.c aacenc.c aacenc.h
by kostya 26 Jan '09

26 Jan '09

Author: kostya Date: Mon Jan 26 15:07:28 2009 New Revision: 4008 Log: Split out from AAC encoder core file all functions with heuristics for selecting scalefactors, band codebooks, etc. Added: aacenc/aaccoder.c - copied, changed from r4006, aacenc/aacenc.c aacenc/aacenc.h - copied, changed from r4006, aacenc/aacenc.c Modified: aacenc/aac_enc.patch aacenc/aacenc.c Modified: aacenc/aac_enc.patch ============================================================================== --- aacenc/aac_enc.patch Mon Jan 26 15:00:57 2009 (r4007) +++ aacenc/aac_enc.patch Mon Jan 26 15:07:28 2009 (r4008) @@ -2,14 +2,14 @@ diff --git a/libavcodec/Makefile b/libav index d4f6d1c..0ed9057 100644 --- a/libavcodec/Makefile +++ b/libavcodec/Makefile -@@ -25,6 +25,7 @@ - - OBJS-$(CONFIG_ENCODERS) += faandct.o jfdctfst.o jfdctint.o +@@ -33,6 +33,7 @@ -+OBJS-$(CONFIG_AAC_ENCODER) += aacenc.o aacpsy.o aactab.o psymodel.o iirfilter.o mdct.o fft.o mpeg4audio.o - OBJS-$(CONFIG_AAC_DECODER) += aac.o aactab.o mdct.o fft.o - OBJS-$(CONFIG_AASC_DECODER) += aasc.o - OBJS-$(CONFIG_AC3_DECODER) += ac3dec.o ac3tab.o ac3dec_data.o ac3.o mdct.o fft.o + # decoders/encoders + OBJS-$(CONFIG_AAC_DECODER) += aac.o aactab.o ++OBJS-$(CONFIG_AAC_ENCODER) += aacenc.o aaccoder.o aacpsy.o aactab.o psymodel.o iirfilter.o mdct.o fft.o mpeg4audio.o + OBJS-$(CONFIG_AASC_DECODER) += aasc.o msrledec.o + OBJS-$(CONFIG_AC3_DECODER) += eac3dec.o ac3dec.o ac3tab.o ac3dec_data.o ac3.o + OBJS-$(CONFIG_AC3_ENCODER) += ac3enc.o ac3tab.o ac3.o diff --git a/libavcodec/allcodecs.c b/libavcodec/allcodecs.c index 33a4242..6871496 100644 --- a/libavcodec/allcodecs.c Copied and modified: aacenc/aaccoder.c (from r4006, aacenc/aacenc.c) ============================================================================== --- aacenc/aacenc.c Mon Jan 26 14:37:36 2009 (r4006, copy source) +++ aacenc/aaccoder.c Mon Jan 26 15:07:28 2009 (r4008) @@ -1,6 +1,6 @@ /* - * AAC encoder - * Copyright (C) 2008 Konstantin Shishkov + * AAC coefficients encoder + * Copyright (C) 2008-2009 Konstantin Shishkov * * This file is part of FFmpeg. * @@ -20,106 +20,16 @@ */ /** - * @file aacenc.c - * AAC encoder + * @file aaccoder.c + * AAC coefficients encoder */ -/*********************************** - * TODOs: - * speedup quantizer selection - * add sane pulse detection - * add temporal noise shaping - ***********************************/ - #include "avcodec.h" #include "bitstream.h" -#include "dsputil.h" -#include "mpeg4audio.h" - #include "aac.h" +#include "aacenc.h" #include "aactab.h" -#include "psymodel.h" - -static const uint8_t swb_size_1024_96[] = { - 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, - 12, 12, 12, 12, 12, 16, 16, 24, 28, 36, 44, - 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64 -}; - -static const uint8_t swb_size_1024_64[] = { - 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, - 12, 12, 12, 16, 16, 16, 20, 24, 24, 28, 36, - 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40 -}; - -static const uint8_t swb_size_1024_48[] = { - 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8, - 12, 12, 12, 12, 16, 16, 20, 20, 24, 24, 28, 28, - 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, - 96 -}; - -static const uint8_t swb_size_1024_32[] = { - 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8, - 12, 12, 12, 12, 16, 16, 20, 20, 24, 24, 28, 28, - 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32 -}; - -static const uint8_t swb_size_1024_24[] = { - 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, - 12, 12, 12, 12, 16, 16, 16, 20, 20, 24, 24, 28, 28, - 32, 36, 36, 40, 44, 48, 52, 52, 64, 64, 64, 64, 64 -}; - -static const uint8_t swb_size_1024_16[] = { - 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, - 12, 12, 12, 12, 12, 12, 12, 12, 12, 16, 16, 16, 16, 20, 20, 20, 24, 24, 28, 28, - 32, 36, 40, 40, 44, 48, 52, 56, 60, 64, 64, 64 -}; - -static const uint8_t swb_size_1024_8[] = { - 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, - 16, 16, 16, 16, 16, 16, 16, 20, 20, 20, 20, 24, 24, 24, 28, 28, - 32, 36, 36, 40, 44, 48, 52, 56, 60, 64, 80 -}; - -static const uint8_t *swb_size_1024[] = { - swb_size_1024_96, swb_size_1024_96, swb_size_1024_64, - swb_size_1024_48, swb_size_1024_48, swb_size_1024_32, - swb_size_1024_24, swb_size_1024_24, swb_size_1024_16, - swb_size_1024_16, swb_size_1024_16, swb_size_1024_8 -}; - -static const uint8_t swb_size_128_96[] = { - 4, 4, 4, 4, 4, 4, 8, 8, 8, 16, 28, 36 -}; - -static const uint8_t swb_size_128_48[] = { - 4, 4, 4, 4, 4, 8, 8, 8, 12, 12, 12, 16, 16, 16 -}; - -static const uint8_t swb_size_128_24[] = { - 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 12, 12, 16, 16, 20 -}; - -static const uint8_t swb_size_128_16[] = { - 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 12, 12, 16, 20, 20 -}; - -static const uint8_t swb_size_128_8[] = { - 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 12, 16, 20, 20 -}; - -static const uint8_t *swb_size_128[] = { - /* the last entry on the following row is swb_size_128_64 but is a - duplicate of swb_size_128_96 */ - swb_size_128_96, swb_size_128_96, swb_size_128_96, - swb_size_128_48, swb_size_128_48, swb_size_128_48, - swb_size_128_24, swb_size_128_24, swb_size_128_16, - swb_size_128_16, swb_size_128_16, swb_size_128_8 -}; - /** bits needed to code codebook run value for long windows */ static const uint8_t run_value_bits_long[64] = { 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, @@ -137,199 +47,6 @@ static const uint8_t* run_value_bits[2] run_value_bits_long, run_value_bits_short }; -/** default channel configurations */ -static const uint8_t aac_chan_configs[6][5] = { - {1, TYPE_SCE}, // 1 channel - single channel element - {1, TYPE_CPE}, // 2 channels - channel pair - {2, TYPE_SCE, TYPE_CPE}, // 3 channels - center + stereo - {3, TYPE_SCE, TYPE_CPE, TYPE_SCE}, // 4 channels - front center + stereo + back center - {3, TYPE_SCE, TYPE_CPE, TYPE_CPE}, // 5 channels - front center + stereo + back stereo - {4, TYPE_SCE, TYPE_CPE, TYPE_CPE, TYPE_LFE}, // 6 channels - front center + stereo + back stereo + LFE -}; - -/** - * AAC encoder context - */ -typedef struct { - PutBitContext pb; - MDCTContext mdct1024; ///< long (1024 samples) frame transform context - MDCTContext mdct128; ///< short (128 samples) frame transform context - DSPContext dsp; - DECLARE_ALIGNED_16(FFTSample, output[2048]); ///< temporary buffer for MDCT input coefficients - int16_t* samples; ///< saved preprocessed input - - int samplerate_index; ///< MPEG-4 samplerate index - - ChannelElement *cpe; ///< channel elements - FFPsyContext psy; - struct FFPsyPreprocessContext* psypp; - int cur_channel; - int last_frame; - float lambda; -} AACEncContext; - -/** - * Make AAC audio config object. - * @see 1.6.2.1 "Syntax - AudioSpecificConfig" - */ -static void put_audio_specific_config(AVCodecContext *avctx) -{ - PutBitContext pb; - AACEncContext *s = avctx->priv_data; - - init_put_bits(&pb, avctx->extradata, avctx->extradata_size*8); - put_bits(&pb, 5, 2); //object type - AAC-LC - put_bits(&pb, 4, s->samplerate_index); //sample rate index - put_bits(&pb, 4, avctx->channels); - //GASpecificConfig - put_bits(&pb, 1, 0); //frame length - 1024 samples - put_bits(&pb, 1, 0); //does not depend on core coder - put_bits(&pb, 1, 0); //is not extension - flush_put_bits(&pb); -} - -static av_cold int aac_encode_init(AVCodecContext *avctx) -{ - AACEncContext *s = avctx->priv_data; - int i; - const uint8_t *sizes[2]; - int lengths[2]; - - avctx->frame_size = 1024; - - for(i = 0; i < 16; i++) - if(avctx->sample_rate == ff_mpeg4audio_sample_rates[i]) - break; - if(i == 16){ - av_log(avctx, AV_LOG_ERROR, "Unsupported sample rate %d\n", avctx->sample_rate); - return -1; - } - if(avctx->channels > 6){ - av_log(avctx, AV_LOG_ERROR, "Unsupported number of channels: %d\n", avctx->channels); - return -1; - } - s->samplerate_index = i; - - dsputil_init(&s->dsp, avctx); - ff_mdct_init(&s->mdct1024, 11, 0); - ff_mdct_init(&s->mdct128, 8, 0); - // window init - ff_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024); - ff_kbd_window_init(ff_aac_kbd_short_128, 6.0, 128); - ff_sine_window_init(ff_sine_1024, 1024); - ff_sine_window_init(ff_sine_128, 128); - - s->samples = av_malloc(2 * 1024 * avctx->channels * sizeof(s->samples[0])); - s->cpe = av_mallocz(sizeof(ChannelElement) * aac_chan_configs[avctx->channels-1][0]); - avctx->extradata = av_malloc(2); - avctx->extradata_size = 2; - put_audio_specific_config(avctx); - - sizes[0] = swb_size_1024[i]; - sizes[1] = swb_size_128[i]; - lengths[0] = ff_aac_num_swb_1024[i]; - lengths[1] = ff_aac_num_swb_128[i]; - ff_psy_init(&s->psy, avctx, 2, sizes, lengths); - s->psypp = ff_psy_preprocess_init(avctx); - -#if !CONFIG_HARDCODED_TABLES - for (i = 0; i < 316; i++) - ff_aac_pow2sf_tab[i] = pow(2, (i - 200)/4.); -#endif /* CONFIG_HARDCODED_TABLES */ - - return 0; -} - -static void apply_window_and_mdct(AVCodecContext *avctx, AACEncContext *s, - SingleChannelElement *sce, short *audio, int channel) -{ - int i, j, k; - const float * lwindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024; - const float * swindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128; - const float * pwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128; - - if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE) { - memcpy(s->output, sce->saved, sizeof(float)*1024); - if(sce->ics.window_sequence[0] == LONG_STOP_SEQUENCE){ - memset(s->output, 0, sizeof(s->output[0]) * 448); - for(i = 448; i < 576; i++) - s->output[i] = sce->saved[i] * pwindow[i - 448]; - for(i = 576; i < 704; i++) - s->output[i] = sce->saved[i]; - } - if(sce->ics.window_sequence[0] != LONG_START_SEQUENCE){ - j = channel; - for (i = 0; i < 1024; i++, j += avctx->channels){ - s->output[i+1024] = audio[j] * lwindow[1024 - i - 1]; - sce->saved[i] = audio[j] * lwindow[i]; - } - }else{ - j = channel; - for(i = 0; i < 448; i++, j += avctx->channels) - s->output[i+1024] = audio[j]; - for(i = 448; i < 576; i++, j += avctx->channels) - s->output[i+1024] = audio[j] * swindow[576 - i - 1]; - memset(s->output+1024+576, 0, sizeof(s->output[0]) * 448); - j = channel; - for(i = 0; i < 1024; i++, j += avctx->channels) - sce->saved[i] = audio[j]; - } - ff_mdct_calc(&s->mdct1024, sce->coeffs, s->output); - }else{ - j = channel; - for (k = 0; k < 1024; k += 128) { - for(i = 448 + k; i < 448 + k + 256; i++) - s->output[i - 448 - k] = (i < 1024) - ? sce->saved[i] - : audio[channel + (i-1024)*avctx->channels]; - s->dsp.vector_fmul (s->output, k ? swindow : pwindow, 128); - s->dsp.vector_fmul_reverse(s->output+128, s->output+128, swindow, 128); - ff_mdct_calc(&s->mdct128, sce->coeffs + k, s->output); - } - j = channel; - for(i = 0; i < 1024; i++, j += avctx->channels) - sce->saved[i] = audio[j]; - } -} - -/** - * Encode ics_info element. - * @see Table 4.6 (syntax of ics_info) - */ -static void put_ics_info(AACEncContext *s, IndividualChannelStream *info) -{ - int w; - - put_bits(&s->pb, 1, 0); // ics_reserved bit - put_bits(&s->pb, 2, info->window_sequence[0]); - put_bits(&s->pb, 1, info->use_kb_window[0]); - if(info->window_sequence[0] != EIGHT_SHORT_SEQUENCE){ - put_bits(&s->pb, 6, info->max_sfb); - put_bits(&s->pb, 1, 0); // no prediction - }else{ - put_bits(&s->pb, 4, info->max_sfb); - for(w = 1; w < 8; w++){ - put_bits(&s->pb, 1, !info->group_len[w]); - } - } -} - -/** - * Encode MS data. - * @see 4.6.8.1 "Joint Coding - M/S Stereo" - */ -static void encode_ms_info(PutBitContext *pb, ChannelElement *cpe) -{ - int i, w; - - put_bits(pb, 2, cpe->ms_mode); - if(cpe->ms_mode == 1){ - for(w = 0; w < cpe->ch[0].ics.num_windows; w += cpe->ch[0].ics.group_len[w]){ - for(i = 0; i < cpe->ch[0].ics.max_sfb; i++) - put_bits(pb, 1, cpe->ms_mask[w*16 + i]); - } - } -} /** * Quantize one coefficient. @@ -381,13 +98,13 @@ static float quantize_band_cost(const fl float mincost; int minidx = 0; int minbits = 0; - int quants[4][2], realquants[4][2]; + int quants[4][2]; mincost = 0.0f; for(j = 0; j < dim; j++){ - realquants[j][0] = quant2(in[i+j], Q); - realquants[j][1] = quant (in[i+j], Q); + quants[j][0] = quant2(in[i+j], Q); + quants[j][1] = quant (in[i+j], Q); for(k = 0; k < 2; k++){ - quants[j][k] = FFMIN(realquants[j][k], maxval); + quants[j][k] = FFMIN(quants[j][k], maxval); if(!IS_CODEBOOK_UNSIGNED(cb) && in[i+j] < 0.0f) quants[j][k] = -quants[j][k]; } @@ -428,7 +145,7 @@ static float quantize_band_cost(const fl di = t - 165140.0f; curbits += 21; }else{ - int c = realquants[k]; + int c = quant(t, Q); di = t - c*cbrt(c)*IQ; curbits += av_log2(c)*2 - 4 + 1; } @@ -664,7 +381,7 @@ static void quantize_and_encode_band(Put const float IQ = ff_aac_pow2sf_tab[200 - scale_idx + SCALE_ONE_POS - SCALE_DIV_512]; int i, j, k; const int dim = cb < FIRST_PAIR_BT ? 4 : 2; - +//START_TIMER if(!cb) return; @@ -731,6 +448,7 @@ static void quantize_and_encode_band(Put } } } +//STOP_TIMER("quantize_and_encode") } #endif @@ -844,65 +562,6 @@ static void encode_window_bands_info(AAC } } -/** - * Produce integer coefficients from scalefactors provided by the model. - */ -static void adjust_frame_information(AACEncContext *apc, ChannelElement *cpe, int chans) -{ - int i, w, w2, g, ch; - int start, sum, maxsfb, cmaxsfb; - - for(ch = 0; ch < chans; ch++){ - IndividualChannelStream *ics = &cpe->ch[ch].ics; - start = 0; - maxsfb = 0; - cpe->ch[ch].pulse.num_pulse = 0; - for(w = 0; w < ics->num_windows*16; w += 16){ - for(g = 0; g < ics->num_swb; g++){ - sum = 0; - //apply M/S - if(!ch && cpe->ms_mask[w + g]){ - for(i = 0; i < ics->swb_sizes[g]; i++){ - cpe->ch[0].coeffs[start+i] = (cpe->ch[0].coeffs[start+i] + cpe->ch[1].coeffs[start+i]) / 2.0; - cpe->ch[1].coeffs[start+i] = cpe->ch[0].coeffs[start+i] - cpe->ch[1].coeffs[start+i]; - } - } - start += ics->swb_sizes[g]; - } - for(cmaxsfb = ics->num_swb; cmaxsfb > 0 && cpe->ch[ch].zeroes[w+cmaxsfb-1]; cmaxsfb--); - maxsfb = FFMAX(maxsfb, cmaxsfb); - } - ics->max_sfb = maxsfb; - - //adjust zero bands for window groups - for(w = 0; w < ics->num_windows; w += ics->group_len[w]){ - for(g = 0; g < ics->max_sfb; g++){ - i = 1; - for(w2 = w; w2 < w + ics->group_len[w]; w2++){ - if(!cpe->ch[ch].zeroes[w2*16 + g]){ - i = 0; - break; - } - } - cpe->ch[ch].zeroes[w*16 + g] = i; - } - } - } - - if(chans > 1 && cpe->common_window){ - IndividualChannelStream *ics0 = &cpe->ch[0].ics; - IndividualChannelStream *ics1 = &cpe->ch[1].ics; - int msc = 0; - ics0->max_sfb = FFMAX(ics0->max_sfb, ics1->max_sfb); - ics1->max_sfb = ics0->max_sfb; - for(w = 0; w < ics0->num_windows*16; w += 16) - for(i = 0; i < ics0->max_sfb; i++) - if(cpe->ms_mask[w+i]) msc++; - if(msc == 0 || ics0->max_sfb == 0) cpe->ms_mode = 0; - else cpe->ms_mode = msc < ics0->max_sfb ? 1 : 2; - } -} - typedef struct TrellisPath { float cost; int prev; @@ -910,7 +569,8 @@ typedef struct TrellisPath { int max_val; } TrellisPath; -static void search_for_quantizers_anmr(AACEncContext *s, SingleChannelElement *sce, const float lambda) +static void search_for_quantizers_anmr(AVCodecContext *avctx, AACEncContext *s, + SingleChannelElement *sce, const float lambda) { int q, w, w2, g, start = 0; int i; @@ -1186,7 +846,8 @@ static void search_for_quantizers_twoloo }while(fflag && its < 10); } -static void search_for_quantizers_faac(AACEncContext *s, SingleChannelElement *sce, const float lambda) +static void search_for_quantizers_faac(AVCodecContext *avctx, AACEncContext *s, + SingleChannelElement *sce, const float lambda) { int start = 0, i, w, w2, g; float uplim[128], maxq[128]; @@ -1351,7 +1012,8 @@ static void search_for_quantizers_faac(A } } -static void search_for_quantizers_fast(AACEncContext *s, SingleChannelElement *sce) +static void search_for_quantizers_fast(AVCodecContext *avctx, AACEncContext *s, + SingleChannelElement *sce, const float lambda) { int start = 0, i, w, w2, g; int minq = 255; @@ -1374,7 +1036,7 @@ static void search_for_quantizers_fast(A } } for(i = 0; i < 128; i++){ - sce->sf_idx[i] = av_clip(sce->sf_idx[i], minq, minq + SCALE_MAX_DIFF - 1); + sce->sf_idx[i] = 140;//av_clip(sce->sf_idx[i], minq, minq + SCALE_MAX_DIFF - 1); } //set the same quantizers inside window groups for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) @@ -1383,15 +1045,6 @@ static void search_for_quantizers_fast(A sce->sf_idx[(w+w2)*16+g] = sce->sf_idx[w*16+g]; } -static void search_for_quantizers(AVCodecContext *avctx, AACEncContext *s, - SingleChannelElement *sce) -{ - search_for_quantizers_anmr(s, sce, s->lambda); -// search_for_quantizers_twoloop(avctx, s, sce, s->lambda); -// search_for_quantizers_faac(s, sce, av_clipf(s->lambda * 2e8, 50.0f, 300.0f)); -// search_for_quantizers_fast(s, sce); -} - static void search_for_ms(AACEncContext *s, ChannelElement *cpe, const float lambda) { int start = 0, i, w, w2, g; @@ -1443,246 +1096,29 @@ static void search_for_ms(AACEncContext } } -/** - * Encode scalefactor band coding type. - */ -static void encode_band_info(AACEncContext *s, SingleChannelElement *sce) -{ - int w; - - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]){ - encode_window_bands_info(s, sce, w, sce->ics.group_len[w]); - } -} - -/** - * Encode scalefactors. - */ -static void encode_scale_factors(AVCodecContext *avctx, AACEncContext *s, SingleChannelElement *sce) -{ - int off = sce->sf_idx[0], diff; - int i, w; - - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]){ - for(i = 0; i < sce->ics.max_sfb; i++){ - if(!sce->zeroes[w*16 + i]){ - diff = sce->sf_idx[w*16 + i] - off + SCALE_DIFF_ZERO; - if(diff < 0 || diff > 120) av_log(avctx, AV_LOG_ERROR, "Scalefactor difference is too big to be coded\n"); - off = sce->sf_idx[w*16 + i]; - put_bits(&s->pb, ff_aac_scalefactor_bits[diff], ff_aac_scalefactor_code[diff]); - } - } - } -} - -/** - * Encode pulse data. - */ -static void encode_pulses(AACEncContext *s, Pulse *pulse) -{ - int i; - - put_bits(&s->pb, 1, !!pulse->num_pulse); - if(!pulse->num_pulse) return; - - put_bits(&s->pb, 2, pulse->num_pulse - 1); - put_bits(&s->pb, 6, pulse->start); - for(i = 0; i < pulse->num_pulse; i++){ - put_bits(&s->pb, 5, pulse->pos[i]); - put_bits(&s->pb, 4, pulse->amp[i]); - } -} - -/** - * Encode spectral coefficients processed by psychoacoustic model. - */ -static void encode_spectral_coeffs(AACEncContext *s, SingleChannelElement *sce) -{ - int start, i, w, w2; - - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]){ - start = 0; - for(i = 0; i < sce->ics.max_sfb; i++){ - if(sce->zeroes[w*16 + i]){ - start += sce->ics.swb_sizes[i]; - continue; - } - for(w2 = w; w2 < w + sce->ics.group_len[w]; w2++){ - quantize_and_encode_band(&s->pb, sce->coeffs + start + w2*128, - sce->ics.swb_sizes[i], - sce->sf_idx[w*16 + i], - sce->band_type[w*16 + i], - s->lambda); - } - start += sce->ics.swb_sizes[i]; - } - } -} - -/** - * Encode one channel of audio data. - */ -static int encode_individual_channel(AVCodecContext *avctx, AACEncContext *s, SingleChannelElement *sce, int common_window) -{ - put_bits(&s->pb, 8, sce->sf_idx[0]); - if(!common_window) put_ics_info(s, &sce->ics); - encode_band_info(s, sce); - encode_scale_factors(avctx, s, sce); - encode_pulses(s, &sce->pulse); - put_bits(&s->pb, 1, 0); //tns - put_bits(&s->pb, 1, 0); //ssr - encode_spectral_coeffs(s, sce); - return 0; -} - -/** - * Write some auxiliary information about the created AAC file. - */ -static void put_bitstream_info(AVCodecContext *avctx, AACEncContext *s, const char *name) -{ - int i, namelen, padbits; - - namelen = strlen(name) + 2; - put_bits(&s->pb, 3, TYPE_FIL); - put_bits(&s->pb, 4, FFMIN(namelen, 15)); - if(namelen >= 15) - put_bits(&s->pb, 8, namelen - 16); - put_bits(&s->pb, 4, 0); //extension type - filler - padbits = 8 - (put_bits_count(&s->pb) & 7); - align_put_bits(&s->pb); - for(i = 0; i < namelen - 2; i++) - put_bits(&s->pb, 8, name[i]); - put_bits(&s->pb, 12 - padbits, 0); -} - -static int aac_encode_frame(AVCodecContext *avctx, - uint8_t *frame, int buf_size, void *data) -{ - AACEncContext *s = avctx->priv_data; - int16_t *samples = s->samples, *samples2, *la; - ChannelElement *cpe; - int i, j, chans, tag, start_ch; - const uint8_t *chan_map = aac_chan_configs[avctx->channels-1]; - int chan_el_counter[4]; - - if(s->last_frame) - return 0; - if(data){ - if(!s->psypp){ - memcpy(s->samples + 1024 * avctx->channels, data, 1024 * avctx->channels * sizeof(s->samples[0])); - }else{ - start_ch = 0; - samples2 = s->samples + 1024 * avctx->channels; - for(i = 0; i < chan_map[0]; i++){ - tag = chan_map[i+1]; - chans = tag == TYPE_CPE ? 2 : 1; - ff_psy_preprocess(s->psypp, (uint16_t*)data + start_ch, samples2 + start_ch, start_ch + i, chans); - start_ch += chans; - } - } - } - if(!avctx->frame_number){ - memcpy(s->samples, s->samples + 1024 * avctx->channels, 1024 * avctx->channels * sizeof(s->samples[0])); - return 0; - } - - init_put_bits(&s->pb, frame, buf_size*8); - if((avctx->frame_number & 0xFF)==1 && !(avctx->flags & CODEC_FLAG_BITEXACT)){ - put_bitstream_info(avctx, s, LIBAVCODEC_IDENT); - } - start_ch = 0; - memset(chan_el_counter, 0, sizeof(chan_el_counter)); - for(i = 0; i < chan_map[0]; i++){ - FFPsyWindowInfo wi[2]; - tag = chan_map[i+1]; - chans = tag == TYPE_CPE ? 2 : 1; - cpe = &s->cpe[i]; - samples2 = samples + start_ch; - la = samples2 + 1024 * avctx->channels + start_ch; - if(!data) la = NULL; - s->lambda = 5e-7f; - for(j = 0; j < chans; j++){ - IndividualChannelStream *ics = &cpe->ch[j].ics; - int k; - wi[j] = ff_psy_suggest_window(&s->psy, samples2, la, start_ch + j, ics->window_sequence[0]); - ics->window_sequence[1] = ics->window_sequence[0]; - ics->window_sequence[0] = wi[j].window_type[0]; - ics->use_kb_window[1] = ics->use_kb_window[0]; - ics->use_kb_window[0] = wi[j].window_shape; - ics->num_windows = wi[j].num_windows; - ics->swb_sizes = s->psy.bands [ics->num_windows == 8]; - ics->num_swb = s->psy.num_bands[ics->num_windows == 8]; - for(k = 0; k < ics->num_windows; k++) - ics->group_len[k] = wi[j].grouping[k]; - - s->cur_channel = start_ch + j; - apply_window_and_mdct(avctx, s, &cpe->ch[j], samples2, j); - search_for_quantizers(avctx, s, &cpe->ch[j]); - } - cpe->common_window = 0; - if(chans > 1 - && wi[0].window_type[0] == wi[1].window_type[0] - && wi[0].window_shape == wi[1].window_shape){ - - cpe->common_window = 1; - for(j = 0; j < wi[0].num_windows; j++){ - if(wi[0].grouping[j] != wi[1].grouping[j]){ - cpe->common_window = 0; - break; - } - } - } -// search_for_ms(s, cpe, s->lambda); - adjust_frame_information(s, cpe, chans); - put_bits(&s->pb, 3, tag); - put_bits(&s->pb, 4, chan_el_counter[tag]++); - if(chans == 2){ - put_bits(&s->pb, 1, cpe->common_window); - if(cpe->common_window){ - put_ics_info(s, &cpe->ch[0].ics); - encode_ms_info(&s->pb, cpe); - } - } - for(j = 0; j < chans; j++){ - s->cur_channel = start_ch + j; - ff_psy_set_band_info(&s->psy, s->cur_channel, cpe->ch[j].coeffs, &wi[j]); - encode_individual_channel(avctx, s, &cpe->ch[j], cpe->common_window); - } - start_ch += chans; - } - - put_bits(&s->pb, 3, TYPE_END); - flush_put_bits(&s->pb); - avctx->frame_bits = put_bits_count(&s->pb); - - if(!data) - s->last_frame = 1; - memcpy(s->samples, s->samples + 1024 * avctx->channels, 1024 * avctx->channels * sizeof(s->samples[0])); - return put_bits_count(&s->pb)>>3; -} - -static av_cold int aac_encode_end(AVCodecContext *avctx) -{ - AACEncContext *s = avctx->priv_data; - - ff_mdct_end(&s->mdct1024); - ff_mdct_end(&s->mdct128); - ff_psy_end(&s->psy); - ff_psy_preprocess_end(s->psypp); - av_freep(&s->samples); - av_freep(&s->cpe); - return 0; -} - -AVCodec aac_encoder = { - "aac", - CODEC_TYPE_AUDIO, - CODEC_ID_AAC, - sizeof(AACEncContext), - aac_encode_init, - aac_encode_frame, - aac_encode_end, - .capabilities = CODEC_CAP_SMALL_LAST_FRAME | CODEC_CAP_DELAY, - .sample_fmts = (enum SampleFormat[]){SAMPLE_FMT_S16,SAMPLE_FMT_NONE}, - .long_name = NULL_IF_CONFIG_SMALL("Advanced Audio Coding"), +AACCoefficientsEncoder ff_aac_coders[] = { + { + search_for_quantizers_anmr, + encode_window_bands_info, + quantize_and_encode_band, +// search_for_ms, + }, + { + search_for_quantizers_faac, + encode_window_bands_info, + quantize_and_encode_band, +// search_for_ms, + }, + { + search_for_quantizers_twoloop, + encode_window_bands_info, + quantize_and_encode_band, +// search_for_ms, + }, + { + search_for_quantizers_fast, + encode_window_bands_info, + quantize_and_encode_band, +// search_for_ms, + }, }; Modified: aacenc/aacenc.c ============================================================================== --- aacenc/aacenc.c Mon Jan 26 15:00:57 2009 (r4007) +++ aacenc/aacenc.c Mon Jan 26 15:07:28 2009 (r4008) @@ -38,6 +38,7 @@ #include "aac.h" #include "aactab.h" +#include "aacenc.h" #include "psymodel.h" @@ -120,23 +121,6 @@ static const uint8_t *swb_size_128[] = { swb_size_128_16, swb_size_128_16, swb_size_128_8 }; -/** bits needed to code codebook run value for long windows */ -static const uint8_t run_value_bits_long[64] = { - 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, - 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 10, - 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, - 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 15 -}; - -/** bits needed to code codebook run value for short windows */ -static const uint8_t run_value_bits_short[16] = { - 3, 3, 3, 3, 3, 3, 3, 6, 6, 6, 6, 6, 6, 6, 6, 9 -}; - -static const uint8_t* run_value_bits[2] = { - run_value_bits_long, run_value_bits_short -}; - /** default channel configurations */ static const uint8_t aac_chan_configs[6][5] = { {1, TYPE_SCE}, // 1 channel - single channel element @@ -148,27 +132,6 @@ static const uint8_t aac_chan_configs[6] }; /** - * AAC encoder context - */ -typedef struct { - PutBitContext pb; - MDCTContext mdct1024; ///< long (1024 samples) frame transform context - MDCTContext mdct128; ///< short (128 samples) frame transform context - DSPContext dsp; - DECLARE_ALIGNED_16(FFTSample, output[2048]); ///< temporary buffer for MDCT input coefficients - int16_t* samples; ///< saved preprocessed input - - int samplerate_index; ///< MPEG-4 samplerate index - - ChannelElement *cpe; ///< channel elements - FFPsyContext psy; - struct FFPsyPreprocessContext* psypp; - int cur_channel; - int last_frame; - float lambda; -} AACEncContext; - -/** * Make AAC audio config object. * @see 1.6.2.1 "Syntax - AudioSpecificConfig" */ @@ -231,6 +194,7 @@ static av_cold int aac_encode_init(AVCod lengths[1] = ff_aac_num_swb_128[i]; ff_psy_init(&s->psy, avctx, 2, sizes, lengths); s->psypp = ff_psy_preprocess_init(avctx); + s->coder = &ff_aac_coders[0]; #if !CONFIG_HARDCODED_TABLES for (i = 0; i < 316; i++) @@ -332,519 +296,6 @@ static void encode_ms_info(PutBitContext } /** - * Quantize one coefficient. - * @return absolute value of the quantized coefficient - * @see 3GPP TS26.403 5.6.2 "Scalefactor determination" - */ -static av_always_inline int quant(float coef, const float Q) -{ - return av_clip((int)(pow(fabsf(coef) * Q, 0.75) + 0.4054), 0, 8191); -} - -#if 1 - -static av_always_inline int quant2(float coef, const float Q) -{ - return av_clip((int)(pow(fabsf(coef) * Q, 0.75)), 0, 8191); -} - -static const float aac_cb_range[12] = { 0, 3, 3, 3, 3, 9, 9, 8, 8, 13, 13, 17}; -static const float aac_cb_maxval[12] = {0, 1, 1, 2, 2, 4, 4, 7, 7, 12, 12, 16}; - -/** - * Calculate rate distortion cost for quantizing with given codebook - * - * @return quantization distortion - */ -static float quantize_band_cost(const float *in, int size, int scale_idx, int cb, - const float lambda, const float uplim, int *bits) -{ - const float IQ = ff_aac_pow2sf_tab[200 + scale_idx - SCALE_ONE_POS + SCALE_DIV_512]; - const float Q = ff_aac_pow2sf_tab[200 - scale_idx + SCALE_ONE_POS - SCALE_DIV_512]; - int i, j, k; - float cost = 0; - const int dim = cb < FIRST_PAIR_BT ? 4 : 2; - int resbits = 0; - const int range = aac_cb_range[cb]; - const int maxval = aac_cb_maxval[cb]; - int offs[4]; - - if(!cb){ - for(i = 0; i < size; i++) - cost += in[i]*in[i]*lambda; - return cost; - } - offs[0] = 1; - for(i = 1; i < dim; i++) - offs[i] = offs[i-1]*range; - for(i = 0; i < size; i += dim){ - float mincost; - int minidx = 0; - int minbits = 0; - int quants[4][2], realquants[4][2]; - mincost = 0.0f; - for(j = 0; j < dim; j++){ - realquants[j][0] = quant2(in[i+j], Q); - realquants[j][1] = quant (in[i+j], Q); - for(k = 0; k < 2; k++){ - quants[j][k] = FFMIN(realquants[j][k], maxval); - if(!IS_CODEBOOK_UNSIGNED(cb) && in[i+j] < 0.0f) - quants[j][k] = -quants[j][k]; - } - mincost += in[i+j]*in[i+j]*lambda; - } - minidx = IS_CODEBOOK_UNSIGNED(cb) ? 0 : 40; - minbits = ff_aac_spectral_bits[cb-1][minidx]; - mincost += minbits; - for(j = 0; j < (1<<dim); j++){ - float rd = 0.0f; - int curbits; - int curidx = IS_CODEBOOK_UNSIGNED(cb) ? 0 : 40; - const float *vec; - int same = 0; - for(k = 0; k < dim; k++){ - if((j & (1 << k)) && quants[k][0] == quants[k][1]){ - same = 1; - break; - } - } - if(same) - continue; - for(k = 0; k < dim; k++) - curidx += quants[k][!!(j & (1 << k))] * offs[dim - 1 - k]; - curbits = ff_aac_spectral_bits[cb-1][curidx]; - vec = &ff_aac_codebook_vectors[cb-1][curidx*dim]; - if(IS_CODEBOOK_UNSIGNED(cb)){ - for(k = 0; k < dim; k++){ - float t = fabsf(in[i+k]); - float di; - //do not code with escape sequence small values - if(vec[k] == 64.0f && t < 39.0f*IQ){ - rd = INFINITY; - break; - } - if(vec[k] == 64.0f){//FIXME: slow - if(t >= 165140.0f*IQ){ // clipped value - di = t - 165140.0f; - curbits += 21; - }else{ - int c = realquants[k]; - di = t - c*cbrt(c)*IQ; - curbits += av_log2(c)*2 - 4 + 1; - } - }else{ - di = t - vec[k]*IQ; - } - if(vec[k] != 0.0f) - curbits++; - rd += di*di*lambda; - } - }else{ - for(k = 0; k < dim; k++){ - float di = in[i+k] - vec[k]*IQ; - rd += di*di*lambda; - } - } - rd += curbits; - if(rd < mincost){ - mincost = rd; - minidx = j; - minbits = curbits; - } - } - cost += mincost; - resbits += minbits; - if(cost >= uplim) - return uplim; - } - - if(bits) - *bits = resbits; - return cost; -} - -static void quantize_and_encode_band(PutBitContext *pb, const float *in, int size, - int scale_idx, int cb, const float lambda) -{ - const float IQ = ff_aac_pow2sf_tab[200 + scale_idx - SCALE_ONE_POS + SCALE_DIV_512]; - const float Q = ff_aac_pow2sf_tab[200 - scale_idx + SCALE_ONE_POS - SCALE_DIV_512]; - const int range = aac_cb_range[cb]; - const int maxval = aac_cb_maxval[cb]; - const int dim = (cb < FIRST_PAIR_BT) ? 4 : 2; - int i, j, k; - int offs[4]; - - if(!cb) - return; - - offs[0] = 1; - for(i = 1; i < dim; i++) - offs[i] = offs[i-1]*range; - for(i = 0; i < size; i += dim){ - float mincost; - int minidx = 0; - int minbits = 0; - int quants[4][2]; - mincost = 0.0f; - for(j = 0; j < dim; j++){ - quants[j][0] = av_clip(quant2(in[i+j], Q), -maxval, maxval); - quants[j][1] = av_clip(quant (in[i+j], Q), -maxval, maxval); - for(k = 0; k < 2; k++){ - quants[j][k] = FFMIN(quants[j][k], maxval); - if(!IS_CODEBOOK_UNSIGNED(cb) && in[i+j] < 0.0f) - quants[j][k] = -quants[j][k]; - } - mincost += in[i+j]*in[i+j]*lambda; - } - minidx = IS_CODEBOOK_UNSIGNED(cb) ? 0 : 40; - minbits = ff_aac_spectral_bits[cb-1][minidx]; - mincost += minbits; - for(j = 0; j < (1<<dim); j++){ - float rd = 0.0f; - int curbits; - int curidx = IS_CODEBOOK_UNSIGNED(cb) ? 0 : 40; - const float *vec; - int same = 0; - for(k = 0; k < dim; k++){ - if((j & (1 << k)) && quants[k][0] == quants[k][1]){ - same = 1; - break; - } - } - if(same) - continue; - for(k = 0; k < dim; k++) - curidx += quants[k][!!(j & (1 << k))] * offs[dim - 1 - k]; - curbits = ff_aac_spectral_bits[cb-1][curidx]; - vec = &ff_aac_codebook_vectors[cb-1][curidx*dim]; - if(IS_CODEBOOK_UNSIGNED(cb)){ - for(k = 0; k < dim; k++){ - float t = fabsf(in[i+k]); - float di; - //do not code with escape sequence small values - if(vec[k] == 64.0f && t < 39.0f*IQ){ - rd = INFINITY; - break; - } - if(vec[k] == 64.0f){//FIXME: slow - if(t >= 165140.0f*IQ){ // clipped value - di = t - 165140.0f; - curbits += 21; - }else{ - int c = quant(t, Q); - di = t - c*cbrt(c)*IQ; - curbits += av_log2(c)*2 - 4 + 1; - } - }else{ - di = t - vec[k]*IQ; - } - if(vec[k] != 0.0f) - curbits++; - rd += di*di*lambda; - } - }else{ - for(k = 0; k < dim; k++){ - float di = in[i+k] - vec[k]*IQ; - rd += di*di*lambda; - } - } - rd += curbits; - if(rd < mincost){ - mincost = rd; - minidx = curidx; - minbits = curbits; - } - } - put_bits(pb, ff_aac_spectral_bits[cb-1][minidx], ff_aac_spectral_codes[cb-1][minidx]); - if(IS_CODEBOOK_UNSIGNED(cb)) - for(j = 0; j < dim; j++) - if(ff_aac_codebook_vectors[cb-1][minidx*dim+j] != 0.0f) - put_bits(pb, 1, in[i+j] < 0.0f); - if(cb == ESC_BT){ - for(j = 0; j < 2; j++){ - if(ff_aac_codebook_vectors[cb-1][minidx*2+j] == 64.0f){ - int coef = quant(in[i+j], Q); - int len = av_log2(coef); - - put_bits(pb, len - 4 + 1, (1 << (len - 4 + 1)) - 2); - put_bits(pb, len, coef & ((1 << len) - 1)); - } - } - } - } -} - -#else - -/** - * Calculate rate distortion cost for quantizing with given codebook - * - * @return quantization distortion - */ -static float quantize_band_cost(const float *in, int size, int scale_idx, int cb, - const float lambda, const float uplim, int *bits) -{ - const float Q = ff_aac_pow2sf_tab[200 + scale_idx - SCALE_ONE_POS + SCALE_DIV_512]; - int i, j, k; - float cost = 0; - const int dim = cb < FIRST_PAIR_BT ? 4 : 2; - int resbits = 0; - - if(!cb){ - for(i = 0; i < size; i++) - cost += in[i]*in[i]*lambda; - return cost; - } - for(i = 0; i < size; i += dim){ - float mincost = INFINITY; - int minidx = 0; - int minbits = 0; - const float *vec = ff_aac_codebook_vectors[cb-1]; - for(j = 0; j < ff_aac_spectral_sizes[cb-1]; j++, vec += dim){ - float rd = 0.0f; - int curbits = ff_aac_spectral_bits[cb-1][j]; - if(IS_CODEBOOK_UNSIGNED(cb)){ - for(k = 0; k < dim; k++){ - float t = fabsf(in[i+k]); - float di; - //do not code with escape sequence small values - if(vec[k] == 64.0f && t < 39.0f*Q){ - rd = INFINITY; - break; - } - if(vec[k] == 64.0f){//FIXME: slow - if(t >= 165140.0f*Q){ // clipped value - di = t - 165140.0f; - curbits += 21; - }else{ - int c = quant(t, 1.0/Q); - di = t - c*cbrt(c)*Q; - curbits += av_log2(c)*2 - 4 + 1; - } - }else{ - di = t - vec[k]*Q; - } - if(vec[k] != 0.0f) - curbits++; - rd += di*di*lambda; - } - }else{ - for(k = 0; k < dim; k++){ - float di = in[i+k] - vec[k]*Q; - rd += di*di*lambda; - } - } - rd += curbits; - if(rd < mincost){ - mincost = rd; - minidx = j; - minbits = curbits; - } - } - cost += mincost; - resbits += minbits; - if(cost >= uplim) - return uplim; - } - - if(bits) - *bits = resbits; - return cost; -} - -/** - * Prepare coefficients for encoding. - * - * @return sum of coefficient absolute values - */ -static void quantize_and_encode_band(PutBitContext *pb, const float *in, int size, - int scale_idx, int cb, const float lambda) -{ - const float Q = ff_aac_pow2sf_tab[200 + scale_idx - SCALE_ONE_POS + SCALE_DIV_512]; - const float IQ = ff_aac_pow2sf_tab[200 - scale_idx + SCALE_ONE_POS - SCALE_DIV_512]; - int i, j, k; - const int dim = cb < FIRST_PAIR_BT ? 4 : 2; - - if(!cb) - return; - - for(i = 0; i < size; i += dim){ - float mincost = INFINITY; - int minidx = 0; - int minbits = 0; - const float *vec = ff_aac_codebook_vectors[cb-1]; - for(j = 0; j < ff_aac_spectral_sizes[cb-1]; j++, vec += dim){ - float rd = 0.0f; - int curbits = ff_aac_spectral_bits[cb-1][j]; - if(IS_CODEBOOK_UNSIGNED(cb)){ - for(k = 0; k < dim; k++){ - float t = fabsf(in[i+k]); - float di; - //do not code with escape sequence small values - if(vec[k] == 64.0f && t < 39.0f*Q){ - rd = INFINITY; - break; - } - if(vec[k] == 64.0f){//FIXME: slow - if(t >= 165140.0f*Q){ // clipped value - di = t - 165140.0f; - curbits += 21; - }else{ - int c = quant(t, IQ); - di = t - c*cbrt(c)*Q; - curbits += av_log2(c)*2 - 4 + 1; - } - }else{ - di = t - vec[k]*Q; - } - if(vec[k] != 0.0f) - curbits++; - rd += di*di*lambda; - } - }else{ - for(k = 0; k < dim; k++){ - float di = in[i+k] - vec[k]*Q; - rd += di*di*lambda; - } - } - rd += curbits; - if(rd < mincost){ - mincost = rd; - minidx = j; - minbits = curbits; - } - } - put_bits(pb, ff_aac_spectral_bits[cb-1][minidx], ff_aac_spectral_codes[cb-1][minidx]); - if(IS_CODEBOOK_UNSIGNED(cb)) - for(j = 0; j < dim; j++) - if(ff_aac_codebook_vectors[cb-1][minidx*dim+j] != 0.0f) - put_bits(pb, 1, in[i+j] < 0.0f); - if(cb == ESC_BT){ - for(j = 0; j < 2; j++){ - if(ff_aac_codebook_vectors[cb-1][minidx*2+j] == 64.0f){ - int coef = quant(in[i+j], IQ); - int len = av_log2(coef); - - put_bits(pb, len - 4 + 1, (1 << (len - 4 + 1)) - 2); - put_bits(pb, len, coef & ((1 << len) - 1)); - } - } - } - } -} - -#endif - -/** - * structure used in optimal codebook search - */ -typedef struct BandCodingPath { - int prev_idx; ///< pointer to the previous path point - int codebook; ///< codebook for coding band run - float cost; ///< path cost - int run; -} BandCodingPath; - -/** - * Encode band info for single window group bands. - */ -static void encode_window_bands_info(AACEncContext *s, SingleChannelElement *sce, - int win, int group_len) -{ - BandCodingPath path[120][12]; - int w, swb, cb, start, start2, size; - int i, j; - const int max_sfb = sce->ics.max_sfb; - const int run_bits = sce->ics.num_windows == 1 ? 5 : 3; - const int run_esc = (1 << run_bits) - 1; - int idx, ppos, count; - int stackrun[120], stackcb[120], stack_len; - - start = win*128; - for(cb = 0; cb < 12; cb++){ - path[0][cb].cost = 0.0f; - path[0][cb].prev_idx = -1; - path[0][cb].run = 0; - } - for(swb = 0; swb < max_sfb; swb++){ - start2 = start; - size = sce->ics.swb_sizes[swb]; - if(sce->zeroes[win*16 + swb]){ - for(cb = 0; cb < 12; cb++){ - path[swb+1][cb].prev_idx = cb; - path[swb+1][cb].cost = path[swb][cb].cost; - path[swb+1][cb].run = path[swb][cb].run + 1; - } - }else{ - float minrd = INFINITY; - int mincb = 0; - for(cb = 0; cb < 12; cb++){ - float rd = 0.0f; - for(w = 0; w < group_len; w++){ - FFPsyBand *band = &s->psy.psy_bands[s->cur_channel*PSY_MAX_BANDS+(win+w)*16+swb]; - rd += quantize_band_cost(sce->coeffs + start + w*128, size, - sce->sf_idx[(win+w)*16+swb], cb, - s->lambda / band->threshold, INFINITY, NULL); - } - if( run_value_bits[sce->ics.num_windows == 8][path[swb][cb].run] - != run_value_bits[sce->ics.num_windows == 8][path[swb][cb].run+1]) - rd += run_bits; - path[swb+1][cb].prev_idx = cb; - path[swb+1][cb].cost = path[swb][cb].cost + rd; - path[swb+1][cb].run = path[swb][cb].run + 1; - if(rd < minrd){ - minrd = rd; - mincb = cb; - } - } - for(cb = 0; cb < 12; cb++){ - float cost = path[swb][cb].cost + minrd + run_bits + 4; - if(cost < path[swb+1][cb].cost){ - path[swb+1][cb].prev_idx = mincb; - path[swb+1][cb].cost = cost; - path[swb+1][cb].run = 1; - } - } - } - start += sce->ics.swb_sizes[swb]; - } - - //convert resulting path from backward-linked list - stack_len = 0; - idx = 0; - for(cb = 1; cb < 12; cb++){ - if(path[max_sfb][cb].cost < path[max_sfb][idx].cost) - idx = cb; - } - ppos = max_sfb; - while(ppos > 0){ - cb = idx; - stackrun[stack_len] = path[ppos][cb].run; - stackcb [stack_len] = cb; - idx = path[ppos][cb].prev_idx; - ppos -= path[ppos][cb].run; - stack_len++; - } - //perform actual band info encoding - start = 0; - for(i = stack_len - 1; i >= 0; i--){ - put_bits(&s->pb, 4, stackcb[i]); - count = stackrun[i]; - memset(sce->zeroes + win*16 + start, !stackcb[i], count); - //XXX: memset when band_type is also uint8_t - for(j = 0; j < count; j++){ - sce->band_type[win*16 + start] = stackcb[i]; - start++; - } - while(count >= run_esc){ - put_bits(&s->pb, run_bits, run_esc); - count -= run_esc; - } - put_bits(&s->pb, run_bits, count); - } -} - -/** * Produce integer coefficients from scalefactors provided by the model. */ static void adjust_frame_information(AACEncContext *apc, ChannelElement *cpe, int chans) @@ -903,546 +354,6 @@ static void adjust_frame_information(AAC } } -typedef struct TrellisPath { - float cost; - int prev; - int min_val; - int max_val; -} TrellisPath; - -static void search_for_quantizers_anmr(AACEncContext *s, SingleChannelElement *sce, const float lambda) -{ - int q, w, w2, g, start = 0; - int i; - int idx; - TrellisPath paths[256*121]; - int bandaddr[121]; - int minq; - float mincost; - - for(i = 0; i < 256; i++){ - paths[i].cost = 0.0f; - paths[i].prev = -1; - paths[i].min_val = i; - paths[i].max_val = i; - } - for(i = 256; i < 256*121; i++){ - paths[i].cost = INFINITY; - paths[i].prev = -2; - paths[i].min_val = INT_MAX; - paths[i].max_val = 0; - } - idx = 256; - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]){ - start = w*128; - for(g = 0; g < sce->ics.num_swb; g++){ - const float *coefs = sce->coeffs + start; - float qmin, qmax; - int nz = 0; - - bandaddr[idx >> 8] = w*16+g; - qmin = INT_MAX; - qmax = 0.0f; - for(w2 = 0; w2 < sce->ics.group_len[w]; w2++){ - FFPsyBand *band = &s->psy.psy_bands[s->cur_channel*PSY_MAX_BANDS+(w+w2)*16+g]; - if(band->energy <= band->threshold || band->threshold == 0.0f){ - sce->zeroes[(w+w2)*16+g] = 1; - continue; - } - sce->zeroes[(w+w2)*16+g] = 0; - nz = 1; - for(i = 0; i < sce->ics.swb_sizes[g]; i++){ - float t = fabsf(coefs[w2*128+i]); - if(t > 0.0f) qmin = fminf(qmin, t); - qmax = fmaxf(qmax, t); - } - } - if(nz){ - int minscale, maxscale; - float minrd = INFINITY; - //minimum scalefactor index is when minimum nonzero coefficient after quantizing is not clipped - minscale = av_clip_uint8(log2(qmin)*4 - 69 + SCALE_ONE_POS - SCALE_DIV_512); - //maximum scalefactor index is when maximum coefficient after quantizing is still not zero - maxscale = av_clip_uint8(log2(qmax)*4 + 6 + SCALE_ONE_POS - SCALE_DIV_512); - for(q = minscale; q < maxscale; q++){ - float dists[12], dist; - memset(dists, 0, sizeof(dists)); - for(w2 = 0; w2 < sce->ics.group_len[w]; w2++){ - FFPsyBand *band = &s->psy.psy_bands[s->cur_channel*PSY_MAX_BANDS+(w+w2)*16+g]; - int cb; - for(cb = 0; cb <= ESC_BT; cb++){ - dists[cb] += quantize_band_cost(coefs + w2*128, sce->ics.swb_sizes[g], - q, cb, s->lambda / band->threshold, INFINITY, NULL); - } - } - dist = dists[0]; - for(i = 1; i <= ESC_BT; i++) - dist = fminf(dist, dists[i]); - minrd = fminf(minrd, dist); - - for(i = FFMAX(q - SCALE_MAX_DIFF, 0); i < FFMIN(q + SCALE_MAX_DIFF, 256); i++){ - float cost; - int minv, maxv; - if(isinf(paths[idx - 256 + i].cost)) - continue; - cost = paths[idx - 256 + i].cost + dist - + ff_aac_scalefactor_bits[q - i + SCALE_DIFF_ZERO]; - minv = FFMIN(paths[idx - 256 + i].min_val, q); - maxv = FFMAX(paths[idx - 256 + i].max_val, q); - if(cost < paths[idx + q].cost && maxv-minv < SCALE_MAX_DIFF){ - paths[idx + q].cost = cost; - paths[idx + q].prev = idx - 256 + i; - paths[idx + q].min_val = minv; - paths[idx + q].max_val = maxv; - } - } - } - }else{ - for(q = 0; q < 256; q++){ - if(!isinf(paths[idx - 256 + q].cost)){ - paths[idx + q].cost = paths[idx - 256 + q].cost + 1; - paths[idx + q].prev = idx - 256 + q; - paths[idx + q].min_val = FFMIN(paths[idx - 256 + q].min_val, q); - paths[idx + q].max_val = FFMAX(paths[idx - 256 + q].max_val, q); - continue; - } - for(i = FFMAX(q - SCALE_MAX_DIFF, 0); i < FFMIN(q + SCALE_MAX_DIFF, 256); i++){ - float cost; - int minv, maxv; - if(isinf(paths[idx - 256 + i].cost)) - continue; - cost = paths[idx - 256 + i].cost + ff_aac_scalefactor_bits[q - i + SCALE_DIFF_ZERO]; - minv = FFMIN(paths[idx - 256 + i].min_val, q); - maxv = FFMAX(paths[idx - 256 + i].max_val, q); - if(cost < paths[idx + q].cost && maxv-minv < SCALE_MAX_DIFF){ - paths[idx + q].cost = cost; - paths[idx + q].prev = idx - 256 + i; - paths[idx + q].min_val = minv; - paths[idx + q].max_val = maxv; - } - } - } - } - sce->zeroes[w*16+g] = !nz; - start += sce->ics.swb_sizes[g]; - idx += 256; - } - } - idx -= 256; - mincost = paths[idx].cost; - minq = idx; - for(i = 1; i < 256; i++){ - if(paths[idx + i].cost < mincost){ - mincost = paths[idx + i].cost; - minq = idx + i; - } - } - while(minq >= 256){ - sce->sf_idx[bandaddr[minq>>8]] = minq & 0xFF; - minq = paths[minq].prev; - } - //set the same quantizers inside window groups - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) - for(g = 0; g < sce->ics.num_swb; g++) - for(w2 = 1; w2 < sce->ics.group_len[w]; w2++) - sce->sf_idx[(w+w2)*16+g] = sce->sf_idx[w*16+g]; -} - -/** - * two-loop quantizers search taken from ISO 13818-7 Appendix C - */ -static void search_for_quantizers_twoloop(AVCodecContext *avctx, AACEncContext *s, - SingleChannelElement *sce, const float lambda) -{ - int start = 0, i, w, w2, g; - int destbits = avctx->bit_rate * 1024.0 / avctx->sample_rate / avctx->channels; - float dists[128], uplims[128]; - int fflag, minscaler; - int its = 0; - int allz = 0; - float minthr = INFINITY; - - //XXX: some heuristic to determine initial quantizers will reduce search time - memset(dists, 0, sizeof(dists)); - //determine zero bands and upper limits - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]){ - for(g = 0; g < sce->ics.num_swb; g++){ - int nz = 0; - float uplim = 0.0f; - for(w2 = 0; w2 < sce->ics.group_len[w]; w2++){ - FFPsyBand *band = &s->psy.psy_bands[s->cur_channel*PSY_MAX_BANDS+(w+w2)*16+g]; - uplim += band->threshold; - if(band->energy <= band->threshold || band->threshold == 0.0f){ - sce->zeroes[(w+w2)*16+g] = 1; - continue; - } - nz = 1; - } - uplims[w*16+g] = uplim *512; - sce->zeroes[w*16+g] = !nz; - if(nz) - minthr = fminf(minthr, uplim); - allz = FFMAX(allz, nz); - } - } - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]){ - for(g = 0; g < sce->ics.num_swb; g++){ - if(sce->zeroes[w*16+g]){ - sce->sf_idx[w*16+g] = SCALE_ONE_POS; - continue; - } - sce->sf_idx[w*16+g] = SCALE_ONE_POS + fminf(log2(uplims[w*16+g]/minthr)*4,59); - } - } - - if(!allz) - return; - //perform two-loop search - //outer loop - improve quality - do{ - int tbits, qstep; - minscaler = sce->sf_idx[0]; - //inner loop - quantize spectrum to fit into given number of bits - qstep = its ? 1 : 32; - do{ - int prev = -1; - tbits = 0; - fflag = 0; - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]){ - start = w*128; - for(g = 0; g < sce->ics.num_swb; g++){ - const float *coefs = sce->coeffs + start; - int bits = 0; - int cb; - float mindist = INFINITY; - int minbits = 0; - - if(sce->zeroes[w*16+g] || sce->sf_idx[w*16+g] >= 218) - continue; - minscaler = FFMIN(minscaler, sce->sf_idx[w*16+g]); - for(cb = 0; cb <= ESC_BT; cb++){ - float dist = 0.0f; - int bb = 0; - for(w2 = 0; w2 < sce->ics.group_len[w]; w2++){ - int b; - dist += quantize_band_cost(coefs + w2*128, - sce->ics.swb_sizes[g], - sce->sf_idx[w*16+g], - ESC_BT, - 1.0, - INFINITY, - &b); - bb += b; - } - if(dist < mindist){ - mindist = dist; - minbits = bb; - } - } - dists[w*16+g] = mindist - minbits; - bits = minbits; - if(prev != -1){ - bits += ff_aac_scalefactor_bits[sce->sf_idx[w*16+g] - prev + SCALE_DIFF_ZERO]; - } - tbits += bits; - start += sce->ics.swb_sizes[g]; - prev = sce->sf_idx[w*16+g]; - } - } - if(tbits > destbits){ - for(i = 0; i < 128; i++){ - if(sce->sf_idx[i] < 218 - qstep){ - sce->sf_idx[i] += qstep; - } - } - }else{ - for(i = 0; i < 128; i++){ - if(sce->sf_idx[i] > 60 - qstep){ - sce->sf_idx[i] -= qstep; - } - } - } - qstep >>= 1; - if(!qstep && tbits > destbits*1.02) - qstep = 1; - if(sce->sf_idx[0] >= 217)break; - }while(qstep); - - fflag = 0; - minscaler = av_clip(minscaler, 60, 255 - SCALE_MAX_DIFF); - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]){ - start = w*128; - for(g = 0; g < sce->ics.num_swb; g++){ - int prevsc = sce->sf_idx[w*16+g]; - if(dists[w*16+g] > uplims[w*16+g] && sce->sf_idx[w*16+g] > 60) - sce->sf_idx[w*16+g]--; - sce->sf_idx[w*16+g] = av_clip(sce->sf_idx[w*16+g], minscaler, minscaler + SCALE_MAX_DIFF); - sce->sf_idx[w*16+g] = FFMIN(sce->sf_idx[w*16+g], 219); - if(sce->sf_idx[w*16+g] != prevsc) - fflag = 1; - } - } - its++; - }while(fflag && its < 10); -} - -static void search_for_quantizers_faac(AACEncContext *s, SingleChannelElement *sce, const float lambda) -{ - int start = 0, i, w, w2, g; - float uplim[128], maxq[128]; - int minq; -#if 0 - float distfact = ((sce->ics.num_windows > 1) ? 85.80 : 147.84) / lambda; - int last = 0, lastband = 0, curband = 0; - float avg_energy = 0.0; - if(sce->ics.num_windows == 1){ - start = 0; - for(i = 0; i < 1024; i++){ - if(i - start >= sce->ics.swb_sizes[curband]){ - start += sce->ics.swb_sizes[curband]; - curband++; - } - if(sce->coeffs[i]){ - avg_energy += sce->coeffs[i] * sce->coeffs[i]; - last = i; - lastband = curband; - } - } - }else{ - for(w = 0; w < 8; w++){ - const float *coeffs = sce->coeffs + w*128; - start = 0; - for(i = 0; i < 128; i++){ - if(i - start >= sce->ics.swb_sizes[curband]){ - start += sce->ics.swb_sizes[curband]; - curband++; - } - if(coeffs[i]){ - avg_energy += coeffs[i] * coeffs[i]; - last = FFMAX(last, i); - lastband = FFMAX(lastband, curband); - } - } - } - } - last++; - avg_energy /= last; - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]){ - start = w*128; - for(g = 0; g < sce->ics.num_swb; g++){ - float *coefs = sce->coeffs + start; - const int size = sce->ics.swb_sizes[g]; - int start2 = start, end2 = start + size, peakpos = start; - float maxval = -1, thr = 0.0f, t; - maxq[w*16+g] = 0.0f; - if(g > lastband){ - maxq[w*16+g] = 0.0f; - start += size; - for(w2 = 0; w2 < sce->ics.group_len[w]; w2++) - memset(coefs + w2*128, 0, sizeof(coefs[0])*size); - continue; - } - for(w2 = 0; w2 < sce->ics.group_len[w]; w2++){ - for(i = 0; i < size; i++){ - float t = coefs[w2*128+i]*coefs[w2*128+i]; - maxq[w*16+g] = fmaxf(maxq[w*16+g], coefs[w2*128 + i]); - thr += t; - if(sce->ics.num_windows == 1 && maxval < t){ - maxval = t; - peakpos = start+i; - } - } - } - if(sce->ics.num_windows == 1){ - start2 = FFMAX(peakpos - 2, start2); - end2 = FFMIN(peakpos + 3, end2); - } - start += size; - thr = pow(thr / (avg_energy * (end2 - start2)), 0.3 + 0.1*(lastband - g) / lastband); - t = 1.0 - (1.0 * start2 / last); - uplim[w*16+g] = distfact / (1.4 * thr + t*t*t + 0.075); - } - } -#else - memset(maxq, 0, sizeof(maxq)); - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]){ - start = w*128; - for(g = 0; g < sce->ics.num_swb; g++){ - float *coefs = sce->coeffs + start; - const int size = sce->ics.swb_sizes[g]; - float thr = 0.0f; - for(w2 = 0; w2 < sce->ics.group_len[w]; w2++){ - FFPsyBand *band = &s->psy.psy_bands[s->cur_channel*PSY_MAX_BANDS+(w+w2)*16+g]; - thr += band->threshold; - for(i = 0; i < size; i++){ - maxq[w*16+g] = fmaxf(maxq[w*16+g], fabsf(coefs[w2*128 + i])); - } - } - uplim[w*16+g] = thr / lambda; - start += size; - } - } -#endif - memset(sce->sf_idx, 0, sizeof(sce->sf_idx)); - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]){ - start = w*128; - for(g = 0; g < sce->ics.num_swb; g++){ - const float *coefs = sce->coeffs + start; - const int size = sce->ics.swb_sizes[g]; - int scf, prev_scf, step; - if(maxq[w*16+g] < 21.544){ - sce->zeroes[w*16+g] = 1; - continue; - } - sce->zeroes[w*16+g] = 0; - scf = prev_scf = av_clip(SCALE_ONE_POS + log2(1/maxq[w*16+g])*16/3, 60, 218); - step = 16; - for(;;){ - float dist = 0.0f, t; - int quant_max; - - for(w2 = 0; w2 < sce->ics.group_len[w]; w2++){ - int b; - dist += quantize_band_cost(coefs + w2*128, - sce->ics.swb_sizes[g], - scf, - ESC_BT, - 1.0, - INFINITY, - &b); - dist -= b; - } - quant_max = quant(maxq[w*16+g], ff_aac_pow2sf_tab[200 - scf + SCALE_ONE_POS]); - if(quant_max >= 8191){ // too much, return to the previous quantizer - sce->sf_idx[w*16+g] = prev_scf; - break; - } - prev_scf = scf; - t = pow(dist / size, -2.0/3.0); - if(FFABS(step) > 4){ - int newstep = 4*log(t / maxq[w*16+g])/log(2) - 2; - step = av_clip(-newstep, -4, 4); - } - if(FFABS(step) >= 4){ - step = 1 - step; - scf += step; - }else if(t >= uplim[w*16+g]){ - scf -= FFABS(step); - }else{ - sce->sf_idx[w*16+g] = scf; - break; - } - } - start += size; - } - } - minq = sce->sf_idx[0] ? sce->sf_idx[0] : INT_MAX; - for(i = 1; i < 128; i++){ - if(!sce->sf_idx[i]) - sce->sf_idx[i] = sce->sf_idx[i-1]; - else - minq = FFMIN(minq, sce->sf_idx[i]); - } - if(minq == INT_MAX) minq = 0; - for(i = 126; i >= 0; i--){ - if(!sce->sf_idx[i]) - sce->sf_idx[i] = sce->sf_idx[i+1]; - sce->sf_idx[i] = av_clip(sce->sf_idx[i], minq, minq + SCALE_MAX_DIFF); - } -} - -static void search_for_quantizers_fast(AACEncContext *s, SingleChannelElement *sce) -{ - int start = 0, i, w, w2, g; - int minq = 255; - - memset(sce->sf_idx, 0, sizeof(sce->sf_idx)); - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]){ - start = w*128; - for(g = 0; g < sce->ics.num_swb; g++){ - for(w2 = 0; w2 < sce->ics.group_len[w]; w2++){ - FFPsyBand *band = &s->psy.psy_bands[s->cur_channel*PSY_MAX_BANDS+(w+w2)*16+g]; - if(band->energy <= band->threshold){ - sce->sf_idx[(w+w2)*16+g] = 218; - sce->zeroes[(w+w2)*16+g] = 1; - }else{ - sce->sf_idx[(w+w2)*16+g] = av_clip(SCALE_ONE_POS - SCALE_DIV_512 + log2(band->threshold), 80, 218); - sce->zeroes[(w+w2)*16+g] = 0; - } - minq = FFMIN(minq, sce->sf_idx[(w+w2)*16+g]); - } - } - } - for(i = 0; i < 128; i++){ - sce->sf_idx[i] = av_clip(sce->sf_idx[i], minq, minq + SCALE_MAX_DIFF - 1); - } - //set the same quantizers inside window groups - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) - for(g = 0; g < sce->ics.num_swb; g++) - for(w2 = 1; w2 < sce->ics.group_len[w]; w2++) - sce->sf_idx[(w+w2)*16+g] = sce->sf_idx[w*16+g]; -} - -static void search_for_quantizers(AVCodecContext *avctx, AACEncContext *s, - SingleChannelElement *sce) -{ - search_for_quantizers_anmr(s, sce, s->lambda); -// search_for_quantizers_twoloop(avctx, s, sce, s->lambda); -// search_for_quantizers_faac(s, sce, av_clipf(s->lambda * 2e8, 50.0f, 300.0f)); -// search_for_quantizers_fast(s, sce); -} - -static void search_for_ms(AACEncContext *s, ChannelElement *cpe, const float lambda) -{ - int start = 0, i, w, w2, g; - float M[128], S[128]; - SingleChannelElement *sce0 = &cpe->ch[0]; - SingleChannelElement *sce1 = &cpe->ch[1]; - if(!cpe->common_window) - return; - for(w = 0; w < sce0->ics.num_windows; w += sce0->ics.group_len[w]){ - for(g = 0; g < sce0->ics.num_swb; g++){ - if(!cpe->ch[0].zeroes[w*16+g] && !cpe->ch[1].zeroes[w*16+g]){ - float dist1 = 0.0f, dist2 = 0.0f; - for(w2 = 0; w2 < sce0->ics.group_len[w]; w2++){ - FFPsyBand *band0 = &s->psy.psy_bands[(s->cur_channel+0)*PSY_MAX_BANDS+(w+w2)*16+g]; - FFPsyBand *band1 = &s->psy.psy_bands[(s->cur_channel+1)*PSY_MAX_BANDS+(w+w2)*16+g]; - float minthr = fminf(band0->threshold, band1->threshold); - float maxthr = fmaxf(band0->threshold, band1->threshold); - for(i = 0; i < sce0->ics.swb_sizes[g]; i++){ - M[i] = (sce0->coeffs[start+w2*128+i] - + sce1->coeffs[start+w2*128+i])*0.5; - S[i] = sce0->coeffs[start+w2*128+i] - - sce1->coeffs[start+w2*128+i]; - } - dist1 += quantize_band_cost(sce0->coeffs + start + w2*128, - sce0->ics.swb_sizes[g], - sce0->sf_idx[(w+w2)*16+g], - sce0->band_type[(w+w2)*16+g], - lambda / band0->threshold, INFINITY, NULL); - dist1 += quantize_band_cost(sce1->coeffs + start + w2*128, - sce1->ics.swb_sizes[g], - sce1->sf_idx[(w+w2)*16+g], - sce1->band_type[(w+w2)*16+g], - lambda / band1->threshold, INFINITY, NULL); - dist2 += quantize_band_cost(M, - sce0->ics.swb_sizes[g], - sce0->sf_idx[(w+w2)*16+g], - sce0->band_type[(w+w2)*16+g], - lambda / maxthr, INFINITY, NULL); - dist2 += quantize_band_cost(S, - sce1->ics.swb_sizes[g], - sce1->sf_idx[(w+w2)*16+g], - sce1->band_type[(w+w2)*16+g], - lambda / minthr, INFINITY, NULL); - } - cpe->ms_mask[w*16+g] = dist2 < dist1; - } - start += sce0->ics.swb_sizes[g]; - } - } -} - /** * Encode scalefactor band coding type. */ @@ -1451,7 +362,7 @@ static void encode_band_info(AACEncConte int w; for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]){ - encode_window_bands_info(s, sce, w, sce->ics.group_len[w]); + s->coder->encode_window_bands_info(s, sce, w, sce->ics.group_len[w]); } } @@ -1508,7 +419,7 @@ static void encode_spectral_coeffs(AACEn continue; } for(w2 = w; w2 < w + sce->ics.group_len[w]; w2++){ - quantize_and_encode_band(&s->pb, sce->coeffs + start + w2*128, + s->coder->quantize_and_encode_band(&s->pb, sce->coeffs + start + w2*128, sce->ics.swb_sizes[i], sce->sf_idx[w*16 + i], sce->band_type[w*16 + i], @@ -1617,7 +528,7 @@ static int aac_encode_frame(AVCodecConte s->cur_channel = start_ch + j; apply_window_and_mdct(avctx, s, &cpe->ch[j], samples2, j); - search_for_quantizers(avctx, s, &cpe->ch[j]); + s->coder->search_for_quantizers(avctx, s, &cpe->ch[j], s->lambda); } cpe->common_window = 0; if(chans > 1 @@ -1632,7 +543,8 @@ static int aac_encode_frame(AVCodecConte } } } -// search_for_ms(s, cpe, s->lambda); + if(cpe->common_window && s->coder->search_for_ms) + s->coder->search_for_ms(s, cpe, s->lambda); adjust_frame_information(s, cpe, chans); put_bits(&s->pb, 3, tag); put_bits(&s->pb, 4, chan_el_counter[tag]++); Copied and modified: aacenc/aacenc.h (from r4006, aacenc/aacenc.c) ============================================================================== --- aacenc/aacenc.c Mon Jan 26 14:37:36 2009 (r4006, copy source) +++ aacenc/aacenc.h Mon Jan 26 15:07:28 2009 (r4008) @@ -19,138 +19,35 @@ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ -/** - * @file aacenc.c - * AAC encoder - */ - -/*********************************** - * TODOs: - * speedup quantizer selection - * add sane pulse detection - * add temporal noise shaping - ***********************************/ +#ifndef AVCODEC_AACENC_H +#define AVCODEC_AACENC_H #include "avcodec.h" #include "bitstream.h" #include "dsputil.h" -#include "mpeg4audio.h" #include "aac.h" -#include "aactab.h" #include "psymodel.h" -static const uint8_t swb_size_1024_96[] = { - 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, - 12, 12, 12, 12, 12, 16, 16, 24, 28, 36, 44, - 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64 -}; - -static const uint8_t swb_size_1024_64[] = { - 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, - 12, 12, 12, 16, 16, 16, 20, 24, 24, 28, 36, - 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40 -}; - -static const uint8_t swb_size_1024_48[] = { - 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8, - 12, 12, 12, 12, 16, 16, 20, 20, 24, 24, 28, 28, - 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, - 96 -}; - -static const uint8_t swb_size_1024_32[] = { - 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8, - 12, 12, 12, 12, 16, 16, 20, 20, 24, 24, 28, 28, - 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32 -}; - -static const uint8_t swb_size_1024_24[] = { - 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, - 12, 12, 12, 12, 16, 16, 16, 20, 20, 24, 24, 28, 28, - 32, 36, 36, 40, 44, 48, 52, 52, 64, 64, 64, 64, 64 -}; - -static const uint8_t swb_size_1024_16[] = { - 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, - 12, 12, 12, 12, 12, 12, 12, 12, 12, 16, 16, 16, 16, 20, 20, 20, 24, 24, 28, 28, - 32, 36, 40, 40, 44, 48, 52, 56, 60, 64, 64, 64 -}; - -static const uint8_t swb_size_1024_8[] = { - 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, - 16, 16, 16, 16, 16, 16, 16, 20, 20, 20, 20, 24, 24, 24, 28, 28, - 32, 36, 36, 40, 44, 48, 52, 56, 60, 64, 80 -}; - -static const uint8_t *swb_size_1024[] = { - swb_size_1024_96, swb_size_1024_96, swb_size_1024_64, - swb_size_1024_48, swb_size_1024_48, swb_size_1024_32, - swb_size_1024_24, swb_size_1024_24, swb_size_1024_16, - swb_size_1024_16, swb_size_1024_16, swb_size_1024_8 -}; - -static const uint8_t swb_size_128_96[] = { - 4, 4, 4, 4, 4, 4, 8, 8, 8, 16, 28, 36 -}; - -static const uint8_t swb_size_128_48[] = { - 4, 4, 4, 4, 4, 8, 8, 8, 12, 12, 12, 16, 16, 16 -}; - -static const uint8_t swb_size_128_24[] = { - 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 12, 12, 16, 16, 20 -}; - -static const uint8_t swb_size_128_16[] = { - 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 12, 12, 16, 20, 20 -}; - -static const uint8_t swb_size_128_8[] = { - 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 12, 16, 20, 20 -}; - -static const uint8_t *swb_size_128[] = { - /* the last entry on the following row is swb_size_128_64 but is a - duplicate of swb_size_128_96 */ - swb_size_128_96, swb_size_128_96, swb_size_128_96, - swb_size_128_48, swb_size_128_48, swb_size_128_48, - swb_size_128_24, swb_size_128_24, swb_size_128_16, - swb_size_128_16, swb_size_128_16, swb_size_128_8 -}; - -/** bits needed to code codebook run value for long windows */ -static const uint8_t run_value_bits_long[64] = { - 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, - 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 10, - 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, - 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 15 -}; - -/** bits needed to code codebook run value for short windows */ -static const uint8_t run_value_bits_short[16] = { - 3, 3, 3, 3, 3, 3, 3, 6, 6, 6, 6, 6, 6, 6, 6, 9 -}; +struct AACEncContext; -static const uint8_t* run_value_bits[2] = { - run_value_bits_long, run_value_bits_short -}; +typedef struct AACCoefficientsEncoder{ + void (*search_for_quantizers)(AVCodecContext *avctx, struct AACEncContext *s, + SingleChannelElement *sce, const float lambda); + void (*encode_window_bands_info)(struct AACEncContext *s, SingleChannelElement *sce, + int win, int group_len); + void (*quantize_and_encode_band)(PutBitContext *pb, const float *in, int size, + int scale_idx, int cb, const float lambda); + void (*search_for_ms)(struct AACEncContext *s, ChannelElement *cpe, const float lambda); +}AACCoefficientsEncoder; -/** default channel configurations */ -static const uint8_t aac_chan_configs[6][5] = { - {1, TYPE_SCE}, // 1 channel - single channel element - {1, TYPE_CPE}, // 2 channels - channel pair - {2, TYPE_SCE, TYPE_CPE}, // 3 channels - center + stereo - {3, TYPE_SCE, TYPE_CPE, TYPE_SCE}, // 4 channels - front center + stereo + back center - {3, TYPE_SCE, TYPE_CPE, TYPE_CPE}, // 5 channels - front center + stereo + back stereo - {4, TYPE_SCE, TYPE_CPE, TYPE_CPE, TYPE_LFE}, // 6 channels - front center + stereo + back stereo + LFE -}; +extern AACCoefficientsEncoder ff_aac_coders[]; /** * AAC encoder context */ -typedef struct { +typedef struct AACEncContext { PutBitContext pb; MDCTContext mdct1024; ///< long (1024 samples) frame transform context MDCTContext mdct128; ///< short (128 samples) frame transform context @@ -163,1526 +60,10 @@ typedef struct { ChannelElement *cpe; ///< channel elements FFPsyContext psy; struct FFPsyPreprocessContext* psypp; + AACCoefficientsEncoder *coder; int cur_channel; int last_frame; float lambda; } AACEncContext; -/** - * Make AAC audio config object. - * @see 1.6.2.1 "Syntax - AudioSpecificConfig" - */ -static void put_audio_specific_config(AVCodecContext *avctx) -{ - PutBitContext pb; - AACEncContext *s = avctx->priv_data; - - init_put_bits(&pb, avctx->extradata, avctx->extradata_size*8); - put_bits(&pb, 5, 2); //object type - AAC-LC - put_bits(&pb, 4, s->samplerate_index); //sample rate index - put_bits(&pb, 4, avctx->channels); - //GASpecificConfig - put_bits(&pb, 1, 0); //frame length - 1024 samples - put_bits(&pb, 1, 0); //does not depend on core coder - put_bits(&pb, 1, 0); //is not extension - flush_put_bits(&pb); -} - -static av_cold int aac_encode_init(AVCodecContext *avctx) -{ - AACEncContext *s = avctx->priv_data; - int i; - const uint8_t *sizes[2]; - int lengths[2]; - - avctx->frame_size = 1024; - - for(i = 0; i < 16; i++) - if(avctx->sample_rate == ff_mpeg4audio_sample_rates[i]) - break; - if(i == 16){ - av_log(avctx, AV_LOG_ERROR, "Unsupported sample rate %d\n", avctx->sample_rate); - return -1; - } - if(avctx->channels > 6){ - av_log(avctx, AV_LOG_ERROR, "Unsupported number of channels: %d\n", avctx->channels); - return -1; - } - s->samplerate_index = i; - - dsputil_init(&s->dsp, avctx); - ff_mdct_init(&s->mdct1024, 11, 0); - ff_mdct_init(&s->mdct128, 8, 0); - // window init - ff_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024); - ff_kbd_window_init(ff_aac_kbd_short_128, 6.0, 128); - ff_sine_window_init(ff_sine_1024, 1024); - ff_sine_window_init(ff_sine_128, 128); - - s->samples = av_malloc(2 * 1024 * avctx->channels * sizeof(s->samples[0])); - s->cpe = av_mallocz(sizeof(ChannelElement) * aac_chan_configs[avctx->channels-1][0]); - avctx->extradata = av_malloc(2); - avctx->extradata_size = 2; - put_audio_specific_config(avctx); - - sizes[0] = swb_size_1024[i]; - sizes[1] = swb_size_128[i]; - lengths[0] = ff_aac_num_swb_1024[i]; - lengths[1] = ff_aac_num_swb_128[i]; - ff_psy_init(&s->psy, avctx, 2, sizes, lengths); - s->psypp = ff_psy_preprocess_init(avctx); - -#if !CONFIG_HARDCODED_TABLES - for (i = 0; i < 316; i++) - ff_aac_pow2sf_tab[i] = pow(2, (i - 200)/4.); -#endif /* CONFIG_HARDCODED_TABLES */ - - return 0; -} - -static void apply_window_and_mdct(AVCodecContext *avctx, AACEncContext *s, - SingleChannelElement *sce, short *audio, int channel) -{ - int i, j, k; - const float * lwindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024; - const float * swindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128; - const float * pwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128; - - if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE) { - memcpy(s->output, sce->saved, sizeof(float)*1024); - if(sce->ics.window_sequence[0] == LONG_STOP_SEQUENCE){ - memset(s->output, 0, sizeof(s->output[0]) * 448); - for(i = 448; i < 576; i++) - s->output[i] = sce->saved[i] * pwindow[i - 448]; - for(i = 576; i < 704; i++) - s->output[i] = sce->saved[i]; - } - if(sce->ics.window_sequence[0] != LONG_START_SEQUENCE){ - j = channel; - for (i = 0; i < 1024; i++, j += avctx->channels){ - s->output[i+1024] = audio[j] * lwindow[1024 - i - 1]; - sce->saved[i] = audio[j] * lwindow[i]; - } - }else{ - j = channel; - for(i = 0; i < 448; i++, j += avctx->channels) - s->output[i+1024] = audio[j]; - for(i = 448; i < 576; i++, j += avctx->channels) - s->output[i+1024] = audio[j] * swindow[576 - i - 1]; - memset(s->output+1024+576, 0, sizeof(s->output[0]) * 448); - j = channel; - for(i = 0; i < 1024; i++, j += avctx->channels) - sce->saved[i] = audio[j]; - } - ff_mdct_calc(&s->mdct1024, sce->coeffs, s->output); - }else{ - j = channel; - for (k = 0; k < 1024; k += 128) { - for(i = 448 + k; i < 448 + k + 256; i++) - s->output[i - 448 - k] = (i < 1024) - ? sce->saved[i] - : audio[channel + (i-1024)*avctx->channels]; - s->dsp.vector_fmul (s->output, k ? swindow : pwindow, 128); - s->dsp.vector_fmul_reverse(s->output+128, s->output+128, swindow, 128); - ff_mdct_calc(&s->mdct128, sce->coeffs + k, s->output); - } - j = channel; - for(i = 0; i < 1024; i++, j += avctx->channels) - sce->saved[i] = audio[j]; - } -} - -/** - * Encode ics_info element. - * @see Table 4.6 (syntax of ics_info) - */ -static void put_ics_info(AACEncContext *s, IndividualChannelStream *info) -{ - int w; - - put_bits(&s->pb, 1, 0); // ics_reserved bit - put_bits(&s->pb, 2, info->window_sequence[0]); - put_bits(&s->pb, 1, info->use_kb_window[0]); - if(info->window_sequence[0] != EIGHT_SHORT_SEQUENCE){ - put_bits(&s->pb, 6, info->max_sfb); - put_bits(&s->pb, 1, 0); // no prediction - }else{ - put_bits(&s->pb, 4, info->max_sfb); - for(w = 1; w < 8; w++){ - put_bits(&s->pb, 1, !info->group_len[w]); - } - } -} - -/** - * Encode MS data. - * @see 4.6.8.1 "Joint Coding - M/S Stereo" - */ -static void encode_ms_info(PutBitContext *pb, ChannelElement *cpe) -{ - int i, w; - - put_bits(pb, 2, cpe->ms_mode); - if(cpe->ms_mode == 1){ - for(w = 0; w < cpe->ch[0].ics.num_windows; w += cpe->ch[0].ics.group_len[w]){ - for(i = 0; i < cpe->ch[0].ics.max_sfb; i++) - put_bits(pb, 1, cpe->ms_mask[w*16 + i]); - } - } -} - -/** - * Quantize one coefficient. - * @return absolute value of the quantized coefficient - * @see 3GPP TS26.403 5.6.2 "Scalefactor determination" - */ -static av_always_inline int quant(float coef, const float Q) -{ - return av_clip((int)(pow(fabsf(coef) * Q, 0.75) + 0.4054), 0, 8191); -} - -#if 1 - -static av_always_inline int quant2(float coef, const float Q) -{ - return av_clip((int)(pow(fabsf(coef) * Q, 0.75)), 0, 8191); -} - -static const float aac_cb_range[12] = { 0, 3, 3, 3, 3, 9, 9, 8, 8, 13, 13, 17}; -static const float aac_cb_maxval[12] = {0, 1, 1, 2, 2, 4, 4, 7, 7, 12, 12, 16}; - -/** - * Calculate rate distortion cost for quantizing with given codebook - * - * @return quantization distortion - */ -static float quantize_band_cost(const float *in, int size, int scale_idx, int cb, - const float lambda, const float uplim, int *bits) -{ - const float IQ = ff_aac_pow2sf_tab[200 + scale_idx - SCALE_ONE_POS + SCALE_DIV_512]; - const float Q = ff_aac_pow2sf_tab[200 - scale_idx + SCALE_ONE_POS - SCALE_DIV_512]; - int i, j, k; - float cost = 0; - const int dim = cb < FIRST_PAIR_BT ? 4 : 2; - int resbits = 0; - const int range = aac_cb_range[cb]; - const int maxval = aac_cb_maxval[cb]; - int offs[4]; - - if(!cb){ - for(i = 0; i < size; i++) - cost += in[i]*in[i]*lambda; - return cost; - } - offs[0] = 1; - for(i = 1; i < dim; i++) - offs[i] = offs[i-1]*range; - for(i = 0; i < size; i += dim){ - float mincost; - int minidx = 0; - int minbits = 0; - int quants[4][2], realquants[4][2]; - mincost = 0.0f; - for(j = 0; j < dim; j++){ - realquants[j][0] = quant2(in[i+j], Q); - realquants[j][1] = quant (in[i+j], Q); - for(k = 0; k < 2; k++){ - quants[j][k] = FFMIN(realquants[j][k], maxval); - if(!IS_CODEBOOK_UNSIGNED(cb) && in[i+j] < 0.0f) - quants[j][k] = -quants[j][k]; - } - mincost += in[i+j]*in[i+j]*lambda; - } - minidx = IS_CODEBOOK_UNSIGNED(cb) ? 0 : 40; - minbits = ff_aac_spectral_bits[cb-1][minidx]; - mincost += minbits; - for(j = 0; j < (1<<dim); j++){ - float rd = 0.0f; - int curbits; - int curidx = IS_CODEBOOK_UNSIGNED(cb) ? 0 : 40; - const float *vec; - int same = 0; - for(k = 0; k < dim; k++){ - if((j & (1 << k)) && quants[k][0] == quants[k][1]){ - same = 1; - break; - } - } - if(same) - continue; - for(k = 0; k < dim; k++) - curidx += quants[k][!!(j & (1 << k))] * offs[dim - 1 - k]; - curbits = ff_aac_spectral_bits[cb-1][curidx]; - vec = &ff_aac_codebook_vectors[cb-1][curidx*dim]; - if(IS_CODEBOOK_UNSIGNED(cb)){ - for(k = 0; k < dim; k++){ - float t = fabsf(in[i+k]); - float di; - //do not code with escape sequence small values - if(vec[k] == 64.0f && t < 39.0f*IQ){ - rd = INFINITY; - break; - } - if(vec[k] == 64.0f){//FIXME: slow - if(t >= 165140.0f*IQ){ // clipped value - di = t - 165140.0f; - curbits += 21; - }else{ - int c = realquants[k]; - di = t - c*cbrt(c)*IQ; - curbits += av_log2(c)*2 - 4 + 1; - } - }else{ - di = t - vec[k]*IQ; - } - if(vec[k] != 0.0f) - curbits++; - rd += di*di*lambda; - } - }else{ - for(k = 0; k < dim; k++){ - float di = in[i+k] - vec[k]*IQ; - rd += di*di*lambda; - } - } - rd += curbits; - if(rd < mincost){ - mincost = rd; - minidx = j; - minbits = curbits; - } - } - cost += mincost; - resbits += minbits; - if(cost >= uplim) - return uplim; - } - - if(bits) - *bits = resbits; - return cost; -} - -static void quantize_and_encode_band(PutBitContext *pb, const float *in, int size, - int scale_idx, int cb, const float lambda) -{ - const float IQ = ff_aac_pow2sf_tab[200 + scale_idx - SCALE_ONE_POS + SCALE_DIV_512]; - const float Q = ff_aac_pow2sf_tab[200 - scale_idx + SCALE_ONE_POS - SCALE_DIV_512]; - const int range = aac_cb_range[cb]; - const int maxval = aac_cb_maxval[cb]; - const int dim = (cb < FIRST_PAIR_BT) ? 4 : 2; - int i, j, k; - int offs[4]; - - if(!cb) - return; - - offs[0] = 1; - for(i = 1; i < dim; i++) - offs[i] = offs[i-1]*range; - for(i = 0; i < size; i += dim){ - float mincost; - int minidx = 0; - int minbits = 0; - int quants[4][2]; - mincost = 0.0f; - for(j = 0; j < dim; j++){ - quants[j][0] = av_clip(quant2(in[i+j], Q), -maxval, maxval); - quants[j][1] = av_clip(quant (in[i+j], Q), -maxval, maxval); - for(k = 0; k < 2; k++){ - quants[j][k] = FFMIN(quants[j][k], maxval); - if(!IS_CODEBOOK_UNSIGNED(cb) && in[i+j] < 0.0f) - quants[j][k] = -quants[j][k]; - } - mincost += in[i+j]*in[i+j]*lambda; - } - minidx = IS_CODEBOOK_UNSIGNED(cb) ? 0 : 40; - minbits = ff_aac_spectral_bits[cb-1][minidx]; - mincost += minbits; - for(j = 0; j < (1<<dim); j++){ - float rd = 0.0f; - int curbits; - int curidx = IS_CODEBOOK_UNSIGNED(cb) ? 0 : 40; - const float *vec; - int same = 0; - for(k = 0; k < dim; k++){ - if((j & (1 << k)) && quants[k][0] == quants[k][1]){ - same = 1; - break; - } - } - if(same) - continue; - for(k = 0; k < dim; k++) - curidx += quants[k][!!(j & (1 << k))] * offs[dim - 1 - k]; - curbits = ff_aac_spectral_bits[cb-1][curidx]; - vec = &ff_aac_codebook_vectors[cb-1][curidx*dim]; - if(IS_CODEBOOK_UNSIGNED(cb)){ - for(k = 0; k < dim; k++){ - float t = fabsf(in[i+k]); - float di; - //do not code with escape sequence small values - if(vec[k] == 64.0f && t < 39.0f*IQ){ - rd = INFINITY; - break; - } - if(vec[k] == 64.0f){//FIXME: slow - if(t >= 165140.0f*IQ){ // clipped value - di = t - 165140.0f; - curbits += 21; - }else{ - int c = quant(t, Q); - di = t - c*cbrt(c)*IQ; - curbits += av_log2(c)*2 - 4 + 1; - } - }else{ - di = t - vec[k]*IQ; - } - if(vec[k] != 0.0f) - curbits++; - rd += di*di*lambda; - } - }else{ - for(k = 0; k < dim; k++){ - float di = in[i+k] - vec[k]*IQ; - rd += di*di*lambda; - } - } - rd += curbits; - if(rd < mincost){ - mincost = rd; - minidx = curidx; - minbits = curbits; - } - } - put_bits(pb, ff_aac_spectral_bits[cb-1][minidx], ff_aac_spectral_codes[cb-1][minidx]); - if(IS_CODEBOOK_UNSIGNED(cb)) - for(j = 0; j < dim; j++) - if(ff_aac_codebook_vectors[cb-1][minidx*dim+j] != 0.0f) - put_bits(pb, 1, in[i+j] < 0.0f); - if(cb == ESC_BT){ - for(j = 0; j < 2; j++){ - if(ff_aac_codebook_vectors[cb-1][minidx*2+j] == 64.0f){ - int coef = quant(in[i+j], Q); - int len = av_log2(coef); - - put_bits(pb, len - 4 + 1, (1 << (len - 4 + 1)) - 2); - put_bits(pb, len, coef & ((1 << len) - 1)); - } - } - } - } -} - -#else - -/** - * Calculate rate distortion cost for quantizing with given codebook - * - * @return quantization distortion - */ -static float quantize_band_cost(const float *in, int size, int scale_idx, int cb, - const float lambda, const float uplim, int *bits) -{ - const float Q = ff_aac_pow2sf_tab[200 + scale_idx - SCALE_ONE_POS + SCALE_DIV_512]; - int i, j, k; - float cost = 0; - const int dim = cb < FIRST_PAIR_BT ? 4 : 2; - int resbits = 0; - - if(!cb){ - for(i = 0; i < size; i++) - cost += in[i]*in[i]*lambda; - return cost; - } - for(i = 0; i < size; i += dim){ - float mincost = INFINITY; - int minidx = 0; - int minbits = 0; - const float *vec = ff_aac_codebook_vectors[cb-1]; - for(j = 0; j < ff_aac_spectral_sizes[cb-1]; j++, vec += dim){ - float rd = 0.0f; - int curbits = ff_aac_spectral_bits[cb-1][j]; - if(IS_CODEBOOK_UNSIGNED(cb)){ - for(k = 0; k < dim; k++){ - float t = fabsf(in[i+k]); - float di; - //do not code with escape sequence small values - if(vec[k] == 64.0f && t < 39.0f*Q){ - rd = INFINITY; - break; - } - if(vec[k] == 64.0f){//FIXME: slow - if(t >= 165140.0f*Q){ // clipped value - di = t - 165140.0f; - curbits += 21; - }else{ - int c = quant(t, 1.0/Q); - di = t - c*cbrt(c)*Q; - curbits += av_log2(c)*2 - 4 + 1; - } - }else{ - di = t - vec[k]*Q; - } - if(vec[k] != 0.0f) - curbits++; - rd += di*di*lambda; - } - }else{ - for(k = 0; k < dim; k++){ - float di = in[i+k] - vec[k]*Q; - rd += di*di*lambda; - } - } - rd += curbits; - if(rd < mincost){ - mincost = rd; - minidx = j; - minbits = curbits; - } - } - cost += mincost; - resbits += minbits; - if(cost >= uplim) - return uplim; - } - - if(bits) - *bits = resbits; - return cost; -} - -/** - * Prepare coefficients for encoding. - * - * @return sum of coefficient absolute values - */ -static void quantize_and_encode_band(PutBitContext *pb, const float *in, int size, - int scale_idx, int cb, const float lambda) -{ - const float Q = ff_aac_pow2sf_tab[200 + scale_idx - SCALE_ONE_POS + SCALE_DIV_512]; - const float IQ = ff_aac_pow2sf_tab[200 - scale_idx + SCALE_ONE_POS - SCALE_DIV_512]; - int i, j, k; - const int dim = cb < FIRST_PAIR_BT ? 4 : 2; - - if(!cb) - return; - - for(i = 0; i < size; i += dim){ - float mincost = INFINITY; - int minidx = 0; - int minbits = 0; - const float *vec = ff_aac_codebook_vectors[cb-1]; - for(j = 0; j < ff_aac_spectral_sizes[cb-1]; j++, vec += dim){ - float rd = 0.0f; - int curbits = ff_aac_spectral_bits[cb-1][j]; - if(IS_CODEBOOK_UNSIGNED(cb)){ - for(k = 0; k < dim; k++){ - float t = fabsf(in[i+k]); - float di; - //do not code with escape sequence small values - if(vec[k] == 64.0f && t < 39.0f*Q){ - rd = INFINITY; - break; - } - if(vec[k] == 64.0f){//FIXME: slow - if(t >= 165140.0f*Q){ // clipped value - di = t - 165140.0f; - curbits += 21; - }else{ - int c = quant(t, IQ); - di = t - c*cbrt(c)*Q; - curbits += av_log2(c)*2 - 4 + 1; - } - }else{ - di = t - vec[k]*Q; - } - if(vec[k] != 0.0f) - curbits++; - rd += di*di*lambda; - } - }else{ - for(k = 0; k < dim; k++){ - float di = in[i+k] - vec[k]*Q; - rd += di*di*lambda; - } - } - rd += curbits; - if(rd < mincost){ - mincost = rd; - minidx = j; - minbits = curbits; - } - } - put_bits(pb, ff_aac_spectral_bits[cb-1][minidx], ff_aac_spectral_codes[cb-1][minidx]); - if(IS_CODEBOOK_UNSIGNED(cb)) - for(j = 0; j < dim; j++) - if(ff_aac_codebook_vectors[cb-1][minidx*dim+j] != 0.0f) - put_bits(pb, 1, in[i+j] < 0.0f); - if(cb == ESC_BT){ - for(j = 0; j < 2; j++){ - if(ff_aac_codebook_vectors[cb-1][minidx*2+j] == 64.0f){ - int coef = quant(in[i+j], IQ); - int len = av_log2(coef); - - put_bits(pb, len - 4 + 1, (1 << (len - 4 + 1)) - 2); - put_bits(pb, len, coef & ((1 << len) - 1)); - } - } - } - } -} - -#endif - -/** - * structure used in optimal codebook search - */ -typedef struct BandCodingPath { - int prev_idx; ///< pointer to the previous path point - int codebook; ///< codebook for coding band run - float cost; ///< path cost - int run; -} BandCodingPath; - -/** - * Encode band info for single window group bands. - */ -static void encode_window_bands_info(AACEncContext *s, SingleChannelElement *sce, - int win, int group_len) -{ - BandCodingPath path[120][12]; - int w, swb, cb, start, start2, size; - int i, j; - const int max_sfb = sce->ics.max_sfb; - const int run_bits = sce->ics.num_windows == 1 ? 5 : 3; - const int run_esc = (1 << run_bits) - 1; - int idx, ppos, count; - int stackrun[120], stackcb[120], stack_len; - - start = win*128; - for(cb = 0; cb < 12; cb++){ - path[0][cb].cost = 0.0f; - path[0][cb].prev_idx = -1; - path[0][cb].run = 0; - } - for(swb = 0; swb < max_sfb; swb++){ - start2 = start; - size = sce->ics.swb_sizes[swb]; - if(sce->zeroes[win*16 + swb]){ - for(cb = 0; cb < 12; cb++){ - path[swb+1][cb].prev_idx = cb; - path[swb+1][cb].cost = path[swb][cb].cost; - path[swb+1][cb].run = path[swb][cb].run + 1; - } - }else{ - float minrd = INFINITY; - int mincb = 0; - for(cb = 0; cb < 12; cb++){ - float rd = 0.0f; - for(w = 0; w < group_len; w++){ - FFPsyBand *band = &s->psy.psy_bands[s->cur_channel*PSY_MAX_BANDS+(win+w)*16+swb]; - rd += quantize_band_cost(sce->coeffs + start + w*128, size, - sce->sf_idx[(win+w)*16+swb], cb, - s->lambda / band->threshold, INFINITY, NULL); - } - if( run_value_bits[sce->ics.num_windows == 8][path[swb][cb].run] - != run_value_bits[sce->ics.num_windows == 8][path[swb][cb].run+1]) - rd += run_bits; - path[swb+1][cb].prev_idx = cb; - path[swb+1][cb].cost = path[swb][cb].cost + rd; - path[swb+1][cb].run = path[swb][cb].run + 1; - if(rd < minrd){ - minrd = rd; - mincb = cb; - } - } - for(cb = 0; cb < 12; cb++){ - float cost = path[swb][cb].cost + minrd + run_bits + 4; - if(cost < path[swb+1][cb].cost){ - path[swb+1][cb].prev_idx = mincb; - path[swb+1][cb].cost = cost; - path[swb+1][cb].run = 1; - } - } - } - start += sce->ics.swb_sizes[swb]; - } - - //convert resulting path from backward-linked list - stack_len = 0; - idx = 0; - for(cb = 1; cb < 12; cb++){ - if(path[max_sfb][cb].cost < path[max_sfb][idx].cost) - idx = cb; - } - ppos = max_sfb; - while(ppos > 0){ - cb = idx; - stackrun[stack_len] = path[ppos][cb].run; - stackcb [stack_len] = cb; - idx = path[ppos][cb].prev_idx; - ppos -= path[ppos][cb].run; - stack_len++; - } - //perform actual band info encoding - start = 0; - for(i = stack_len - 1; i >= 0; i--){ - put_bits(&s->pb, 4, stackcb[i]); - count = stackrun[i]; - memset(sce->zeroes + win*16 + start, !stackcb[i], count); - //XXX: memset when band_type is also uint8_t - for(j = 0; j < count; j++){ - sce->band_type[win*16 + start] = stackcb[i]; - start++; - } - while(count >= run_esc){ - put_bits(&s->pb, run_bits, run_esc); - count -= run_esc; - } - put_bits(&s->pb, run_bits, count); - } -} - -/** - * Produce integer coefficients from scalefactors provided by the model. - */ -static void adjust_frame_information(AACEncContext *apc, ChannelElement *cpe, int chans) -{ - int i, w, w2, g, ch; - int start, sum, maxsfb, cmaxsfb; - - for(ch = 0; ch < chans; ch++){ - IndividualChannelStream *ics = &cpe->ch[ch].ics; - start = 0; - maxsfb = 0; - cpe->ch[ch].pulse.num_pulse = 0; - for(w = 0; w < ics->num_windows*16; w += 16){ - for(g = 0; g < ics->num_swb; g++){ - sum = 0; - //apply M/S - if(!ch && cpe->ms_mask[w + g]){ - for(i = 0; i < ics->swb_sizes[g]; i++){ - cpe->ch[0].coeffs[start+i] = (cpe->ch[0].coeffs[start+i] + cpe->ch[1].coeffs[start+i]) / 2.0; - cpe->ch[1].coeffs[start+i] = cpe->ch[0].coeffs[start+i] - cpe->ch[1].coeffs[start+i]; - } - } - start += ics->swb_sizes[g]; - } - for(cmaxsfb = ics->num_swb; cmaxsfb > 0 && cpe->ch[ch].zeroes[w+cmaxsfb-1]; cmaxsfb--); - maxsfb = FFMAX(maxsfb, cmaxsfb); - } - ics->max_sfb = maxsfb; - - //adjust zero bands for window groups - for(w = 0; w < ics->num_windows; w += ics->group_len[w]){ - for(g = 0; g < ics->max_sfb; g++){ - i = 1; - for(w2 = w; w2 < w + ics->group_len[w]; w2++){ - if(!cpe->ch[ch].zeroes[w2*16 + g]){ - i = 0; - break; - } - } - cpe->ch[ch].zeroes[w*16 + g] = i; - } - } - } - - if(chans > 1 && cpe->common_window){ - IndividualChannelStream *ics0 = &cpe->ch[0].ics; - IndividualChannelStream *ics1 = &cpe->ch[1].ics; - int msc = 0; - ics0->max_sfb = FFMAX(ics0->max_sfb, ics1->max_sfb); - ics1->max_sfb = ics0->max_sfb; - for(w = 0; w < ics0->num_windows*16; w += 16) - for(i = 0; i < ics0->max_sfb; i++) - if(cpe->ms_mask[w+i]) msc++; - if(msc == 0 || ics0->max_sfb == 0) cpe->ms_mode = 0; - else cpe->ms_mode = msc < ics0->max_sfb ? 1 : 2; - } -} - -typedef struct TrellisPath { - float cost; - int prev; - int min_val; - int max_val; -} TrellisPath; - -static void search_for_quantizers_anmr(AACEncContext *s, SingleChannelElement *sce, const float lambda) -{ - int q, w, w2, g, start = 0; - int i; - int idx; - TrellisPath paths[256*121]; - int bandaddr[121]; - int minq; - float mincost; - - for(i = 0; i < 256; i++){ - paths[i].cost = 0.0f; - paths[i].prev = -1; - paths[i].min_val = i; - paths[i].max_val = i; - } - for(i = 256; i < 256*121; i++){ - paths[i].cost = INFINITY; - paths[i].prev = -2; - paths[i].min_val = INT_MAX; - paths[i].max_val = 0; - } - idx = 256; - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]){ - start = w*128; - for(g = 0; g < sce->ics.num_swb; g++){ - const float *coefs = sce->coeffs + start; - float qmin, qmax; - int nz = 0; - - bandaddr[idx >> 8] = w*16+g; - qmin = INT_MAX; - qmax = 0.0f; - for(w2 = 0; w2 < sce->ics.group_len[w]; w2++){ - FFPsyBand *band = &s->psy.psy_bands[s->cur_channel*PSY_MAX_BANDS+(w+w2)*16+g]; - if(band->energy <= band->threshold || band->threshold == 0.0f){ - sce->zeroes[(w+w2)*16+g] = 1; - continue; - } - sce->zeroes[(w+w2)*16+g] = 0; - nz = 1; - for(i = 0; i < sce->ics.swb_sizes[g]; i++){ - float t = fabsf(coefs[w2*128+i]); - if(t > 0.0f) qmin = fminf(qmin, t); - qmax = fmaxf(qmax, t); - } - } - if(nz){ - int minscale, maxscale; - float minrd = INFINITY; - //minimum scalefactor index is when minimum nonzero coefficient after quantizing is not clipped - minscale = av_clip_uint8(log2(qmin)*4 - 69 + SCALE_ONE_POS - SCALE_DIV_512); - //maximum scalefactor index is when maximum coefficient after quantizing is still not zero - maxscale = av_clip_uint8(log2(qmax)*4 + 6 + SCALE_ONE_POS - SCALE_DIV_512); - for(q = minscale; q < maxscale; q++){ - float dists[12], dist; - memset(dists, 0, sizeof(dists)); - for(w2 = 0; w2 < sce->ics.group_len[w]; w2++){ - FFPsyBand *band = &s->psy.psy_bands[s->cur_channel*PSY_MAX_BANDS+(w+w2)*16+g]; - int cb; - for(cb = 0; cb <= ESC_BT; cb++){ - dists[cb] += quantize_band_cost(coefs + w2*128, sce->ics.swb_sizes[g], - q, cb, s->lambda / band->threshold, INFINITY, NULL); - } - } - dist = dists[0]; - for(i = 1; i <= ESC_BT; i++) - dist = fminf(dist, dists[i]); - minrd = fminf(minrd, dist); - - for(i = FFMAX(q - SCALE_MAX_DIFF, 0); i < FFMIN(q + SCALE_MAX_DIFF, 256); i++){ - float cost; - int minv, maxv; - if(isinf(paths[idx - 256 + i].cost)) - continue; - cost = paths[idx - 256 + i].cost + dist - + ff_aac_scalefactor_bits[q - i + SCALE_DIFF_ZERO]; - minv = FFMIN(paths[idx - 256 + i].min_val, q); - maxv = FFMAX(paths[idx - 256 + i].max_val, q); - if(cost < paths[idx + q].cost && maxv-minv < SCALE_MAX_DIFF){ - paths[idx + q].cost = cost; - paths[idx + q].prev = idx - 256 + i; - paths[idx + q].min_val = minv; - paths[idx + q].max_val = maxv; - } - } - } - }else{ - for(q = 0; q < 256; q++){ - if(!isinf(paths[idx - 256 + q].cost)){ - paths[idx + q].cost = paths[idx - 256 + q].cost + 1; - paths[idx + q].prev = idx - 256 + q; - paths[idx + q].min_val = FFMIN(paths[idx - 256 + q].min_val, q); - paths[idx + q].max_val = FFMAX(paths[idx - 256 + q].max_val, q); - continue; - } - for(i = FFMAX(q - SCALE_MAX_DIFF, 0); i < FFMIN(q + SCALE_MAX_DIFF, 256); i++){ - float cost; - int minv, maxv; - if(isinf(paths[idx - 256 + i].cost)) - continue; - cost = paths[idx - 256 + i].cost + ff_aac_scalefactor_bits[q - i + SCALE_DIFF_ZERO]; - minv = FFMIN(paths[idx - 256 + i].min_val, q); - maxv = FFMAX(paths[idx - 256 + i].max_val, q); - if(cost < paths[idx + q].cost && maxv-minv < SCALE_MAX_DIFF){ - paths[idx + q].cost = cost; - paths[idx + q].prev = idx - 256 + i; - paths[idx + q].min_val = minv; - paths[idx + q].max_val = maxv; - } - } - } - } - sce->zeroes[w*16+g] = !nz; - start += sce->ics.swb_sizes[g]; - idx += 256; - } - } - idx -= 256; - mincost = paths[idx].cost; - minq = idx; - for(i = 1; i < 256; i++){ - if(paths[idx + i].cost < mincost){ - mincost = paths[idx + i].cost; - minq = idx + i; - } - } - while(minq >= 256){ - sce->sf_idx[bandaddr[minq>>8]] = minq & 0xFF; - minq = paths[minq].prev; - } - //set the same quantizers inside window groups - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) - for(g = 0; g < sce->ics.num_swb; g++) - for(w2 = 1; w2 < sce->ics.group_len[w]; w2++) - sce->sf_idx[(w+w2)*16+g] = sce->sf_idx[w*16+g]; -} - -/** - * two-loop quantizers search taken from ISO 13818-7 Appendix C - */ -static void search_for_quantizers_twoloop(AVCodecContext *avctx, AACEncContext *s, - SingleChannelElement *sce, const float lambda) -{ - int start = 0, i, w, w2, g; - int destbits = avctx->bit_rate * 1024.0 / avctx->sample_rate / avctx->channels; - float dists[128], uplims[128]; - int fflag, minscaler; - int its = 0; - int allz = 0; - float minthr = INFINITY; - - //XXX: some heuristic to determine initial quantizers will reduce search time - memset(dists, 0, sizeof(dists)); - //determine zero bands and upper limits - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]){ - for(g = 0; g < sce->ics.num_swb; g++){ - int nz = 0; - float uplim = 0.0f; - for(w2 = 0; w2 < sce->ics.group_len[w]; w2++){ - FFPsyBand *band = &s->psy.psy_bands[s->cur_channel*PSY_MAX_BANDS+(w+w2)*16+g]; - uplim += band->threshold; - if(band->energy <= band->threshold || band->threshold == 0.0f){ - sce->zeroes[(w+w2)*16+g] = 1; - continue; - } - nz = 1; - } - uplims[w*16+g] = uplim *512; - sce->zeroes[w*16+g] = !nz; - if(nz) - minthr = fminf(minthr, uplim); - allz = FFMAX(allz, nz); - } - } - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]){ - for(g = 0; g < sce->ics.num_swb; g++){ - if(sce->zeroes[w*16+g]){ - sce->sf_idx[w*16+g] = SCALE_ONE_POS; - continue; - } - sce->sf_idx[w*16+g] = SCALE_ONE_POS + fminf(log2(uplims[w*16+g]/minthr)*4,59); - } - } - - if(!allz) - return; - //perform two-loop search - //outer loop - improve quality - do{ - int tbits, qstep; - minscaler = sce->sf_idx[0]; - //inner loop - quantize spectrum to fit into given number of bits - qstep = its ? 1 : 32; - do{ - int prev = -1; - tbits = 0; - fflag = 0; - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]){ - start = w*128; - for(g = 0; g < sce->ics.num_swb; g++){ - const float *coefs = sce->coeffs + start; - int bits = 0; - int cb; - float mindist = INFINITY; - int minbits = 0; - - if(sce->zeroes[w*16+g] || sce->sf_idx[w*16+g] >= 218) - continue; - minscaler = FFMIN(minscaler, sce->sf_idx[w*16+g]); - for(cb = 0; cb <= ESC_BT; cb++){ - float dist = 0.0f; - int bb = 0; - for(w2 = 0; w2 < sce->ics.group_len[w]; w2++){ - int b; - dist += quantize_band_cost(coefs + w2*128, - sce->ics.swb_sizes[g], - sce->sf_idx[w*16+g], - ESC_BT, - 1.0, - INFINITY, - &b); - bb += b; - } - if(dist < mindist){ - mindist = dist; - minbits = bb; - } - } - dists[w*16+g] = mindist - minbits; - bits = minbits; - if(prev != -1){ - bits += ff_aac_scalefactor_bits[sce->sf_idx[w*16+g] - prev + SCALE_DIFF_ZERO]; - } - tbits += bits; - start += sce->ics.swb_sizes[g]; - prev = sce->sf_idx[w*16+g]; - } - } - if(tbits > destbits){ - for(i = 0; i < 128; i++){ - if(sce->sf_idx[i] < 218 - qstep){ - sce->sf_idx[i] += qstep; - } - } - }else{ - for(i = 0; i < 128; i++){ - if(sce->sf_idx[i] > 60 - qstep){ - sce->sf_idx[i] -= qstep; - } - } - } - qstep >>= 1; - if(!qstep && tbits > destbits*1.02) - qstep = 1; - if(sce->sf_idx[0] >= 217)break; - }while(qstep); - - fflag = 0; - minscaler = av_clip(minscaler, 60, 255 - SCALE_MAX_DIFF); - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]){ - start = w*128; - for(g = 0; g < sce->ics.num_swb; g++){ - int prevsc = sce->sf_idx[w*16+g]; - if(dists[w*16+g] > uplims[w*16+g] && sce->sf_idx[w*16+g] > 60) - sce->sf_idx[w*16+g]--; - sce->sf_idx[w*16+g] = av_clip(sce->sf_idx[w*16+g], minscaler, minscaler + SCALE_MAX_DIFF); - sce->sf_idx[w*16+g] = FFMIN(sce->sf_idx[w*16+g], 219); - if(sce->sf_idx[w*16+g] != prevsc) - fflag = 1; - } - } - its++; - }while(fflag && its < 10); -} - -static void search_for_quantizers_faac(AACEncContext *s, SingleChannelElement *sce, const float lambda) -{ - int start = 0, i, w, w2, g; - float uplim[128], maxq[128]; - int minq; -#if 0 - float distfact = ((sce->ics.num_windows > 1) ? 85.80 : 147.84) / lambda; - int last = 0, lastband = 0, curband = 0; - float avg_energy = 0.0; - if(sce->ics.num_windows == 1){ - start = 0; - for(i = 0; i < 1024; i++){ - if(i - start >= sce->ics.swb_sizes[curband]){ - start += sce->ics.swb_sizes[curband]; - curband++; - } - if(sce->coeffs[i]){ - avg_energy += sce->coeffs[i] * sce->coeffs[i]; - last = i; - lastband = curband; - } - } - }else{ - for(w = 0; w < 8; w++){ - const float *coeffs = sce->coeffs + w*128; - start = 0; - for(i = 0; i < 128; i++){ - if(i - start >= sce->ics.swb_sizes[curband]){ - start += sce->ics.swb_sizes[curband]; - curband++; - } - if(coeffs[i]){ - avg_energy += coeffs[i] * coeffs[i]; - last = FFMAX(last, i); - lastband = FFMAX(lastband, curband); - } - } - } - } - last++; - avg_energy /= last; - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]){ - start = w*128; - for(g = 0; g < sce->ics.num_swb; g++){ - float *coefs = sce->coeffs + start; - const int size = sce->ics.swb_sizes[g]; - int start2 = start, end2 = start + size, peakpos = start; - float maxval = -1, thr = 0.0f, t; - maxq[w*16+g] = 0.0f; - if(g > lastband){ - maxq[w*16+g] = 0.0f; - start += size; - for(w2 = 0; w2 < sce->ics.group_len[w]; w2++) - memset(coefs + w2*128, 0, sizeof(coefs[0])*size); - continue; - } - for(w2 = 0; w2 < sce->ics.group_len[w]; w2++){ - for(i = 0; i < size; i++){ - float t = coefs[w2*128+i]*coefs[w2*128+i]; - maxq[w*16+g] = fmaxf(maxq[w*16+g], coefs[w2*128 + i]); - thr += t; - if(sce->ics.num_windows == 1 && maxval < t){ - maxval = t; - peakpos = start+i; - } - } - } - if(sce->ics.num_windows == 1){ - start2 = FFMAX(peakpos - 2, start2); - end2 = FFMIN(peakpos + 3, end2); - } - start += size; - thr = pow(thr / (avg_energy * (end2 - start2)), 0.3 + 0.1*(lastband - g) / lastband); - t = 1.0 - (1.0 * start2 / last); - uplim[w*16+g] = distfact / (1.4 * thr + t*t*t + 0.075); - } - } -#else - memset(maxq, 0, sizeof(maxq)); - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]){ - start = w*128; - for(g = 0; g < sce->ics.num_swb; g++){ - float *coefs = sce->coeffs + start; - const int size = sce->ics.swb_sizes[g]; - float thr = 0.0f; - for(w2 = 0; w2 < sce->ics.group_len[w]; w2++){ - FFPsyBand *band = &s->psy.psy_bands[s->cur_channel*PSY_MAX_BANDS+(w+w2)*16+g]; - thr += band->threshold; - for(i = 0; i < size; i++){ - maxq[w*16+g] = fmaxf(maxq[w*16+g], fabsf(coefs[w2*128 + i])); - } - } - uplim[w*16+g] = thr / lambda; - start += size; - } - } -#endif - memset(sce->sf_idx, 0, sizeof(sce->sf_idx)); - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]){ - start = w*128; - for(g = 0; g < sce->ics.num_swb; g++){ - const float *coefs = sce->coeffs + start; - const int size = sce->ics.swb_sizes[g]; - int scf, prev_scf, step; - if(maxq[w*16+g] < 21.544){ - sce->zeroes[w*16+g] = 1; - continue; - } - sce->zeroes[w*16+g] = 0; - scf = prev_scf = av_clip(SCALE_ONE_POS + log2(1/maxq[w*16+g])*16/3, 60, 218); - step = 16; - for(;;){ - float dist = 0.0f, t; - int quant_max; - - for(w2 = 0; w2 < sce->ics.group_len[w]; w2++){ - int b; - dist += quantize_band_cost(coefs + w2*128, - sce->ics.swb_sizes[g], - scf, - ESC_BT, - 1.0, - INFINITY, - &b); - dist -= b; - } - quant_max = quant(maxq[w*16+g], ff_aac_pow2sf_tab[200 - scf + SCALE_ONE_POS]); - if(quant_max >= 8191){ // too much, return to the previous quantizer - sce->sf_idx[w*16+g] = prev_scf; - break; - } - prev_scf = scf; - t = pow(dist / size, -2.0/3.0); - if(FFABS(step) > 4){ - int newstep = 4*log(t / maxq[w*16+g])/log(2) - 2; - step = av_clip(-newstep, -4, 4); - } - if(FFABS(step) >= 4){ - step = 1 - step; - scf += step; - }else if(t >= uplim[w*16+g]){ - scf -= FFABS(step); - }else{ - sce->sf_idx[w*16+g] = scf; - break; - } - } - start += size; - } - } - minq = sce->sf_idx[0] ? sce->sf_idx[0] : INT_MAX; - for(i = 1; i < 128; i++){ - if(!sce->sf_idx[i]) - sce->sf_idx[i] = sce->sf_idx[i-1]; - else - minq = FFMIN(minq, sce->sf_idx[i]); - } - if(minq == INT_MAX) minq = 0; - for(i = 126; i >= 0; i--){ - if(!sce->sf_idx[i]) - sce->sf_idx[i] = sce->sf_idx[i+1]; - sce->sf_idx[i] = av_clip(sce->sf_idx[i], minq, minq + SCALE_MAX_DIFF); - } -} - -static void search_for_quantizers_fast(AACEncContext *s, SingleChannelElement *sce) -{ - int start = 0, i, w, w2, g; - int minq = 255; - - memset(sce->sf_idx, 0, sizeof(sce->sf_idx)); - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]){ - start = w*128; - for(g = 0; g < sce->ics.num_swb; g++){ - for(w2 = 0; w2 < sce->ics.group_len[w]; w2++){ - FFPsyBand *band = &s->psy.psy_bands[s->cur_channel*PSY_MAX_BANDS+(w+w2)*16+g]; - if(band->energy <= band->threshold){ - sce->sf_idx[(w+w2)*16+g] = 218; - sce->zeroes[(w+w2)*16+g] = 1; - }else{ - sce->sf_idx[(w+w2)*16+g] = av_clip(SCALE_ONE_POS - SCALE_DIV_512 + log2(band->threshold), 80, 218); - sce->zeroes[(w+w2)*16+g] = 0; - } - minq = FFMIN(minq, sce->sf_idx[(w+w2)*16+g]); - } - } - } - for(i = 0; i < 128; i++){ - sce->sf_idx[i] = av_clip(sce->sf_idx[i], minq, minq + SCALE_MAX_DIFF - 1); - } - //set the same quantizers inside window groups - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) - for(g = 0; g < sce->ics.num_swb; g++) - for(w2 = 1; w2 < sce->ics.group_len[w]; w2++) - sce->sf_idx[(w+w2)*16+g] = sce->sf_idx[w*16+g]; -} - -static void search_for_quantizers(AVCodecContext *avctx, AACEncContext *s, - SingleChannelElement *sce) -{ - search_for_quantizers_anmr(s, sce, s->lambda); -// search_for_quantizers_twoloop(avctx, s, sce, s->lambda); -// search_for_quantizers_faac(s, sce, av_clipf(s->lambda * 2e8, 50.0f, 300.0f)); -// search_for_quantizers_fast(s, sce); -} - -static void search_for_ms(AACEncContext *s, ChannelElement *cpe, const float lambda) -{ - int start = 0, i, w, w2, g; - float M[128], S[128]; - SingleChannelElement *sce0 = &cpe->ch[0]; - SingleChannelElement *sce1 = &cpe->ch[1]; - if(!cpe->common_window) - return; - for(w = 0; w < sce0->ics.num_windows; w += sce0->ics.group_len[w]){ - for(g = 0; g < sce0->ics.num_swb; g++){ - if(!cpe->ch[0].zeroes[w*16+g] && !cpe->ch[1].zeroes[w*16+g]){ - float dist1 = 0.0f, dist2 = 0.0f; - for(w2 = 0; w2 < sce0->ics.group_len[w]; w2++){ - FFPsyBand *band0 = &s->psy.psy_bands[(s->cur_channel+0)*PSY_MAX_BANDS+(w+w2)*16+g]; - FFPsyBand *band1 = &s->psy.psy_bands[(s->cur_channel+1)*PSY_MAX_BANDS+(w+w2)*16+g]; - float minthr = fminf(band0->threshold, band1->threshold); - float maxthr = fmaxf(band0->threshold, band1->threshold); - for(i = 0; i < sce0->ics.swb_sizes[g]; i++){ - M[i] = (sce0->coeffs[start+w2*128+i] - + sce1->coeffs[start+w2*128+i])*0.5; - S[i] = sce0->coeffs[start+w2*128+i] - - sce1->coeffs[start+w2*128+i]; - } - dist1 += quantize_band_cost(sce0->coeffs + start + w2*128, - sce0->ics.swb_sizes[g], - sce0->sf_idx[(w+w2)*16+g], - sce0->band_type[(w+w2)*16+g], - lambda / band0->threshold, INFINITY, NULL); - dist1 += quantize_band_cost(sce1->coeffs + start + w2*128, - sce1->ics.swb_sizes[g], - sce1->sf_idx[(w+w2)*16+g], - sce1->band_type[(w+w2)*16+g], - lambda / band1->threshold, INFINITY, NULL); - dist2 += quantize_band_cost(M, - sce0->ics.swb_sizes[g], - sce0->sf_idx[(w+w2)*16+g], - sce0->band_type[(w+w2)*16+g], - lambda / maxthr, INFINITY, NULL); - dist2 += quantize_band_cost(S, - sce1->ics.swb_sizes[g], - sce1->sf_idx[(w+w2)*16+g], - sce1->band_type[(w+w2)*16+g], - lambda / minthr, INFINITY, NULL); - } - cpe->ms_mask[w*16+g] = dist2 < dist1; - } - start += sce0->ics.swb_sizes[g]; - } - } -} - -/** - * Encode scalefactor band coding type. - */ -static void encode_band_info(AACEncContext *s, SingleChannelElement *sce) -{ - int w; - - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]){ - encode_window_bands_info(s, sce, w, sce->ics.group_len[w]); - } -} - -/** - * Encode scalefactors. - */ -static void encode_scale_factors(AVCodecContext *avctx, AACEncContext *s, SingleChannelElement *sce) -{ - int off = sce->sf_idx[0], diff; - int i, w; - - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]){ - for(i = 0; i < sce->ics.max_sfb; i++){ - if(!sce->zeroes[w*16 + i]){ - diff = sce->sf_idx[w*16 + i] - off + SCALE_DIFF_ZERO; - if(diff < 0 || diff > 120) av_log(avctx, AV_LOG_ERROR, "Scalefactor difference is too big to be coded\n"); - off = sce->sf_idx[w*16 + i]; - put_bits(&s->pb, ff_aac_scalefactor_bits[diff], ff_aac_scalefactor_code[diff]); - } - } - } -} - -/** - * Encode pulse data. - */ -static void encode_pulses(AACEncContext *s, Pulse *pulse) -{ - int i; - - put_bits(&s->pb, 1, !!pulse->num_pulse); - if(!pulse->num_pulse) return; - - put_bits(&s->pb, 2, pulse->num_pulse - 1); - put_bits(&s->pb, 6, pulse->start); - for(i = 0; i < pulse->num_pulse; i++){ - put_bits(&s->pb, 5, pulse->pos[i]); - put_bits(&s->pb, 4, pulse->amp[i]); - } -} - -/** - * Encode spectral coefficients processed by psychoacoustic model. - */ -static void encode_spectral_coeffs(AACEncContext *s, SingleChannelElement *sce) -{ - int start, i, w, w2; - - for(w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]){ - start = 0; - for(i = 0; i < sce->ics.max_sfb; i++){ - if(sce->zeroes[w*16 + i]){ - start += sce->ics.swb_sizes[i]; - continue; - } - for(w2 = w; w2 < w + sce->ics.group_len[w]; w2++){ - quantize_and_encode_band(&s->pb, sce->coeffs + start + w2*128, - sce->ics.swb_sizes[i], - sce->sf_idx[w*16 + i], - sce->band_type[w*16 + i], - s->lambda); - } - start += sce->ics.swb_sizes[i]; - } - } -} - -/** - * Encode one channel of audio data. - */ -static int encode_individual_channel(AVCodecContext *avctx, AACEncContext *s, SingleChannelElement *sce, int common_window) -{ - put_bits(&s->pb, 8, sce->sf_idx[0]); - if(!common_window) put_ics_info(s, &sce->ics); - encode_band_info(s, sce); - encode_scale_factors(avctx, s, sce); - encode_pulses(s, &sce->pulse); - put_bits(&s->pb, 1, 0); //tns - put_bits(&s->pb, 1, 0); //ssr - encode_spectral_coeffs(s, sce); - return 0; -} - -/** - * Write some auxiliary information about the created AAC file. - */ -static void put_bitstream_info(AVCodecContext *avctx, AACEncContext *s, const char *name) -{ - int i, namelen, padbits; - - namelen = strlen(name) + 2; - put_bits(&s->pb, 3, TYPE_FIL); - put_bits(&s->pb, 4, FFMIN(namelen, 15)); - if(namelen >= 15) - put_bits(&s->pb, 8, namelen - 16); - put_bits(&s->pb, 4, 0); //extension type - filler - padbits = 8 - (put_bits_count(&s->pb) & 7); - align_put_bits(&s->pb); - for(i = 0; i < namelen - 2; i++) - put_bits(&s->pb, 8, name[i]); - put_bits(&s->pb, 12 - padbits, 0); -} - -static int aac_encode_frame(AVCodecContext *avctx, - uint8_t *frame, int buf_size, void *data) -{ - AACEncContext *s = avctx->priv_data; - int16_t *samples = s->samples, *samples2, *la; - ChannelElement *cpe; - int i, j, chans, tag, start_ch; - const uint8_t *chan_map = aac_chan_configs[avctx->channels-1]; - int chan_el_counter[4]; - - if(s->last_frame) - return 0; - if(data){ - if(!s->psypp){ - memcpy(s->samples + 1024 * avctx->channels, data, 1024 * avctx->channels * sizeof(s->samples[0])); - }else{ - start_ch = 0; - samples2 = s->samples + 1024 * avctx->channels; - for(i = 0; i < chan_map[0]; i++){ - tag = chan_map[i+1]; - chans = tag == TYPE_CPE ? 2 : 1; - ff_psy_preprocess(s->psypp, (uint16_t*)data + start_ch, samples2 + start_ch, start_ch + i, chans); - start_ch += chans; - } - } - } - if(!avctx->frame_number){ - memcpy(s->samples, s->samples + 1024 * avctx->channels, 1024 * avctx->channels * sizeof(s->samples[0])); - return 0; - } - - init_put_bits(&s->pb, frame, buf_size*8); - if((avctx->frame_number & 0xFF)==1 && !(avctx->flags & CODEC_FLAG_BITEXACT)){ - put_bitstream_info(avctx, s, LIBAVCODEC_IDENT); - } - start_ch = 0; - memset(chan_el_counter, 0, sizeof(chan_el_counter)); - for(i = 0; i < chan_map[0]; i++){ - FFPsyWindowInfo wi[2]; - tag = chan_map[i+1]; - chans = tag == TYPE_CPE ? 2 : 1; - cpe = &s->cpe[i]; - samples2 = samples + start_ch; - la = samples2 + 1024 * avctx->channels + start_ch; - if(!data) la = NULL; - s->lambda = 5e-7f; - for(j = 0; j < chans; j++){ - IndividualChannelStream *ics = &cpe->ch[j].ics; - int k; - wi[j] = ff_psy_suggest_window(&s->psy, samples2, la, start_ch + j, ics->window_sequence[0]); - ics->window_sequence[1] = ics->window_sequence[0]; - ics->window_sequence[0] = wi[j].window_type[0]; - ics->use_kb_window[1] = ics->use_kb_window[0]; - ics->use_kb_window[0] = wi[j].window_shape; - ics->num_windows = wi[j].num_windows; - ics->swb_sizes = s->psy.bands [ics->num_windows == 8]; - ics->num_swb = s->psy.num_bands[ics->num_windows == 8]; - for(k = 0; k < ics->num_windows; k++) - ics->group_len[k] = wi[j].grouping[k]; - - s->cur_channel = start_ch + j; - apply_window_and_mdct(avctx, s, &cpe->ch[j], samples2, j); - search_for_quantizers(avctx, s, &cpe->ch[j]); - } - cpe->common_window = 0; - if(chans > 1 - && wi[0].window_type[0] == wi[1].window_type[0] - && wi[0].window_shape == wi[1].window_shape){ - - cpe->common_window = 1; - for(j = 0; j < wi[0].num_windows; j++){ - if(wi[0].grouping[j] != wi[1].grouping[j]){ - cpe->common_window = 0; - break; - } - } - } -// search_for_ms(s, cpe, s->lambda); - adjust_frame_information(s, cpe, chans); - put_bits(&s->pb, 3, tag); - put_bits(&s->pb, 4, chan_el_counter[tag]++); - if(chans == 2){ - put_bits(&s->pb, 1, cpe->common_window); - if(cpe->common_window){ - put_ics_info(s, &cpe->ch[0].ics); - encode_ms_info(&s->pb, cpe); - } - } - for(j = 0; j < chans; j++){ - s->cur_channel = start_ch + j; - ff_psy_set_band_info(&s->psy, s->cur_channel, cpe->ch[j].coeffs, &wi[j]); - encode_individual_channel(avctx, s, &cpe->ch[j], cpe->common_window); - } - start_ch += chans; - } - - put_bits(&s->pb, 3, TYPE_END); - flush_put_bits(&s->pb); - avctx->frame_bits = put_bits_count(&s->pb); - - if(!data) - s->last_frame = 1; - memcpy(s->samples, s->samples + 1024 * avctx->channels, 1024 * avctx->channels * sizeof(s->samples[0])); - return put_bits_count(&s->pb)>>3; -} - -static av_cold int aac_encode_end(AVCodecContext *avctx) -{ - AACEncContext *s = avctx->priv_data; - - ff_mdct_end(&s->mdct1024); - ff_mdct_end(&s->mdct128); - ff_psy_end(&s->psy); - ff_psy_preprocess_end(s->psypp); - av_freep(&s->samples); - av_freep(&s->cpe); - return 0; -} - -AVCodec aac_encoder = { - "aac", - CODEC_TYPE_AUDIO, - CODEC_ID_AAC, - sizeof(AACEncContext), - aac_encode_init, - aac_encode_frame, - aac_encode_end, - .capabilities = CODEC_CAP_SMALL_LAST_FRAME | CODEC_CAP_DELAY, - .sample_fmts = (enum SampleFormat[]){SAMPLE_FMT_S16,SAMPLE_FMT_NONE}, - .long_name = NULL_IF_CONFIG_SMALL("Advanced Audio Coding"), -}; +#endif /* AVCODEC_AACENC_H */

1 0

[soc]: r4007 - aacenc/aac.h
by kostya 26 Jan '09

26 Jan '09

Author: kostya Date: Mon Jan 26 15:00:57 2009 New Revision: 4007 Log: Sync aac.h to the current FFmpeg SVN Modified: aacenc/aac.h Modified: aacenc/aac.h ============================================================================== --- aacenc/aac.h Mon Jan 26 14:37:36 2009 (r4006) +++ aacenc/aac.h Mon Jan 26 15:00:57 2009 (r4007) @@ -46,7 +46,6 @@ #define MAX_ELEM_ID 16 #define TNS_MAX_ORDER 20 -#define PNS_MEAN_ENERGY 3719550720.0f // sqrt(3.0) * 1<<31 enum AudioObjectType { AOT_NULL, @@ -133,6 +132,20 @@ enum CouplingPoint { AFTER_IMDCT = 3, }; +/** + * Predictor State + */ +typedef struct { + float cor0; + float cor1; + float var0; + float var1; + float r0; + float r1; +} PredictorState; + +#define MAX_PREDICTORS 672 + #define SCALE_DIV_512 36 ///< scalefactor difference that corresponds to scale difference in 512 times #define SCALE_ONE_POS 140 ///< scalefactor index that corresponds to scale=1.0 #define SCALE_MAX_POS 255 ///< scalefactor index maximum value @@ -153,6 +166,10 @@ typedef struct { int num_swb; ///< number of scalefactor window bands int num_windows; int tns_max_bands; + int predictor_present; + int predictor_initialized; + int predictor_reset_group; + uint8_t prediction_used[41]; } IndividualChannelStream; /** @@ -198,8 +215,8 @@ typedef struct { int num_coupled; ///< number of target elements enum RawDataBlockType type[8]; ///< Type of channel element to be coupled - SCE or CPE. int id_select[8]; ///< element id - int ch_select[8]; /**< [0] shared list of gains; [1] list of gains for left channel; - * [2] list of gains for right channel; [3] lists of gains for both channels + int ch_select[8]; /**< [0] shared list of gains; [1] list of gains for right channel; + * [2] list of gains for left channel; [3] lists of gains for both channels */ float gain[16][120]; } ChannelCoupling; @@ -220,6 +237,7 @@ typedef struct { DECLARE_ALIGNED_16(float, saved[1024]); ///< overlap DECLARE_ALIGNED_16(float, ret[1024]); ///< PCM output DECLARE_ALIGNED_16(int, icoefs[1024]); ///< integer coefficients for encoding + PredictorState predictor_state[MAX_PREDICTORS]; } SingleChannelElement; /** @@ -284,6 +302,7 @@ typedef struct { int sf_offset; ///< offset into pow2sf_tab as appropriate for dsp.float_to_int16 /** @} */ + DECLARE_ALIGNED(16, float, temp[128]); } AACContext; #endif /* AVCODEC_AAC_H */

1 0

[soc]: r4006 - aacenc/aacenc.c
by kostya 26 Jan '09

26 Jan '09

Author: kostya Date: Mon Jan 26 14:37:36 2009 New Revision: 4006 Log: Sync AAC encoder to current FFmpeg CONFIG_ handling policy Modified: aacenc/aacenc.c Modified: aacenc/aacenc.c ============================================================================== --- aacenc/aacenc.c Sun Jan 25 16:56:43 2009 (r4005) +++ aacenc/aacenc.c Mon Jan 26 14:37:36 2009 (r4006) @@ -232,7 +232,7 @@ static av_cold int aac_encode_init(AVCod ff_psy_init(&s->psy, avctx, 2, sizes, lengths); s->psypp = ff_psy_preprocess_init(avctx); -#ifndef CONFIG_HARDCODED_TABLES +#if !CONFIG_HARDCODED_TABLES for (i = 0; i < 316; i++) ff_aac_pow2sf_tab[i] = pow(2, (i - 200)/4.); #endif /* CONFIG_HARDCODED_TABLES */

1 0

[soc]: r4005 - in jpeg2000: j2kdec.c j2kenc.c
by diego 25 Jan '09

25 Jan '09

Author: diego Date: Sun Jan 25 16:56:43 2009 New Revision: 4005 Log: Add av_cold to initialization functions. Modified: jpeg2000/j2kdec.c jpeg2000/j2kenc.c Modified: jpeg2000/j2kdec.c ============================================================================== --- jpeg2000/j2kdec.c Sun Jan 25 16:55:31 2009 (r4004) +++ jpeg2000/j2kdec.c Sun Jan 25 16:56:43 2009 (r4005) @@ -924,7 +924,7 @@ static int decode_frame(AVCodecContext * return s->buf - s->buf_start; } -static int j2kdec_init(AVCodecContext *avctx) +static av_cold int j2kdec_init(AVCodecContext *avctx) { J2kDecoderContext *s = avctx->priv_data; Modified: jpeg2000/j2kenc.c ============================================================================== --- jpeg2000/j2kenc.c Sun Jan 25 16:55:31 2009 (r4004) +++ jpeg2000/j2kenc.c Sun Jan 25 16:56:43 2009 (r4005) @@ -960,7 +960,7 @@ static int encode_frame(AVCodecContext * return s->buf - s->buf_start; } -static int j2kenc_init(AVCodecContext *avctx) +static av_cold int j2kenc_init(AVCodecContext *avctx) { int i, ret; J2kEncoderContext *s = avctx->priv_data;

1 0

[soc]: r4004 - in dirac/libavcodec: diracdec.c diracenc.c
by diego 25 Jan '09

25 Jan '09

Author: diego Date: Sun Jan 25 16:55:31 2009 New Revision: 4004 Log: Add av_cold to initialization functions. Modified: dirac/libavcodec/diracdec.c dirac/libavcodec/diracenc.c Modified: dirac/libavcodec/diracdec.c ============================================================================== --- dirac/libavcodec/diracdec.c Sun Jan 25 16:54:19 2009 (r4003) +++ dirac/libavcodec/diracdec.c Sun Jan 25 16:55:31 2009 (r4004) @@ -79,7 +79,7 @@ static int add_frame(AVFrame *(*framelis return -1; } -static int decode_init(AVCodecContext *avctx) +static av_cold int decode_init(AVCodecContext *avctx) { DiracContext *s = avctx->priv_data; Modified: dirac/libavcodec/diracenc.c ============================================================================== --- dirac/libavcodec/diracenc.c Sun Jan 25 16:54:19 2009 (r4003) +++ dirac/libavcodec/diracenc.c Sun Jan 25 16:55:31 2009 (r4004) @@ -37,7 +37,7 @@ #include "dirac_wavelet.h" #include "mpeg12data.h" -static int encode_init(AVCodecContext *avctx) +static av_cold int encode_init(AVCodecContext *avctx) { DiracContext *s = avctx->priv_data; av_log_set_level(AV_LOG_DEBUG);

1 0

[soc]: r4003 - in amr: amrnbdec.c amrnbfloatdec.c
by diego 25 Jan '09

25 Jan '09

Author: diego Date: Sun Jan 25 16:54:19 2009 New Revision: 4003 Log: Add av_cold to initialization functions. Modified: amr/amrnbdec.c amr/amrnbfloatdec.c Modified: amr/amrnbdec.c ============================================================================== --- amr/amrnbdec.c Sun Jan 25 16:41:13 2009 (r4002) +++ amr/amrnbdec.c Sun Jan 25 16:54:19 2009 (r4003) @@ -1286,7 +1286,7 @@ static int amrnb_decode_frame(AVCodecCon } -static int amrnb_decode_close(AVCodecContext *avctx) +static av_cold int amrnb_decode_close(AVCodecContext *avctx) { AMRContext *p = avctx->priv_data; Modified: amr/amrnbfloatdec.c ============================================================================== --- amr/amrnbfloatdec.c Sun Jan 25 16:41:13 2009 (r4002) +++ amr/amrnbfloatdec.c Sun Jan 25 16:54:19 2009 (r4003) @@ -122,7 +122,7 @@ static void reset_state(AMRContext *p) } } -static int amrnb_decode_init(AVCodecContext *avctx) +static av_cold int amrnb_decode_init(AVCodecContext *avctx) { AMRContext *p = avctx->priv_data;

1 0

[soc]: r4002 - amr/amrnbdec.c
by diego 25 Jan '09

25 Jan '09

Author: diego Date: Sun Jan 25 16:41:13 2009 New Revision: 4002 Log: Add missing codec long_name declaration. Modified: amr/amrnbdec.c Modified: amr/amrnbdec.c ============================================================================== --- amr/amrnbdec.c Sun Jan 25 16:40:55 2009 (r4001) +++ amr/amrnbdec.c Sun Jan 25 16:41:13 2009 (r4002) @@ -1308,4 +1308,5 @@ AVCodec amrnb_decoder = { .init = amrnb_decode_init, .close = amrnb_decode_close, .decode = amrnb_decode_frame, + .long_name = NULL_IF_CONFIG_SMALL("Adaptive Multi-Rate NarrowBand"), };

1 0

[soc]: r4001 - amr/amrnbfloatdec.c
by diego 25 Jan '09

25 Jan '09

Author: diego Date: Sun Jan 25 16:40:55 2009 New Revision: 4001 Log: Add missing NULL_IF_CONFIG_SMALL to codec long_name declaration. Modified: amr/amrnbfloatdec.c Modified: amr/amrnbfloatdec.c ============================================================================== --- amr/amrnbfloatdec.c Sun Jan 25 16:38:52 2009 (r4000) +++ amr/amrnbfloatdec.c Sun Jan 25 16:40:55 2009 (r4001) @@ -1330,6 +1330,6 @@ AVCodec amrnb_decoder = { .init = amrnb_decode_init, .close = amrnb_decode_close, .decode = amrnb_decode_frame, - .long_name = "Adaptive Multi-Rate NarrowBand", + .long_name = NULL_IF_CONFIG_SMALL("Adaptive Multi-Rate NarrowBand"), };

1 0

[soc]: r4000 - in amr: amrdata.h amrnbdec.c amrnbfloatdata.h amrnbfloatdec.c
by diego 25 Jan '09

25 Jan '09

Author: diego Date: Sun Jan 25 16:38:52 2009 New Revision: 4000 Log: spelling/wording/grammar cosmetics Modified: amr/amrdata.h amr/amrnbdec.c amr/amrnbfloatdata.h amr/amrnbfloatdec.c Modified: amr/amrdata.h ============================================================================== --- amr/amrdata.h Sun Jan 25 16:35:48 2009 (r3999) +++ amr/amrdata.h Sun Jan 25 16:38:52 2009 (r4000) @@ -108,7 +108,8 @@ int16_t cos_table[65]; // cosine table u #define TRACKS_MODE_102 4 // number of tracks for MODE_102 -// FIXME - not quite sure how i'm going to implement this yet and some other variables may be moved around later +/* FIXME - not quite sure how i'm going to implement this yet + * and some other variables may be moved around later */ typedef struct AMRDecoderState { // Bad frame int state; @@ -244,7 +245,7 @@ static const AMROrder order_MODE_DTX[35] // encoder homing frame pattern #define EHF_MASK 0x0008 -// Homing frames for the decoder +// homing frames for the decoder static const int16_t dhf_MODE_475[PRMS_MODE_475] = { 0x00F8,0x009D,0x001C,0x0066,0x0000,0x0003,0x0028,0x000F,0x0038,0x0001, 0x000F,0x0031,0x0002,0x0008,0x000F,0x0026,0x0003 @@ -376,7 +377,7 @@ static const uint16_t pred_fac[10] = { 11651, 10588, 9767, 8593, 6484 }; -// vector quantised lsf tables +// vector quantized lsf tables static const int16_t lsf_3_MODE_515[128][4] = { { 419, 163, -30, -262}, { -455, -789,-1430, -721}, { 1006, 664, 269, 25}, { 619, 260, 183, 96}, { -968,-1358, -388, 135}, { -693, 835, 456, 154}, { 1105, 703, 569, 363}, { 1625, 1326, 985, 748}, @@ -1157,7 +1158,7 @@ static const uint16_t pitch_gain_attenua 6553 }; -// quantisation table for the adaptive codebook gain (MODE_795, MODE_122) +// quantization table for the adaptive codebook gain (MODE_795, MODE_122) #define N_PITCH_QUANTS 16 static const uint16_t pitch_gain_quant[N_PITCH_QUANTS] = { 0, Modified: amr/amrnbdec.c ============================================================================== --- amr/amrnbdec.c Sun Jan 25 16:35:48 2009 (r3999) +++ amr/amrnbdec.c Sun Jan 25 16:38:52 2009 (r4000) @@ -41,18 +41,18 @@ typedef struct AMRContext { GetBitContext gb; int16_t *sample_buffer; - int16_t *amr_prms; ///< pointer to the decoded amr parameters (lsf coefficients, codebook indices, etc) + int16_t *amr_prms; ///< pointer to the decoded AMR parameters (lsf coefficients, codebook indexes, etc) int bad_frame_indicator; ///< bad frame ? 1 : 0 int prev_frame_homing; ///< previous frame was a homing frame ? 1 : 0 enum RXFrameType prev_frame_type; ///< frame type of previous frame enum Mode prev_frame_mode; ///< mode of previous frame - int *prev_lsf_dq; ///< previous dequantised lsfs - int *prev_residual_q; ///< previous quantised residual + int *prev_lsf_dq; ///< previous dequantized lsfs + int *prev_residual_q; ///< previous quantized residual - int lsp1_q[LP_FILTER_ORDER]; ///< vector of quantised lsps - int lsp2_q[LP_FILTER_ORDER]; ///< vector of quantised lsps + int lsp1_q[LP_FILTER_ORDER]; ///< vector of quantized lsps + int lsp2_q[LP_FILTER_ORDER]; ///< vector of quantized lsps int cur_frame_homing; ///< current frame homing ? 1 : 0 enum RXFrameType cur_frame_type; ///< current frame type @@ -96,7 +96,7 @@ static int amrnb_decode_init(AVCodecCont // allocate and zero the sample buffer p->sample_buffer = av_mallocz(sizeof(int16_t)*AMR_BLOCK_SIZE); - // allocate and zero the amr parameters + // allocate and zero the AMR parameters p->amr_prms = av_mallocz(sizeof(int16_t)*PRMS_MODE_122); // allocate and zero the decoder state p->state = av_mallocz(sizeof(AMRDecoderState)); @@ -117,13 +117,13 @@ static int amrnb_decode_init(AVCodecCont /** - * Decode the bitstream into the AMR parameters and discover the frame mode + * Decode the bitstream into the AMR parameters and discover the frame mode. * * @param buf pointer to the input buffer * @param buf_size size of the input buffer * @param speech_mode pointer to the speech mode * - * @return Returns the frame mode + * @return the frame mode */ enum Mode decode_bitstream(AVCodecContext *avctx, uint8_t *buf, int buf_size, @@ -134,7 +134,7 @@ enum Mode decode_bitstream(AVCodecContex int i; const AMROrder *order; - // initialise get_bits + // initialize get_bits init_get_bits(&p->gb, buf, buf_size*8); skip_bits1(&p->gb); // set the mode @@ -207,8 +207,8 @@ enum Mode decode_bitstream(AVCodecContex /** - * DESCRIPTION FROM REF SOURCE: - * Make sure that the LSFs are properly ordered and to keep a certain minimum + * DESCRIPTION FROM REFERENCE SOURCE: + * Make sure that the LSFs are properly ordered and keep a certain minimum * distance between adjacent LSFs. * * @param lsf a vector of lsfs (range: 0<=val<=0.5) @@ -230,7 +230,7 @@ static void reorder_lsf(int *lsf, int mi /** - * Convert a vector of lsfs into the corresponding, cosine domain lsps + * Convert a vector of lsfs into the corresponding cosine domain lsps. * * @param lsf a vector of lsfs * @param lsp a vector of lsps @@ -250,7 +250,7 @@ static void lsf2lsp(int *lsf, int *lsp) /** - * Convert a set of 3 lsf quantisation indices into lsp parameters + * Convert a set of 3 lsf quantization indexes into lsp parameters. * * @param avctx pointer to the AVCodecContext for AMR */ @@ -260,12 +260,13 @@ static void decode_lsf2lsp_3(AVCodecCont AMRContext *p = avctx->priv_data; int lsf1_r[LP_FILTER_ORDER]; // vector of residual lsfs - int lsf1_q[LP_FILTER_ORDER]; // vector of quantised lsfs + int lsf1_q[LP_FILTER_ORDER]; // vector of quantized lsfs const int16_t (*lsf_3_temp1)[3], (*lsf_3_temp3)[4]; // temp ptrs for switching tables depending on mode int index_temp; // temp lsf index int i; // counter - // if the current frame is bad estimate the past quantised residual based on the past lsf shifted slightly towards the mean + /* if the current frame is bad, estimate the past quantized residual + * based on the past lsf shifted slightly towards the mean */ if(p->bad_frame_indicator) { for(i=0; i<LP_FILTER_ORDER; i++) { lsf1_q[i] = ( (p->prev_lsf_dq[i]*ALPHA)>>15 ) + ( (lsf_3_mean[i]*ONE_ALPHA)>>15 ); @@ -292,13 +293,14 @@ static void decode_lsf2lsp_3(AVCodecCont lsf_3_temp3 = lsf_3_3; } - // decode lsf residuals from lsf tables using the indices from the bitstream + // decode lsf residuals from lsf tables using the bitstream indexes lsf1_r[0] = lsf_3_temp1[ p->amr_prms[0] ][0]; lsf1_r[1] = lsf_3_temp1[ p->amr_prms[0] ][1]; lsf1_r[2] = lsf_3_temp1[ p->amr_prms[0] ][2]; index_temp = p->amr_prms[1]; - // MODE_475, MODE_515 only use every second entry - NOTE : not quite sure what they mean and the spec doesn't mention this + /* MODE_475, MODE_515 only use every second entry + * NOTE: unsure what they mean and the spec doesn't mention this */ if((p->cur_frame_mode == MODE_475) || (p->cur_frame_mode == MODE_515)) { index_temp = index_temp << 1; } @@ -325,7 +327,7 @@ static void decode_lsf2lsp_3(AVCodecCont } } - /* verification that LSFs has minimum distance of LSF_GAP Hz */ + /* verification that LSFs have minimum distance of LSF_GAP Hz */ reorder_lsf(lsf1_q, LSF_GAP); memcpy(p->prev_lsf_dq, lsf1_q, LP_FILTER_ORDER<<2); @@ -335,7 +337,7 @@ static void decode_lsf2lsp_3(AVCodecCont /** - * Convert a set of 5 lsf quantisation indices into lsp parameters + * Convert a set of 5 lsf quantization indexes into lsp parameters. * * @param avctx pointer to the AVCodecContext for AMR */ @@ -345,11 +347,12 @@ static void decode_lsf2lsp_5(AVCodecCont AMRContext *p = avctx->priv_data; int lsf1_r[LP_FILTER_ORDER], lsf2_r[LP_FILTER_ORDER]; // vectors of residual lsfs - int lsf1_q[LP_FILTER_ORDER], lsf2_q[LP_FILTER_ORDER]; // vectors of quantised lsfs + int lsf1_q[LP_FILTER_ORDER], lsf2_q[LP_FILTER_ORDER]; // vectors of quantized lsfs int temp; int i, sign; // counter and sign of 3rd lsf table - // if the current frame is bad estimate the past quantised residual based on the past lsf shifted slightly towards the mean + /* if the current frame is bad, estimate the past quantized residual + * based on the past lsf shifted slightly towards the mean */ if(p->bad_frame_indicator) { for(i=0; i<LP_FILTER_ORDER; i++) { lsf1_q[i] = ( (p->prev_lsf_dq[i]*ALPHA_122)>>15 ) + ( (lsf_5_mean[i]*ONE_ALPHA_122)>>15 ); @@ -359,7 +362,7 @@ static void decode_lsf2lsp_5(AVCodecCont p->prev_residual_q[i] = lsf2_q[i] - lsf_5_mean[i] - ( (p->prev_residual_q[i]*LSP_PRED_FAC_MODE_122)>>15 ); } }else { - /* decode prediction residuals from 5 received indices */ + /* decode prediction residuals from 5 received indexes */ lsf1_r[0] = lsf_5_1[ p->amr_prms[0] ][0]; lsf1_r[1] = lsf_5_1[ p->amr_prms[0] ][1]; lsf2_r[0] = lsf_5_1[ p->amr_prms[0] ][2]; @@ -436,7 +439,8 @@ static void lsp2poly(int *lsp, int *f) f[1] -= lsp[2*i-2]<<10; } } -/* FIXME - check rounding sensitivity of F1(z) and F2(z) as the code below is faster but rounds differently +/* FIXME - check rounding sensitivity of F1(z) and F2(z) + * as the code below is faster but rounds differently static void lsp2poly(int *lsp, int *f) { int i, j; f[0] = 1<<24; @@ -452,7 +456,7 @@ static void lsp2poly(int *lsp, int *f) { /** - * Convert a vector of lsps to lpc coefficients for a 10th order filter + * Convert a vector of lsps to lpc coefficients for a 10th order filter. * * @param lsp vector of lsps * @param lpc_coeffs pointer to a subframe of lpc coefficients @@ -494,7 +498,7 @@ static void lsp2lpc(int *lsp, int *lpc_c /** - * Interpolate lsps for subframes 1 and 3 and convert them to lpc coefficients + * Interpolate lsps for subframes 1 and 3 and convert them to lpc coefficients. * * @param avctx pointer to AVCodecContext * @param lpc_coeffs array of lpc coefficients for the four subframes @@ -531,7 +535,7 @@ static void lpc_interp_13(AVCodecContext /** - * Interpolate lsps for subframes 1, 2 and 3 and convert them to lpc coefficients + * Interpolate lsps for subframes 1, 2 and 3 and convert them to lpc coefficients. * * @param avctx pointer to AVCodecContext * @param lpc_coeffs array of lpc coefficients for the four subframes @@ -573,10 +577,10 @@ static void lpc_interp_123(AVCodecContex /** - * Decode the adaptive codebook index to the integer and fractional pitch lags for one subframe - * at 1/3 resolution + * Decode the adaptive codebook index to the integer and fractional + * pitch lags for one subframe at 1/3 resolution. * - * Description from ref source: + * Description from reference source: * The fractional lag in 1st and 3rd subframes is encoded with 8 bits * while that in 2nd and 4th subframes is relatively encoded with 4, 5 * and 6 bits depending on the mode. @@ -593,16 +597,17 @@ static void decode_pitch_lag_3(AVCodecCo AMRContext *p = avctx->priv_data; int tmp_lag; - // FIXME - find out where these constants come from and add appropriate comments - // such that people reading the code can understand why these particular values - // are used + // FIXME: find out where these constants come from and add appropriate + // comments such that people reading the code can understand why these + // particular values are used // subframe 1 or 3 if(p->cur_subframe & 1) { if(pitch_index < 197) { *pitch_lag_int = (( (pitch_index + 2)*10923 )>>15) + 19; *pitch_lag_frac = pitch_index - *pitch_lag_int*3 + 58; - // For the sake of deciphering where this comes from, the above code probably means: + // For the sake of deciphering where this comes from, + // the above code probably means: // *pitch_lag_int = (pitch_index + 2)/3 + 19; // *pitch_lag_frac = (pitch_index + 2)%3 + 56; }else { @@ -639,16 +644,16 @@ static void decode_pitch_lag_3(AVCodecCo /** - * Decode the adaptive codebook index to the integer and fractional pitch lag for one subframe - * at 1/6 resolution + * Decode the adaptive codebook index to the integer and fractional + * pitch lag for one subframe at 1/6 resolution. * - * Description from ref source: + * Description from reference source: * The fractional lag in 1st and 3rd subframes is encoded with 9 bits * while that in 2nd and 4th subframes is relatively encoded with 6 bits. * Note that in relative encoding only 61 values are used. If the * decoder receives 61, 62, or 63 as the relative pitch index, it means * that a transmission error occurred. In this case, the pitch lag from - * previous subframe (actually from previous frame) is used. + * the previous subframe (actually from the previous frame) is used. * * @param avctx pointer to AVCodecContext * @param pitch_index received adaptive codebook (pitch) index @@ -662,9 +667,9 @@ static void decode_pitch_lag_6(AVCodecCo AMRContext *p = avctx->priv_data; int temp; - // FIXME - find out where these constants come from and add appropriate comments - // such that people reading the code can understand why these particular values - // are used + // FIXME - find out where these constants come from and add appropriate + // comments such that people reading the code can understand why these + // particular values are used. // subframe 1 or 3 if(p->cur_subframe & 1) { @@ -694,11 +699,12 @@ static void decode_pitch_lag_6(AVCodecCo /** - * Find the adaptive codebook (pitch) vector from interpolating the past excitation at the - * decoded pitch lag with corresponding resolution of 1/3 or 1/6 + * Find the adaptive codebook (pitch) vector from interpolating the past + * excitation at the decoded pitch lag with corresponding resolution + * of 1/3 or 1/6. * * See the specification for details of the underlying mathematics and the used - * FIR filter + * FIR filter. * * @param avctx pointer to AVCodecContext * @param excitation pointer to the excitation buffer @@ -732,7 +738,7 @@ static void decode_pitch_vector(AVCodecC /** - * Reconstruct the algebraic codebook vector + * Reconstruct the algebraic codebook vector. * * @param fixed_code algebraic codebook vector * @param pulse_position vector of pulse positions @@ -755,7 +761,7 @@ static void reconstruct_fixed_code(int * /** - * Decode the algebraic codebook index to pulse position indices for MODE_102 + * Decode the algebraic codebook index to pulse position indexes for MODE_102. * * @param fixed_index positions of the eight pulses (encoded) * @param position_index position index of the eight pulses @@ -765,7 +771,7 @@ static void fixed2position(int16_t *fixe { int MSBs, LSBs, MSBs0_24, divMSB; - // indices from track 1 (7+3 bits) + // indexes from track 1 (7+3 bits) MSBs= FFMIN(fixed_index[0] >> 3, 124); // LSBs = fixed_index[0]%8; LSBs = fixed_index[0] & 0x7; @@ -777,7 +783,7 @@ static void fixed2position(int16_t *fixe position_index[4] = (( (MSBs - 25*divMSB)/5 )<<1) + ((LSBs & 0x2)>>1); position_index[1] = (divMSB<<1) + (LSBs>>2); - // indices from track 2 (7+3 bits) + // indexes from track 2 (7+3 bits) MSBs= FFMIN(fixed_index[1] >> 3, 124); // LSBs = fixed_index[1]%8; LSBs = fixed_index[1] & 0x7; @@ -789,7 +795,7 @@ static void fixed2position(int16_t *fixe position_index[6] = (( (MSBs - 25*divMSB)/5 )<<1) + ((LSBs & 0x2)>>1); position_index[5] = (divMSB<<1) + (LSBs>>2); - // indices from track 3 (5+2 bits) + // indexes from track 3 (5+2 bits) MSBs = fixed_index[2] >> 2; // LSBs = fixed_index[2]%4; LSBs = fixed_index[2] & 0x3; @@ -806,8 +812,8 @@ static void fixed2position(int16_t *fixe /** - * Decode the algebraic codebook index to pulse positions and signs and construct - * the algebraic codebook vector for MODE_475 and MODE_515 + * Decode the algebraic codebook index to pulse positions and signs and + * construct the algebraic codebook vector for MODE_475 and MODE_515. * * @param avctx pointer to AVCodecContext * @param sign signs of the two pulses @@ -835,8 +841,8 @@ static void decode_2_pulses_9bits(AVCode /** - * Decode the algebraic codebook index to pulse positions and signs and construct - * the algebraic codebook vector for MODE_59 + * Decode the algebraic codebook index to pulse positions and signs + * and construct the algebraic codebook vector for MODE_59. * * @param sign signs of the two pulses * @param fixed_index positions of the two pulses @@ -867,8 +873,8 @@ static void decode_2_pulses_11bits(int s /** - * Decode the algebraic codebook index to pulse positions and signs and construct - * the algebraic codebook vector for MODE_67 + * Decode the algebraic codebook index to pulse positions and signs + * and construct the algebraic codebook vector for MODE_67. * * @param sign signs of the three pulses * @param fixed_index positions of the three pulses @@ -897,8 +903,8 @@ static void decode_3_pulses_14bits(int s /** - * Decode the algebraic codebook index to pulse positions and signs and construct - * the algebraic codebook vector for MODE_74 and MODE_795 + * Decode the algebraic codebook index to pulse positions and signs and + * construct the algebraic codebook vector for MODE_74 and MODE_795. * * @param sign signs of the four pulses * @param fixed_index positions of the four pulses @@ -927,8 +933,8 @@ static void decode_4_pulses_17bits(int s /** - * Decode the algebraic codebook index to pulse positions and signs and construct - * the algebraic codebook vector for MODE_102 + * Decode the algebraic codebook index to pulse positions and signs + * and construct the algebraic codebook vector for MODE_102. * * @param fixed_index positions of the eight pulses * @param fixed_code pointer to the algebraic codebook vector @@ -961,8 +967,8 @@ static void decode_8_pulses_31bits(int16 /** - * Decode the algebraic codebook index to pulse positions and signs and construct - * the algebraic codebook vector for MODE_122 + * Decode the algebraic codebook index to pulse positions and signs + * and construct the algebraic codebook vector for MODE_122. * * @param fixed_index positions of the ten pulses * @param fixed_code pointer to the algebraic codebook vector @@ -979,7 +985,7 @@ static void decode_10_pulses_35bits(int1 // find the position of the ith pulse pos1 = dgray[fixed_index[i ] & 7]*5 + i; // find the sign of the ith pulse - sign = (fixed_index[i] & 8) ? -4096 : 4096; // +/-1 : 4096 is used here in the ref source + sign = (fixed_index[i] & 8) ? -4096 : 4096; // +/-1 : 4096 is used here in the reference source // find the position of the i+5th pulse pos2 = dgray[fixed_index[i+5] & 7]*5 + i; // assign the ith pulse (+/-1) to its appropriate position @@ -995,10 +1001,10 @@ static void decode_10_pulses_35bits(int1 // general functions FIXME - useful enough to put into libavutil? /** - * Comparison function for use with qsort + * Comparison function for use with qsort. * - * @param a First value for comparison - * @param b Second value for comparison + * @param a first value for comparison + * @param b second value for comparison * @return a-b : the result of the comparison */ @@ -1008,11 +1014,11 @@ int qsort_compare(const int *a, const in } /** - * Find the median of some values + * Find the median of some values. * * @param values pointer to the values of which to find the median * @param n number of values - * @return Returns the median value + * @return the median value */ static int median(int *values, int n) @@ -1030,10 +1036,10 @@ static int median(int *values, int n) // gain functions /** - * Calculate the pitch gain from previous values + * Calculate the pitch gain from previous values. * * @param state_ptr pointer to the current state - * @return Returns the pitch gain + * @return the pitch gain */ static int find_pitch_gain(AMRDecoderState *state_ptr) @@ -1052,11 +1058,11 @@ static int find_pitch_gain(AMRDecoderSta /** - * Decode the pitch gain using the received index + * Decode the pitch gain using the received index. * * @param mode current mode - * @param index quantisation index - * @return Returns the pitch gain + * @param index quantization index + * @return the pitch gain */ static int decode_pitch_gain(enum Mode mode, int index) @@ -1075,7 +1081,7 @@ static int decode_pitch_gain(enum Mode m /** - * Update the pitch gain and limit pitch_gain if the previous frame was bad + * Update the pitch gain and limit pitch_gain if the previous frame was bad. * * @param state_ptr pointer to the current state * @param bad_frame_indicator bad frame indicator @@ -1113,7 +1119,7 @@ static void pitch_gain_update(AMRDecoder /** - * Reset the AMR frame parameters + * Reset the AMR frame parameters. * * @param avctx pointer to the AVCodecContext for AMR */ @@ -1139,7 +1145,7 @@ static int amrnb_decode_frame(AVCodecCon const int16_t *homing_frame; // pointer to the homing frame int homing_frame_size; // homing frame size - // decode the bitstream to amr parameters + // decode the bitstream to AMR parameters p->cur_frame_mode = decode_bitstream(avctx, buf, buf_size, &speech_mode); // guess the mode from the previous frame if no data or bad data @@ -1267,14 +1273,15 @@ static int amrnb_decode_frame(AVCodecCon p->prev_frame_type = p->cur_frame_type; p->prev_frame_mode = p->cur_frame_mode; - /* To make it easy the stream can only be 16 bits mono, so let's convert it to that */ + /* To make it easy the stream can only be 16 bits mono, + * so let's convert it to that */ for (i=0 ; i<buf_size; i++) outbuffer[i] = (int16_t)p->sample_buffer[i]; /* Report how many samples we got */ *data_size = buf_size; - /* Return the amount of bytes consumed if everything was ok */ + /* Return the amount of bytes consumed if everything was OK */ return *data_size*sizeof(int16_t); } @@ -1288,7 +1295,7 @@ static int amrnb_decode_close(AVCodecCon av_free(p->amr_prms); av_free(p->state); - /* Return 0 if everything is ok, -1 if not */ + /* Return 0 if everything is OK, -1 if not */ return 0; } Modified: amr/amrnbfloatdata.h ============================================================================== --- amr/amrnbfloatdata.h Sun Jan 25 16:35:48 2009 (r3999) +++ amr/amrnbfloatdata.h Sun Jan 25 16:38:52 2009 (r4000) @@ -98,9 +98,9 @@ typedef struct AMROrder { /**************************** tables *****************************/ -// The following order* tables are used to reorder the bitstream into the bit allocation -// format as in the specification. The bits are grouped into three classes of importance. -// See the specification, for more information. +// The following order* tables are used to reorder the bitstream into the bit +// allocation format as in the specification. The bits are grouped into three +// classes of importance. See the specification for more information. static const AMROrder order_MODE_475[95] = { { 0, 7}, { 0, 6}, { 0, 5}, { 0, 4}, { 0, 3}, { 0, 2}, { 0, 1}, { 0, 0}, @@ -1814,7 +1814,7 @@ static const uint8_t gray_decode[8] = { #define MIN_ENERGY -14.0 -// scalar quantised pitch gain table for 7.95 and 12.2 kbps modes +// scalar quantized pitch gain table for 7.95 and 12.2 kbps modes static const float qua_gain_pit[16] = { 0.0 , 0.20001220703125, 0.400146484375 , 0.5 , 0.5999755859375 , 0.70001220703125, 0.75 , 0.79998779296875, @@ -1822,10 +1822,11 @@ static const float qua_gain_pit[16] = { 1.04998779296875, 1.0999755859375 , 1.1500244140625 , 1.20001220703125, }; -// scalar quantised fixed gain table for 7.95 and 12.2 kbps modes -// note: this table was generated from the integer qua_gain_code tables in the -// ref source that store Q10 qua_ener = 20*log10(g_fac). The table below stores -// g_fac as a float from printf("%10.6f,\n", powf(10.0, qua_ener/(20.0*(1<<10)))) +// scalar quantized fixed gain table for 7.95 and 12.2 kbps modes +// NOTE: This table was generated from the integer qua_gain_code tables +// in the reference source that store Q10 qua_ener = 20*log10(g_fac). +// The table below stores g_fac as a float from +// printf("%10.6f,\n", powf(10.0, qua_ener/(20.0*(1<<10)))). static const float qua_gain_code[32] = { 0.077640, 0.100586, 0.130857, 0.170410, 0.204592, 0.235358, 0.270507, 0.311044, 0.357898, 0.411116, 0.473151, 0.543935, 0.625519, 0.719258, 0.827138, 0.951199, @@ -1841,7 +1842,7 @@ static const float energy_mean[8] = { 33 static const float energy_pred_fac[4] = { 0.68, 0.58, 0.34, 0.19 }; // gain table for 4.75 kbps mode -// first index has even/odd indices for subframes 0,2/1,3 +// first index has even/odd indexes for subframes 0,2/1,3 // second index is {pitch_gain, fixed_gain_factor} static const float gains_MODE_475[512][2] = { {0.049561, 0.031250}, {0.033081, 0.034180}, {0.175354, 0.277100}, {0.138306, 0.830566}, Modified: amr/amrnbfloatdec.c ============================================================================== --- amr/amrnbfloatdec.c Sun Jan 25 16:35:48 2009 (r3999) +++ amr/amrnbfloatdec.c Sun Jan 25 16:38:52 2009 (r4000) @@ -41,7 +41,7 @@ typedef struct AMRContext { GetBitContext gb; - int16_t amr_prms[PRMS_MODE_122]; ///< decoded amr parameters (lsf coefficients, codebook indices, etc) + int16_t amr_prms[PRMS_MODE_122]; ///< decoded AMR parameters (lsf coefficients, codebook indexes, etc) int bad_frame_indicator; ///< bad frame ? 1 : 0 int cur_frame_mode; ///< current frame mode int cur_frame_type; ///< current frame type @@ -52,7 +52,7 @@ typedef struct AMRContext { float lsp_avg[LP_FILTER_ORDER]; ///< vector of averaged lsp coefficients - float lpc[4][LP_FILTER_ORDER]; ///< vectors of lpc coefficients for 4 subframes + float lpc[4][LP_FILTER_ORDER]; ///< lpc coefficient vectors for 4 subframes int search_range_min; ///< minimum pitch lag search range int search_range_max; ///< maximum pitch lag search range @@ -90,9 +90,9 @@ static void reset_state(AMRContext *p) { int i; - // initialise values for the lsp vector from the 4th subframe of the - // previous subframe values - // taken from Decoder_amr_reset using val/(float)(1<<15) + // Initialize values for the lsp vector from the 4th subframe of the + // previous subframe values. + // Taken from Decoder_amr_reset using val/(float)(1<<15). p->prev_lsp_sub4[0] = 0.91552734375; p->prev_lsp_sub4[1] = 0.79345703125; p->prev_lsp_sub4[2] = 0.640869140625; @@ -104,8 +104,8 @@ static void reset_state(AMRContext *p) p->prev_lsp_sub4[8] = -0.640869140625; p->prev_lsp_sub4[9] = -0.79345703125; - // initialise mean lsp values - // taken from Decoder_amr_reset using val/(float)(1<<15) + // Initialize mean lsp values. + // Taken from Decoder_amr_reset using val/(float)(1<<15). p->lsp_avg[0] = 0.042236328125; p->lsp_avg[1] = 0.063385009765625; p->lsp_avg[2] = 0.1043701171875; @@ -148,13 +148,13 @@ static int amrnb_decode_init(AVCodecCont /** - * Decode the bitstream into the AMR parameters and discover the frame mode + * Decode the bitstream into the AMR parameters and discover the frame mode. * * @param buf pointer to the input buffer * @param buf_size size of the input buffer * @param speech_mode pointer to the speech mode * - * @return Returns the frame mode + * @return the frame mode */ enum Mode decode_bitstream(AMRContext *p, uint8_t *buf, int buf_size, @@ -167,7 +167,7 @@ enum Mode decode_bitstream(AMRContext *p // reset amr_prms memset(p->amr_prms, 0, PRMS_MODE_122 * sizeof(int16_t)); - // initialise get_bits + // initialize get_bits init_get_bits(&p->gb, buf, buf_size*8); skip_bits(&p->gb, 1); // set the mode @@ -242,7 +242,7 @@ enum Mode decode_bitstream(AMRContext *p /*** LPC coefficient decoding functions ***/ /** - * Convert an lsf vector into an lsp vector + * Convert an lsf vector into an lsp vector. * * @param lsf input lsf vector * @param lsp output lsp vector @@ -258,7 +258,7 @@ static void lsf2lsp(float *lsf, float *l } /** - * Decode a set of 5 split-matrix quantised lsf indices into an lsp vector + * Decode a set of 5 split-matrix quantized lsf indexes into an lsp vector. * * @param p pointer to the AMRContext */ @@ -319,7 +319,7 @@ static void lsf2lsp_5(AMRContext *p) } /** - * Decode a set of 3 split-matrix quantised lsf indices into an lsp vector + * Decode a set of 3 split-matrix quantized lsf indexes into an lsp vector. * * @param p pointer to the AMRContext */ @@ -351,8 +351,8 @@ static void lsf2lsp_3(AMRContext *p) lsf_r[2] = lsf_3_1_tmp[ idx ][2]; idx = p->amr_prms[1]; - // MODE_475, MODE_515 only use every other entry as their indices are stored - // using 1 less bit (8-bits vs 9-bits) + // MODE_475, MODE_515 only use every other entry as their indexes + // are stored using 1 less bit (8 bits vs 9 bits) if((p->cur_frame_mode == MODE_475) || (p->cur_frame_mode == MODE_515)) { idx <<= 1; } @@ -377,7 +377,7 @@ static void lsf2lsp_3(AMRContext *p) } /** - * Interpolate lsp vectors for subframes 1 and 3 + * Interpolate lsp vectors for subframes 1 and 3. * * @param p pointer to the AMRContext */ @@ -393,7 +393,7 @@ static void interp_lsp_13(AMRContext *p) } /** - * Interpolate lsp vectors for subframes 1, 2 and 3 + * Interpolate lsp vectors for subframes 1, 2 and 3. * * @param p pointer to the AMRContext */ @@ -410,7 +410,7 @@ static void interp_lsp_123(AMRContext *p } /** - * Find the polynomial F1(z) or F2(z) from the lsp vectors + * Find the polynomial F1(z) or F2(z) from the lsp vectors. * * @param lsp input lsp vector * @param f pointer to the polynomial F1(z) or F2(z) @@ -433,7 +433,7 @@ static void lsp2poly(float *lsp, float * } /** - * Convert an lsp vector to lpc coefficients + * Convert an lsp vector to lpc coefficients. * * @param lsp input lsp vector * @param lpc output lpc coefficients @@ -457,8 +457,8 @@ static void lsp2lpc(float *lsp, float *l // A(z) = ( F1'(z) + F2'(z) )/2 // note f1 and f2 are actually f1' and f2' for(i=0; i<5; i++) { - lpc_coeffs[i] = 0.5*(f1[i+2] + f2[i+2]); // lpc 0..4 uses indices to f, 2..6 - lpc_coeffs[i+5] = 0.5*(f1[6-i] - f2[6-i]); // lpc 5..9 uses indices to f, 6..2 + lpc_coeffs[i] = 0.5*(f1[i+2] + f2[i+2]); // lpc 0..4 uses indexes to f, 2..6 + lpc_coeffs[i+5] = 0.5*(f1[6-i] - f2[6-i]); // lpc 5..9 uses indexes to f, 6..2 } } @@ -468,8 +468,8 @@ static void lsp2lpc(float *lsp, float *l /*** pitch vector decoding functions ***/ /** - * Decode the adaptive codebook index to the integer and fractional parts of the - * pitch lag for one subframe at 1/3 resolution + * Decode the adaptive codebook index to the integer and fractional parts + * of the pitch lag for one subframe at 1/3 resolution. * * @param p pointer to the AMRContext * @param pitch_index parsed adaptive codebook (pitch) index @@ -492,7 +492,7 @@ static void decode_pitch_lag_3(AMRContex }else { if( (p->cur_frame_mode == MODE_475) || (p->cur_frame_mode == MODE_515) || (p->cur_frame_mode == MODE_59) || (p->cur_frame_mode == MODE_67) ) { - // decoding with 4 bit resolution + // decoding with 4-bit resolution int t1_temp = FFMAX(FFMIN(p->prev_pitch_lag_int, p->search_range_max-4), p->search_range_min+5); if(pitch_index < 4) { @@ -519,8 +519,8 @@ static void decode_pitch_lag_3(AMRContex } /** - * Decode the adaptive codebook index to the integer and fractional parts of the - * pitch lag for one subframe at 1/6 resolution + * Decode the adaptive codebook index to the integer and fractional parts + * of the pitch lag for one subframe at 1/6 resolution. * * @param p pointer to the AMRContext * @param pitch_index parsed adaptive codebook (pitch) index @@ -550,7 +550,7 @@ static void decode_pitch_lag_6(AMRContex /** * Calculate the pitch vector by interpolating the past excitation at the pitch - * pitch lag using a b60 hamming windowed sinc function + * pitch lag using a b60 hamming windowed sinc function. * * @param prev_excitation pointer to the element after the previous excitations * @param lag_int integer part of pitch lag @@ -599,7 +599,7 @@ static void interp_pitch_vector(float *p /*** algebraic code book (fixed) vector decoding functions ***/ /** - * Reconstruct the algebraic codebook vector + * Reconstruct the algebraic codebook vector. * * @param pulse_position vector of pulse positions * @param sign signs of the pulses @@ -620,8 +620,8 @@ static void reconstruct_fixed_vector(int } /** - * Decode the algebraic codebook index to pulse positions and signs and construct - * the algebraic codebook vector for MODE_475 and MODE_515 + * Decode the algebraic codebook index to pulse positions and signs and + * construct the algebraic codebook vector for MODE_475 and MODE_515. * * @param fixed_index positions of the two pulses * @param sign signs of the two pulses @@ -647,8 +647,8 @@ static void decode_2_pulses_9bits(int fi } /** - * Decode the algebraic codebook index to pulse positions and signs and construct - * the algebraic codebook vector for MODE_59 + * Decode the algebraic codebook index to pulse positions and signs and + * construct the algebraic codebook vector for MODE_59. * * @param fixed_index positions of the two pulses * @param sign signs of the two pulses @@ -676,8 +676,8 @@ static void decode_2_pulses_11bits(int f } /** - * Decode the algebraic codebook index to pulse positions and signs and construct - * the algebraic codebook vector for MODE_67 + * Decode the algebraic codebook index to pulse positions and signs and + * construct the algebraic codebook vector for MODE_67. * * @param fixed_index positions of the three pulses * @param sign signs of the three pulses @@ -706,8 +706,8 @@ static void decode_3_pulses_14bits(int f } /** - * Decode the algebraic codebook index to pulse positions and signs and construct - * the algebraic codebook vector for MODE_74 and MODE_795 + * Decode the algebraic codebook index to pulse positions and signs and + * construct the algebraic codebook vector for MODE_74 and MODE_795. * * @param fixed_index positions of the four pulses * @param sign signs of the four pulses @@ -736,8 +736,8 @@ static void decode_4_pulses_17bits(int f } /** - * Decode the algebraic codebook index to pulse positions and signs and construct - * the algebraic codebook vector for MODE_102 + * Decode the algebraic codebook index to pulse positions and signs and + * construct the algebraic codebook vector for MODE_102. * * @param fixed_index positions of the eight pulses * @param fixed_vector pointer to the algebraic codebook vector @@ -791,8 +791,8 @@ static void decode_8_pulses_31bits(int16 } /** - * Decode the algebraic codebook index to pulse positions and signs and construct - * the algebraic codebook vector for MODE_122 + * Decode the algebraic codebook index to pulse positions and signs and + * construct the algebraic codebook vector for MODE_122. * * @param fixed_index positions of the ten pulses * @param fixed_vector pointer to the algebraic codebook vector @@ -827,12 +827,12 @@ static void decode_10_pulses_35bits(int1 /*** gain decoding functions ***/ /** - * Predict the fixed gain + * Predict the fixed gain. * * @param fixed_vector pointer to the algebraic codebook vector * @param prev_pred_error pointer to the quantified prediction errors from the previous four subframes * - * @return Returns the predicted fixed gain + * @return the predicted fixed gain */ static float fixed_gain_prediction(float *fixed_vector, float *prev_pred_error, @@ -863,10 +863,10 @@ static float fixed_gain_prediction(float /*** pre-processing functions ***/ /** - * Comparison function for use with qsort + * Comparison function for use with qsort. * - * @param a First value for comparison - * @param b Second value for comparison + * @param a first value for comparison + * @param b second value for comparison * @return a-b : the result of the comparison */ @@ -881,11 +881,11 @@ int qsort_compare(const float *a, const } /** - * Find the median of some float values + * Find the median of some float values. * * @param values pointer to the values of which to find the median * @param n number of values - * @return Returns the median value + * @return the median value */ static float medianf(float *values, int n) @@ -905,7 +905,7 @@ static float medianf(float *values, int /** * Circularly convolve the fixed vector with a phase dispersion impulse response - * filter + * filter. * * @param fixed_vector pointer to the fixed vector * @param ir_filter pointer to the impulse response filter @@ -945,7 +945,7 @@ static void convolve_circ(float *fixed_v /*** synthesis functions ***/ /** - * Conduct 10th order linear predictive coding synthesis + * Conduct 10th order linear predictive coding synthesis. * * @param p pointer to the AMRContext * @param excitation pointer to the excitation vector @@ -980,7 +980,7 @@ static int synthesis(AMRContext *p, floa float eta, temp1 = 0.0, temp2 = 0.0; for(i=0; i<AMR_SUBFRAME_SIZE; i++) { - // emphasise pitch vector contribution + // emphasize pitch vector contribution excitation_temp[i] = excitation[i] + pitch_factor*p->pitch_vector[i]; // find gain scale temp1 += excitation[i]*excitation[i]; @@ -999,7 +999,7 @@ static int synthesis(AMRContext *p, floa for(j=0; j<LP_FILTER_ORDER; j++) { samples[i] -= lpc[j]*samples[i-j-1]; } - // Detect overflow + // detect overflow if(fabsf(samples[i])>1.0) { overflow_temp = 1; samples[i] = av_clipf(samples[i], -1.0, 1.0); @@ -1015,7 +1015,7 @@ static int synthesis(AMRContext *p, floa /*** update functions ***/ /** - * Update buffers and history at the end of decoding a subframe + * Update buffers and history at the end of decoding a subframe. * * @param p pointer to the AMRContext */ @@ -1066,20 +1066,20 @@ static int amrnb_decode_frame(AVCodecCon int gains_index_MODE_475 = 0; // MODE_475 gains index coded every other subframe enum Mode speech_mode = MODE_475; // ??? - // decode the bitstream to amr parameters + // decode the bitstream to AMR parameters p->cur_frame_mode = decode_bitstream(p, buf, buf_size, &speech_mode); /*** LPC coefficient decoding ***/ if(p->cur_frame_mode == MODE_122) { - // decode split-matrix quantised lsf vector indices to lsp vectors + // decode split-matrix quantized lsf vector indexes to lsp vectors lsf2lsp_5(p); // interpolate LSP vectors at subframes 1 and 3 interp_lsp_13(p); // advance index into amr_prms index += 5; }else { - // decode split-matrix quantised lsf vector indices to an lsp vector + // decode split-matrix quantized lsf vector indexes to an lsp vector lsf2lsp_3(p); // interpolate LSP vectors at subframes 1, 2 and 3 interp_lsp_123(p); @@ -1305,10 +1305,10 @@ static int amrnb_decode_frame(AVCodecCon } - /* Report how many samples we got */ + /* report how many samples we got */ *data_size = AMR_BLOCK_SIZE * sizeof(int16_t); - /* Return the amount of bytes consumed if everything was ok */ + /* return the amount of bytes consumed if everything was OK */ return (mode_bits[p->cur_frame_mode] + 15)>>3; // +7 for rounding and +8 for TOC } @@ -1317,7 +1317,7 @@ static int amrnb_decode_close(AVCodecCon { AMRContext *p = avctx->priv_data; - /* Return 0 if everything is ok, -1 if not */ + /* return 0 if everything is OK, -1 if not */ return 0; }

1 0

[soc]: r3999 - in amr: amrnbdec.c amrnbfloatdec.c
by diego 25 Jan '09

25 Jan '09

Author: diego Date: Sun Jan 25 16:35:48 2009 New Revision: 3999 Log: Remove nonsensical '@return void' Doxygen comments. Modified: amr/amrnbdec.c amr/amrnbfloatdec.c Modified: amr/amrnbdec.c ============================================================================== --- amr/amrnbdec.c Sun Jan 25 16:27:50 2009 (r3998) +++ amr/amrnbdec.c Sun Jan 25 16:35:48 2009 (r3999) @@ -420,8 +420,6 @@ static void decode_lsf2lsp_5(AVCodecCont * * @param lsp vector of lsps * @param f pointer to the polynomial F1(z) or F2(z) - * - * @return void */ static void lsp2poly(int *lsp, int *f) @@ -458,8 +456,6 @@ static void lsp2poly(int *lsp, int *f) { * * @param lsp vector of lsps * @param lpc_coeffs pointer to a subframe of lpc coefficients - * - * @return void */ static void lsp2lpc(int *lsp, int *lpc_coeffs) @@ -502,8 +498,6 @@ static void lsp2lpc(int *lsp, int *lpc_c * * @param avctx pointer to AVCodecContext * @param lpc_coeffs array of lpc coefficients for the four subframes - * - * @return void */ static void lpc_interp_13(AVCodecContext *avctx, int **lpc_coeffs) @@ -541,8 +535,6 @@ static void lpc_interp_13(AVCodecContext * * @param avctx pointer to AVCodecContext * @param lpc_coeffs array of lpc coefficients for the four subframes - * - * @return void */ static void lpc_interp_123(AVCodecContext *avctx, int **lpc_coeffs) @@ -593,8 +585,6 @@ static void lpc_interp_123(AVCodecContex * @param pitch_index received adaptive codebook (pitch) index * @param pitch_lag_int pointer to integer part of the subframe pitch lag * @param pitch_lag_frac pointer to fractional part of the subframe pitch lag - * - * @return void */ static void decode_pitch_lag_3(AVCodecContext *avctx, int pitch_index, @@ -664,8 +654,6 @@ static void decode_pitch_lag_3(AVCodecCo * @param pitch_index received adaptive codebook (pitch) index * @param pitch_lag_int pointer to integer part of the subframe pitch lag * @param pitch_lag_frac pointer to fractional part of the subframe pitch lag - * - * @return void */ static void decode_pitch_lag_6(AVCodecContext *avctx, int pitch_index, @@ -714,8 +702,6 @@ static void decode_pitch_lag_6(AVCodecCo * * @param avctx pointer to AVCodecContext * @param excitation pointer to the excitation buffer - * - * @return void */ static void decode_pitch_vector(AVCodecContext *avctx, int *excitation) @@ -752,8 +738,6 @@ static void decode_pitch_vector(AVCodecC * @param pulse_position vector of pulse positions * @param sign signs of the pulses * @param nr_pulses number of pulses - * - * @return void */ static void reconstruct_fixed_code(int *fixed_code, int *pulse_position, @@ -775,8 +759,6 @@ static void reconstruct_fixed_code(int * * * @param fixed_index positions of the eight pulses (encoded) * @param position_index position index of the eight pulses - * - * @return void */ static void fixed2position(int16_t *fixed_index, int *position_index) @@ -831,8 +813,6 @@ static void fixed2position(int16_t *fixe * @param sign signs of the two pulses * @param fixed_index positions of the two pulses * @param fixed_code pointer to the algebraic codebook vector - * - * @return void */ static void decode_2_pulses_9bits(AVCodecContext *avctx, int sign, @@ -861,8 +841,6 @@ static void decode_2_pulses_9bits(AVCode * @param sign signs of the two pulses * @param fixed_index positions of the two pulses * @param fixed_code pointer to the algebraic codebook vector - * - * @return void */ static void decode_2_pulses_11bits(int sign, int fixed_index, int *fixed_code) @@ -895,8 +873,6 @@ static void decode_2_pulses_11bits(int s * @param sign signs of the three pulses * @param fixed_index positions of the three pulses * @param fixed_code pointer to the algebraic codebook vector - * - * @return void */ static void decode_3_pulses_14bits(int sign, int fixed_index, int *fixed_code) @@ -927,8 +903,6 @@ static void decode_3_pulses_14bits(int s * @param sign signs of the four pulses * @param fixed_index positions of the four pulses * @param fixed_code pointer to the algebraic codebook vector - * - * @return void */ static void decode_4_pulses_17bits(int sign, int fixed_index, int *fixed_code) @@ -958,8 +932,6 @@ static void decode_4_pulses_17bits(int s * * @param fixed_index positions of the eight pulses * @param fixed_code pointer to the algebraic codebook vector - * - * @return void */ static void decode_8_pulses_31bits(int16_t *fixed_index, int *fixed_code) @@ -994,8 +966,6 @@ static void decode_8_pulses_31bits(int16 * * @param fixed_index positions of the ten pulses * @param fixed_code pointer to the algebraic codebook vector - * - * @return void */ static void decode_10_pulses_35bits(int16_t *fixed_index, int *fixed_code) Modified: amr/amrnbfloatdec.c ============================================================================== --- amr/amrnbfloatdec.c Sun Jan 25 16:27:50 2009 (r3998) +++ amr/amrnbfloatdec.c Sun Jan 25 16:35:48 2009 (r3999) @@ -246,8 +246,6 @@ enum Mode decode_bitstream(AMRContext *p * * @param lsf input lsf vector * @param lsp output lsp vector - * - * @return void */ static void lsf2lsp(float *lsf, float *lsp) @@ -263,8 +261,6 @@ static void lsf2lsp(float *lsf, float *l * Decode a set of 5 split-matrix quantised lsf indices into an lsp vector * * @param p pointer to the AMRContext - * - * @return void */ static void lsf2lsp_5(AMRContext *p) @@ -326,8 +322,6 @@ static void lsf2lsp_5(AMRContext *p) * Decode a set of 3 split-matrix quantised lsf indices into an lsp vector * * @param p pointer to the AMRContext - * - * @return void */ static void lsf2lsp_3(AMRContext *p) @@ -386,8 +380,6 @@ static void lsf2lsp_3(AMRContext *p) * Interpolate lsp vectors for subframes 1 and 3 * * @param p pointer to the AMRContext - * - * @return void */ static void interp_lsp_13(AMRContext *p) @@ -404,8 +396,6 @@ static void interp_lsp_13(AMRContext *p) * Interpolate lsp vectors for subframes 1, 2 and 3 * * @param p pointer to the AMRContext - * - * @return void */ static void interp_lsp_123(AMRContext *p) @@ -424,8 +414,6 @@ static void interp_lsp_123(AMRContext *p * * @param lsp input lsp vector * @param f pointer to the polynomial F1(z) or F2(z) - * - * @return void */ static void lsp2poly(float *lsp, float *f) @@ -449,8 +437,6 @@ static void lsp2poly(float *lsp, float * * * @param lsp input lsp vector * @param lpc output lpc coefficients - * - * @return void */ static void lsp2lpc(float *lsp, float *lpc_coeffs) @@ -488,8 +474,6 @@ static void lsp2lpc(float *lsp, float *l * @param p pointer to the AMRContext * @param pitch_index parsed adaptive codebook (pitch) index * @param subframe current subframe - * - * @return void */ static void decode_pitch_lag_3(AMRContext *p, int pitch_index, int subframe) @@ -541,8 +525,6 @@ static void decode_pitch_lag_3(AMRContex * @param p pointer to the AMRContext * @param pitch_index parsed adaptive codebook (pitch) index * @param subframe current subframe - * - * @return void */ static void decode_pitch_lag_6(AMRContext *p, int pitch_index, int subframe) @@ -575,8 +557,6 @@ static void decode_pitch_lag_6(AMRContex * @param lag_frac fractional part of pitch lag * @param mode current frame mode * @param pitch_vector pointer to the pitch vector - * - * @return void */ static void interp_pitch_vector(float *prev_excitation, int lag_int, @@ -625,8 +605,6 @@ static void interp_pitch_vector(float *p * @param sign signs of the pulses * @param nr_pulses number of pulses * @param fixed_vector algebraic codebook vector - * - * @return void */ static void reconstruct_fixed_vector(int *pulse_position, int sign, @@ -649,8 +627,6 @@ static void reconstruct_fixed_vector(int * @param sign signs of the two pulses * @param subframe current subframe * @param fixed_vector pointer to the algebraic codebook vector - * - * @return void */ static void decode_2_pulses_9bits(int fixed_index, int sign, int subframe, @@ -677,8 +653,6 @@ static void decode_2_pulses_9bits(int fi * @param fixed_index positions of the two pulses * @param sign signs of the two pulses * @param fixed_vector pointer to the algebraic codebook vector - * - * @return void */ static void decode_2_pulses_11bits(int fixed_index, int sign, @@ -708,8 +682,6 @@ static void decode_2_pulses_11bits(int f * @param fixed_index positions of the three pulses * @param sign signs of the three pulses * @param fixed_vector pointer to the algebraic codebook vector - * - * @return void */ static void decode_3_pulses_14bits(int fixed_index, int sign, @@ -740,8 +712,6 @@ static void decode_3_pulses_14bits(int f * @param fixed_index positions of the four pulses * @param sign signs of the four pulses * @param fixed_vector pointer to the algebraic codebook vector - * - * @return void */ static void decode_4_pulses_17bits(int fixed_index, int sign, @@ -771,8 +741,6 @@ static void decode_4_pulses_17bits(int f * * @param fixed_index positions of the eight pulses * @param fixed_vector pointer to the algebraic codebook vector - * - * @return void */ static void decode_8_pulses_31bits(int16_t *fixed_index, float *fixed_vector) @@ -828,8 +796,6 @@ static void decode_8_pulses_31bits(int16 * * @param fixed_index positions of the ten pulses * @param fixed_vector pointer to the algebraic codebook vector - * - * @return void */ static void decode_10_pulses_35bits(int16_t *fixed_index, float *fixed_vector)

1 0

[soc]: r3998 - in wmapro: wma3.h wma3dec.c
by diego 25 Jan '09

25 Jan '09

Author: diego Date: Sun Jan 25 16:27:50 2009 New Revision: 3998 Log: Use American English throughout, even for variable names. Modified: wmapro/wma3.h wmapro/wma3dec.c Modified: wmapro/wma3.h ============================================================================== --- wmapro/wma3.h Sun Jan 25 16:27:18 2009 (r3997) +++ wmapro/wma3.h Sun Jan 25 16:27:50 2009 (r3998) @@ -136,7 +136,7 @@ typedef struct WMA3DecodeContext { /** subframe/block decode state */ int subframe_len; //< current subframe length int channels_for_cur_subframe; - int channel_indices_for_cur_subframe[MAX_CHANNELS]; + int channel_indexes_for_cur_subframe[MAX_CHANNELS]; int subwoofer_cutoff; //< subwoofer cutoff value int cValidBarkBand; int* rgiBarkIndex; Modified: wmapro/wma3dec.c ============================================================================== --- wmapro/wma3dec.c Sun Jan 25 16:27:18 2009 (r3997) +++ wmapro/wma3dec.c Sun Jan 25 16:27:50 2009 (r3998) @@ -613,7 +613,7 @@ static int wma_decode_channel_transform( if(remaining_channels > 2){ for(i=0;i<s->channels_for_cur_subframe;i++){ - int channel_idx = s->channel_indices_for_cur_subframe[i]; + int channel_idx = s->channel_indexes_for_cur_subframe[i]; if(!s->channel[channel_idx].grouped && get_bits(gb,1)){ ++chgroup->nb_channels; s->channel[channel_idx].grouped = 1; @@ -870,7 +870,7 @@ static int wma_decode_scale_factors(WMA3 { int i; for(i=0;i<s->channels_for_cur_subframe;i++){ - int c = s->channel_indices_for_cur_subframe[i]; + int c = s->channel_indexes_for_cur_subframe[i]; /** resample scale factors for the new block size */ if(s->channel[c].reuse_sf){ @@ -1033,7 +1033,7 @@ static void wma_inverse_channel_transfor wma_calc_decorrelation_matrix(s,&s->chgroup[i]); for(x=0;x<s->channels_for_cur_subframe;x++){ - int chan = s->channel_indices_for_cur_subframe[x]; + int chan = s->channel_indexes_for_cur_subframe[x]; if(s->chgroup[i].use_channel[chan] == 1){ // assign ptrs ch_data[cnt] = s->channel[chan].coeffs; ++cnt; @@ -1072,7 +1072,7 @@ static void wma_window(WMA3DecodeContext { int i; for(i=0;i<s->channels_for_cur_subframe;i++){ - int c = s->channel_indices_for_cur_subframe[i]; + int c = s->channel_indexes_for_cur_subframe[i]; int j = s->channel[c].cur_subframe; int x; // FIXME: use dsp.vector_fmul_window @@ -1150,7 +1150,7 @@ static int wma_decode_subframe(WMA3Decod if(offset == s->channel[i].decoded_samples && subframe_len == s->channel[i].subframe_len[s->channel[i].cur_subframe]){ total_samples -= s->channel[i].subframe_len[s->channel[i].cur_subframe]; s->channel[i].decoded_samples += s->channel[i].subframe_len[s->channel[i].cur_subframe]; - s->channel_indices_for_cur_subframe[s->channels_for_cur_subframe] = i; + s->channel_indexes_for_cur_subframe[s->channels_for_cur_subframe] = i; ++s->channels_for_cur_subframe; } } @@ -1164,7 +1164,7 @@ static int wma_decode_subframe(WMA3Decod /** configure the decoder for the current subframe */ for(i=0;i<s->channels_for_cur_subframe;i++){ - int c = s->channel_indices_for_cur_subframe[i]; + int c = s->channel_indexes_for_cur_subframe[i]; if(s->channel[c].num_subframes <= 1){ s->cValidBarkBand = s->num_sfb[0]; @@ -1214,7 +1214,7 @@ static int wma_decode_subframe(WMA3Decod for(i=0;i<s->channels_for_cur_subframe;i++){ - int c = s->channel_indices_for_cur_subframe[i]; + int c = s->channel_indexes_for_cur_subframe[i]; if((s->channel[c].transmit_coefs = get_bits(gb,1))) transmit_coeffs = 1; } @@ -1255,11 +1255,11 @@ static int wma_decode_subframe(WMA3Decod /** decode quantization step modifiers for every channel */ if(s->channels_for_cur_subframe == 1) - s->channel[s->channel_indices_for_cur_subframe[0]].quant_step_modifier = 0; + s->channel[s->channel_indexes_for_cur_subframe[0]].quant_step_modifier = 0; else{ int modifier_len = get_bits(gb,3); for(i=0;i<s->channels_for_cur_subframe;i++){ - int c = s->channel_indices_for_cur_subframe[i]; + int c = s->channel_indexes_for_cur_subframe[i]; s->channel[c].quant_step_modifier = 0; if(get_bits(gb,1)){ if(modifier_len) @@ -1281,7 +1281,7 @@ static int wma_decode_subframe(WMA3Decod /** parse coefficients */ for(i=0;i<s->channels_for_cur_subframe;i++){ - int c = s->channel_indices_for_cur_subframe[i]; + int c = s->channel_indexes_for_cur_subframe[i]; if(s->channel[c].transmit_coefs) decode_coeffs(s,gb,c); } @@ -1291,7 +1291,7 @@ static int wma_decode_subframe(WMA3Decod if(transmit_coeffs){ wma_inverse_channel_transform(s); for(i=0;i<s->channels_for_cur_subframe;i++){ - int c = s->channel_indices_for_cur_subframe[i]; + int c = s->channel_indexes_for_cur_subframe[i]; int b; float* dst; if(c == s->lfe_channel) @@ -1328,7 +1328,7 @@ static int wma_decode_subframe(WMA3Decod /** handled one subframe */ for(i=0;i<s->channels_for_cur_subframe;i++){ - int c = s->channel_indices_for_cur_subframe[i]; + int c = s->channel_indexes_for_cur_subframe[i]; if(s->channel[c].cur_subframe >= s->channel[c].num_subframes){ av_log(s->avctx,AV_LOG_ERROR,"broken subframe\n"); return 0;

1 0

[soc]: r3997 - jpeg2000/j2k.h
by diego 25 Jan '09

25 Jan '09

Author: diego Date: Sun Jan 25 16:27:18 2009 New Revision: 3997 Log: small wording/spellling fix Modified: jpeg2000/j2k.h Modified: jpeg2000/j2k.h ============================================================================== --- jpeg2000/j2k.h Fri Jan 23 01:38:23 2009 (r3996) +++ jpeg2000/j2k.h Sun Jan 25 16:27:18 2009 (r3997) @@ -136,7 +136,7 @@ typedef struct { } J2kCblk; ///< code block typedef struct { - uint16_t xi0, xi1, yi0, yi1; ///< indices of codeblocks ([xi0, xi1)) + uint16_t xi0, xi1, yi0, yi1; ///< codeblock indexes ([xi0, xi1)) J2kTgtNode *zerobits; J2kTgtNode *cblkincl; } J2kPrec; ///< precinct

1 0

[soc]: r3964 - in wmapro: wma3.h wma3data.h wma3dec.c
by faust3 25 Jan '09

25 Jan '09

Author: faust3 Date: Sat Jan 17 12:48:58 2009 New Revision: 3964 Log: renamed huffman tables of the vector coder so that their names contain the vectorsize Modified: wmapro/wma3.h wmapro/wma3data.h wmapro/wma3dec.c Modified: wmapro/wma3.h ============================================================================== --- wmapro/wma3.h Sat Jan 17 12:03:28 2009 (r3963) +++ wmapro/wma3.h Sat Jan 17 12:48:58 2009 (r3964) @@ -116,9 +116,9 @@ typedef struct WMA3DecodeContext { int* rgiBarkIndex; VLC exp_vlc; - VLC huff_0_vlc; - VLC huff_1_vlc; - VLC huff_2_vlc; + VLC vec4_vlc; + VLC vec2_vlc; + VLC vec1_vlc; VLC coef_vlc[2]; VLC rlc_vlc; int coef_max[2]; Modified: wmapro/wma3data.h ============================================================================== --- wmapro/wma3data.h Sat Jan 17 12:03:28 2009 (r3963) +++ wmapro/wma3data.h Sat Jan 17 12:48:58 2009 (r3964) @@ -239,9 +239,9 @@ static const uint8_t ff_wma3_coef1_huffb }; -#define FF_WMA3_HUFF_0_SIZE 127 -#define FF_WMA3_HUFF_0_MAXBITS 14 -static const uint32_t ff_wma3_huff_0_codes[FF_WMA3_HUFF_0_SIZE] = { +#define FF_WMA3_HUFF_VEC4_SIZE 127 +#define FF_WMA3_HUFF_VEC4_MAXBITS 14 +static const uint32_t ff_wma3_vec4_huffcodes[FF_WMA3_HUFF_VEC4_SIZE] = { 0x0019, 0x0027, 0x00F2, 0x03BA, 0x0930, 0x1267, 0x0031, 0x0030, 0x0097, 0x0221, 0x058B, 0x0124, 0x00EB, 0x01D4, 0x03D8, 0x0584, 0x0364, 0x045F, 0x0F66, 0x0931, 0x24CD, 0x002F, 0x0039, 0x00E8, @@ -260,7 +260,7 @@ static const uint32_t ff_wma3_huff_0_cod 0x03C4, 0x08B8, 0x078B, 0x0755, 0x0581, 0x0F67, 0x0000, }; -static const uint8_t ff_wma3_huff_0_bits[FF_WMA3_HUFF_0_SIZE] = { +static const uint8_t ff_wma3_vec4_huffbits[FF_WMA3_HUFF_VEC4_SIZE] = { 5, 6, 8, 10, 12, 13, 6, 6, 8, 10, 11, 9, 8, 9, 10, 11, 10, 11, 12, 12, 14, 6, 6, 8, @@ -280,9 +280,9 @@ static const uint8_t ff_wma3_huff_0_bits }; -#define FF_WMA3_HUFF_1_SIZE 137 -#define FF_WMA3_HUFF_1_MAXBITS 12 -static const uint32_t ff_wma3_huff_1_codes[FF_WMA3_HUFF_1_SIZE] = { +#define FF_WMA3_HUFF_VEC2_SIZE 137 +#define FF_WMA3_HUFF_VEC2_MAXBITS 12 +static const uint32_t ff_wma3_vec2_huffcodes[FF_WMA3_HUFF_VEC2_SIZE] = { 0x055, 0x01C, 0x01A, 0x02B, 0x028, 0x067, 0x08B, 0x039, 0x170, 0x10D, 0x2A5, 0x047, 0x464, 0x697, 0x523, 0x8CB, 0x01B, 0x00E, 0x000, 0x010, 0x012, 0x036, 0x048, 0x04C, @@ -303,7 +303,7 @@ static const uint32_t ff_wma3_huff_1_cod 0x001, }; -static const uint8_t ff_wma3_huff_1_bits[FF_WMA3_HUFF_1_SIZE] = { +static const uint8_t ff_wma3_vec2_huffbits[FF_WMA3_HUFF_VEC2_SIZE] = { 7, 6, 6, 6, 7, 7, 8, 9, 9, 10, 10, 11, 11, 11, 12, 12, 6, 4, 5, 5, 6, 6, 7, 8, @@ -325,9 +325,9 @@ static const uint8_t ff_wma3_huff_1_bits }; -#define FF_WMA3_HUFF_2_SIZE 101 -#define FF_WMA3_HUFF_2_MAXBITS 11 -static const uint32_t ff_wma3_huff_2_codes[FF_WMA3_HUFF_2_SIZE] = { +#define FF_WMA3_HUFF_VEC1_SIZE 101 +#define FF_WMA3_HUFF_VEC1_MAXBITS 11 +static const uint32_t ff_wma3_vec1_huffcodes[FF_WMA3_HUFF_VEC1_SIZE] = { 0x01A, 0x003, 0x017, 0x010, 0x00C, 0x009, 0x005, 0x000, 0x00D, 0x00A, 0x009, 0x00C, 0x00F, 0x002, 0x004, 0x007, 0x00B, 0x00F, 0x01C, 0x006, 0x010, 0x015, 0x01C, 0x022, @@ -343,7 +343,7 @@ static const uint32_t ff_wma3_huff_2_cod 0x297, 0x355, 0x372, 0x373, 0x016, }; -static const uint8_t ff_wma3_huff_2_bits[FF_WMA3_HUFF_2_SIZE] = { +static const uint8_t ff_wma3_vec1_huffbits[FF_WMA3_HUFF_VEC1_SIZE] = { 7, 6, 5, 5, 5, 5, 5, 5, 4, 4, 4, 4, 4, 5, 5, 5, 5, 5, 5, 6, 6, 6, 6, 6, Modified: wmapro/wma3dec.c ============================================================================== --- wmapro/wma3dec.c Sat Jan 17 12:03:28 2009 (r3963) +++ wmapro/wma3dec.c Sat Jan 17 12:48:58 2009 (r3964) @@ -112,9 +112,9 @@ static av_cold int wma3_decode_end(AVCod free_vlc(&s->exp_vlc); free_vlc(&s->coef_vlc[0]); free_vlc(&s->coef_vlc[1]); - free_vlc(&s->huff_0_vlc); - free_vlc(&s->huff_1_vlc); - free_vlc(&s->huff_2_vlc); + free_vlc(&s->vec4_vlc); + free_vlc(&s->vec2_vlc); + free_vlc(&s->vec1_vlc); free_vlc(&s->rlc_vlc); for(i=0 ; i<BLOCK_NB_SIZES ; i++) @@ -228,17 +228,17 @@ static av_cold int wma3_decode_init(AVCo s->coef_max[1] = ((21+VLCBITS-1)/VLCBITS); - init_vlc(&s->huff_0_vlc, VLCBITS, FF_WMA3_HUFF_0_SIZE, - ff_wma3_huff_0_bits, 1, 1, - ff_wma3_huff_0_codes, 4, 4, 0); + init_vlc(&s->vec4_vlc, VLCBITS, FF_WMA3_HUFF_VEC4_SIZE, + ff_wma3_vec4_huffbits, 1, 1, + ff_wma3_vec4_huffcodes, 4, 4, 0); - init_vlc(&s->huff_1_vlc, VLCBITS, FF_WMA3_HUFF_1_SIZE, - ff_wma3_huff_1_bits, 1, 1, - ff_wma3_huff_1_codes, 4, 4, 0); + init_vlc(&s->vec2_vlc, VLCBITS, FF_WMA3_HUFF_VEC2_SIZE, + ff_wma3_vec2_huffbits, 1, 1, + ff_wma3_vec2_huffcodes, 4, 4, 0); - init_vlc(&s->huff_2_vlc, VLCBITS, FF_WMA3_HUFF_2_SIZE, - ff_wma3_huff_2_bits, 1, 1, - ff_wma3_huff_2_codes, 4, 4, 0); + init_vlc(&s->vec1_vlc, VLCBITS, FF_WMA3_HUFF_VEC1_SIZE, + ff_wma3_vec1_huffbits, 1, 1, + ff_wma3_vec1_huffcodes, 4, 4, 0); init_vlc(&s->rlc_vlc, VLCBITS, FF_WMA3_HUFF_RLC_SIZE, ff_wma3_huff_rlc_bits, 1, 1, @@ -725,18 +725,18 @@ static inline void wma_get_vec4(WMA3Deco unsigned int idx; int i = 0; // read 4 values - idx = get_vlc2(s->getbit, s->huff_0_vlc.table, VLCBITS, ((FF_WMA3_HUFF_0_MAXBITS+VLCBITS-1)/VLCBITS)); + idx = get_vlc2(s->getbit, s->vec4_vlc.table, VLCBITS, ((FF_WMA3_HUFF_VEC4_MAXBITS+VLCBITS-1)/VLCBITS)); if ( idx == 126 ) { while(i < 4){ - idx = get_vlc2(s->getbit, s->huff_1_vlc.table, VLCBITS, ((FF_WMA3_HUFF_1_MAXBITS+VLCBITS-1)/VLCBITS)); + idx = get_vlc2(s->getbit, s->vec2_vlc.table, VLCBITS, ((FF_WMA3_HUFF_VEC2_MAXBITS+VLCBITS-1)/VLCBITS)); if ( idx == 136 ){ - vals[i] = get_vlc2(s->getbit, s->huff_2_vlc.table, VLCBITS, ((FF_WMA3_HUFF_2_MAXBITS+VLCBITS-1)/VLCBITS)); + vals[i] = get_vlc2(s->getbit, s->vec1_vlc.table, VLCBITS, ((FF_WMA3_HUFF_VEC1_MAXBITS+VLCBITS-1)/VLCBITS)); if(vals[i] == 100) vals[i] += wma_get_large_val(s); - vals[i+1] = get_vlc2(s->getbit, s->huff_2_vlc.table, VLCBITS, ((FF_WMA3_HUFF_2_MAXBITS+VLCBITS-1)/VLCBITS)); + vals[i+1] = get_vlc2(s->getbit, s->vec1_vlc.table, VLCBITS, ((FF_WMA3_HUFF_VEC1_MAXBITS+VLCBITS-1)/VLCBITS)); if(vals[i+1] == 100) vals[i+1] += wma_get_large_val(s); }else{

5 7

[soc]: r3970 - in wmapro: wma3data.h wma3dec.c
by faust3 24 Jan '09

24 Jan '09

Author: faust3 Date: Sat Jan 17 13:55:52 2009 New Revision: 3970 Log: moved scale factor run length level and run tables next to the huffman tables and renamed them to ff_wma3_scale_rl_level / run Modified: wmapro/wma3data.h wmapro/wma3dec.c Modified: wmapro/wma3data.h ============================================================================== --- wmapro/wma3data.h Sat Jan 17 13:46:01 2009 (r3969) +++ wmapro/wma3data.h Sat Jan 17 13:55:52 2009 (r3970) @@ -118,6 +118,27 @@ static const uint8_t ff_wma3_scale_rl_hu }; +static const uint8_t ff_wma3_scale_rl_run[] = { + 0, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, + 18, 19, 20, 21, 22, 23, 24, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, + 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 0, 1, 2, 3, 4, 5, + 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, + 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, + 19, 20, 21, 22, 23, 24, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 0, 1, + 0, 1, 0, 1, +}; + +static const uint8_t ff_wma3_scale_rl_level[] = { + 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, + 2, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, + 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, + 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, + 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, + 5, 5, 5, 5, 5, 5, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 7, 7, + 8, 8, 9, 9, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, +}; + + #define FF_WMA3_COEF0_SIZE 244 #define FF_WMA3_COEF0_MAXBITS 22 static const uint32_t ff_wma3_coef0_huffcodes[FF_WMA3_COEF0_SIZE] = { @@ -521,27 +542,6 @@ static const uint8_t ff_wma3_level_1[] = }; -static const uint8_t ff_wma3_run_mask[] = { - 0, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, - 18, 19, 20, 21, 22, 23, 24, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, - 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 0, 1, 2, 3, 4, 5, - 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, - 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, - 19, 20, 21, 22, 23, 24, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 0, 1, - 0, 1, 0, 1, -}; - -static const uint8_t ff_wma3_level_mask[] = { - 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, - 2, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, - 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, - 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, - 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, - 5, 5, 5, 5, 5, 5, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 7, 7, - 8, 8, 9, 9, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, -}; - - static const float ff_wma3_default_decorrelation_matrices[] = { 1.000000, 0.707031, -0.707031, 0.707031, 0.707031, 0.578125, 0.707031, 0.410156, 0.578125, -0.707031, 0.410156, 0.578125, 0.000000, -0.816406, Modified: wmapro/wma3dec.c ============================================================================== --- wmapro/wma3dec.c Sat Jan 17 13:46:01 2009 (r3969) +++ wmapro/wma3dec.c Sat Jan 17 13:55:52 2009 (r3970) @@ -914,8 +914,8 @@ static int wma_decode_scale_factors(WMA3 }else if(idx == 1){ break; }else{ - skip = ff_wma3_run_mask[idx-2]; - level_mask = ff_wma3_level_mask[idx-2]; + skip = ff_wma3_scale_rl_run[idx-2]; + level_mask = ff_wma3_scale_rl_level[idx-2]; val = get_bits(gb,1)-1; }

3 3

[soc]: r3977 - in wmapro: wma3.h wma3dec.c
by faust3 24 Jan '09

24 Jan '09

Author: faust3 Date: Sat Jan 17 16:56:55 2009 New Revision: 3977 Log: removed channel mask from the context Modified: wmapro/wma3.h wmapro/wma3dec.c Modified: wmapro/wma3.h ============================================================================== --- wmapro/wma3.h Sat Jan 17 16:29:58 2009 (r3976) +++ wmapro/wma3.h Sat Jan 17 16:56:55 2009 (r3977) @@ -143,7 +143,6 @@ typedef struct WMA3DecodeContext { /** extradata */ unsigned int decode_flags; //< used compression features - unsigned int dwChannelMask; uint8_t sample_bit_depth; //< bits per sample /** general frame info */ Modified: wmapro/wma3dec.c ============================================================================== --- wmapro/wma3dec.c Sat Jan 17 16:29:58 2009 (r3976) +++ wmapro/wma3dec.c Sat Jan 17 16:56:55 2009 (r3977) @@ -37,7 +37,6 @@ static void dump_context(WMA3DecodeConte #define PRINT(a,b) av_log(s->avctx,AV_LOG_ERROR," %s = %d\n", a, b); #define PRINT_HEX(a,b) av_log(s->avctx,AV_LOG_ERROR," %s = %x\n", a, b); - PRINT_HEX("ed channelmask",s->dwChannelMask); PRINT("ed sample bit depth",s->sample_bit_depth); PRINT_HEX("ed decode flags",s->decode_flags); PRINT("samples per frame",s->samples_per_frame); @@ -133,6 +132,7 @@ static av_cold int wma3_decode_init(AVCo WMA3DecodeContext *s = avctx->priv_data; uint8_t *edata_ptr = avctx->extradata; int* sfb_offsets; + unsigned int channel_mask; int i; s->avctx = avctx; @@ -142,7 +142,7 @@ static av_cold int wma3_decode_init(AVCo s->sample_bit_depth = 16; //avctx->bits_per_sample; if (avctx->extradata_size >= 18) { s->decode_flags = AV_RL16(edata_ptr+14); - s->dwChannelMask = AV_RL32(edata_ptr+2); + channel_mask = AV_RL32(edata_ptr+2); // s->sample_bit_depth = AV_RL16(edata_ptr); /** dump the extradata */ @@ -195,10 +195,10 @@ static av_cold int wma3_decode_init(AVCo /** extract lfe channel position */ s->lfe_channel = -1; - if(s->dwChannelMask & 8){ + if(channel_mask & 8){ unsigned int mask = 1; for(i=0;i<32;i++){ - if(s->dwChannelMask & mask) + if(channel_mask & mask) ++s->lfe_channel; if(mask & 8) break;

3 2

[soc]: r3996 - dvbmuxer/mpegenc.c
by diego 23 Jan '09

23 Jan '09

Author: diego Date: Fri Jan 23 01:38:23 2009 New Revision: 3996 Log: #ifdef --> #if where appropriate to sync with build system changes Modified: dvbmuxer/mpegenc.c Modified: dvbmuxer/mpegenc.c ============================================================================== --- dvbmuxer/mpegenc.c Fri Jan 23 01:35:26 2009 (r3995) +++ dvbmuxer/mpegenc.c Fri Jan 23 01:38:23 2009 (r3996) @@ -860,7 +860,7 @@ static int mpeg_mux_end(AVFormatContext return 0; } -#ifdef CONFIG_MPEG1SYSTEM_MUXER +#if CONFIG_MPEG1SYSTEM_MUXER AVOutputFormat mpeg1system_muxer = { "mpeg", NULL_IF_CONFIG_SMALL("MPEG-1 System format"), @@ -874,7 +874,8 @@ AVOutputFormat mpeg1system_muxer = { mpeg_mux_end, }; #endif -#ifdef CONFIG_MPEG1VCD_MUXER + +#if CONFIG_MPEG1VCD_MUXER AVOutputFormat mpeg1vcd_muxer = { "vcd", NULL_IF_CONFIG_SMALL("MPEG-1 System format (VCD)"), @@ -888,7 +889,8 @@ AVOutputFormat mpeg1vcd_muxer = { mpeg_mux_end, }; #endif -#ifdef CONFIG_MPEG2VOB_MUXER + +#if CONFIG_MPEG2VOB_MUXER AVOutputFormat mpeg2vob_muxer = { "vob", NULL_IF_CONFIG_SMALL("MPEG-2 PS format (VOB)"), @@ -904,7 +906,7 @@ AVOutputFormat mpeg2vob_muxer = { #endif /* Same as mpeg2vob_mux except that the pack size is 2324 */ -#ifdef CONFIG_MPEG2SVCD_MUXER +#if CONFIG_MPEG2SVCD_MUXER AVOutputFormat mpeg2svcd_muxer = { "svcd", NULL_IF_CONFIG_SMALL("MPEG-2 PS format (VOB)"), @@ -920,7 +922,7 @@ AVOutputFormat mpeg2svcd_muxer = { #endif /* Same as mpeg2vob_mux except the 'is_dvd' flag is set to produce NAV pkts */ -#ifdef CONFIG_MPEG2DVD_MUXER +#if CONFIG_MPEG2DVD_MUXER AVOutputFormat mpeg2dvd_muxer = { "dvd", NULL_IF_CONFIG_SMALL("MPEG-2 PS format (DVD VOB)"),

1 0

[soc]: r3995 - dirac/libavcodec/diracenc.c
by diego 23 Jan '09

23 Jan '09

Author: diego Date: Fri Jan 23 01:35:26 2009 New Revision: 3995 Log: Remove useless #ifdef. Modified: dirac/libavcodec/diracenc.c Modified: dirac/libavcodec/diracenc.c ============================================================================== --- dirac/libavcodec/diracenc.c Fri Jan 23 01:12:17 2009 (r3994) +++ dirac/libavcodec/diracenc.c Fri Jan 23 01:35:26 2009 (r3995) @@ -953,7 +953,6 @@ static int encode_frame(AVCodecContext * return size; } -#ifdef CONFIG_ENCODERS AVCodec dirac_encoder = { "dirac", CODEC_TYPE_VIDEO, @@ -964,4 +963,3 @@ AVCodec dirac_encoder = { encode_end, .pix_fmts = (enum PixelFormat[]) {PIX_FMT_YUV420P, -1} }; -#endif

1 0

[PATCH] [soc: dvbmuxer] Comply to trunk r16600
by Tomer Barletz 23 Jan '09

23 Jan '09

In r16600 all ENABLE_* was renamed to CONFIG_* in file mpegenc.c. Apply to soc branch too. Tomer

2 1