Subversion Repositories Kolibri OS

Rev

Blame | Last modification | View Log | RSS feed

  1. /*
  2.  * audio resampling
  3.  * Copyright (c) 2004-2012 Michael Niedermayer <michaelni@gmx.at>
  4.  *
  5.  * This file is part of FFmpeg.
  6.  *
  7.  * FFmpeg is free software; you can redistribute it and/or
  8.  * modify it under the terms of the GNU Lesser General Public
  9.  * License as published by the Free Software Foundation; either
  10.  * version 2.1 of the License, or (at your option) any later version.
  11.  *
  12.  * FFmpeg is distributed in the hope that it will be useful,
  13.  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14.  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15.  * Lesser General Public License for more details.
  16.  *
  17.  * You should have received a copy of the GNU Lesser General Public
  18.  * License along with FFmpeg; if not, write to the Free Software
  19.  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  20.  */
  21.  
  22. /**
  23.  * @file
  24.  * audio resampling
  25.  * @author Michael Niedermayer <michaelni@gmx.at>
  26.  */
  27.  
  28. #include "libavutil/avassert.h"
  29. #include "resample.h"
  30.  
  31. /**
  32.  * 0th order modified bessel function of the first kind.
  33.  */
  34. static double bessel(double x){
  35.     double v=1;
  36.     double lastv=0;
  37.     double t=1;
  38.     int i;
  39.     static const double inv[100]={
  40.  1.0/( 1* 1), 1.0/( 2* 2), 1.0/( 3* 3), 1.0/( 4* 4), 1.0/( 5* 5), 1.0/( 6* 6), 1.0/( 7* 7), 1.0/( 8* 8), 1.0/( 9* 9), 1.0/(10*10),
  41.  1.0/(11*11), 1.0/(12*12), 1.0/(13*13), 1.0/(14*14), 1.0/(15*15), 1.0/(16*16), 1.0/(17*17), 1.0/(18*18), 1.0/(19*19), 1.0/(20*20),
  42.  1.0/(21*21), 1.0/(22*22), 1.0/(23*23), 1.0/(24*24), 1.0/(25*25), 1.0/(26*26), 1.0/(27*27), 1.0/(28*28), 1.0/(29*29), 1.0/(30*30),
  43.  1.0/(31*31), 1.0/(32*32), 1.0/(33*33), 1.0/(34*34), 1.0/(35*35), 1.0/(36*36), 1.0/(37*37), 1.0/(38*38), 1.0/(39*39), 1.0/(40*40),
  44.  1.0/(41*41), 1.0/(42*42), 1.0/(43*43), 1.0/(44*44), 1.0/(45*45), 1.0/(46*46), 1.0/(47*47), 1.0/(48*48), 1.0/(49*49), 1.0/(50*50),
  45.  1.0/(51*51), 1.0/(52*52), 1.0/(53*53), 1.0/(54*54), 1.0/(55*55), 1.0/(56*56), 1.0/(57*57), 1.0/(58*58), 1.0/(59*59), 1.0/(60*60),
  46.  1.0/(61*61), 1.0/(62*62), 1.0/(63*63), 1.0/(64*64), 1.0/(65*65), 1.0/(66*66), 1.0/(67*67), 1.0/(68*68), 1.0/(69*69), 1.0/(70*70),
  47.  1.0/(71*71), 1.0/(72*72), 1.0/(73*73), 1.0/(74*74), 1.0/(75*75), 1.0/(76*76), 1.0/(77*77), 1.0/(78*78), 1.0/(79*79), 1.0/(80*80),
  48.  1.0/(81*81), 1.0/(82*82), 1.0/(83*83), 1.0/(84*84), 1.0/(85*85), 1.0/(86*86), 1.0/(87*87), 1.0/(88*88), 1.0/(89*89), 1.0/(90*90),
  49.  1.0/(91*91), 1.0/(92*92), 1.0/(93*93), 1.0/(94*94), 1.0/(95*95), 1.0/(96*96), 1.0/(97*97), 1.0/(98*98), 1.0/(99*99), 1.0/(10000)
  50.     };
  51.  
  52.     x= x*x/4;
  53.     for(i=0; v != lastv; i++){
  54.         lastv=v;
  55.         t *= x*inv[i];
  56.         v += t;
  57.         av_assert2(i<99);
  58.     }
  59.     return v;
  60. }
  61.  
  62. /**
  63.  * builds a polyphase filterbank.
  64.  * @param factor resampling factor
  65.  * @param scale wanted sum of coefficients for each filter
  66.  * @param filter_type  filter type
  67.  * @param kaiser_beta  kaiser window beta
  68.  * @return 0 on success, negative on error
  69.  */
  70. static int build_filter(ResampleContext *c, void *filter, double factor, int tap_count, int alloc, int phase_count, int scale,
  71.                         int filter_type, int kaiser_beta){
  72.     int ph, i;
  73.     double x, y, w;
  74.     double *tab = av_malloc_array(tap_count,  sizeof(*tab));
  75.     const int center= (tap_count-1)/2;
  76.  
  77.     if (!tab)
  78.         return AVERROR(ENOMEM);
  79.  
  80.     /* if upsampling, only need to interpolate, no filter */
  81.     if (factor > 1.0)
  82.         factor = 1.0;
  83.  
  84.     for(ph=0;ph<phase_count;ph++) {
  85.         double norm = 0;
  86.         for(i=0;i<tap_count;i++) {
  87.             x = M_PI * ((double)(i - center) - (double)ph / phase_count) * factor;
  88.             if (x == 0) y = 1.0;
  89.             else        y = sin(x) / x;
  90.             switch(filter_type){
  91.             case SWR_FILTER_TYPE_CUBIC:{
  92.                 const float d= -0.5; //first order derivative = -0.5
  93.                 x = fabs(((double)(i - center) - (double)ph / phase_count) * factor);
  94.                 if(x<1.0) y= 1 - 3*x*x + 2*x*x*x + d*(            -x*x + x*x*x);
  95.                 else      y=                       d*(-4 + 8*x - 5*x*x + x*x*x);
  96.                 break;}
  97.             case SWR_FILTER_TYPE_BLACKMAN_NUTTALL:
  98.                 w = 2.0*x / (factor*tap_count) + M_PI;
  99.                 y *= 0.3635819 - 0.4891775 * cos(w) + 0.1365995 * cos(2*w) - 0.0106411 * cos(3*w);
  100.                 break;
  101.             case SWR_FILTER_TYPE_KAISER:
  102.                 w = 2.0*x / (factor*tap_count*M_PI);
  103.                 y *= bessel(kaiser_beta*sqrt(FFMAX(1-w*w, 0)));
  104.                 break;
  105.             default:
  106.                 av_assert0(0);
  107.             }
  108.  
  109.             tab[i] = y;
  110.             norm += y;
  111.         }
  112.  
  113.         /* normalize so that an uniform color remains the same */
  114.         switch(c->format){
  115.         case AV_SAMPLE_FMT_S16P:
  116.             for(i=0;i<tap_count;i++)
  117.                 ((int16_t*)filter)[ph * alloc + i] = av_clip(lrintf(tab[i] * scale / norm), INT16_MIN, INT16_MAX);
  118.             break;
  119.         case AV_SAMPLE_FMT_S32P:
  120.             for(i=0;i<tap_count;i++)
  121.                 ((int32_t*)filter)[ph * alloc + i] = av_clipl_int32(llrint(tab[i] * scale / norm));
  122.             break;
  123.         case AV_SAMPLE_FMT_FLTP:
  124.             for(i=0;i<tap_count;i++)
  125.                 ((float*)filter)[ph * alloc + i] = tab[i] * scale / norm;
  126.             break;
  127.         case AV_SAMPLE_FMT_DBLP:
  128.             for(i=0;i<tap_count;i++)
  129.                 ((double*)filter)[ph * alloc + i] = tab[i] * scale / norm;
  130.             break;
  131.         }
  132.     }
  133. #if 0
  134.     {
  135. #define LEN 1024
  136.         int j,k;
  137.         double sine[LEN + tap_count];
  138.         double filtered[LEN];
  139.         double maxff=-2, minff=2, maxsf=-2, minsf=2;
  140.         for(i=0; i<LEN; i++){
  141.             double ss=0, sf=0, ff=0;
  142.             for(j=0; j<LEN+tap_count; j++)
  143.                 sine[j]= cos(i*j*M_PI/LEN);
  144.             for(j=0; j<LEN; j++){
  145.                 double sum=0;
  146.                 ph=0;
  147.                 for(k=0; k<tap_count; k++)
  148.                     sum += filter[ph * tap_count + k] * sine[k+j];
  149.                 filtered[j]= sum / (1<<FILTER_SHIFT);
  150.                 ss+= sine[j + center] * sine[j + center];
  151.                 ff+= filtered[j] * filtered[j];
  152.                 sf+= sine[j + center] * filtered[j];
  153.             }
  154.             ss= sqrt(2*ss/LEN);
  155.             ff= sqrt(2*ff/LEN);
  156.             sf= 2*sf/LEN;
  157.             maxff= FFMAX(maxff, ff);
  158.             minff= FFMIN(minff, ff);
  159.             maxsf= FFMAX(maxsf, sf);
  160.             minsf= FFMIN(minsf, sf);
  161.             if(i%11==0){
  162.                 av_log(NULL, AV_LOG_ERROR, "i:%4d ss:%f ff:%13.6e-%13.6e sf:%13.6e-%13.6e\n", i, ss, maxff, minff, maxsf, minsf);
  163.                 minff=minsf= 2;
  164.                 maxff=maxsf= -2;
  165.             }
  166.         }
  167.     }
  168. #endif
  169.  
  170.     av_free(tab);
  171.     return 0;
  172. }
  173.  
  174. static ResampleContext *resample_init(ResampleContext *c, int out_rate, int in_rate, int filter_size, int phase_shift, int linear,
  175.                                     double cutoff0, enum AVSampleFormat format, enum SwrFilterType filter_type, int kaiser_beta,
  176.                                     double precision, int cheby)
  177. {
  178.     double cutoff = cutoff0? cutoff0 : 0.97;
  179.     double factor= FFMIN(out_rate * cutoff / in_rate, 1.0);
  180.     int phase_count= 1<<phase_shift;
  181.  
  182.     if (!c || c->phase_shift != phase_shift || c->linear!=linear || c->factor != factor
  183.            || c->filter_length != FFMAX((int)ceil(filter_size/factor), 1) || c->format != format
  184.            || c->filter_type != filter_type || c->kaiser_beta != kaiser_beta) {
  185.         c = av_mallocz(sizeof(*c));
  186.         if (!c)
  187.             return NULL;
  188.  
  189.         c->format= format;
  190.  
  191.         c->felem_size= av_get_bytes_per_sample(c->format);
  192.  
  193.         switch(c->format){
  194.         case AV_SAMPLE_FMT_S16P:
  195.             c->filter_shift = 15;
  196.             break;
  197.         case AV_SAMPLE_FMT_S32P:
  198.             c->filter_shift = 30;
  199.             break;
  200.         case AV_SAMPLE_FMT_FLTP:
  201.         case AV_SAMPLE_FMT_DBLP:
  202.             c->filter_shift = 0;
  203.             break;
  204.         default:
  205.             av_log(NULL, AV_LOG_ERROR, "Unsupported sample format\n");
  206.             av_assert0(0);
  207.         }
  208.  
  209.         if (filter_size/factor > INT32_MAX/256) {
  210.             av_log(NULL, AV_LOG_ERROR, "Filter length too large\n");
  211.             goto error;
  212.         }
  213.  
  214.         c->phase_shift   = phase_shift;
  215.         c->phase_mask    = phase_count - 1;
  216.         c->linear        = linear;
  217.         c->factor        = factor;
  218.         c->filter_length = FFMAX((int)ceil(filter_size/factor), 1);
  219.         c->filter_alloc  = FFALIGN(c->filter_length, 8);
  220.         c->filter_bank   = av_calloc(c->filter_alloc, (phase_count+1)*c->felem_size);
  221.         c->filter_type   = filter_type;
  222.         c->kaiser_beta   = kaiser_beta;
  223.         if (!c->filter_bank)
  224.             goto error;
  225.         if (build_filter(c, (void*)c->filter_bank, factor, c->filter_length, c->filter_alloc, phase_count, 1<<c->filter_shift, filter_type, kaiser_beta))
  226.             goto error;
  227.         memcpy(c->filter_bank + (c->filter_alloc*phase_count+1)*c->felem_size, c->filter_bank, (c->filter_alloc-1)*c->felem_size);
  228.         memcpy(c->filter_bank + (c->filter_alloc*phase_count  )*c->felem_size, c->filter_bank + (c->filter_alloc - 1)*c->felem_size, c->felem_size);
  229.     }
  230.  
  231.     c->compensation_distance= 0;
  232.     if(!av_reduce(&c->src_incr, &c->dst_incr, out_rate, in_rate * (int64_t)phase_count, INT32_MAX/2))
  233.         goto error;
  234.     while (c->dst_incr < (1<<20) && c->src_incr < (1<<20)) {
  235.         c->dst_incr *= 2;
  236.         c->src_incr *= 2;
  237.     }
  238.     c->ideal_dst_incr = c->dst_incr;
  239.     c->dst_incr_div   = c->dst_incr / c->src_incr;
  240.     c->dst_incr_mod   = c->dst_incr % c->src_incr;
  241.  
  242.     c->index= -phase_count*((c->filter_length-1)/2);
  243.     c->frac= 0;
  244.  
  245.     swri_resample_dsp_init(c);
  246.  
  247.     return c;
  248. error:
  249.     av_freep(&c->filter_bank);
  250.     av_free(c);
  251.     return NULL;
  252. }
  253.  
  254. static void resample_free(ResampleContext **c){
  255.     if(!*c)
  256.         return;
  257.     av_freep(&(*c)->filter_bank);
  258.     av_freep(c);
  259. }
  260.  
  261. static int set_compensation(ResampleContext *c, int sample_delta, int compensation_distance){
  262.     c->compensation_distance= compensation_distance;
  263.     if (compensation_distance)
  264.         c->dst_incr = c->ideal_dst_incr - c->ideal_dst_incr * (int64_t)sample_delta / compensation_distance;
  265.     else
  266.         c->dst_incr = c->ideal_dst_incr;
  267.  
  268.     c->dst_incr_div   = c->dst_incr / c->src_incr;
  269.     c->dst_incr_mod   = c->dst_incr % c->src_incr;
  270.  
  271.     return 0;
  272. }
  273.  
  274. static int swri_resample(ResampleContext *c,
  275.                          uint8_t *dst, const uint8_t *src, int *consumed,
  276.                          int src_size, int dst_size, int update_ctx)
  277. {
  278.     if (c->filter_length == 1 && c->phase_shift == 0) {
  279.         int index= c->index;
  280.         int frac= c->frac;
  281.         int64_t index2= (1LL<<32)*c->frac/c->src_incr + (1LL<<32)*index;
  282.         int64_t incr= (1LL<<32) * c->dst_incr / c->src_incr;
  283.         int new_size = (src_size * (int64_t)c->src_incr - frac + c->dst_incr - 1) / c->dst_incr;
  284.  
  285.         dst_size= FFMIN(dst_size, new_size);
  286.         c->dsp.resample_one(dst, src, dst_size, index2, incr);
  287.  
  288.         index += dst_size * c->dst_incr_div;
  289.         index += (frac + dst_size * (int64_t)c->dst_incr_mod) / c->src_incr;
  290.         av_assert2(index >= 0);
  291.         *consumed= index;
  292.         if (update_ctx) {
  293.             c->frac   = (frac + dst_size * (int64_t)c->dst_incr_mod) % c->src_incr;
  294.             c->index = 0;
  295.         }
  296.     } else {
  297.         int64_t end_index = (1LL + src_size - c->filter_length) << c->phase_shift;
  298.         int64_t delta_frac = (end_index - c->index) * c->src_incr - c->frac;
  299.         int delta_n = (delta_frac + c->dst_incr - 1) / c->dst_incr;
  300.  
  301.         dst_size = FFMIN(dst_size, delta_n);
  302.         if (dst_size > 0) {
  303.             *consumed = c->dsp.resample(c, dst, src, dst_size, update_ctx);
  304.         } else {
  305.             *consumed = 0;
  306.         }
  307.     }
  308.  
  309.     return dst_size;
  310. }
  311.  
  312. static int multiple_resample(ResampleContext *c, AudioData *dst, int dst_size, AudioData *src, int src_size, int *consumed){
  313.     int i, ret= -1;
  314.     int av_unused mm_flags = av_get_cpu_flags();
  315.     int need_emms = c->format == AV_SAMPLE_FMT_S16P && ARCH_X86_32 &&
  316.                     (mm_flags & (AV_CPU_FLAG_MMX2 | AV_CPU_FLAG_SSE2)) == AV_CPU_FLAG_MMX2;
  317.     int64_t max_src_size = (INT64_MAX >> (c->phase_shift+1)) / c->src_incr;
  318.  
  319.     if (c->compensation_distance)
  320.         dst_size = FFMIN(dst_size, c->compensation_distance);
  321.     src_size = FFMIN(src_size, max_src_size);
  322.  
  323.     for(i=0; i<dst->ch_count; i++){
  324.         ret= swri_resample(c, dst->ch[i], src->ch[i],
  325.                            consumed, src_size, dst_size, i+1==dst->ch_count);
  326.     }
  327.     if(need_emms)
  328.         emms_c();
  329.  
  330.     if (c->compensation_distance) {
  331.         c->compensation_distance -= ret;
  332.         if (!c->compensation_distance) {
  333.             c->dst_incr     = c->ideal_dst_incr;
  334.             c->dst_incr_div = c->dst_incr / c->src_incr;
  335.             c->dst_incr_mod = c->dst_incr % c->src_incr;
  336.         }
  337.     }
  338.  
  339.     return ret;
  340. }
  341.  
  342. static int64_t get_delay(struct SwrContext *s, int64_t base){
  343.     ResampleContext *c = s->resample;
  344.     int64_t num = s->in_buffer_count - (c->filter_length-1)/2;
  345.     num *= 1 << c->phase_shift;
  346.     num -= c->index;
  347.     num *= c->src_incr;
  348.     num -= c->frac;
  349.     return av_rescale(num, base, s->in_sample_rate*(int64_t)c->src_incr << c->phase_shift);
  350. }
  351.  
  352. static int64_t get_out_samples(struct SwrContext *s, int in_samples) {
  353.     ResampleContext *c = s->resample;
  354.     // The + 2 are added to allow implementations to be slightly inaccurate, they should not be needed currently.
  355.     // They also make it easier to proof that changes and optimizations do not
  356.     // break the upper bound.
  357.     int64_t num = s->in_buffer_count + 2LL + in_samples;
  358.     num *= 1 << c->phase_shift;
  359.     num -= c->index;
  360.     num = av_rescale_rnd(num, s->out_sample_rate, ((int64_t)s->in_sample_rate) << c->phase_shift, AV_ROUND_UP) + 2;
  361.  
  362.     if (c->compensation_distance) {
  363.         if (num > INT_MAX)
  364.             return AVERROR(EINVAL);
  365.  
  366.         num = FFMAX(num, (num * c->ideal_dst_incr - 1) / c->dst_incr + 1);
  367.     }
  368.     return num;
  369. }
  370.  
  371. static int resample_flush(struct SwrContext *s) {
  372.     AudioData *a= &s->in_buffer;
  373.     int i, j, ret;
  374.     if((ret = swri_realloc_audio(a, s->in_buffer_index + 2*s->in_buffer_count)) < 0)
  375.         return ret;
  376.     av_assert0(a->planar);
  377.     for(i=0; i<a->ch_count; i++){
  378.         for(j=0; j<s->in_buffer_count; j++){
  379.             memcpy(a->ch[i] + (s->in_buffer_index+s->in_buffer_count+j  )*a->bps,
  380.                 a->ch[i] + (s->in_buffer_index+s->in_buffer_count-j-1)*a->bps, a->bps);
  381.         }
  382.     }
  383.     s->in_buffer_count += (s->in_buffer_count+1)/2;
  384.     return 0;
  385. }
  386.  
  387. // in fact the whole handle multiple ridiculously small buffers might need more thinking...
  388. static int invert_initial_buffer(ResampleContext *c, AudioData *dst, const AudioData *src,
  389.                                  int in_count, int *out_idx, int *out_sz)
  390. {
  391.     int n, ch, num = FFMIN(in_count + *out_sz, c->filter_length + 1), res;
  392.  
  393.     if (c->index >= 0)
  394.         return 0;
  395.  
  396.     if ((res = swri_realloc_audio(dst, c->filter_length * 2 + 1)) < 0)
  397.         return res;
  398.  
  399.     // copy
  400.     for (n = *out_sz; n < num; n++) {
  401.         for (ch = 0; ch < src->ch_count; ch++) {
  402.             memcpy(dst->ch[ch] + ((c->filter_length + n) * c->felem_size),
  403.                    src->ch[ch] + ((n - *out_sz) * c->felem_size), c->felem_size);
  404.         }
  405.     }
  406.  
  407.     // if not enough data is in, return and wait for more
  408.     if (num < c->filter_length + 1) {
  409.         *out_sz = num;
  410.         *out_idx = c->filter_length;
  411.         return INT_MAX;
  412.     }
  413.  
  414.     // else invert
  415.     for (n = 1; n <= c->filter_length; n++) {
  416.         for (ch = 0; ch < src->ch_count; ch++) {
  417.             memcpy(dst->ch[ch] + ((c->filter_length - n) * c->felem_size),
  418.                    dst->ch[ch] + ((c->filter_length + n) * c->felem_size),
  419.                    c->felem_size);
  420.         }
  421.     }
  422.  
  423.     res = num - *out_sz;
  424.     *out_idx = c->filter_length + (c->index >> c->phase_shift);
  425.     *out_sz = FFMAX(*out_sz + c->filter_length,
  426.                     1 + c->filter_length * 2) - *out_idx;
  427.     c->index &= c->phase_mask;
  428.  
  429.     return FFMAX(res, 0);
  430. }
  431.  
  432. struct Resampler const swri_resampler={
  433.   resample_init,
  434.   resample_free,
  435.   multiple_resample,
  436.   resample_flush,
  437.   set_compensation,
  438.   get_delay,
  439.   invert_initial_buffer,
  440.   get_out_samples,
  441. };
  442.