layer3.c

mpg123 是 MPEG 1.0/2.0/2.5 的实时播放软件

/*
	leyer3.c: the layer 3 decoder

	copyright 1995-2006 by the mpg123 project - free software under the terms of the LGPL 2.1
	see COPYING and AUTHORS files in distribution or http://mpg123.de
	initially written by Michael Hipp

	Optimize-TODO: put short bands into the band-field without the stride of 3 reals
	Length-optimze: unify long and short band code where it is possible

	The int-vs-pointer situation has to be cleaned up.
*/

#include <stdlib.h>
#include "config.h"
#include "mpg123.h"
#include "huffman.h"

#include "common.h"
#include "debug.h"

#include "getbits.h"

static real ispow[8207];
static real aa_ca[8],aa_cs[8];
static real COS1[12][6];
static real win[4][36];
static real win1[4][36];
static real gainpow2[256+118+4];
#ifdef USE_3DNOW
real COS9[9];
static real COS6_1,COS6_2;
real tfcos36[9];
#else
static real COS9[9];
static real COS6_1,COS6_2;
static real tfcos36[9];
#endif
static real tfcos12[3];
#define NEW_DCT9
#ifdef NEW_DCT9
static real cos9[3],cos18[3];
#endif

struct bandInfoStruct {
  int longIdx[23];
  int longDiff[22];
  int shortIdx[14];
  int shortDiff[13];
};

int longLimit[9][23];
int shortLimit[9][14];

struct bandInfoStruct bandInfo[9] = { 

/* MPEG 1.0 */
 { {0,4,8,12,16,20,24,30,36,44,52,62,74, 90,110,134,162,196,238,288,342,418,576},
   {4,4,4,4,4,4,6,6,8, 8,10,12,16,20,24,28,34,42,50,54, 76,158},
   {0,4*3,8*3,12*3,16*3,22*3,30*3,40*3,52*3,66*3, 84*3,106*3,136*3,192*3},
   {4,4,4,4,6,8,10,12,14,18,22,30,56} } ,

 { {0,4,8,12,16,20,24,30,36,42,50,60,72, 88,106,128,156,190,230,276,330,384,576},
   {4,4,4,4,4,4,6,6,6, 8,10,12,16,18,22,28,34,40,46,54, 54,192},
   {0,4*3,8*3,12*3,16*3,22*3,28*3,38*3,50*3,64*3, 80*3,100*3,126*3,192*3},
   {4,4,4,4,6,6,10,12,14,16,20,26,66} } ,

 { {0,4,8,12,16,20,24,30,36,44,54,66,82,102,126,156,194,240,296,364,448,550,576} ,
   {4,4,4,4,4,4,6,6,8,10,12,16,20,24,30,38,46,56,68,84,102, 26} ,
   {0,4*3,8*3,12*3,16*3,22*3,30*3,42*3,58*3,78*3,104*3,138*3,180*3,192*3} ,
   {4,4,4,4,6,8,12,16,20,26,34,42,12} }  ,

/* MPEG 2.0 */
 { {0,6,12,18,24,30,36,44,54,66,80,96,116,140,168,200,238,284,336,396,464,522,576},
   {6,6,6,6,6,6,8,10,12,14,16,20,24,28,32,38,46,52,60,68,58,54 } ,
   {0,4*3,8*3,12*3,18*3,24*3,32*3,42*3,56*3,74*3,100*3,132*3,174*3,192*3} ,
   {4,4,4,6,6,8,10,14,18,26,32,42,18 } } ,

/* mhipp trunk has 330 -> 332 without further explanation ... */
 { {0,6,12,18,24,30,36,44,54,66,80,96,114,136,162,194,232,278,330,394,464,540,576},
   {6,6,6,6,6,6,8,10,12,14,16,18,22,26,32,38,46,52,64,70,76,36 } ,
   {0,4*3,8*3,12*3,18*3,26*3,36*3,48*3,62*3,80*3,104*3,136*3,180*3,192*3} ,
   {4,4,4,6,8,10,12,14,18,24,32,44,12 } } ,

 { {0,6,12,18,24,30,36,44,54,66,80,96,116,140,168,200,238,284,336,396,464,522,576},
   {6,6,6,6,6,6,8,10,12,14,16,20,24,28,32,38,46,52,60,68,58,54 },
   {0,4*3,8*3,12*3,18*3,26*3,36*3,48*3,62*3,80*3,104*3,134*3,174*3,192*3},
   {4,4,4,6,8,10,12,14,18,24,30,40,18 } } ,
/* MPEG 2.5 */
 { {0,6,12,18,24,30,36,44,54,66,80,96,116,140,168,200,238,284,336,396,464,522,576} ,
   {6,6,6,6,6,6,8,10,12,14,16,20,24,28,32,38,46,52,60,68,58,54},
   {0,12,24,36,54,78,108,144,186,240,312,402,522,576},
   {4,4,4,6,8,10,12,14,18,24,30,40,18} },
 { {0,6,12,18,24,30,36,44,54,66,80,96,116,140,168,200,238,284,336,396,464,522,576} ,
   {6,6,6,6,6,6,8,10,12,14,16,20,24,28,32,38,46,52,60,68,58,54},
   {0,12,24,36,54,78,108,144,186,240,312,402,522,576},
   {4,4,4,6,8,10,12,14,18,24,30,40,18} },
 { {0,12,24,36,48,60,72,88,108,132,160,192,232,280,336,400,476,566,568,570,572,574,576},
   {12,12,12,12,12,12,16,20,24,28,32,40,48,56,64,76,90,2,2,2,2,2},
   {0, 24, 48, 72,108,156,216,288,372,480,486,492,498,576},
   {8,8,8,12,16,20,24,28,36,2,2,2,26} } ,
};

static int mapbuf0[9][152];
static int mapbuf1[9][156];
static int mapbuf2[9][44];
static int *map[9][3];
static int *mapend[9][3];

static unsigned int n_slen2[512]; /* MPEG 2.0 slen for 'normal' mode */
static unsigned int i_slen2[256]; /* MPEG 2.0 slen for intensity stereo */

static real tan1_1[16],tan2_1[16],tan1_2[16],tan2_2[16];
static real pow1_1[2][16],pow2_1[2][16],pow1_2[2][16],pow2_2[2][16];

#ifdef GAPLESS
/* still a dirty hack, places in bytes (zero-based)... */
static unsigned long position; /* position in raw decoder bytestream */
static unsigned long begin; /* first byte to play == number to skip */
static unsigned long end; /* last byte to play */
static unsigned long ignore; /* forcedly ignore stuff in between */
static int bytified;

/* input in bytes already */
void layer3_gapless_init(unsigned long b, unsigned long e)
{
	bytified = 0;
	position = 0;
	ignore = 0;
	begin = b;
	end = e;
	debug2("layer3_gapless_init: from %lu to %lu samples", begin, end);
}

void layer3_gapless_set_position(unsigned long frames, struct frame* fr, struct audio_info_struct *ai)
{
	position = samples_to_bytes(frames*spf(fr), fr, ai);
	debug1("set; position now %lu", position);
}

void layer3_gapless_bytify(struct frame *fr, struct audio_info_struct *ai)
{
	if(!bytified)
	{
		begin = samples_to_bytes(begin, fr, ai);
		end = samples_to_bytes(end, fr, ai);
		bytified = 1;
		debug2("bytified: begin=%lu; end=%5lu", begin, end);
	}
}

/* I need initialized fr here! */
void layer3_gapless_set_ignore(unsigned long frames, struct frame *fr, struct audio_info_struct *ai)
{
	ignore = samples_to_bytes(frames*spf(fr), fr, ai);
}

/*
	take the (partially or fully) filled and remove stuff for gapless mode if needed
	pcm_point may then be smaller than before...
*/
void layer3_gapless_buffercheck()
{
	/* pcm_point bytes added since last position... */
	unsigned long new_pos = position + pcm_point;
	if(begin && (position < begin))
	{
		debug4("new_pos %lu (old: %lu), begin %lu, pcm_point %i", new_pos, position, begin, pcm_point);
		if(new_pos < begin)
		{
			if(ignore > pcm_point) ignore -= pcm_point;
			else ignore = 0;
			pcm_point = 0; /* full of padding/delay */
		}
		else
		{
			unsigned long ignored = begin-position;
			/* we need to shift the memory to the left... */
			debug3("old pcm_point: %i, begin %lu; good bytes: %i", pcm_point, begin, (int)(new_pos-begin));
			if(ignore > ignored) ignore -= ignored;
			else ignore = 0;
			pcm_point -= ignored;
			debug3("shifting %i bytes from %p to %p", pcm_point, pcm_sample+(int)(begin-position), pcm_sample);
			memmove(pcm_sample, pcm_sample+(int)(begin-position), pcm_point);
		}
	}
	/* I don't cover the case with both end and begin in chunk! */
	else if(end && (new_pos > end))
	{
		ignore = 0;
		/* either end in current chunk or chunk totally out */
		debug2("ending at position %lu / point %i", new_pos, pcm_point);
		if(position < end)	pcm_point -= new_pos-end;
		else pcm_point = 0;
		debug1("set pcm_point to %i", pcm_point);
	}
	else if(ignore)
	{
		if(pcm_point < ignore)
		{
			ignore -= pcm_point;
			debug2("ignored %i bytes; pcm_point = 0; %lu bytes left", pcm_point, ignore);
			pcm_point = 0;
		}
		else
		{
			/* we need to shift the memory to the left... */
			debug3("old pcm_point: %i, to ignore: %lu; good bytes: %i", pcm_point, ignore, pcm_point-(int)ignore);
			pcm_point -= ignore;
			debug3("shifting %i bytes from %p to %p", pcm_point, pcm_sample+ignore, pcm_sample);
			memmove(pcm_sample, pcm_sample+ignore, pcm_point);
			ignore = 0;
		}
	}
	position = new_pos;
}
#endif

/* 
 * init tables for layer-3 
 */
void init_layer3(int down_sample_sblimit)
{
  int i,j,k,l;

  for(i=-256;i<118+4;i++)
#ifdef USE_MMX
    if(!param.down_sample)
      gainpow2[i+256] = 16384.0 * pow((double)2.0,-0.25 * (double) (i+210) );
    else
#endif
    gainpow2[i+256] = DOUBLE_TO_REAL(pow((double)2.0,-0.25 * (double) (i+210)));

  for(i=0;i<8207;i++)
    ispow[i] = DOUBLE_TO_REAL(pow((double)i,(double)4.0/3.0));

  for (i=0;i<8;i++) {
    static double Ci[8]={-0.6,-0.535,-0.33,-0.185,-0.095,-0.041,-0.0142,-0.0037};
    double sq=sqrt(1.0+Ci[i]*Ci[i]);
    aa_cs[i] = DOUBLE_TO_REAL(1.0/sq);
    aa_ca[i] = DOUBLE_TO_REAL(Ci[i]/sq);
  }

  for(i=0;i<18;i++) {
    win[0][i]    = win[1][i]    = DOUBLE_TO_REAL(0.5 * sin( M_PI / 72.0 * (double) (2*(i+0) +1) ) / cos ( M_PI * (double) (2*(i+0) +19) / 72.0 ));
    win[0][i+18] = win[3][i+18] = DOUBLE_TO_REAL(0.5 * sin( M_PI / 72.0 * (double) (2*(i+18)+1) ) / cos ( M_PI * (double) (2*(i+18)+19) / 72.0 ));
  }
  for(i=0;i<6;i++) {
    win[1][i+18] = DOUBLE_TO_REAL(0.5 / cos ( M_PI * (double) (2*(i+18)+19) / 72.0 ));
    win[3][i+12] = DOUBLE_TO_REAL(0.5 / cos ( M_PI * (double) (2*(i+12)+19) / 72.0 ));
    win[1][i+24] = DOUBLE_TO_REAL(0.5 * sin( M_PI / 24.0 * (double) (2*i+13) ) / cos ( M_PI * (double) (2*(i+24)+19) / 72.0 ));
    win[1][i+30] = win[3][i] = DOUBLE_TO_REAL(0.0);
    win[3][i+6 ] = DOUBLE_TO_REAL(0.5 * sin( M_PI / 24.0 * (double) (2*i+1) ) / cos ( M_PI * (double) (2*(i+6 )+19) / 72.0 ));
  }

  for(i=0;i<9;i++)
    COS9[i] = DOUBLE_TO_REAL(cos( M_PI / 18.0 * (double) i));

  for(i=0;i<9;i++)
    tfcos36[i] = DOUBLE_TO_REAL(0.5 / cos ( M_PI * (double) (i*2+1) / 36.0 ));
  for(i=0;i<3;i++)
    tfcos12[i] = DOUBLE_TO_REAL(0.5 / cos ( M_PI * (double) (i*2+1) / 12.0 ));

  COS6_1 = DOUBLE_TO_REAL(cos( M_PI / 6.0 * (double) 1));
  COS6_2 = DOUBLE_TO_REAL(cos( M_PI / 6.0 * (double) 2));

#ifdef NEW_DCT9
  cos9[0]  = DOUBLE_TO_REAL(cos(1.0*M_PI/9.0));
  cos9[1]  = DOUBLE_TO_REAL(cos(5.0*M_PI/9.0));
  cos9[2]  = DOUBLE_TO_REAL(cos(7.0*M_PI/9.0));
  cos18[0] = DOUBLE_TO_REAL(cos(1.0*M_PI/18.0));
  cos18[1] = DOUBLE_TO_REAL(cos(11.0*M_PI/18.0));
  cos18[2] = DOUBLE_TO_REAL(cos(13.0*M_PI/18.0));
#endif

  for(i=0;i<12;i++) {
    win[2][i]  = DOUBLE_TO_REAL(0.5 * sin( M_PI / 24.0 * (double) (2*i+1) ) / cos ( M_PI * (double) (2*i+7) / 24.0 ));
    for(j=0;j<6;j++)
      COS1[i][j] = DOUBLE_TO_REAL(cos( M_PI / 24.0 * (double) ((2*i+7)*(2*j+1)) ));
  }

  for(j=0;j<4;j++) {
    static int len[4] = { 36,36,12,36 };
    for(i=0;i<len[j];i+=2)
      win1[j][i] = + win[j][i];
    for(i=1;i<len[j];i+=2)
      win1[j][i] = - win[j][i];
  }

  for(i=0;i<16;i++) {
    double t = tan( (double) i * M_PI / 12.0 );
    tan1_1[i] = DOUBLE_TO_REAL(t / (1.0+t));
    tan2_1[i] = DOUBLE_TO_REAL(1.0 / (1.0 + t));
    tan1_2[i] = DOUBLE_TO_REAL(M_SQRT2 * t / (1.0+t));
    tan2_2[i] = DOUBLE_TO_REAL(M_SQRT2 / (1.0 + t));

    for(j=0;j<2;j++) {
      double base = pow(2.0,-0.25*(j+1.0));
      double p1=1.0,p2=1.0;
      if(i > 0) {
        if( i & 1 )
          p1 = pow(base,(i+1.0)*0.5);
        else
          p2 = pow(base,i*0.5);
      }
      pow1_1[j][i] = DOUBLE_TO_REAL(p1);
      pow2_1[j][i] = DOUBLE_TO_REAL(p2);
      pow1_2[j][i] = DOUBLE_TO_REAL(M_SQRT2 * p1);
      pow2_2[j][i] = DOUBLE_TO_REAL(M_SQRT2 * p2);
    }
  }

  for(j=0;j<9;j++) {
   struct bandInfoStruct *bi = &bandInfo[j];
   int *mp;
   int cb,lwin;
   int *bdf;

   mp = map[j][0] = mapbuf0[j];
   bdf = bi->longDiff;
   for(i=0,cb = 0; cb < 8 ; cb++,i+=*bdf++) {
     *mp++ = (*bdf) >> 1;
     *mp++ = i;
     *mp++ = 3;
     *mp++ = cb;
   }
   bdf = bi->shortDiff+3;
   for(cb=3;cb<13;cb++) {
     int l = (*bdf++) >> 1;
     for(lwin=0;lwin<3;lwin++) {
       *mp++ = l;
       *mp++ = i + lwin;
       *mp++ = lwin;
       *mp++ = cb;
     }
     i += 6*l;
   }
   mapend[j][0] = mp;

   mp = map[j][1] = mapbuf1[j];
   bdf = bi->shortDiff+0;
   for(i=0,cb=0;cb<13;cb++) {
     int l = (*bdf++) >> 1;
     for(lwin=0;lwin<3;lwin++) {
       *mp++ = l;
       *mp++ = i + lwin;
       *mp++ = lwin;
       *mp++ = cb;
     }
     i += 6*l;
   }
   mapend[j][1] = mp;

   mp = map[j][2] = mapbuf2[j];
   bdf = bi->longDiff;
   for(cb = 0; cb < 22 ; cb++) {
     *mp++ = (*bdf++) >> 1;
     *mp++ = cb;
   }
   mapend[j][2] = mp;

  }

  for(j=0;j<9;j++) {
    for(i=0;i<23;i++) {
      longLimit[j][i] = (bandInfo[j].longIdx[i] - 1 + 8) / 18 + 1;
      if(longLimit[j][i] > (down_sample_sblimit) )
        longLimit[j][i] = down_sample_sblimit;
    }
    for(i=0;i<14;i++) {
      shortLimit[j][i] = (bandInfo[j].shortIdx[i] - 1) / 18 + 1;
      if(shortLimit[j][i] > (down_sample_sblimit) )
        shortLimit[j][i] = down_sample_sblimit;
    }
  }

  for(i=0;i<5;i++) {
    for(j=0;j<6;j++) {
      for(k=0;k<6;k++) {
        int n = k + j * 6 + i * 36;
        i_slen2[n] = i|(j<<3)|(k<<6)|(3<<12);
      }
    }
  }
  for(i=0;i<4;i++) {
    for(j=0;j<4;j++) {
      for(k=0;k<4;k++) {
        int n = k + j * 4 + i * 16;
        i_slen2[n+180] = i|(j<<3)|(k<<6)|(4<<12);
      }
    }
  }
  for(i=0;i<4;i++) {
    for(j=0;j<3;j++) {
      int n = j + i * 3;
      i_slen2[n+244] = i|(j<<3) | (5<<12);
      n_slen2[n+500] = i|(j<<3) | (2<<12) | (1<<15);
    }
  }

  for(i=0;i<5;i++) {
    for(j=0;j<5;j++) {
      for(k=0;k<4;k++) {
        for(l=0;l<4;l++) {
          int n = l + k * 4 + j * 16 + i * 80;
          n_slen2[n] = i|(j<<3)|(k<<6)|(l<<9)|(0<<12);
        }
      }
    }
  }
  for(i=0;i<5;i++) {
    for(j=0;j<5;j++) {
      for(k=0;k<4;k++) {
        int n = k + j * 4 + i * 20;
        n_slen2[n+400] = i|(j<<3)|(k<<6)|(1<<12);
      }
    }
  }
}

/*
 * read additional side information (for MPEG 1 and MPEG 2)
 */
static int III_get_side_info(struct III_sideinfo *si,int stereo,
 int ms_stereo,long sfreq,int single,int lsf)
{
   int ch, gr;
   int powdiff = (single == 3) ? 4 : 0;

   static const int tabs[2][5] = { { 2,9,5,3,4 } , { 1,8,1,2,9 } };
   const int *tab = tabs[lsf];
   
   si->main_data_begin = getbits(tab[1]);
   if (stereo == 1)
     si->private_bits = getbits_fast(tab[2]);
   else 
     si->private_bits = getbits_fast(tab[3]);

   if(!lsf) {
     for (ch=0; ch<stereo; ch++) {
         si->ch[ch].gr[0].scfsi = -1;
         si->ch[ch].gr[1].scfsi = getbits_fast(4);
     }
   }

   for (gr=0; gr<tab[0]; gr++) {
     for (ch=0; ch<stereo; ch++) {
       register struct gr_info_s *gr_info = &(si->ch[ch].gr[gr]);

       gr_info->part2_3_length = getbits(12);
       gr_info->big_values = getbits(9);
       if(gr_info->big_values > 288) {
          error("big_values too large!");
          gr_info->big_values = 288;
       }
       gr_info->pow2gain = gainpow2+256 - getbits_fast(8) + powdiff;
       if(ms_stereo)
         gr_info->pow2gain += 2;
       gr_info->scalefac_compress = getbits(tab[4]);

       if(get1bit()) { /* window switch flag  */
         int i;
         gr_info->block_type       = getbits_fast(2);
         gr_info->mixed_block_flag = get1bit();
         gr_info->table_select[0]  = getbits_fast(5);
         gr_info->table_select[1]  = getbits_fast(5);
         /*
          * table_select[2] not needed, because there is no region2,
          * but to satisfy some verifications tools we set it either.
          */
         gr_info->table_select[2] = 0;
         for(i=0;i<3;i++)
           gr_info->full_gain[i] = gr_info->pow2gain + (getbits_fast(3)<<3);

         if(gr_info->block_type == 0) {
           error("Blocktype == 0 and window-switching == 1 not allowed.");
           /* exit(1); */
           return 1;
         }
      
         /* region_count/start parameters are implicit in this case. */       
         if(!lsf || gr_info->block_type == 2)
           gr_info->region1start = 36>>1;
         else {
/* check this again for 2.5 and sfreq=8 */
           if(sfreq == 8)
             gr_info->region1start = 108>>1;
           else
             gr_info->region1start = 54>>1;
         }
         gr_info->region2start = 576>>1;