layer3.c

神龙卡开发原代码

          }
        }
      } /* ... */
}

static void III_antialias(real xr[SBLIMIT][SSLIMIT],struct gr_info_s *gr_info)
{
   int sblim;

   if(gr_info->block_type == 2)
   {
      if(!gr_info->mixed_block_flag) 
        return;
      sblim = 1; 
   }
   else {
     sblim = gr_info->maxb-1;
   }

   /* 31 alias-reduction operations between each pair of sub-bands */
   /* with 8 butterflies between each pair                         */

   {
     int sb;
     real *xr1=(real *) xr[1];

     for(sb=sblim;sb;sb--,xr1+=10)
     {
       int ss;
       real *cs=aa_cs,*ca=aa_ca;
       real *xr2 = xr1;

       for(ss=7;ss>=0;ss--)
       {       /* upper and lower butterfly inputs */
         register real bu = *--xr2,bd = *xr1;
         *xr2   = (bu * (*cs)   ) - (bd * (*ca)   );
         *xr1++ = (bd * (*cs++) ) + (bu * (*ca++) );
       }
     }
  }
}

/*
 DCT insipired by Jeff Tsay's DCT from the maplay package
 this is an optimized version with manual unroll.

 References:
 [1] S. Winograd: "On Computing the Discrete Fourier Transform",
     Mathematics of Computation, Volume 32, Number 141, January 1978,
     Pages 175-199
*/

static void dct36(real *inbuf,real *o1,real *o2,real *wintab,real *tsbuf)
{
  {
    register real *in = inbuf;

    in[17]+=in[16]; in[16]+=in[15]; in[15]+=in[14];
    in[14]+=in[13]; in[13]+=in[12]; in[12]+=in[11];
    in[11]+=in[10]; in[10]+=in[9];  in[9] +=in[8];
    in[8] +=in[7];  in[7] +=in[6];  in[6] +=in[5];
    in[5] +=in[4];  in[4] +=in[3];  in[3] +=in[2];
    in[2] +=in[1];  in[1] +=in[0];

    in[17]+=in[15]; in[15]+=in[13]; in[13]+=in[11]; in[11]+=in[9];
    in[9] +=in[7];  in[7] +=in[5];  in[5] +=in[3];  in[3] +=in[1];


  {

#define MACRO0(v) { \
    real tmp; \
    out2[9+(v)] = (tmp = sum0 + sum1) * w[27+(v)]; \
    out2[8-(v)] = tmp * w[26-(v)];  } \
    sum0 -= sum1; \
    ts[SBLIMIT*(8-(v))] = out1[8-(v)] + sum0 * w[8-(v)]; \
    ts[SBLIMIT*(9+(v))] = out1[9+(v)] + sum0 * w[9+(v)]; 
#define MACRO1(v) { \
	real sum0,sum1; \
    sum0 = tmp1a + tmp2a; \
	sum1 = (tmp1b + tmp2b) * tfcos36[(v)]; \
	MACRO0(v); }
#define MACRO2(v) { \
    real sum0,sum1; \
    sum0 = tmp2a - tmp1a; \
    sum1 = (tmp2b - tmp1b) * tfcos36[(v)]; \
	MACRO0(v); }

    register const real *c = COS9;
    register real *out2 = o2;
	register real *w = wintab;
	register real *out1 = o1;
	register real *ts = tsbuf;

    real ta33,ta66,tb33,tb66;

    ta33 = in[2*3+0] * c[3];
    ta66 = in[2*6+0] * c[6];
    tb33 = in[2*3+1] * c[3];
    tb66 = in[2*6+1] * c[6];

    { 
      real tmp1a,tmp2a,tmp1b,tmp2b;
      tmp1a =             in[2*1+0] * c[1] + ta33 + in[2*5+0] * c[5] + in[2*7+0] * c[7];
      tmp1b =             in[2*1+1] * c[1] + tb33 + in[2*5+1] * c[5] + in[2*7+1] * c[7];
      tmp2a = in[2*0+0] + in[2*2+0] * c[2] + in[2*4+0] * c[4] + ta66 + in[2*8+0] * c[8];
      tmp2b = in[2*0+1] + in[2*2+1] * c[2] + in[2*4+1] * c[4] + tb66 + in[2*8+1] * c[8];

      MACRO1(0);
      MACRO2(8);
    }

    {
      real tmp1a,tmp2a,tmp1b,tmp2b;
      tmp1a = ( in[2*1+0] - in[2*5+0] - in[2*7+0] ) * c[3];
      tmp1b = ( in[2*1+1] - in[2*5+1] - in[2*7+1] ) * c[3];
      tmp2a = ( in[2*2+0] - in[2*4+0] - in[2*8+0] ) * c[6] - in[2*6+0] + in[2*0+0];
      tmp2b = ( in[2*2+1] - in[2*4+1] - in[2*8+1] ) * c[6] - in[2*6+1] + in[2*0+1];

      MACRO1(1);
      MACRO2(7);
    }

    {
      real tmp1a,tmp2a,tmp1b,tmp2b;
      tmp1a =             in[2*1+0] * c[5] - ta33 - in[2*5+0] * c[7] + in[2*7+0] * c[1];
      tmp1b =             in[2*1+1] * c[5] - tb33 - in[2*5+1] * c[7] + in[2*7+1] * c[1];
      tmp2a = in[2*0+0] - in[2*2+0] * c[8] - in[2*4+0] * c[2] + ta66 + in[2*8+0] * c[4];
      tmp2b = in[2*0+1] - in[2*2+1] * c[8] - in[2*4+1] * c[2] + tb66 + in[2*8+1] * c[4];

      MACRO1(2);
      MACRO2(6);
    }

    {
      real tmp1a,tmp2a,tmp1b,tmp2b;
      tmp1a =             in[2*1+0] * c[7] - ta33 + in[2*5+0] * c[1] - in[2*7+0] * c[5];
      tmp1b =             in[2*1+1] * c[7] - tb33 + in[2*5+1] * c[1] - in[2*7+1] * c[5];
      tmp2a = in[2*0+0] - in[2*2+0] * c[4] + in[2*4+0] * c[8] + ta66 - in[2*8+0] * c[2];
      tmp2b = in[2*0+1] - in[2*2+1] * c[4] + in[2*4+1] * c[8] + tb66 - in[2*8+1] * c[2];

      MACRO1(3);
      MACRO2(5);
    }

	{
		real sum0,sum1;
    	sum0 =  in[2*0+0] - in[2*2+0] + in[2*4+0] - in[2*6+0] + in[2*8+0];
    	sum1 = (in[2*0+1] - in[2*2+1] + in[2*4+1] - in[2*6+1] + in[2*8+1] ) * tfcos36[4];
		MACRO0(4);
	}
  }

  }
}

/*
 * new DCT12
 */
static void dct12(real *in,real *rawout1,real *rawout2,register real *wi,register real *ts)
{
#define DCT12_PART1 \
             in5 = in[5*3];  \
     in5 += (in4 = in[4*3]); \
     in4 += (in3 = in[3*3]); \
     in3 += (in2 = in[2*3]); \
     in2 += (in1 = in[1*3]); \
     in1 += (in0 = in[0*3]); \
                             \
     in5 += in3; in3 += in1; \
                             \
     in2 *= COS6_1; \
     in3 *= COS6_1; \

#define DCT12_PART2 \
     in0 += in4 * COS6_2; \
                          \
     in4 = in0 + in2;     \
     in0 -= in2;          \
                          \
     in1 += in5 * COS6_2; \
                          \
     in5 = (in1 + in3) * tfcos12[0]; \
     in1 = (in1 - in3) * tfcos12[2]; \
                         \
     in3 = in4 + in5;    \
     in4 -= in5;         \
                         \
     in2 = in0 + in1;    \
     in0 -= in1;


   {
     real in0,in1,in2,in3,in4,in5;
     register real *out1 = rawout1;
     ts[SBLIMIT*0] = out1[0]; ts[SBLIMIT*1] = out1[1]; ts[SBLIMIT*2] = out1[2];
     ts[SBLIMIT*3] = out1[3]; ts[SBLIMIT*4] = out1[4]; ts[SBLIMIT*5] = out1[5];
 
     DCT12_PART1

     {
       real tmp0,tmp1 = (in0 - in4);
       {
         real tmp2 = (in1 - in5) * tfcos12[1];
         tmp0 = tmp1 + tmp2;
         tmp1 -= tmp2;
       }
       ts[(17-1)*SBLIMIT] = out1[17-1] + tmp0 * wi[11-1];
       ts[(12+1)*SBLIMIT] = out1[12+1] + tmp0 * wi[6+1];
       ts[(6 +1)*SBLIMIT] = out1[6 +1] + tmp1 * wi[1];
       ts[(11-1)*SBLIMIT] = out1[11-1] + tmp1 * wi[5-1];
     }

     DCT12_PART2

     ts[(17-0)*SBLIMIT] = out1[17-0] + in2 * wi[11-0];
     ts[(12+0)*SBLIMIT] = out1[12+0] + in2 * wi[6+0];
     ts[(12+2)*SBLIMIT] = out1[12+2] + in3 * wi[6+2];
     ts[(17-2)*SBLIMIT] = out1[17-2] + in3 * wi[11-2];

     ts[(6+0)*SBLIMIT]  = out1[6+0] + in0 * wi[0];
     ts[(11-0)*SBLIMIT] = out1[11-0] + in0 * wi[5-0];
     ts[(6+2)*SBLIMIT]  = out1[6+2] + in4 * wi[2];
     ts[(11-2)*SBLIMIT] = out1[11-2] + in4 * wi[5-2];
  }

  in++;

  {
     real in0,in1,in2,in3,in4,in5;
     register real *out2 = rawout2;
 
     DCT12_PART1

     {
       real tmp0,tmp1 = (in0 - in4);
       {
         real tmp2 = (in1 - in5) * tfcos12[1];
         tmp0 = tmp1 + tmp2;
         tmp1 -= tmp2;
       }
       out2[5-1] = tmp0 * wi[11-1];
       out2[0+1] = tmp0 * wi[6+1];
       ts[(12+1)*SBLIMIT] += tmp1 * wi[1];
       ts[(17-1)*SBLIMIT] += tmp1 * wi[5-1];
     }

     DCT12_PART2

     out2[5-0] = in2 * wi[11-0];
     out2[0+0] = in2 * wi[6+0];
     out2[0+2] = in3 * wi[6+2];
     out2[5-2] = in3 * wi[11-2];

     ts[(12+0)*SBLIMIT] += in0 * wi[0];
     ts[(17-0)*SBLIMIT] += in0 * wi[5-0];
     ts[(12+2)*SBLIMIT] += in4 * wi[2];
     ts[(17-2)*SBLIMIT] += in4 * wi[5-2];
  }

  in++; 

  {
     real in0,in1,in2,in3,in4,in5;
     register real *out2 = rawout2;
     out2[12]=out2[13]=out2[14]=out2[15]=out2[16]=out2[17]=0.0;

     DCT12_PART1

     {
       real tmp0,tmp1 = (in0 - in4);
       {
         real tmp2 = (in1 - in5) * tfcos12[1];
         tmp0 = tmp1 + tmp2;
         tmp1 -= tmp2;
       }
       out2[11-1] = tmp0 * wi[11-1];
       out2[6 +1] = tmp0 * wi[6+1];
       out2[0+1] += tmp1 * wi[1];
       out2[5-1] += tmp1 * wi[5-1];
     }

     DCT12_PART2

     out2[11-0] = in2 * wi[11-0];
     out2[6 +0] = in2 * wi[6+0];
     out2[6 +2] = in3 * wi[6+2];
     out2[11-2] = in3 * wi[11-2];

     out2[0+0] += in0 * wi[0];
     out2[5-0] += in0 * wi[5-0];
     out2[0+2] += in4 * wi[2];
     out2[5-2] += in4 * wi[5-2];
  }
}

/*
 * III_hybrid
 */
static void III_hybrid(real fsIn[SBLIMIT][SSLIMIT],real tsOut[SSLIMIT][SBLIMIT],
   int ch,struct gr_info_s *gr_info)
{
   real *tspnt = (real *) tsOut;
   real (*block)[2][SBLIMIT*SSLIMIT] = gmp->hybrid_block;
   int *blc = gmp->hybrid_blc;
   real *rawout1,*rawout2;
   int bt;
   int sb = 0;

   {
     int b = blc[ch];
     rawout1=block[b][ch];
     b=-b+1;
     rawout2=block[b][ch];
     blc[ch] = b;
   }

  
   if(gr_info->mixed_block_flag) {
     sb = 2;
     dct36(fsIn[0],rawout1,rawout2,win[0],tspnt);
     dct36(fsIn[1],rawout1+18,rawout2+18,win1[0],tspnt+1);
     rawout1 += 36; rawout2 += 36; tspnt += 2;
   }
 
   bt = gr_info->block_type;
   if(bt == 2) {
     for (; sb<gr_info->maxb; sb+=2,tspnt+=2,rawout1+=36,rawout2+=36) {
       dct12(fsIn[sb],rawout1,rawout2,win[2],tspnt);
       dct12(fsIn[sb+1],rawout1+18,rawout2+18,win1[2],tspnt+1);
     }
   }
   else {
     for (; sb<gr_info->maxb; sb+=2,tspnt+=2,rawout1+=36,rawout2+=36) {
       dct36(fsIn[sb],rawout1,rawout2,win[bt],tspnt);
       dct36(fsIn[sb+1],rawout1+18,rawout2+18,win1[bt],tspnt+1);
     }
   }

   for(;sb<SBLIMIT;sb++,tspnt++) {
     int i;
     for(i=0;i<SSLIMIT;i++) {
       tspnt[i*SBLIMIT] = *rawout1++;
       *rawout2++ = 0.0;
     }
   }
}

/*
 * main layer3 handler
 */
int do_layer3(struct frame *fr,unsigned char *pcm_sample,int *pcm_point)
{
  int gr, ch, ss,clip=0;
  int scalefacs[39]; /* max 39 for short[13][3] mode, mixed: 38, long: 22 */
  struct III_sideinfo sideinfo;
  int stereo = fr->stereo;
  int single = fr->single;
  int ms_stereo,i_stereo;
  int sfreq = fr->sampling_frequency;
  int stereo1,granules;
  int res;

  if(stereo == 1) { /* stream is mono */
    stereo1 = 1;
    single = 0;
  }
  else if(single >= 0) /* stream is stereo, but force to mono */
    stereo1 = 1;
  else
    stereo1 = 2;

  if(fr->mode == MPG_MD_JOINT_STEREO) {
    ms_stereo = fr->mode_ext & 0x2;
    i_stereo  = fr->mode_ext & 0x1;
  }
  else
    ms_stereo = i_stereo = 0;

  if(fr->lsf) {
    granules = 1;
    III_get_side_info_2(&sideinfo,stereo,ms_stereo,sfreq,single);
  }
  else {
    granules = 2;
#ifdef MPEG1
    III_get_side_info_1(&sideinfo,stereo,ms_stereo,sfreq,single);
#else
    //fprintf(stderr,"Not supported\n");
#endif
  }

  if(set_pointer(sideinfo.main_data_begin) == MP3_ERR)
    return -(sideinfo.main_data_begin);

  for (gr=0;gr<granules;gr++) 
  {
    static real hybridIn[2][SBLIMIT][SSLIMIT];
    static real hybridOut[2][SSLIMIT][SBLIMIT];

    {
      struct gr_info_s *gr_info = &(sideinfo.ch[0].gr[gr]);
      long part2bits;
      if(fr->lsf)
        part2bits = III_get_scale_factors_2(scalefacs,gr_info,0);
      else {
#ifdef MPEG1
        part2bits = III_get_scale_factors_1(scalefacs,gr_info);
#else
	//fprintf(stderr,"Not supported\n");
#endif
      }
      
      res = III_dequantize_sample(hybridIn[0], scalefacs,gr_info,sfreq,part2bits);
      if(res < 0) {
	      //fprintf(stderr, "error in III_dequantize_sample, res = %d\n", res);
	      return res;
      }
    }
    if(stereo == 2) {
      struct gr_info_s *gr_info = &(sideinfo.ch[1].gr[gr]);
      long part2bits;
      if(fr->lsf) 
        part2bits = III_get_scale_factors_2(scalefacs,gr_info,i_stereo);
      else {
#ifdef MPEG1
        part2bits = III_get_scale_factors_1(scalefacs,gr_info);
#else
	//fprintf(stderr,"Not supported\n");
#endif
      }

      res = III_dequantize_sample(hybridIn[1],scalefacs,gr_info,sfreq,part2bits);
      if(res < 0){
	      //fprintf(stderr, "error in III_dequantize_sample, res = %d (2)\n", res);
	      return res;
      }

      if(ms_stereo) {
        int i;
        for(i=0;i<SBLIMIT*SSLIMIT;i++) {
          real tmp0,tmp1;
          tmp0 = ((real *) hybridIn[0])[i];
          tmp1 = ((real *) hybridIn[1])[i];
          ((real *) hybridIn[1])[i] = tmp0 - tmp1;  
          ((real *) hybridIn[0])[i] = tmp0 + tmp1;
        }
      }

      if(i_stereo)
        III_i_stereo(hybridIn,scalefacs,gr_info,sfreq,ms_stereo,fr->lsf);

      if(ms_stereo || i_stereo || (single == 3) ) {
        if(gr_info->maxb > sideinfo.ch[0].gr[gr].maxb) 
          sideinfo.ch[0].gr[gr].maxb = gr_info->maxb;
        else
          gr_info->maxb = sideinfo.ch[0].gr[gr].maxb;
      }

      switch(single) {
        case 3:
          {
            register int i;
            register real *in0 = (real *) hybridIn[0],*in1 = (real *) hybridIn[1];
            for(i=0;i<SSLIMIT*gr_info->maxb;i++,in0++)
              *in0 = (*in0 + *in1++); /* *0.5 done by pow-scale */ 
          }

          break;
        case 1:
          {
            register int i;
            register real *in0 = (real *) hybridIn[0],*in1 = (real *) hybridIn[1];
            for(i=0;i<SSLIMIT*gr_info->maxb;i++)
              *in0++ = *in1++;
          }
          break;
      }
    }

    for(ch=0;ch<stereo1;ch++) {
      struct gr_info_s *gr_info = &(sideinfo.ch[ch].gr[gr]);
      III_antialias(hybridIn[ch],gr_info);
      III_hybrid(hybridIn[ch], hybridOut[ch], ch,gr_info);
    }

    for(ss=0;ss<SSLIMIT;ss++) {
      if(single >= 0) {
        clip += synth_1to1_mono(hybridOut[0][ss],pcm_sample,pcm_point);
      }
      else {
        int p1 = *pcm_point;
        clip += synth_1to1(hybridOut[0][ss],0,pcm_sample,&p1);
        clip += synth_1to1(hybridOut[1][ss],1,pcm_sample,pcm_point);
      }
    }
  }
  
  return clip;
}