layer3.c

mips上编译过的mpg 运行正常 环境：AU12

    real tmpval; \
    tmpval = tmp[(v)] + tmp[17-(v)]; \
    out2[9+(v)] = REAL_MUL(tmpval, w[27+(v)]); \
    out2[8-(v)] = REAL_MUL(tmpval, w[26-(v)]); \
    tmpval = tmp[(v)] - tmp[17-(v)]; \
    ts[SBLIMIT*(8-(v))] = out1[8-(v)] + REAL_MUL(tmpval, w[8-(v)]); \
    ts[SBLIMIT*(9+(v))] = out1[9+(v)] + REAL_MUL(tmpval, w[9+(v)]); }

{
   register real *out2 = o2;
   register real *w = wintab;
   register real *out1 = o1;
   register real *ts = tsbuf;

   MACRO(0);
   MACRO(1);
   MACRO(2);
   MACRO(3);
   MACRO(4);
   MACRO(5);
   MACRO(6);
   MACRO(7);
   MACRO(8);
}

#else

  {

#define MACRO0(v) { \
    real tmp; \
    out2[9+(v)] = REAL_MUL((tmp = sum0 + sum1), w[27+(v)]); \
    out2[8-(v)] = REAL_MUL(tmp, w[26-(v)]);   } \
    sum0 -= sum1; \
    ts[SBLIMIT*(8-(v))] = out1[8-(v)] + REAL_MUL(sum0, w[8-(v)]); \
    ts[SBLIMIT*(9+(v))] = out1[9+(v)] + REAL_MUL(sum0, w[9+(v)]);
#define MACRO1(v) { \
	real sum0,sum1; \
    sum0 = tmp1a + tmp2a; \
	sum1 = REAL_MUL((tmp1b + tmp2b), tfcos36[(v)]); \
	MACRO0(v); }
#define MACRO2(v) { \
    real sum0,sum1; \
    sum0 = tmp2a - tmp1a; \
    sum1 = REAL_MUL((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 = REAL_MUL(in[2*3+0], c[3]);
    ta66 = REAL_MUL(in[2*6+0], c[6]);
    tb33 = REAL_MUL(in[2*3+1], c[3]);
    tb66 = REAL_MUL(in[2*6+1], c[6]);

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

      MACRO1(0);
      MACRO2(8);
    }

    {
      real tmp1a,tmp2a,tmp1b,tmp2b;
      tmp1a = REAL_MUL(( in[2*1+0] - in[2*5+0] - in[2*7+0] ), c[3]);
      tmp1b = REAL_MUL(( in[2*1+1] - in[2*5+1] - in[2*7+1] ), c[3]);
      tmp2a = REAL_MUL(( in[2*2+0] - in[2*4+0] - in[2*8+0] ), c[6]) - in[2*6+0] + in[2*0+0];
      tmp2b = REAL_MUL(( 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 =   REAL_MUL(in[2*1+0], c[5]) - ta33 - REAL_MUL(in[2*5+0], c[7]) + REAL_MUL(in[2*7+0], c[1]);
      tmp1b =   REAL_MUL(in[2*1+1], c[5]) - tb33 - REAL_MUL(in[2*5+1], c[7]) + REAL_MUL(in[2*7+1], c[1]);
      tmp2a = - REAL_MUL(in[2*2+0], c[8]) - REAL_MUL(in[2*4+0], c[2]) + ta66 + REAL_MUL(in[2*8+0], c[4]);
      tmp2b = - REAL_MUL(in[2*2+1], c[8]) - REAL_MUL(in[2*4+1], c[2]) + tb66 + REAL_MUL(in[2*8+1], c[4]);

      MACRO1(2);
      MACRO2(6);
    }

    {
      real tmp1a,tmp2a,tmp1b,tmp2b;
      tmp1a =   REAL_MUL(in[2*1+0], c[7]) - ta33 + REAL_MUL(in[2*5+0], c[1]) - REAL_MUL(in[2*7+0], c[5]);
      tmp1b =   REAL_MUL(in[2*1+1], c[7]) - tb33 + REAL_MUL(in[2*5+1], c[1]) - REAL_MUL(in[2*7+1], c[5]);
      tmp2a = - REAL_MUL(in[2*2+0], c[4]) + REAL_MUL(in[2*4+0], c[8]) + ta66 - REAL_MUL(in[2*8+0], c[2]);
      tmp2b = - REAL_MUL(in[2*2+1], c[4]) + REAL_MUL(in[2*4+1], c[8]) + tb66 - REAL_MUL(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 = REAL_MUL((in[2*0+1] - in[2*2+1] + in[2*4+1] - in[2*6+1] + in[2*8+1] ), tfcos36[4]);
		MACRO0(4);
	}
  }
#endif

  }
}

/*
 * 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 = REAL_MUL(in2, COS6_1); \
     in3 = REAL_MUL(in3, COS6_1); \

#define DCT12_PART2 \
     in0 += REAL_MUL(in4, COS6_2); \
                          \
     in4 = in0 + in2;     \
     in0 -= in2;          \
                          \
     in1 += REAL_MUL(in5, COS6_2); \
                          \
     in5 = REAL_MUL((in1 + in3), tfcos12[0]); \
     in1 = REAL_MUL((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 = REAL_MUL((in1 - in5), tfcos12[1]);
         tmp0 = tmp1 + tmp2;
         tmp1 -= tmp2;
       }
       ts[(17-1)*SBLIMIT] = out1[17-1] + REAL_MUL(tmp0, wi[11-1]);
       ts[(12+1)*SBLIMIT] = out1[12+1] + REAL_MUL(tmp0, wi[6+1]);
       ts[(6 +1)*SBLIMIT] = out1[6 +1] + REAL_MUL(tmp1, wi[1]);
       ts[(11-1)*SBLIMIT] = out1[11-1] + REAL_MUL(tmp1, wi[5-1]);
     }

     DCT12_PART2

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

     ts[(6 +0)*SBLIMIT]  = out1[6+0] + REAL_MUL(in0, wi[0]);
     ts[(11-0)*SBLIMIT] = out1[11-0] + REAL_MUL(in0, wi[5-0]);
     ts[(6 +2)*SBLIMIT]  = out1[6+2] + REAL_MUL(in4, wi[2]);
     ts[(11-2)*SBLIMIT] = out1[11-2] + REAL_MUL(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 = REAL_MUL((in1 - in5), tfcos12[1]);
         tmp0 = tmp1 + tmp2;
         tmp1 -= tmp2;
       }
       out2[5-1] = REAL_MUL(tmp0, wi[11-1]);
       out2[0+1] = REAL_MUL(tmp0, wi[6+1]);
       ts[(12+1)*SBLIMIT] += REAL_MUL(tmp1, wi[1]);
       ts[(17-1)*SBLIMIT] += REAL_MUL(tmp1, wi[5-1]);
     }

     DCT12_PART2

     out2[5-0] = REAL_MUL(in2, wi[11-0]);
     out2[0+0] = REAL_MUL(in2, wi[6+0]);
     out2[0+2] = REAL_MUL(in3, wi[6+2]);
     out2[5-2] = REAL_MUL(in3, wi[11-2]);

     ts[(12+0)*SBLIMIT] += REAL_MUL(in0, wi[0]);
     ts[(17-0)*SBLIMIT] += REAL_MUL(in0, wi[5-0]);
     ts[(12+2)*SBLIMIT] += REAL_MUL(in4, wi[2]);
     ts[(17-2)*SBLIMIT] += REAL_MUL(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 = REAL_MUL((in1 - in5), tfcos12[1]);
         tmp0 = tmp1 + tmp2;
         tmp1 -= tmp2;
       }
       out2[11-1] = REAL_MUL(tmp0, wi[11-1]);
       out2[6 +1] = REAL_MUL(tmp0, wi[6+1]);
       out2[0+1] += REAL_MUL(tmp1, wi[1]);
       out2[5-1] += REAL_MUL(tmp1, wi[5-1]);
     }

     DCT12_PART2

     out2[11-0] = REAL_MUL(in2, wi[11-0]);
     out2[6 +0] = REAL_MUL(in2, wi[6+0]);
     out2[6 +2] = REAL_MUL(in3, wi[6+2]);
     out2[11-2] = REAL_MUL(in3, wi[11-2]);

     out2[0+0] += REAL_MUL(in0, wi[0]);
     out2[5-0] += REAL_MUL(in0, wi[5-0]);
     out2[0+2] += REAL_MUL(in4, wi[2]);
     out2[5-2] += REAL_MUL(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, mpg123_handle *fr)
{
   real (*block)[2][SBLIMIT*SSLIMIT] = fr->hybrid_block;
   int *blc = fr->hybrid_blc;

   real *tspnt = (real *) tsOut;
   real *rawout1,*rawout2;
   int bt,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;
     opt_dct36(fr)(fsIn[0],rawout1,rawout2,win[0],tspnt);
     opt_dct36(fr)(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) {
       opt_dct36(fr)(fsIn[sb],rawout1,rawout2,win[bt],tspnt);
       opt_dct36(fr)(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++ = DOUBLE_TO_REAL(0.0);
     }
   }
}


/*
 * main layer3 handler
 */
int do_layer3(mpg123_handle *fr)
{
  int gr, ch, ss,clip=0;
  int scalefacs[2][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;

  if(stereo == 1) { /* stream is mono */
    stereo1 = 1;
    single = SINGLE_LEFT;
  }
  else if(single != SINGLE_STEREO) /* 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)>>1;
    i_stereo  = fr->mode_ext & 0x1;
  }
  else
    ms_stereo = i_stereo = 0;

  if(fr->lsf) {
    granules = 1;
  }
  else {
    granules = 2;
  }
  /* quick hack to keep the music playing */
  /* after having seen this nasty test file... */
  if(III_get_side_info(fr, &sideinfo,stereo,ms_stereo,sfreq,single))
  {
    error("bad frame - unable to get valid sideinfo");
    return clip;
  }

  set_pointer(fr,sideinfo.main_data_begin);

  for (gr=0;gr<granules;gr++) {
    ALIGNED(16) real hybridIn[2][SBLIMIT][SSLIMIT];
    ALIGNED(16) 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(fr, scalefacs[0],gr_info,0);
      else
        part2bits = III_get_scale_factors_1(fr, scalefacs[0],gr_info,0,gr);

      if(III_dequantize_sample(fr, hybridIn[0], scalefacs[0],gr_info,sfreq,part2bits))
        return clip;
    }

    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(fr, scalefacs[1],gr_info,i_stereo);
      else
        part2bits = III_get_scale_factors_1(fr, scalefacs[1],gr_info,1,gr);

      if(III_dequantize_sample(fr, hybridIn[1],scalefacs[1],gr_info,sfreq,part2bits))
          return clip;

      if(ms_stereo) {
        int i;
        int maxb = sideinfo.ch[0].gr[gr].maxb;
        if(sideinfo.ch[1].gr[gr].maxb > maxb)
            maxb = sideinfo.ch[1].gr[gr].maxb;
        for(i=0;i<SSLIMIT*maxb;i++) {
          real tmp0 = ((real *)hybridIn[0])[i];
          real tmp1 = ((real *)hybridIn[1])[i];
          ((real *)hybridIn[0])[i] = tmp0 + tmp1;
          ((real *)hybridIn[1])[i] = tmp0 - tmp1;
        }
      }

      if(i_stereo)
        III_i_stereo(hybridIn,scalefacs[1],gr_info,sfreq,ms_stereo,fr->lsf);

      if(ms_stereo || i_stereo || (single == SINGLE_MIX) ) {
        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 SINGLE_MIX:
          {
            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 SINGLE_RIGHT:
          {
            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, fr);
    }

#ifdef OPT_I486
    if (fr->synth != opt_synth_1to1(fr) || single != SINGLE_STEREO) {
#endif
    for(ss=0;ss<SSLIMIT;ss++) {
      if(single != SINGLE_STEREO) {
        clip += (fr->synth_mono)(hybridOut[0][ss], fr);
      }
      else
      {
        clip += (fr->synth)(hybridOut[0][ss], 0, fr, 0);
        clip += (fr->synth)(hybridOut[1][ss], 1, fr, 1);
      }

#ifdef VARMODESUPPORT
      if (playlimit < 128) {
        fr->buffer.fill -= playlimit >> 1;
        playlimit = 0;
      }
      else
        playlimit -= 128;
#endif
    }
#ifdef OPT_I486
    } else {
      /* Only stereo, 16 bits benefit from the 486 optimization. */
      ss=0;
      while (ss < SSLIMIT) {
        int n;
        n=(fr->buffer.size - fr->buffer.fill) / (2*2*32);
        if (n > (SSLIMIT-ss)) n=SSLIMIT-ss;
        
        synth_1to1_486(hybridOut[0][ss], 0, fr, n);
        synth_1to1_486(hybridOut[1][ss], 1, fr, n);
        ss+=n;
        fr->buffer.fill+=(2*2*32)*n;
      }
    }
#endif
  }
  
  return clip;
}