view mp3lib/layer1.c @ 11007:48b7d7aa444d

configure altivec patch by Magnus Damm <[e-mail]> * CC is not checked for Altivec support (see above). The patch adds checks for FSF-style flags and Darwin-style flags. The check is performed regardless of the gcc version. * Disabling of Altivec. --disable-altivec is broken today if /proc/cpuinfo shows that your cpu supports altivec. The patch takes care of that. * "GCC & CPU optimization abilities" always show that it is optimizing for the cpu configure is running on, it should show the optimization that is enabled for gcc instead. Cosmetic change only, but confusing as it is today IMHO. * Runtime CPU-detection now enables altivec for powerpc. Now with the patch it should be possible to use --enable-altivec, --disable-altivec, --enable-runtime-cpudetection regardless of powerpc cpu type. The configure script handles altivec support in the following order: 1. Altivec is enabled by default if your cpu supports it. 2. --enable-runtime-cpudetection will enable altivec support. 3. If you have forced altivec on/off with --enable-altivec/--disable-altivec, then your selection will override the previous altivec configuration. 4. If altivec is enabled but the compiler doesn't support it, altivec gets turned off.

/* 
 * Mpeg Layer-1 audio decoder 
 * --------------------------
 * copyright (c) 1995 by Michael Hipp, All rights reserved. See also 'README'
 * near unoptimzed ...
 *
 * may have a few bugs after last optimization ... 
 *
 */

//#include "mpg123.h"

static void I_step_one(unsigned int balloc[], unsigned int scale_index[2][SBLIMIT],struct frame *fr)
{
  unsigned int *ba=balloc;
  unsigned int *sca = (unsigned int *) scale_index;

  if(fr->stereo) {
    int i;
    int jsbound = fr->jsbound;
    for (i=0;i<jsbound;i++) { 
      *ba++ = getbits(4);
      *ba++ = getbits(4);
    }
    for (i=jsbound;i<SBLIMIT;i++)
      *ba++ = getbits(4);

    ba = balloc;

    for (i=0;i<jsbound;i++) {
      if ((*ba++))
        *sca++ = getbits(6);
      if ((*ba++))
        *sca++ = getbits(6);
    }
    for (i=jsbound;i<SBLIMIT;i++)
      if ((*ba++)) {
        *sca++ =  getbits(6);
        *sca++ =  getbits(6);
      }
  }
  else {
    int i;
    for (i=0;i<SBLIMIT;i++)
      *ba++ = getbits(4);
    ba = balloc;
    for (i=0;i<SBLIMIT;i++)
      if ((*ba++))
        *sca++ = getbits(6);
  }
}

static void I_step_two(real fraction[2][SBLIMIT],unsigned int balloc[2*SBLIMIT],
	unsigned int scale_index[2][SBLIMIT],struct frame *fr)
{
  int i,n;
  int smpb[2*SBLIMIT]; /* values: 0-65535 */
  int *sample;
  register unsigned int *ba;
  register unsigned int *sca = (unsigned int *) scale_index;

  if(fr->stereo) {
    int jsbound = fr->jsbound;
    register real *f0 = fraction[0];
    register real *f1 = fraction[1];
    ba = balloc;
    for (sample=smpb,i=0;i<jsbound;i++)  {
      if ((n = *ba++))
        *sample++ = getbits(n+1);
      if ((n = *ba++))
        *sample++ = getbits(n+1);
    }
    for (i=jsbound;i<SBLIMIT;i++) 
      if ((n = *ba++))
        *sample++ = getbits(n+1);

    ba = balloc;
    for (sample=smpb,i=0;i<jsbound;i++) {
      if((n=*ba++))
        *f0++ = (real) ( ((-1)<<n) + (*sample++) + 1) * muls[n+1][*sca++];
      else
        *f0++ = 0.0;
      if((n=*ba++))
        *f1++ = (real) ( ((-1)<<n) + (*sample++) + 1) * muls[n+1][*sca++];
      else
        *f1++ = 0.0;
    }
    for (i=jsbound;i<SBLIMIT;i++) {
      if ((n=*ba++)) {
        real samp = ( ((-1)<<n) + (*sample++) + 1);
        *f0++ = samp * muls[n+1][*sca++];
        *f1++ = samp * muls[n+1][*sca++];
      }
      else
        *f0++ = *f1++ = 0.0;
    }
    for(i=fr->down_sample_sblimit;i<32;i++)
      fraction[0][i] = fraction[1][i] = 0.0;
  }
  else {
    register real *f0 = fraction[0];
    ba = balloc;
    for (sample=smpb,i=0;i<SBLIMIT;i++)
      if ((n = *ba++))
        *sample++ = getbits(n+1);
    ba = balloc;
    for (sample=smpb,i=0;i<SBLIMIT;i++) {
      if((n=*ba++))
        *f0++ = (real) ( ((-1)<<n) + (*sample++) + 1) * muls[n+1][*sca++];
      else
        *f0++ = 0.0;
    }
    for(i=fr->down_sample_sblimit;i<32;i++)
      fraction[0][i] = 0.0;
  }
}

static int do_layer1(struct frame *fr,int single)
{
  int clip=0;
  int i,stereo = fr->stereo;
  unsigned int balloc[2*SBLIMIT];
  unsigned int scale_index[2][SBLIMIT];
  real fraction[2][SBLIMIT];
//  int single = fr->single;

//  printf("do_layer1(0x%02X 0x%02X 0x%02X 0x%02X 0x%02X 0x%02X 0x%02X 0x%02X )\n",
//    wordpointer[0],wordpointer[1],wordpointer[2],wordpointer[3],wordpointer[4],wordpointer[5],wordpointer[6],wordpointer[7]);

  fr->jsbound = (fr->mode == MPG_MD_JOINT_STEREO) ? 
                         (fr->mode_ext<<2)+4 : 32;

  if(stereo == 1 || single == 3)
    single = 0;

  I_step_one(balloc,scale_index,fr);

  for (i=0;i<SCALE_BLOCK;i++)
  {
    I_step_two(fraction,balloc,scale_index,fr);

    if(single >= 0)
    {
      clip += (fr->synth_mono)( (real *) fraction[single],pcm_sample,&pcm_point);
    }
    else {
        int p1 = pcm_point;
        clip += (fr->synth)( (real *) fraction[0],0,pcm_sample,&p1);
        clip += (fr->synth)( (real *) fraction[1],1,pcm_sample,&pcm_point);
    }

  }

  return clip;
}