view liba52/resample_altivec.c @ 30884:53901d222e8a

Announce SMP support for Win32. Don't hardcode dwNumberOfProcessors=1 for Win32 anymore; the mutex/event code is still far from perfect, but now good enough that I can't find any codecs that breaks with this (tested on a quad with various codecs). This tells codecs they can use more than one core if they want to (some already did, by launching multiple threads even when told there was only a single core).

/*
 * resample.c
 * Copyright (C) 2004 Romain Dolbeau <[e-mail]>
 *
 * This file is part of a52dec, a free ATSC A-52 stream decoder.
 * See http://liba52.sourceforge.net/ for updates.
 *
 * File added for use with MPlayer and not part of original a52dec.
 *
 * a52dec is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * a52dec is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 */

#ifdef HAVE_ALTIVEC_H
#include <altivec.h>
#endif

const vector signed int magic = {0x43c00000,0x43c00000,0x43c00000,0x43c00000};

static inline vector signed short convert16_altivec(vector signed int v1, vector signed int v2)
{
  register vector signed short result;
  v1 = vec_subs(v1, magic);
  v2 = vec_subs(v2, magic);
  result = vec_packs(v1, v2);

  return result;
}

static void unaligned_store(vector signed short value, int off, int16_t *dst)
{
    register vector unsigned char align = vec_lvsr(0, dst),
                                  mask = vec_lvsl(0, dst);
    register vector signed short t0,t1, edges;

    t0 = vec_ld(0+off, dst);
    t1 = vec_ld(15+off, dst);
    edges = vec_perm(t1 ,t0, mask);
    t1 = vec_perm(value, edges, align);
    t0 = vec_perm(edges, value, align);
    vec_st(t1, 15+off, dst);
    vec_st(t0, 0+off, dst);
}

static int a52_resample_STEREO_to_2_altivec(float * _f, int16_t * s16){
#if 0
  int i;
  int32_t * f = (int32_t *) _f;
  for (i = 0; i < 256; i++) {
    s16[2*i] = convert (f[i]);
    s16[2*i+1] = convert (f[i+256]);
  }
  return 2*256;
#else
  int i = 0;
  int32_t * f = (int32_t *) _f;
  register vector signed int f0, f4, f256, f260;
  register vector signed short reven, rodd, r0, r1;

  for (i = 0; i < 256; i+= 8) {
    f0 = vec_ld(0, f);
    f4 = vec_ld(16, f);

    f256 = vec_ld(1024, f);
    f260 = vec_ld(1040, f);

    reven = convert16_altivec(f0, f4);
    rodd = convert16_altivec(f256, f260);

    r0 = vec_mergeh(reven, rodd);
    r1 = vec_mergel(reven, rodd);
    // FIXME can be merged to spare some I/O
    unaligned_store(r0, 0, s16);
    unaligned_store(r1, 16, s16);

    f += 8;
    s16 += 16;
  }
  return(2*256);
#endif
}

static void* a52_resample_altivec(int flags, int ch){
fprintf(stderr, "Checking for AltiVec resampler : 0x%08x, %d\n", flags, ch);

  switch (flags) {
  case A52_CHANNEL:
  case A52_STEREO:
  case A52_DOLBY:
    if(ch==2) return a52_resample_STEREO_to_2_altivec;
    break;

  default:
	fprintf(stderr, "Unsupported flags: 0x%08x (%d channels)\n", flags, ch);
	break;
  }
  return NULL;
}