liba52/resample_altivec.c

   1 /*
   2  * resample.c
   3  * Copyright (C) 2004 Romain Dolbeau <romain@dolbeau.org>
   4  *
   5  * This file is part of a52dec, a free ATSC A-52 stream decoder.
   6  * See http://liba52.sourceforge.net/ for updates.
   7  *
   8  * File added for use with MPlayer and not part of original a52dec.
   9  *
  10  * a52dec is free software; you can redistribute it and/or modify
  11  * it under the terms of the GNU General Public License as published by
  12  * the Free Software Foundation; either version 2 of the License, or
  13  * (at your option) any later version.
  14  *
  15  * a52dec is distributed in the hope that it will be useful,
  16  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  17  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  18  * GNU General Public License for more details.
  19  *
  20  * You should have received a copy of the GNU General Public License
  21  * along with this program; if not, write to the Free Software
  22  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  23  */
  24
  25 #ifdef HAVE_ALTIVEC_H
  26 #include <altivec.h>
  27 #endif
  28
  29 const vector signed int magic = {0x43c00000,0x43c00000,0x43c00000,0x43c00000};
  30
  31 static inline vector signed short convert16_altivec(vector signed int v1, vector signed int v2)
  32 {
  33   register vector signed short result;
  34   v1 = vec_subs(v1, magic);
  35   v2 = vec_subs(v2, magic);
  36   result = vec_packs(v1, v2);
  37
  38   return result;
  39 }
  40
  41 static void unaligned_store(vector signed short value, int off, int16_t *dst)
  42 {
  43     register vector unsigned char align = vec_lvsr(0, dst),
  44                                   mask = vec_lvsl(0, dst);
  45     register vector signed short t0,t1, edges;
  46
  47     t0 = vec_ld(0+off, dst);
  48     t1 = vec_ld(15+off, dst);
  49     edges = vec_perm(t1 ,t0, mask);
  50     t1 = vec_perm(value, edges, align);
  51     t0 = vec_perm(edges, value, align);
  52     vec_st(t1, 15+off, dst);
  53     vec_st(t0, 0+off, dst);
  54 }
  55
  56 static int a52_resample_STEREO_to_2_altivec(float * _f, int16_t * s16){
  57 #if 0
  58   int i;
  59   int32_t * f = (int32_t *) _f;
  60   for (i = 0; i < 256; i++) {
  61     s16[2*i] = convert (f[i]);
  62     s16[2*i+1] = convert (f[i+256]);
  63   }
  64   return 2*256;
  65 #else
  66   int i = 0;
  67   int32_t * f = (int32_t *) _f;
  68   register vector signed int f0, f4, f256, f260;
  69   register vector signed short reven, rodd, r0, r1;
  70
  71   for (i = 0; i < 256; i+= 8) {
  72     f0 = vec_ld(0, f);
  73     f4 = vec_ld(16, f);
  74
  75     f256 = vec_ld(1024, f);
  76     f260 = vec_ld(1040, f);
  77
  78     reven = convert16_altivec(f0, f4);
  79     rodd = convert16_altivec(f256, f260);
  80
  81     r0 = vec_mergeh(reven, rodd);
  82     r1 = vec_mergel(reven, rodd);
  83     // FIXME can be merged to spare some I/O
  84     unaligned_store(r0, 0, s16);
  85     unaligned_store(r1, 16, s16);
  86
  87     f += 8;
  88     s16 += 16;
  89   }
  90   return(2*256);
  91 #endif
  92 }
  93
  94 static void* a52_resample_altivec(int flags, int ch){
  95 fprintf(stderr, "Checking for AltiVec resampler : 0x%08x, %d\n", flags, ch);
  96
  97   switch (flags) {
  98   case A52_CHANNEL:
  99   case A52_STEREO:
 100   case A52_DOLBY:
 101     if(ch==2) return a52_resample_STEREO_to_2_altivec;
 102     break;
 103
 104   default:
 105         fprintf(stderr, "Unsupported flags: 0x%08x (%d channels)\n", flags, ch);
 106         break;
 107   }
 108   return NULL;
 109 }
 110