libavcodec/ppc/fft_altivec.c

   1 /*
   2  * FFT/IFFT transforms
   3  * AltiVec-enabled
   4  * Copyright (c) 2003 Romain Dolbeau <romain@dolbeau.org>
   5  * Based on code Copyright (c) 2002 Fabrice Bellard.
   6  *
   7  * This library is free software; you can redistribute it and/or
   8  * modify it under the terms of the GNU Lesser General Public
   9  * License as published by the Free Software Foundation; either
  10  * version 2 of the License, or (at your option) any later version.
  11  *
  12  * This library is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15  * Lesser General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU Lesser General Public
  18  * License along with this library; if not, write to the Free Software
  19  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  20  */
  21 #include "../dsputil.h"
  22
  23 #include "dsputil_altivec.h"
  24
  25 /*
  26   those three macros are from libavcodec/fft.c
  27   and are required for the reference C code
  28 */
  29 /* butter fly op */
  30 #define BF(pre, pim, qre, qim, pre1, pim1, qre1, qim1) \
  31 {\
  32   FFTSample ax, ay, bx, by;\
  33   bx=pre1;\
  34   by=pim1;\
  35   ax=qre1;\
  36   ay=qim1;\
  37   pre = (bx + ax);\
  38   pim = (by + ay);\
  39   qre = (bx - ax);\
  40   qim = (by - ay);\
  41 }
  42 #define MUL16(a,b) ((a) * (b))
  43 #define CMUL(pre, pim, are, aim, bre, bim) \
  44 {\
  45    pre = (MUL16(are, bre) - MUL16(aim, bim));\
  46    pim = (MUL16(are, bim) + MUL16(bre, aim));\
  47 }
  48
  49
  50 /**
  51  * Do a complex FFT with the parameters defined in fft_init(). The
  52  * input data must be permuted before with s->revtab table. No
  53  * 1.0/sqrt(n) normalization is done.
  54  * AltiVec-enabled
  55  * This code assumes that the 'z' pointer is 16 bytes-aligned
  56  * It also assumes all FFTComplex are 8 bytes-aligned pair of float
  57  * The code is exactly the same as the SSE version, except
  58  * that successive MUL + ADD/SUB have been merged into
  59  * fused multiply-add ('vec_madd' in altivec)
  60  */
  61 void fft_calc_altivec(FFTContext *s, FFTComplex *z)
  62 {
  63 POWERPC_TBL_DECLARE(altivec_fft_num, s->nbits >= 6);
  64 #ifdef ALTIVEC_USE_REFERENCE_C_CODE
  65     int ln = s->nbits;
  66     int j, np, np2;
  67     int nblocks, nloops;
  68     register FFTComplex *p, *q;
  69     FFTComplex *exptab = s->exptab;
  70     int l;
  71     FFTSample tmp_re, tmp_im;
  72
  73 POWERPC_TBL_START_COUNT(altivec_fft_num, s->nbits >= 6);
  74
  75     np = 1 << ln;
  76
  77     /* pass 0 */
  78
  79     p=&z[0];
  80     j=(np >> 1);
  81     do {
  82         BF(p[0].re, p[0].im, p[1].re, p[1].im,
  83            p[0].re, p[0].im, p[1].re, p[1].im);
  84         p+=2;
  85     } while (--j != 0);
  86
  87     /* pass 1 */
  88
  89
  90     p=&z[0];
  91     j=np >> 2;
  92     if (s->inverse) {
  93         do {
  94             BF(p[0].re, p[0].im, p[2].re, p[2].im,
  95                p[0].re, p[0].im, p[2].re, p[2].im);
  96             BF(p[1].re, p[1].im, p[3].re, p[3].im,
  97                p[1].re, p[1].im, -p[3].im, p[3].re);
  98             p+=4;
  99         } while (--j != 0);
 100     } else {
 101         do {
 102             BF(p[0].re, p[0].im, p[2].re, p[2].im,
 103                p[0].re, p[0].im, p[2].re, p[2].im);
 104             BF(p[1].re, p[1].im, p[3].re, p[3].im,
 105                p[1].re, p[1].im, p[3].im, -p[3].re);
 106             p+=4;
 107         } while (--j != 0);
 108     }
 109     /* pass 2 .. ln-1 */
 110
 111     nblocks = np >> 3;
 112     nloops = 1 << 2;
 113     np2 = np >> 1;
 114     do {
 115         p = z;
 116         q = z + nloops;
 117         for (j = 0; j < nblocks; ++j) {
 118             BF(p->re, p->im, q->re, q->im,
 119                p->re, p->im, q->re, q->im);
 120
 121             p++;
 122             q++;
 123             for(l = nblocks; l < np2; l += nblocks) {
 124                 CMUL(tmp_re, tmp_im, exptab[l].re, exptab[l].im, q->re, q->im);
 125                 BF(p->re, p->im, q->re, q->im,
 126                    p->re, p->im, tmp_re, tmp_im);
 127                 p++;
 128                 q++;
 129             }
 130
 131             p += nloops;
 132             q += nloops;
 133         }
 134         nblocks = nblocks >> 1;
 135         nloops = nloops << 1;
 136     } while (nblocks != 0);
 137
 138 POWERPC_TBL_STOP_COUNT(altivec_fft_num, s->nbits >= 6);
 139
 140 #else /* ALTIVEC_USE_REFERENCE_C_CODE */
 141 #ifdef CONFIG_DARWIN
 142     register const vector float vczero = (const vector float)(0.);
 143 #else
 144     register const vector float vczero = (const vector float){0.,0.,0.,0.};
 145 #endif
 146
 147     int ln = s->nbits;
 148     int j, np, np2;
 149     int nblocks, nloops;
 150     register FFTComplex *p, *q;
 151     FFTComplex *cptr, *cptr1;
 152     int k;
 153
 154 POWERPC_TBL_START_COUNT(altivec_fft_num, s->nbits >= 6);
 155
 156     np = 1 << ln;
 157
 158     {
 159         vector float *r, a, b, a1, c1, c2;
 160
 161         r = (vector float *)&z[0];
 162
 163         c1 = vcii(p,p,n,n);
 164
 165         if (s->inverse)
 166             {
 167                 c2 = vcii(p,p,n,p);
 168             }
 169         else
 170             {
 171                 c2 = vcii(p,p,p,n);
 172             }
 173
 174         j = (np >> 2);
 175         do {
 176             a = vec_ld(0, r);
 177             a1 = vec_ld(sizeof(vector float), r);
 178
 179             b = vec_perm(a,a,vcprmle(1,0,3,2));
 180             a = vec_madd(a,c1,b);
 181             /* do the pass 0 butterfly */
 182
 183             b = vec_perm(a1,a1,vcprmle(1,0,3,2));
 184             b = vec_madd(a1,c1,b);
 185             /* do the pass 0 butterfly */
 186
 187             /* multiply third by -i */
 188             b = vec_perm(b,b,vcprmle(2,3,1,0));
 189
 190             /* do the pass 1 butterfly */
 191             vec_st(vec_madd(b,c2,a), 0, r);
 192             vec_st(vec_nmsub(b,c2,a), sizeof(vector float), r);
 193
 194             r += 2;
 195         } while (--j != 0);
 196     }
 197     /* pass 2 .. ln-1 */
 198
 199     nblocks = np >> 3;
 200     nloops = 1 << 2;
 201     np2 = np >> 1;
 202
 203     cptr1 = s->exptab1;
 204     do {
 205         p = z;
 206         q = z + nloops;
 207         j = nblocks;
 208         do {
 209             cptr = cptr1;
 210             k = nloops >> 1;
 211             do {
 212                 vector float a,b,c,t1;
 213
 214                 a = vec_ld(0, (float*)p);
 215                 b = vec_ld(0, (float*)q);
 216
 217                 /* complex mul */
 218                 c = vec_ld(0, (float*)cptr);
 219                 /*  cre*re cim*re */
 220                 t1 = vec_madd(c, vec_perm(b,b,vcprmle(2,2,0,0)),vczero);
 221                 c = vec_ld(sizeof(vector float), (float*)cptr);
 222                 /*  -cim*im cre*im */
 223                 b = vec_madd(c, vec_perm(b,b,vcprmle(3,3,1,1)),t1);
 224
 225                 /* butterfly */
 226                 vec_st(vec_add(a,b), 0, (float*)p);
 227                 vec_st(vec_sub(a,b), 0, (float*)q);
 228
 229                 p += 2;
 230                 q += 2;
 231                 cptr += 4;
 232             } while (--k);
 233
 234             p += nloops;
 235             q += nloops;
 236         } while (--j);
 237         cptr1 += nloops * 2;
 238         nblocks = nblocks >> 1;
 239         nloops = nloops << 1;
 240     } while (nblocks != 0);
 241
 242 POWERPC_TBL_STOP_COUNT(altivec_fft_num, s->nbits >= 6);
 243
 244 #endif /* ALTIVEC_USE_REFERENCE_C_CODE */
 245 }