svq1enc_altivec.c

   1 /*
   2  * Copyright (c) 2007 Luca Barbato <lu_zero@gentoo.org>
   3  *
   4  * This file is part of FFmpeg.
   5  *
   6  * FFmpeg is free software; you can redistribute it and/or
   7  * modify it under the terms of the GNU Lesser General Public
   8  * License as published by the Free Software Foundation; either
   9  * version 2.1 of the License, or (at your option) any later version.
  10  *
  11  * FFmpeg is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  14  * Lesser General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU Lesser General Public
  17  * License along with FFmpeg; if not, write to the Free Software
  18  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  19  */
  20
  21 #include <stdint.h>
  22
  23 #include "config.h"
  24 #if HAVE_ALTIVEC_H
  25 #include <altivec.h>
  26 #endif
  27
  28 #include "libavutil/attributes.h"
  29 #include "libavutil/cpu.h"
  30 #include "libavutil/ppc/cpu.h"
  31 #include "libavutil/ppc/types_altivec.h"
  32 #include "libavutil/ppc/util_altivec.h"
  33 #include "libavcodec/svq1enc.h"
  34
  35 #if HAVE_ALTIVEC
  36 static int ssd_int8_vs_int16_altivec(const int8_t *pix1, const int16_t *pix2,
  37                                      int size)
  38 {
  39     int i, size16 = size >> 4;
  40     vector signed char vpix1;
  41     vector signed short vpix2, vdiff, vpix1l, vpix1h;
  42     union {
  43         vector signed int vscore;
  44         int32_t score[4];
  45     } u = { .vscore = vec_splat_s32(0) };
  46
  47     while (size16) {
  48         // score += (pix1[i] - pix2[i]) * (pix1[i] - pix2[i]);
  49         // load pix1 and the first batch of pix2
  50
  51         vpix1 = vec_unaligned_load(pix1);
  52         vpix2 = vec_unaligned_load(pix2);
  53         pix2 += 8;
  54         // unpack
  55         vpix1h = vec_unpackh(vpix1);
  56         vdiff  = vec_sub(vpix1h, vpix2);
  57         vpix1l = vec_unpackl(vpix1);
  58         // load another batch from pix2
  59         vpix2    = vec_unaligned_load(pix2);
  60         u.vscore = vec_msum(vdiff, vdiff, u.vscore);
  61         vdiff    = vec_sub(vpix1l, vpix2);
  62         u.vscore = vec_msum(vdiff, vdiff, u.vscore);
  63         pix1    += 16;
  64         pix2    += 8;
  65         size16--;
  66     }
  67     u.vscore = vec_sums(u.vscore, vec_splat_s32(0));
  68
  69     size %= 16;
  70     for (i = 0; i < size; i++)
  71         u.score[3] += (pix1[i] - pix2[i]) * (pix1[i] - pix2[i]);
  72
  73     return u.score[3];
  74 }
  75 #endif /* HAVE_ALTIVEC */
  76
  77 av_cold void ff_svq1enc_init_ppc(SVQ1EncContext *c)
  78 {
  79 #if HAVE_ALTIVEC
  80     if (!PPC_ALTIVEC(av_get_cpu_flags()))
  81         return;
  82
  83     c->ssd_int8_vs_int16 = ssd_int8_vs_int16_altivec;
  84 #endif /* HAVE_ALTIVEC */
  85 }