libavcodec/sh4/dsputil_sh4.c

   1 /*
   2  *  sh4 dsputil
   3  *
   4  * Copyright (c) 2003 BERO <bero@geocities.co.jp>
   5  *
   6  * This file is part of FFmpeg.
   7  *
   8  * FFmpeg is free software; you can redistribute it and/or
   9  * modify it under the terms of the GNU Lesser General Public
  10  * License as published by the Free Software Foundation; either
  11  * version 2.1 of the License, or (at your option) any later version.
  12  *
  13  * FFmpeg is distributed in the hope that it will be useful,
  14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  16  * Lesser General Public License for more details.
  17  *
  18  * You should have received a copy of the GNU Lesser General Public
  19  * License along with FFmpeg; if not, write to the Free Software
  20  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  21  */
  22
  23 #include "libavcodec/avcodec.h"
  24 #include "libavcodec/dsputil.h"
  25
  26 static void memzero_align8(void *dst,size_t size)
  27 {
  28 #if defined(__SH4__) || defined(__SH4_SINGLE__) || defined(__SH4_SINGLE_ONLY__)
  29         (char*)dst+=size;
  30         size/=8*4;
  31         asm(
  32 #if defined(__SH4__)
  33         " fschg\n"  //single float mode
  34 #endif
  35         " fldi0 fr0\n"
  36         " fldi0 fr1\n"
  37         " fschg\n"  // double
  38         "1: \n" \
  39         " dt %1\n"
  40         " fmov  dr0,@-%0\n"
  41         " fmov  dr0,@-%0\n"
  42         " fmov  dr0,@-%0\n"
  43         " bf.s 1b\n"
  44         " fmov  dr0,@-%0\n"
  45 #if defined(__SH4_SINGLE__) || defined(__SH4_SINGLE_ONLY__)
  46         " fschg" //back to single
  47 #endif
  48         : : "r"(dst),"r"(size): "memory" );
  49 #else
  50         double *d = dst;
  51         size/=8*4;
  52         do {
  53                 d[0] = 0.0;
  54                 d[1] = 0.0;
  55                 d[2] = 0.0;
  56                 d[3] = 0.0;
  57                 d+=4;
  58         } while(--size);
  59 #endif
  60 }
  61
  62 static void clear_blocks_sh4(DCTELEM *blocks)
  63 {
  64 //        if (((int)blocks&7)==0)
  65         memzero_align8(blocks,sizeof(DCTELEM)*6*64);
  66 }
  67
  68 extern void idct_sh4(DCTELEM *block);
  69 static void idct_put(uint8_t *dest, int line_size, DCTELEM *block)
  70 {
  71         idct_sh4(block);
  72         int i;
  73         uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
  74         for(i=0;i<8;i++) {
  75                 dest[0] = cm[block[0]];
  76                 dest[1] = cm[block[1]];
  77                 dest[2] = cm[block[2]];
  78                 dest[3] = cm[block[3]];
  79                 dest[4] = cm[block[4]];
  80                 dest[5] = cm[block[5]];
  81                 dest[6] = cm[block[6]];
  82                 dest[7] = cm[block[7]];
  83                 dest+=line_size;
  84                 block+=8;
  85         }
  86 }
  87 static void idct_add(uint8_t *dest, int line_size, DCTELEM *block)
  88 {
  89         idct_sh4(block);
  90         int i;
  91         uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
  92         for(i=0;i<8;i++) {
  93                 dest[0] = cm[dest[0]+block[0]];
  94                 dest[1] = cm[dest[1]+block[1]];
  95                 dest[2] = cm[dest[2]+block[2]];
  96                 dest[3] = cm[dest[3]+block[3]];
  97                 dest[4] = cm[dest[4]+block[4]];
  98                 dest[5] = cm[dest[5]+block[5]];
  99                 dest[6] = cm[dest[6]+block[6]];
 100                 dest[7] = cm[dest[7]+block[7]];
 101                 dest+=line_size;
 102                 block+=8;
 103         }
 104 }
 105
 106 extern void dsputil_init_align(DSPContext* c, AVCodecContext *avctx);
 107
 108 void dsputil_init_sh4(DSPContext* c, AVCodecContext *avctx)
 109 {
 110         const int idct_algo= avctx->idct_algo;
 111         dsputil_init_align(c,avctx);
 112
 113         c->clear_blocks = clear_blocks_sh4;
 114         if(idct_algo==FF_IDCT_AUTO || idct_algo==FF_IDCT_SH4){
 115                 c->idct_put = idct_put;
 116                 c->idct_add = idct_add;
 117                c->idct     = idct_sh4;
 118                 c->idct_permutation_type= FF_NO_IDCT_PERM; //FF_SIMPLE_IDCT_PERM; //FF_LIBMPEG2_IDCT_PERM;
 119         }
 120 }