modules/altivec/memcpy.c

   1 /*****************************************************************************
   2  * memcpy.c : AltiVec memcpy module
   3  *****************************************************************************
   4  * Copyright (C) 2001 the VideoLAN team
   5  * $Id$
   6  *
   7  * Author: Christophe Massiot <massiot@via.ecp.fr>
   8  *
   9  * This program is free software; you can redistribute it and/or modify
  10  * it under the terms of the GNU General Public License as published by
  11  * the Free Software Foundation; either version 2 of the License, or
  12  * (at your option) any later version.
  13  *
  14  * This program is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  17  * GNU General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU General Public License
  20  * along with this program; if not, write to the Free Software
  21  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston MA 02110-1301, USA.
  22  *****************************************************************************/
  23
  24 #ifndef __BUILD_ALTIVEC_ASM__
  25
  26 /*****************************************************************************
  27  * Preamble
  28  *****************************************************************************/
  29
  30 #ifdef HAVE_CONFIG_H
  31 # include "config.h"
  32 #endif
  33
  34 #include <vlc_common.h>
  35 #include <vlc_plugin.h>
  36 #include <vlc_cpu.h>
  37
  38 #ifdef HAVE_ALTIVEC_H
  39 #   include <altivec.h>
  40 #endif
  41
  42 /*****************************************************************************
  43  * Local prototypes.
  44  *****************************************************************************/
  45 static void * fast_memcpy ( void * to, const void * from, size_t len );
  46
  47 /*****************************************************************************
  48  * Module initializer.
  49  *****************************************************************************/
  50 static int Activate ( vlc_object_t *p_this )
  51 {
  52     if( !(vlc_CPU() & CPU_CAPABILITY_ALTIVEC) )
  53         return VLC_EGENERIC;
  54
  55     VLC_UNUSED(p_this);
  56     vlc_fastmem_register( fast_memcpy );
  57     return VLC_SUCCESS;
  58 }
  59
  60 /*****************************************************************************
  61  * Module descriptor.
  62  *****************************************************************************/
  63 vlc_module_begin ()
  64     set_description( N_("AltiVec memcpy") )
  65     set_category( CAT_ADVANCED )
  66     set_subcategory( SUBCAT_ADVANCED_MISC )
  67     set_capability( "memcpy", 100 )
  68     set_callbacks( Activate, NULL )
  69     add_shortcut( "altivec" )
  70 vlc_module_end ()
  71
  72 #else
  73 typedef unsigned long size_t;
  74 #endif /* __BUILD_ALTIVEC_ASM__ */
  75
  76 #if defined(CAN_COMPILE_C_ALTIVEC) || defined( __BUILD_ALTIVEC_ASM__ )
  77
  78 #define vector_s16_t vector signed short
  79 #define vector_u16_t vector unsigned short
  80 #define vector_s8_t vector signed char
  81 #define vector_u8_t vector unsigned char
  82 #define vector_s32_t vector signed int
  83 #define vector_u32_t vector unsigned int
  84 #define MMREG_SIZE 16
  85
  86 #define SMALL_MEMCPY(to, from, len)                                         \
  87 {                                                                           \
  88     unsigned char * end = to + len;                                         \
  89     while( to < end )                                                       \
  90     {                                                                       \
  91         *to++ = *from++;                                                    \
  92     }                                                                       \
  93 }
  94
  95 static void * fast_memcpy( void * _to, const void * _from, size_t len )
  96 {
  97     void * retval = _to;
  98     unsigned char * to = (unsigned char *)_to;
  99     unsigned char * from = (unsigned char *)_from;
 100
 101     if( len > 16 )
 102     {
 103         /* Align destination to MMREG_SIZE -boundary */
 104         register unsigned long int delta;
 105
 106         delta = ((unsigned long)to)&(MMREG_SIZE-1);
 107         if( delta )
 108         {
 109             delta = MMREG_SIZE - delta;
 110             len -= delta;
 111             SMALL_MEMCPY(to, from, delta);
 112         }
 113
 114         if( len & ~(MMREG_SIZE-1) )
 115         {
 116             vector_u8_t perm, ref0, ref1, tmp;
 117
 118             perm = vec_lvsl( 0, from );
 119             ref0 = vec_ld( 0, from );
 120             ref1 = vec_ld( 15, from );
 121             from += 16;
 122             len -= 16;
 123             tmp = vec_perm( ref0, ref1, perm );
 124             while( len & ~(MMREG_SIZE-1) )
 125             {
 126                 ref0 = vec_ld( 0, from );
 127                 ref1 = vec_ld( 15, from );
 128                 from += 16;
 129                 len -= 16;
 130                 vec_st( tmp, 0, to );
 131                 tmp = vec_perm( ref0, ref1, perm );
 132                 to += 16;
 133             }
 134             vec_st( tmp, 0, to );
 135             to += 16;
 136         }
 137     }
 138
 139     if( len )
 140     {
 141         SMALL_MEMCPY( to, from, len );
 142     }
 143
 144     return retval;
 145 }
 146
 147 #endif
 148
 149 #if !defined(CAN_COMPILE_C_ALTIVEC) && !defined(__BUILD_ALTIVEC_ASM__)
 150
 151 /*
 152  * The asm code is generated with:
 153  *
 154  * gcc-2.95 -fvec -D__BUILD_ALTIVEC_ASM__ -O9 -fomit-frame-pointer -mregnames -S *      memcpyaltivec.c
 155  *
 156  * sed 's/.L/._L/g' memcpyaltivec.s |
 157  * awk '{args=""; len=split ($2, arg, ",");
 158  *      for (i=1; i<=len; i++) { a=arg[i]; if (i<len) a=a",";
 159  *                               args = args sprintf ("%-6s", a) }
 160  *      printf ("\t\"\t%-16s%-24s\\n\"\n", $1, args) }' |
 161  * unexpand -a
 162  */
 163
 164 static void * fast_memcpy( void * _to, const void * _from, size_t len )
 165 {
 166     asm ("                                              \n"
 167     "    cmplwi        %cr0, %r5,  16        \n"
 168     "    mr        %r9,  %r3        \n"
 169     "    bc        4,    1,    ._L3    \n"
 170     "    andi.        %r0,  %r3,  15        \n"
 171     "    bc        12,   2,    ._L4    \n"
 172     "    subfic        %r0,  %r0,  16        \n"
 173     "    add        %r11, %r3,  %r0        \n"
 174     "    cmplw        %cr0, %r3,  %r11    \n"
 175     "    subf        %r5,  %r0,  %r5        \n"
 176     "    bc        4,    0,    ._L4    \n"
 177     "    ._L7:                    \n"
 178     "    lbz        %r0,  0(%r4)        \n"
 179     "    stb        %r0,  0(%r9)        \n"
 180     "    addi        %r9,  %r9,  1        \n"
 181     "    cmplw        %cr0, %r9,  %r11    \n"
 182     "    addi        %r4,  %r4,  1        \n"
 183     "    bc        12,   0,    ._L7    \n"
 184     "    ._L4:                    \n"
 185     "    rlwinm.        %r0,  %r5,  0,      0,    27    \n"
 186     "    bc        12,   2,    ._L3    \n"
 187     "    addi        %r5,  %r5,  -16        \n"
 188     "    li        %r11, 15        \n"
 189     "    lvsl        %v12, 0,    %r4        \n"
 190     "    lvx        %v1,  0,    %r4        \n"
 191     "    lvx        %v0,  %r11, %r4        \n"
 192     "    rlwinm.        %r0,  %r5,  0,      0,    27    \n"
 193     "    vperm        %v13, %v1,  %v0,  %v12    \n"
 194     "    addi        %r4,  %r4,  16        \n"
 195     "    bc        12,   2,    ._L11    \n"
 196     "    ._L12:                    \n"
 197     "    addi        %r5,  %r5,  -16        \n"
 198     "    li        %r11, 15        \n"
 199     "    lvx        %v1,  0,    %r4        \n"
 200     "    lvx        %v0,  %r11, %r4        \n"
 201     "    rlwinm.        %r0,  %r5,  0,      0,    27    \n"
 202     "    stvx        %v13, 0,    %r9        \n"
 203     "    vperm        %v13, %v1,  %v0,  %v12    \n"
 204     "    addi        %r4,  %r4,  16        \n"
 205     "    addi        %r9,  %r9,  16        \n"
 206     "    bc        4,    2,    ._L12    \n"
 207     "    ._L11:                    \n"
 208     "    stvx        %v13, 0,    %r9        \n"
 209     "    addi        %r9,  %r9,  16        \n"
 210     "    ._L3:                    \n"
 211     "    cmpwi        %cr0, %r5,  0        \n"
 212     "    bclr        12,   2            \n"
 213     "    add        %r5,  %r9,  %r5        \n"
 214     "    cmplw        %cr0, %r9,  %r5        \n"
 215     "    bclr        4,    0            \n"
 216     "    ._L17:                    \n"
 217     "    lbz        %r0,  0(%r4)        \n"
 218     "    stb        %r0,  0(%r9)        \n"
 219     "    addi        %r9,  %r9,  1        \n"
 220     "    cmplw        %cr0, %r9,  %r5        \n"
 221     "    addi        %r4,  %r4,  1        \n"
 222     "    bc        12,   0,    ._L17    \n"
 223         );
 224 }
 225
 226 #endif