sysdeps/i386/i586/memcopy.h

   1 /* memcopy.h -- definitions for memory copy functions.  Pentium version.
   2    Copyright (C) 1994, 1995, 1996, 1997 Free Software Foundation, Inc.
   3    Contributed by Torbjorn Granlund (tege@sics.se).
   4    This file is part of the GNU C Library.
   5
   6    The GNU C Library is free software; you can redistribute it and/or
   7    modify it under the terms of the GNU Library General Public License as
   8    published by the Free Software Foundation; either version 2 of the
   9    License, or (at your option) any later version.
  10
  11    The GNU C Library 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    Library General Public License for more details.
  15
  16    You should have received a copy of the GNU Library General Public
  17    License along with the GNU C Library; see the file COPYING.LIB.  If not,
  18    write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  19    Boston, MA 02111-1307, USA.  */
  20
  21 /* Get the i386 definitions.  We will override some of them below.  */
  22 #include_next <memcopy.h>
  23
  24
  25 /* Written like this, the Pentium pipeline can execute the loop at a
  26    sustained rate of 2 instructions/clock, or asymptotically 480
  27    Mbytes/second at 60Mhz.  */
  28
  29 #undef  WORD_COPY_FWD
  30 #define WORD_COPY_FWD(dst_bp, src_bp, nbytes_left, nbytes)              \
  31   do                                                                    \
  32     {                                                                   \
  33       asm volatile ("subl       $32,%2\n"                               \
  34                     "js         2f\n"                                   \
  35                     "movl       0(%0),%%edx\n"  /* alloc dest line */   \
  36                     "1:\n"                                              \
  37                     "movl       28(%0),%%eax\n" /* alloc dest line */   \
  38                     "subl       $32,%2\n"       /* decr loop count */   \
  39                     "movl       0(%1),%%eax\n"  /* U pipe */            \
  40                     "movl       4(%1),%%edx\n"  /* V pipe */            \
  41                     "movl       %%eax,0(%0)\n"  /* U pipe */            \
  42                     "movl       %%edx,4(%0)\n"  /* V pipe */            \
  43                     "movl       8(%1),%%eax\n"                          \
  44                     "movl       12(%1),%%edx\n"                         \
  45                     "movl       %%eax,8(%0)\n"                          \
  46                     "movl       %%edx,12(%0)\n"                         \
  47                     "movl       16(%1),%%eax\n"                         \
  48                     "movl       20(%1),%%edx\n"                         \
  49                     "movl       %%eax,16(%0)\n"                         \
  50                     "movl       %%edx,20(%0)\n"                         \
  51                     "movl       24(%1),%%eax\n"                         \
  52                     "movl       28(%1),%%edx\n"                         \
  53                     "movl       %%eax,24(%0)\n"                         \
  54                     "movl       %%edx,28(%0)\n"                         \
  55                     "leal       32(%1),%1\n"    /* update src ptr */    \
  56                     "leal       32(%0),%0\n"    /* update dst ptr */    \
  57                     "jns        1b\n"                                   \
  58                     "2: addl    $32,%2" :                               \
  59                     "=r" (dst_bp), "=r" (src_bp), "=r" (nbytes_left) :  \
  60                     "0" (dst_bp), "1" (src_bp), "2" (nbytes) :          \
  61                     "ax", "dx");                                        \
  62     } while (0)
  63
  64 #undef  WORD_COPY_BWD
  65 #define WORD_COPY_BWD(dst_ep, src_ep, nbytes_left, nbytes)              \
  66   do                                                                    \
  67     {                                                                   \
  68       asm volatile ("subl       $32,%2\n"                               \
  69                     "js         2f\n"                                   \
  70                     "movl       -4(%0),%%edx\n"                         \
  71                     "1:\n"                                              \
  72                     "movl       -32(%0),%%eax\n"                        \
  73                     "subl       $32,%2\n"                               \
  74                     "movl       -4(%1),%%eax\n"                         \
  75                     "movl       -8(%1),%%edx\n"                         \
  76                     "movl       %%eax,-4(%0)\n"                         \
  77                     "movl       %%edx,-8(%0)\n"                         \
  78                     "movl       -12(%1),%%eax\n"                        \
  79                     "movl       -16(%1),%%edx\n"                        \
  80                     "movl       %%eax,-12(%0)\n"                        \
  81                     "movl       %%edx,-16(%0)\n"                        \
  82                     "movl       -20(%1),%%eax\n"                        \
  83                     "movl       -24(%1),%%edx\n"                        \
  84                     "movl       %%eax,-20(%0)\n"                        \
  85                     "movl       %%edx,-24(%0)\n"                        \
  86                     "movl       -28(%1),%%eax\n"                        \
  87                     "movl       -32(%1),%%edx\n"                        \
  88                     "movl       %%eax,-28(%0)\n"                        \
  89                     "movl       %%edx,-32(%0)\n"                        \
  90                     "leal       -32(%1),%1\n"                           \
  91                     "leal       -32(%0),%0\n"                           \
  92                     "jns        1b\n"                                   \
  93                     "2: addl    $32,%2" :                               \
  94                     "=r" (dst_ep), "=r" (src_ep), "=r" (nbytes_left) :  \
  95                     "0" (dst_ep), "1" (src_ep), "2" (nbytes) :          \
  96                     "ax", "dx");                                        \
  97     } while (0)