source/libs/gmp/gmp-src/mpn/x86/pentium/copyd.asm

   1 dnl  Intel Pentium mpn_copyd -- copy limb vector, decrementing.
   2
   3 dnl  Copyright 1996, 2001, 2002, 2006 Free Software Foundation, Inc.
   4
   5 dnl  This file is part of the GNU MP Library.
   6 dnl
   7 dnl  The GNU MP Library is free software; you can redistribute it and/or modify
   8 dnl  it under the terms of either:
   9 dnl
  10 dnl    * the GNU Lesser General Public License as published by the Free
  11 dnl      Software Foundation; either version 3 of the License, or (at your
  12 dnl      option) any later version.
  13 dnl
  14 dnl  or
  15 dnl
  16 dnl    * the GNU General Public License as published by the Free Software
  17 dnl      Foundation; either version 2 of the License, or (at your option) any
  18 dnl      later version.
  19 dnl
  20 dnl  or both in parallel, as here.
  21 dnl
  22 dnl  The GNU MP Library is distributed in the hope that it will be useful, but
  23 dnl  WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
  24 dnl  or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
  25 dnl  for more details.
  26 dnl
  27 dnl  You should have received copies of the GNU General Public License and the
  28 dnl  GNU Lesser General Public License along with the GNU MP Library.  If not,
  29 dnl  see https://www.gnu.org/licenses/.
  30
  31 include(`../config.m4')
  32
  33
  34 C P5: 1.25 cycles/limb
  35
  36
  37 C void mpn_copyd (mp_ptr dst, mp_srcptr src, mp_size_t size);
  38 C
  39 C See comments in copyi.asm.
  40
  41 defframe(PARAM_SIZE,12)
  42 defframe(PARAM_SRC, 8)
  43 defframe(PARAM_DST, 4)
  44
  45         TEXT
  46         ALIGN(8)
  47 PROLOGUE(mpn_copyd)
  48 deflit(`FRAME',0)
  49
  50         movl    PARAM_SRC, %eax
  51         movl    PARAM_SIZE, %ecx
  52
  53         pushl   %esi    FRAME_pushl()
  54         pushl   %edi    FRAME_pushl()
  55
  56         leal    -4(%eax,%ecx,4), %eax           C &src[size-1]
  57         movl    PARAM_DST, %edx
  58
  59         subl    $7, %ecx                        C size-7
  60         jle     L(end)
  61
  62         movl    28-4(%edx,%ecx,4), %esi         C prefetch cache, dst[size-1]
  63         nop
  64
  65 L(top):
  66         C eax   src, decrementing
  67         C ebx
  68         C ecx   counter, limbs
  69         C edx   dst
  70         C esi   scratch
  71         C edi   scratch
  72         C ebp
  73
  74         movl    28-32(%edx,%ecx,4), %esi        C prefetch dst cache line
  75         subl    $8, %ecx
  76
  77         movl    (%eax), %esi                    C read words pairwise
  78         movl    -4(%eax), %edi
  79         movl    %esi, 56(%edx,%ecx,4)           C store words pairwise
  80         movl    %edi, 52(%edx,%ecx,4)
  81
  82         movl    -8(%eax), %esi
  83         movl    -12(%eax), %edi
  84         movl    %esi, 48(%edx,%ecx,4)
  85         movl    %edi, 44(%edx,%ecx,4)
  86
  87         movl    -16(%eax), %esi
  88         movl    -20(%eax), %edi
  89         movl    %esi, 40(%edx,%ecx,4)
  90         movl    %edi, 36(%edx,%ecx,4)
  91
  92         movl    -24(%eax), %esi
  93         movl    -28(%eax), %edi
  94         movl    %esi, 32(%edx,%ecx,4)
  95         movl    %edi, 28(%edx,%ecx,4)
  96
  97         leal    -32(%eax), %eax
  98         jg      L(top)
  99
 100
 101 L(end):
 102         C ecx   -7 to 0, representing respectively 0 to 7 limbs remaining
 103         C eax   src end
 104         C edx   dst, next location to store
 105
 106         addl    $4, %ecx
 107         jle     L(no4)
 108
 109         movl    (%eax), %esi
 110         movl    -4(%eax), %edi
 111         movl    %esi, 8(%edx,%ecx,4)
 112         movl    %edi, 4(%edx,%ecx,4)
 113
 114         movl    -8(%eax), %esi
 115         movl    -12(%eax), %edi
 116         movl    %esi, (%edx,%ecx,4)
 117         movl    %edi, -4(%edx,%ecx,4)
 118
 119         subl    $16, %eax
 120         subl    $4, %ecx
 121 L(no4):
 122
 123         addl    $2, %ecx
 124         jle     L(no2)
 125
 126         movl    (%eax), %esi
 127         movl    -4(%eax), %edi
 128         movl    %esi, (%edx,%ecx,4)
 129         movl    %edi, -4(%edx,%ecx,4)
 130
 131         subl    $8, %eax
 132         subl    $2, %ecx
 133 L(no2):
 134
 135         jnz     L(done)
 136
 137         movl    (%eax), %ecx
 138         movl    %ecx, (%edx)    C risk of cache bank clash here
 139
 140 L(done):
 141         popl    %edi
 142         popl    %esi
 143
 144         ret
 145
 146 EPILOGUE()