Return to com_n.asm CVS log

Up to [local] / OpenXM_contrib / gmp / mpn / x86 / k6 / mmx

Annotation of OpenXM_contrib/gmp/mpn/x86/k6/mmx/com_n.asm, Revision 1.1.1.1

1.1       maekawa     1: dnl  AMD K6-2 mpn_com_n -- mpn bitwise one's complement.
                      2: dnl
                      3: dnl     alignment dst/src, A=0mod8 N=4mod8
                      4: dnl        A/A   A/N   N/A   N/N
                      5: dnl  K6-2  1.0   1.18  1.18  1.18  cycles/limb
                      6: dnl  K6    1.5   1.85  1.75  1.85
                      7:
                      8:
                      9: dnl  Copyright (C) 1999, 2000 Free Software Foundation, Inc.
                     10: dnl
                     11: dnl  This file is part of the GNU MP Library.
                     12: dnl
                     13: dnl  The GNU MP Library is free software; you can redistribute it and/or
                     14: dnl  modify it under the terms of the GNU Lesser General Public License as
                     15: dnl  published by the Free Software Foundation; either version 2.1 of the
                     16: dnl  License, or (at your option) any later version.
                     17: dnl
                     18: dnl  The GNU MP Library is distributed in the hope that it will be useful,
                     19: dnl  but WITHOUT ANY WARRANTY; without even the implied warranty of
                     20: dnl  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
                     21: dnl  Lesser General Public License for more details.
                     22: dnl
                     23: dnl  You should have received a copy of the GNU Lesser General Public
                     24: dnl  License along with the GNU MP Library; see the file COPYING.LIB.  If
                     25: dnl  not, write to the Free Software Foundation, Inc., 59 Temple Place -
                     26: dnl  Suite 330, Boston, MA 02111-1307, USA.
                     27:
                     28:
                     29: include(`../config.m4')
                     30:
                     31:
                     32: C void mpn_com_n (mp_ptr dst, mp_srcptr src, mp_size_t size);
                     33: C
                     34: C Take the bitwise ones-complement of src,size and write it to dst,size.
                     35:
                     36: defframe(PARAM_SIZE,12)
                     37: defframe(PARAM_SRC, 8)
                     38: defframe(PARAM_DST, 4)
                     39:
                     40:        .text
                     41:        ALIGN(32)
                     42: PROLOGUE(mpn_com_n)
                     43: deflit(`FRAME',0)
                     44:
                     45:        movl    PARAM_SIZE, %ecx
                     46:        movl    PARAM_SRC, %eax
                     47:        movl    PARAM_DST, %edx
                     48:        shrl    %ecx
                     49:        jnz     L(two_or_more)
                     50:
                     51:        movl    (%eax), %eax
                     52:        notl    %eax
                     53:        movl    %eax, (%edx)
                     54:        ret
                     55:
                     56:
                     57: L(two_or_more):
                     58:        pushl   %ebx
                     59: FRAME_pushl()
                     60:        movl    %ecx, %ebx
                     61:
                     62:        pcmpeqd %mm7, %mm7      C all ones
                     63:
                     64:
                     65:        ALIGN(16)
                     66: L(top):
                     67:        C eax   src
                     68:        C ebx   floor(size/2)
                     69:        C ecx   counter
                     70:        C edx   dst
                     71:        C esi
                     72:        C edi
                     73:        C ebp
                     74:
                     75:        movq    -8(%eax,%ecx,8), %mm0
                     76:        pxor    %mm7, %mm0
                     77:        movq    %mm0, -8(%edx,%ecx,8)
                     78:        loop    L(top)
                     79:
                     80:
                     81:        jnc     L(no_extra)
                     82:        movl    (%eax,%ebx,8), %eax
                     83:        notl    %eax
                     84:        movl    %eax, (%edx,%ebx,8)
                     85: L(no_extra):
                     86:
                     87:        popl    %ebx
                     88:        emms_or_femms
                     89:        ret
                     90:
                     91: EPILOGUE()