1
dnl Intel P6 mpn_add_n, mpn_sub_n -- mpn add or subtract.
3
dnl Copyright 2003 Free Software Foundation, Inc.
5
dnl This file is part of the GNU MP Library.
7
dnl The GNU MP Library is free software; you can redistribute it and/or
8
dnl modify it under the terms of the GNU Lesser General Public License as
9
dnl published by the Free Software Foundation; either version 2.1 of the
10
dnl License, or (at your option) any later version.
12
dnl The GNU MP Library is distributed in the hope that it will be useful,
13
dnl but WITHOUT ANY WARRANTY; without even the implied warranty of
14
dnl MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15
dnl Lesser General Public License for more details.
17
dnl You should have received a copy of the GNU Lesser General Public
18
dnl License along with the GNU MP Library; see the file COPYING.LIB. If
19
dnl not, write to the Free Software Foundation, Inc., 59 Temple Place -
20
dnl Suite 330, Boston, MA 02111-1307, USA.
22
include(`../config.m4')
28
C The K7 code runs quite well on P6, but this seems mainly due to the larger
29
C amount of unrolling than in mpn/x86/aors_n.asm.
31
C P6 apparently doesn't separately rename the carry flag, or something, so a
32
C loop holding a carry across decl or incl takes 4 cycles for the loop
33
C control. Perhaps it's more when relying on out-of-order execution to hide
36
C Not sure what the best approach would be. sbbl then addl to save and
37
C restore the carry across the loop control would be a possibility. A
38
C couple of experiments didn't get much joy, but such an approach would at
39
C least avoid serialization, presumably.
42
MULFUNC_PROLOGUE(mpn_add_n mpn_sub_n)
43
include_mpn(`x86/k7/aors_n.asm')