1 dnl HP-PA 1.1 mpn_submul_1 -- Multiply a limb vector with a limb and subtract
2 dnl the result from a second limb vector.
4 dnl Copyright 1992, 1993, 1994, 2000, 2001, 2002 Free Software Foundation,
7 dnl This file is part of the GNU MP Library.
9 dnl The GNU MP Library is free software; you can redistribute it and/or modify
10 dnl it under the terms of the GNU Lesser General Public License as published
11 dnl by the Free Software Foundation; either version 3 of the License, or (at
12 dnl your option) any later version.
14 dnl The GNU MP Library is distributed in the hope that it will be useful, but
15 dnl WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
16 dnl or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public
17 dnl License for more details.
19 dnl You should have received a copy of the GNU Lesser General Public License
20 dnl along with the GNU MP Library. If not, see http://www.gnu.org/licenses/.
22 include(`../config.m4')
30 C This runs at 12 cycles/limb on a PA7000. With the used instructions, it can
31 C not become faster due to data cache contention after a store. On the PA7100
32 C it runs at 11 cycles/limb.
34 C There are some ideas described in mul_1.asm that applies to this code too.
36 C It seems possible to make this run as fast as mpn_addmul_1, if we use
37 C sub,>>= %r29,%r19,%r22
39 C but that requires reworking the hairy software pipeline...
42 PROLOGUE(mpn_submul_1)
43 C .callinfo frame=64,no_calls
47 stw %r23,-16(%r30) C move s2_limb ...
48 addib,= -1,%r24,L(just_one_limb)
49 fldws -16(%r30),%fr4 C ... into fr4
50 add %r0,%r0,%r0 C clear carry
55 ldw -12(%r30),%r19 C least significant limb in product
59 addib,= -1,%r24,L(end)
74 addib,<> -1,%r24,L(loop)