1 ; HP-PA-
1.1 __mpn_submul_1
-- Multiply
a limb vector with
a limb
and
2 ; subtract the result from
a second limb vector.
4 ; Copyright
(C
) 1992, 1993, 1994 Free Software Foundation
, Inc.
6 ; This file is part of the GNU MP Library.
8 ; The GNU MP Library is free software; you can redistribute it
and/or modify
9 ; it under the terms of the GNU Library General Public License as published by
10 ; the Free Software Foundation; either version
2 of the License
, or (at your
11 ; option
) any later version.
13 ; The GNU MP Library is distributed in the hope that it will
be useful
, but
14 ; WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
15 ;
or FITNESS FOR
A PARTICULAR PURPOSE. See the GNU Library General Public
16 ; License for more details.
18 ; You should have received
a copy of the GNU Library General Public License
19 ; along with the GNU MP Library; see the file COPYING.LIB. If
not, write to
20 ; the Free Software Foundation
, Inc.
, 59 Temple Place
- Suite
330, Boston
,
30 ; This runs at
12 cycles
/limb on
a PA7000. With the used instructions
, it
31 ; can
not become faster due to data cache contention after
a store. On the
32 ; PA7100 it runs at
11 cycles
/limb
, and that can
not be improved either
,
33 ; since only the xmpyu does
not need the integer pipeline
, so the only
34 ; dual-issue we will get are addc+xmpyu. Unrolling could gain
a cycle
/limb
37 ; There are some ideas described in mul_1.s that applies to this code too.
39 ; It seems possible to make this run as fast as __mpn_addmul_1
, if we use
40 ;
sub,>>= %r29,%r19,%r22
42 ; but that requires reworking the hairy software pipeline.
..
45 .export __mpn_submul_1
48 .callinfo frame=64,no_calls
53 stw %r23,-16(%r30) ; move s2_limb
...
54 addib
,= -1,%r24,L$just_one_limb
55 fldws
-16(%r30),%fr4 ;
... into fr4
56 add %r0,%r0,%r0 ; clear carry
61 ldw
-12(%r30),%r19 ; least significant limb in product
69 L$loop ldws
0(%r26),%r29
79 addib
,<> -1,%r24,L$loop
82 L$end ldw
0(%r26),%r29