1 dnl PowerPC
-32 mpn_mod_34lsub1
-- mpn remainder
mod 2^
24-1.
3 dnl Copyright
2002, 2003, 2005 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 modify
8 dnl it under the terms of
either:
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.
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
20 dnl
or both
in parallel
, as here.
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
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
/.
31 include(`..
/config.m4
')
43 C mp_limb_t mpn_mod_34lsub1 (mp_srcptr src, mp_size_t size)
45 C There seems no need to schedule the loads back, the code is still 3.0 c/l
46 C on 750/7400 no matter where they're placed.
50 C Fetching half words would allow
add instead for accumulating
, instead of
51 C adde
and its serialization. An outer
loop would be required though
, since
52 C
2^
16 halfwords can overflow. lhz
+add would be
2.0 c
/l
, but if there
's
53 C also a bdz or bdnz for each and a pointer update say every three limbs
54 C then the total would be 2.67 c/l which isn't much faster than the current
58 PROLOGUE
(mpn_mod_34lsub1
)
64 addic r6
, r3
, 8 C
&src
[2], and clear CA
66 lwz r3
, 0(r3
) C acc0
= src
[0]
69 lwz r4
, -4(r6
) C acc1
= src
[1]
72 lwz r5
, 0(r6
) C acc2
= src
[2]
73 lis r7
, 0 C no carry if just three limbs
76 lis r7
, 1 C
0x10000 carry pos
82 C r6 src
, incrementing
102 subfe r0
, r0
, r0 C
-1 if
not CA
104 andc r7
, r7
, r0 C final carry
, 0x10000, 0x100, 1 or 0
106 rlwinm r6
, r3
, 0,8,31 C acc0
low
109 rlwinm r6
, r3
, 8,24,31 C acc0
high
112 rlwinm r6
, r4
, 8,8,23 C acc1
low
115 rlwinm r6
, r4
, 16,16,31 C acc1
high
118 rlwinm r6
, r5
, 16,8,15 C acc2
low
121 rlwinm r6
, r5
, 24,8,31 C acc2
high
132 rlwinm r5
, r3
, 8,24,31 C acc0
high
133 rlwinm r3
, r3
, 0,8,31 C acc0
low
135 add r3
, r3
, r5 C acc0
high + low
136 rlwinm r5
, r4
, 16,16,31 C acc1
high
138 add r3
, r3
, r5 C
add acc1
high
139 rlwinm r5
, r4
, 8,8,23 C acc1
low
141 add r3
, r3
, r5 C
add acc1
low