1 dnl ARM v6t2 mpn_gcd_1.
3 dnl Based on the K7 gcd_1.asm
, by Kevin Ryde. Rehacked for ARM by Torbjorn
6 dnl Copyright
2000-2002, 2005, 2009, 2011-2013 Free Software Foundation
, Inc.
8 dnl
This file is part of the GNU MP Library.
10 dnl The GNU MP Library is free software
; you can redistribute it and/or modify
11 dnl it under the terms of
either:
13 dnl
* the GNU Lesser General
Public License as published by the Free
14 dnl Software Foundation
; either version 3 of the License, or (at your
15 dnl option
) any later version.
19 dnl
* the GNU General
Public License as published by the Free Software
20 dnl Foundation
; either version 2 of the License, or (at your option) any
23 dnl
or both
in parallel
, as here.
25 dnl The GNU MP Library is distributed
in the hope that it will be useful
, but
26 dnl WITHOUT ANY WARRANTY
; without even the implied warranty of MERCHANTABILITY
27 dnl
or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General
Public License
30 dnl You should have received copies of the GNU General
Public License
and the
31 dnl GNU Lesser General
Public License along with the GNU MP Library. If
not,
32 dnl see
https://www.gnu.
org/licenses
/.
34 include(`..
/config.m4
')
43 C * Optimise inner-loop better.
44 C * Push saving/restoring of callee-user regs into call code
46 C Threshold of when to call bmod when U is one limb. Should be about
47 C (time_in_cycles(bmod_1,1) + call_overhead) / (cycles/bit).
48 define(`BMOD_THRES_LOG2', 7)
55 ifdef
(`BMOD_1_TO_MOD_1_THRESHOLD
',,
56 `define(`BMOD_1_TO_MOD_1_THRESHOLD',30)')
62 stp x29, x30, [sp,#-32]!
63 ldr x3, [up] C U low limb
64 stp x19, x20, [sp,#16]
68 clz x20, x4 C min(ctz(u0),ctz(v0))
79 C Both U and V are single limbs, reduce with bmod if u0 >> v0.
81 cmp v0, x3, lsr #BMOD_THRES_LOG2
84 L(bmod):mov x3, #0 C carry argument
85 bl mpn_modexact_1c_odd
88 L(nby1):cmp n, #BMOD_1_TO_MOD_1_THRESHOLD
103 C This shorter variant makes full use of armv8 insns
104 csneg x3, x1, x1, cs C if x-y < 0
105 csel x19, x4, x19, cs C use x,y-x
106 L(mid): lsr x4, x3, x12 C
109 C
This variant is akin to the
32-bit v6t2 code
110 csel x3
, x1
, x3
, cs C if x
-y
< 0
111 csel x19
, x0
, x19
, cs C use x
,y
-x
112 L
(mid
): lsr x3
, x3
, x12 C
121 L(end): lsl x0, x19, x20
122 ldp x19, x20, [sp,#16]
123 ldp x29, x30, [sp],#32