xref: /netbsd-src/external/lgpl3/gmp/dist/mpn/arm/mod_34lsub1.asm (revision 212397c69a103ae7e5eafa8731ddfae671d2dee7)
1dnl  ARM mpn_mod_34lsub1 -- remainder modulo 2^24-1.
2
3dnl  Copyright 2012 Free Software Foundation, Inc.
4
5dnl  This file is part of the GNU MP Library.
6
7dnl  The GNU MP Library is free software; you can redistribute it and/or modify
8dnl  it under the terms of the GNU Lesser General Public License as published
9dnl  by the Free Software Foundation; either version 3 of the License, or (at
10dnl  your option) any later version.
11
12dnl  The GNU MP Library is distributed in the hope that it will be useful, but
13dnl  WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
14dnl  or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU Lesser General Public
15dnl  License for more details.
16
17dnl  You should have received a copy of the GNU Lesser General Public License
18dnl  along with the GNU MP Library.  If not, see http://www.gnu.org/licenses/.
19
20include(`../config.m4')
21
22C	     cycles/limb
23C StrongARM	 ?
24C XScale	 ?
25C Cortex-A8	 ?
26C Cortex-A9	 1.33
27C Cortex-A15	 ?
28
29define(`ap',	r0)
30define(`n',	r1)
31
32C mp_limb_t mpn_mod_34lsub1 (mp_srcptr up, mp_size_t n)
33
34C TODO
35C  * Write cleverer summation code.
36C  * Consider loading 6 64-bit aligned registers at a time, to approach 1 c/l.
37
38ASM_START()
39	TEXT
40	ALIGN(32)
41PROLOGUE(mpn_mod_34lsub1)
42	push	{ r4, r5, r6, r7 }
43
44	subs	n, n, #3
45	mov	r7, #0
46	blt	L(le2)			C n <= 2
47
48	ldmia	ap!, { r2, r3, r12 }
49	subs	n, n, #3
50	blt	L(sum)			C n <= 5
51	adds	r0, r0, #0		C clear carry
52	sub	n, n, #3
53	b	L(mid)
54
55L(top):	adcs	r2, r2, r4
56	adcs	r3, r3, r5
57	adcs	r12, r12, r6
58L(mid):	ldmia	ap!, { r4, r5, r6 }
59	tst	n, n
60	sub	n, n, #3
61	bpl	L(top)
62
63	add	n, n, #3
64
65	adcs	r2, r2, r4
66	adcs	r3, r3, r5
67	adcs	r12, r12, r6
68	movcs	r7, #1			C r7 <= 1
69
70L(sum):	cmn	n, #2
71	movlo	r4, #0
72	ldrhs	r4, [ap], #4
73	movls	r5, #0
74	ldrhi	r5, [ap], #4
75
76	adds	r2, r2, r4
77	adcs	r3, r3, r5
78	adcs	r12, r12, #0
79	adc	r7, r7, #0		C r7 <= 2
80
81L(sum2):
82	bic	r0, r2, #0xff000000
83	add	r0, r0, r2, lsr #24
84	add	r0, r0, r7
85
86	lsl	r7, r3, #8
87	bic	r1, r7, #0xff000000
88	add	r0, r0, r1
89	add	r0, r0, r3, lsr #16
90
91	lsl	r7, r12, #16
92	bic	r1, r7, #0xff000000
93	add	r0, r0, r1
94	add	r0, r0, r12, lsr #8
95
96	pop	{ r4, r5, r6, r7 }
97ifdef(`ARM_THUMB_MODE',
98`	bx	lr
99',`	mov	pc, lr
100')
101
102L(le2):	cmn	n, #1
103	bne	L(1)
104	ldmia	ap!, { r2, r3 }
105	mov	r12, #0
106	b	L(sum2)
107L(1):	ldr	r2, [ap]
108	bic	r0, r2, #0xff000000
109	add	r0, r0, r2, lsr #24
110	pop	{ r4, r5, r6, r7 }
111ifdef(`ARM_THUMB_MODE',
112`	bx	lr
113',`	mov	pc, lr
114')
115EPILOGUE()
116