xref: /netbsd-src/external/lgpl3/gmp/dist/mpn/arm/dive_1.asm (revision 72c7faa4dbb41dbb0238d6b4a109da0d4b236dd4)
1dnl  ARM v4 mpn_divexact_1.
2
3dnl  Contributed to the GNU project by Torbjorn Granlund.
4
5dnl  Copyright 2012 Free Software Foundation, Inc.
6
7dnl  This file is part of the GNU MP Library.
8dnl
9dnl  The GNU MP Library is free software; you can redistribute it and/or modify
10dnl  it under the terms of either:
11dnl
12dnl    * the GNU Lesser General Public License as published by the Free
13dnl      Software Foundation; either version 3 of the License, or (at your
14dnl      option) any later version.
15dnl
16dnl  or
17dnl
18dnl    * the GNU General Public License as published by the Free Software
19dnl      Foundation; either version 2 of the License, or (at your option) any
20dnl      later version.
21dnl
22dnl  or both in parallel, as here.
23dnl
24dnl  The GNU MP Library is distributed in the hope that it will be useful, but
25dnl  WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
26dnl  or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
27dnl  for more details.
28dnl
29dnl  You should have received copies of the GNU General Public License and the
30dnl  GNU Lesser General Public License along with the GNU MP Library.  If not,
31dnl  see https://www.gnu.org/licenses/.
32
33include(`../config.m4')
34
35C               cycles/limb       cycles/limb
36C               norm    unorm    modexact_1c_odd
37C StrongARM	 ?
38C XScale	 ?
39C Cortex-A7	 ?
40C Cortex-A8	 ?
41C Cortex-A9	10	12
42C Cortex-A15	 9	 9
43
44C Architecture requirements:
45C v5	-
46C v5t	-
47C v5te	-
48C v6	-
49C v6t2	-
50C v7a	-
51
52define(`rp', `r0')
53define(`up', `r1')
54define(`n',  `r2')
55define(`d',  `r3')
56
57define(`cy', `r7')
58define(`cnt', `r6')
59define(`tnc', `r8')
60
61ASM_START()
62PROLOGUE(mpn_divexact_1)
63	tst	d, #1
64	push	{r4-r9}
65	mov	cnt, #0
66	bne	L(inv)
67
68C count trailing zeros
69	movs	r4, d, lsl #16
70	moveq	d, d, lsr #16
71	moveq	cnt, #16
72	tst	d, #0xff
73	moveq	d, d, lsr #8
74	addeq	cnt, cnt, #8
75	LEA(	r4, ctz_tab)
76	and	r5, d, #0xff
77	ldrb	r4, [r4, r5]
78	mov	d, d, lsr r4
79	add	cnt, cnt, r4
80
81C binvert limb
82L(inv):	LEA(	r4, binvert_limb_table)
83	and	r12, d, #254
84	ldrb	r4, [r4, r12, lsr #1]
85	mul	r12, r4, r4
86	mul	r12, d, r12
87	rsb	r12, r12, r4, lsl #1
88	mul	r4, r12, r12
89	mul	r4, d, r4
90	rsb	r4, r4, r12, lsl #1	C r4 = inverse
91
92	tst	cnt, cnt
93	ldr	r5, [up], #4		C up[0]
94	mov	cy, #0
95	bne	L(unnorm)
96
97L(norm):
98	subs	n, n, #1		C set carry as side-effect
99	beq	L(edn)
100
101	ALIGN(16)
102L(tpn):	sbcs	cy, r5, cy
103	ldr	r5, [up], #4
104	sub	n, n, #1
105	mul	r9, r4, cy
106	tst	n, n
107	umull	r12, cy, d, r9
108	str	r9, [rp], #4
109	bne	L(tpn)
110
111L(edn):	sbc	cy, r5, cy
112	mul	r9, r4, cy
113	str	r9, [rp]
114	pop	{r4-r9}
115	return	r14
116
117L(unnorm):
118	rsb	tnc, cnt, #32
119	mov	r5, r5, lsr cnt
120	subs	n, n, #1		C set carry as side-effect
121	beq	L(edu)
122
123	ALIGN(16)
124L(tpu):	ldr	r12, [up], #4
125	orr	r9, r5, r12, lsl tnc
126	mov	r5, r12, lsr cnt
127	sbcs	cy, r9, cy		C critical path ->cy->cy->
128	sub	n, n, #1
129	mul	r9, r4, cy		C critical path ->cy->r9->
130	tst	n, n
131	umull	r12, cy, d, r9		C critical path ->r9->cy->
132	str	r9, [rp], #4
133	bne	L(tpu)
134
135L(edu):	sbc	cy, r5, cy
136	mul	r9, r4, cy
137	str	r9, [rp]
138	pop	{r4-r9}
139	return	r14
140EPILOGUE()
141
142	RODATA
143ctz_tab:
144	.byte	8,0,1,0,2,0,1,0,3,0,1,0,2,0,1,0,4,0,1,0,2,0,1,0,3,0,1,0,2,0,1,0
145	.byte	5,0,1,0,2,0,1,0,3,0,1,0,2,0,1,0,4,0,1,0,2,0,1,0,3,0,1,0,2,0,1,0
146	.byte	6,0,1,0,2,0,1,0,3,0,1,0,2,0,1,0,4,0,1,0,2,0,1,0,3,0,1,0,2,0,1,0
147	.byte	5,0,1,0,2,0,1,0,3,0,1,0,2,0,1,0,4,0,1,0,2,0,1,0,3,0,1,0,2,0,1,0
148	.byte	7,0,1,0,2,0,1,0,3,0,1,0,2,0,1,0,4,0,1,0,2,0,1,0,3,0,1,0,2,0,1,0
149	.byte	5,0,1,0,2,0,1,0,3,0,1,0,2,0,1,0,4,0,1,0,2,0,1,0,3,0,1,0,2,0,1,0
150	.byte	6,0,1,0,2,0,1,0,3,0,1,0,2,0,1,0,4,0,1,0,2,0,1,0,3,0,1,0,2,0,1,0
151	.byte	5,0,1,0,2,0,1,0,3,0,1,0,2,0,1,0,4,0,1,0,2,0,1,0,3,0,1,0,2,0,1,0
152