xref: /netbsd-src/external/lgpl3/gmp/dist/mpn/arm/mul_1.asm (revision ce54336801cf28877c3414aa2fcb251dddd543a2)
1dnl  ARM mpn_mul_1 -- Multiply a limb vector with a limb and store the result
2dnl  in a second limb vector.
3dnl  Contributed by Robert Harley.
4
5dnl  Copyright 1998, 2000, 2001, 2003, 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
36C StrongARM	6-8
37C XScale	 ?
38C Cortex-A7	 ?
39C Cortex-A8	 ?
40C Cortex-A9	 4.75
41C Cortex-A15	 ?
42
43C We should rewrite this along the lines of addmul_1.asm.  That should save a
44C cycle on StrongARM, and several cycles on XScale.
45
46define(`rp',`r0')
47define(`up',`r1')
48define(`n',`r2')
49define(`vl',`r3')
50
51
52ASM_START()
53PROLOGUE(mpn_mul_1)
54	stmfd	sp!, { r8, r9, lr }
55	ands	r12, n, #1
56	beq	L(skip1)
57	ldr	lr, [up], #4
58	umull	r9, r12, lr, vl
59	str	r9, [rp], #4
60L(skip1):
61	tst	n, #2
62	beq	L(skip2)
63	mov	r8, r12
64	ldmia	up!, { r12, lr }
65	mov	r9, #0
66	umlal	r8, r9, r12, vl
67	mov	r12, #0
68	umlal	r9, r12, lr, vl
69	stmia	rp!, { r8, r9 }
70L(skip2):
71	bics	n, n, #3
72	beq	L(rtn)
73	stmfd	sp!, { r6, r7 }
74
75L(top):	mov	r6, r12
76	ldmia	up!, { r8, r9, r12, lr }
77	ldr	r7, [rp, #12]			C cache allocate
78	mov	r7, #0
79	umlal	r6, r7, r8, vl
80	mov	r8, #0
81	umlal	r7, r8, r9, vl
82	mov	r9, #0
83	umlal	r8, r9, r12, vl
84	mov	r12, #0
85	umlal	r9, r12, lr, vl
86	subs	n, n, #4
87	stmia	rp!, { r6, r7, r8, r9 }
88	bne	L(top)
89
90	ldmfd	sp!, { r6, r7 }
91
92L(rtn):	mov	r0, r12
93	ldmfd	sp!, { r8, r9, pc }
94EPILOGUE()
95