1dnl  PowerPC-32 mpn_divexact_by3 -- mpn by 3 exact division
2
3dnl  Copyright 2002, 2003, 2005, 2006 Free Software Foundation, Inc.
4dnl
5dnl  This file is part of the GNU MP Library.
6dnl
7dnl  The GNU MP Library is free software; you can redistribute it and/or
8dnl  modify it under the terms of the GNU Lesser General Public License as
9dnl  published by the Free Software Foundation; either version 3 of the
10dnl  License, or (at your option) any later version.
11dnl
12dnl  The GNU MP Library is distributed in the hope that it will be useful,
13dnl  but WITHOUT ANY WARRANTY; without even the implied warranty of
14dnl  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
15dnl  Lesser General Public License for more details.
16dnl
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 603e:              ?
24C 604e:              5
25C 75x (G3):          ?
26C 7400,7410 (G4):    8
27C 744x,745x (G4+):   6
28C power4/ppc970:    12
29C power5:            ?
30
31C void mpn_divexact_by3 (mp_ptr dst, mp_srcptr src, mp_size_t size);
32C
33C We avoid the slow subfe instruction and instead rely on an extremely unlikely
34C branch.
35C
36C The mullw has the inverse in the first operand, since 0xAA..AB won't allow
37C any early-out.  The src[] data normally won't either, but there's at least
38C a chance, whereas 0xAA..AB never will.  If, for instance, src[] is all
39C zeros (not a sensible input of course) we run at 7.0 c/l on ppc750.
40C
41C The mulhwu has the "3" multiplier in the second operand, which lets 750 and
42C 7400 use an early-out.
43
44C INPUT PARAMETERS
45define(`rp', `r3')
46define(`up', `r4')
47define(`n',  `r5')
48define(`cy', `r6')
49
50ASM_START()
51PROLOGUE(mpn_divexact_by3c)
52	lwz	r11, 0(up)
53	mtctr	n
54	lis	r12, 0xAAAA
55	ori	r12, r12, 0xAAAB
56	li	r10, 3
57
58	cmplw	cr7, cy, r11
59	subf	r11, cy, r11
60
61	mullw	r0, r11, r12
62	stw	r0, 0(rp)
63	bdz	L(one)
64
65L(top):	lwzu	r9, 4(up)
66	mulhwu	r7, r0, r10
67	bgt-	cr7, L(adj)		C very unlikely branch
68L(bko):	cmplw	cr7, r7, r9
69	subf	r0, r7, r9
70	mullw	r0, r12, r0
71	stwu	r0, 4(rp)
72	bdnz	L(top)
73
74L(one):	mulhwu	r3, r0, r10
75	blelr+	cr7
76	addi	r3, r3, 1
77	blr
78
79L(adj):	addi	r7, r7, 1
80	b	L(bko)
81EPILOGUE()
82ASM_END()
83