1/* IBM POWER submul_1 -- Multiply a limb vector with a limb and subtract
2 *			 the result from a second limb vector.
3 *
4 *      Copyright (C) 1992, 1994, 1999, 2002 Free Software Foundation, Inc.
5 *
6 * This file is part of Libgcrypt.
7 *
8 * Libgcrypt is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU Lesser General Public License as
10 * published by the Free Software Foundation; either version 2.1 of
11 * the License, or (at your option) any later version.
12 *
13 * Libgcrypt is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
16 * GNU Lesser General Public License for more details.
17 *
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with this program; if not, write to the Free Software
20 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA
21 */
22
23#include "sysdep.h"
24#include "asm-syntax.h"
25
26
27/*
28
29# INPUT PARAMETERS
30# res_ptr	r3
31# s1_ptr	r4
32# size		r5
33# s2_limb	r6
34
35# The RS/6000 has no unsigned 32x32->64 bit multiplication instruction.  To
36# obtain that operation, we have to use the 32x32->64 signed multiplication
37# instruction, and add the appropriate compensation to the high limb of the
38# result.  We add the multiplicand if the multiplier has its most significant
39# bit set, and we add the multiplier if the multiplicand has its most
40# significant bit set.	We need to preserve the carry flag between each
41# iteration, so we have to compute the compensation carefully (the natural,
42# srai+and doesn't work).  Since the POWER architecture has a branch unit
43# we can branch in zero cycles, so that's how we perform the additions.
44 */
45
46	.toc
47	.csect ._gcry_mpih_submul_1[PR]
48	.align 2
49	.globl _gcry_mpih_submul_1
50	.globl ._gcry_mpih_submul_1
51	.csect _gcry_mpih_submul_1[DS]
52_gcry_mpih_submul_1:
53	.long ._gcry_mpih_submul_1[PR], TOC[tc0], 0
54	.csect ._gcry_mpih_submul_1[PR]
55._gcry_mpih_submul_1:
56
57	cal	3,-4(3)
58	l	0,0(4)
59	cmpi	0,6,0
60	mtctr	5
61	mul	9,0,6
62	srai	7,0,31
63	and	7,7,6
64	mfmq	11
65	cax	9,9,7
66	l	7,4(3)
67	sf	8,11,7		# add res_limb
68	a	11,8,11 	# invert cy (r11 is junk)
69	blt	Lneg
70Lpos:	bdz	Lend
71
72Lploop: lu	0,4(4)
73	stu	8,4(3)
74	cmpi	0,0,0
75	mul	10,0,6
76	mfmq	0
77	ae	11,0,9		# low limb + old_cy_limb + old cy
78	l	7,4(3)
79	aze	10,10		# propagate cy to new cy_limb
80	sf	8,11,7		# add res_limb
81	a	11,8,11 	# invert cy (r11 is junk)
82	bge	Lp0
83	cax	10,10,6 	# adjust high limb for negative limb from s1
84Lp0:	bdz	Lend0
85	lu	0,4(4)
86	stu	8,4(3)
87	cmpi	0,0,0
88	mul	9,0,6
89	mfmq	0
90	ae	11,0,10
91	l	7,4(3)
92	aze	9,9
93	sf	8,11,7
94	a	11,8,11 	# invert cy (r11 is junk)
95	bge	Lp1
96	cax	9,9,6		# adjust high limb for negative limb from s1
97Lp1:	bdn	Lploop
98
99	b	Lend
100
101Lneg:	cax	9,9,0
102	bdz	Lend
103Lnloop: lu	0,4(4)
104	stu	8,4(3)
105	cmpi	0,0,0
106	mul	10,0,6
107	mfmq	7
108	ae	11,7,9
109	l	7,4(3)
110	ae	10,10,0 	# propagate cy to new cy_limb
111	sf	8,11,7		# add res_limb
112	a	11,8,11 	# invert cy (r11 is junk)
113	bge	Ln0
114	cax	10,10,6 	# adjust high limb for negative limb from s1
115Ln0:	bdz	Lend0
116	lu	0,4(4)
117	stu	8,4(3)
118	cmpi	0,0,0
119	mul	9,0,6
120	mfmq	7
121	ae	11,7,10
122	l	7,4(3)
123	ae	9,9,0		# propagate cy to new cy_limb
124	sf	8,11,7		# add res_limb
125	a	11,8,11 	# invert cy (r11 is junk)
126	bge	Ln1
127	cax	9,9,6		# adjust high limb for negative limb from s1
128Ln1:	bdn	Lnloop
129	b	Lend
130
131Lend0:	cal	9,0(10)
132Lend:	st	8,4(3)
133	aze	3,9
134	br
135
136