File: /usr/src/linux/arch/mips/math-emu/dp_mul.c
1 /* IEEE754 floating point arithmetic
2 * double precision: common utilities
3 */
4 /*
5 * MIPS floating point support
6 * Copyright (C) 1994-2000 Algorithmics Ltd. All rights reserved.
7 * http://www.algor.co.uk
8 *
9 * ########################################################################
10 *
11 * This program is free software; you can distribute it and/or modify it
12 * under the terms of the GNU General Public License (Version 2) as
13 * published by the Free Software Foundation.
14 *
15 * This program is distributed in the hope it will be useful, but WITHOUT
16 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
17 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
18 * for more details.
19 *
20 * You should have received a copy of the GNU General Public License along
21 * with this program; if not, write to the Free Software Foundation, Inc.,
22 * 59 Temple Place - Suite 330, Boston MA 02111-1307, USA.
23 *
24 * ########################################################################
25 */
26
27
28 #include "ieee754dp.h"
29
30 ieee754dp ieee754dp_mul(ieee754dp x, ieee754dp y)
31 {
32 COMPXDP;
33 COMPYDP;
34
35 CLEARCX;
36
37 EXPLODEXDP;
38 EXPLODEYDP;
39
40 switch (CLPAIR(xc, yc)) {
41 case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_SNAN):
42 return ieee754dp_nanxcpt(ieee754dp_bestnan(x, y), "mul", x,
43 y);
44
45 case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_SNAN):
46 case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_SNAN):
47 case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_SNAN):
48 case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_SNAN):
49 case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_SNAN):
50 return ieee754dp_nanxcpt(y, "mul", x, y);
51
52 case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_QNAN):
53 case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_ZERO):
54 case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_NORM):
55 case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_DNORM):
56 case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_INF):
57 return ieee754dp_nanxcpt(x, "mul", x, y);
58
59 case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_QNAN):
60 return ieee754dp_bestnan(x, y);
61
62 case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_QNAN):
63 case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_QNAN):
64 case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_QNAN):
65 case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_QNAN):
66 return y;
67
68 case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_ZERO):
69 case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_NORM):
70 case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_DNORM):
71 case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_INF):
72 return x;
73
74
75 /* Infinity handeling */
76
77 case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_ZERO):
78 case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_INF):
79 SETCX(IEEE754_INVALID_OPERATION);
80 return ieee754dp_xcpt(ieee754dp_indef(), "mul", x, y);
81
82 case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_INF):
83 case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_INF):
84 case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_NORM):
85 case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_DNORM):
86 case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_INF):
87 return ieee754dp_inf(xs ^ ys);
88
89 case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_ZERO):
90 case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_NORM):
91 case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_DNORM):
92 case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_ZERO):
93 case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_ZERO):
94 return ieee754dp_zero(xs ^ ys);
95
96
97 case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_DNORM):
98 DPDNORMX;
99
100 case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_DNORM):
101 DPDNORMY;
102 break;
103
104 case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_NORM):
105 DPDNORMX;
106 break;
107
108 case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_NORM):
109 break;
110 }
111 /* rm = xm * ym, re = xe+ye basicly */
112 assert(xm & DP_HIDDEN_BIT);
113 assert(ym & DP_HIDDEN_BIT);
114 {
115 int re = xe + ye;
116 int rs = xs ^ ys;
117 unsigned long long rm;
118
119 /* shunt to top of word */
120 xm <<= 64 - (DP_MBITS + 1);
121 ym <<= 64 - (DP_MBITS + 1);
122
123 /* multiply 32bits xm,ym to give high 32bits rm with stickness
124 */
125
126 /* 32 * 32 => 64 */
127 #define DPXMULT(x,y) ((unsigned long long)(x) * (unsigned long long)y)
128
129 {
130 unsigned lxm = xm;
131 unsigned hxm = xm >> 32;
132 unsigned lym = ym;
133 unsigned hym = ym >> 32;
134 unsigned long long lrm;
135 unsigned long long hrm;
136
137 lrm = DPXMULT(lxm, lym);
138 hrm = DPXMULT(hxm, hym);
139
140 {
141 unsigned long long t = DPXMULT(lxm, hym);
142 {
143 unsigned long long at =
144 lrm + (t << 32);
145 hrm += at < lrm;
146 lrm = at;
147 }
148 hrm = hrm + (t >> 32);
149 }
150
151 {
152 unsigned long long t = DPXMULT(hxm, lym);
153 {
154 unsigned long long at =
155 lrm + (t << 32);
156 hrm += at < lrm;
157 lrm = at;
158 }
159 hrm = hrm + (t >> 32);
160 }
161 rm = hrm | (lrm != 0);
162 }
163
164 /*
165 * sticky shift down to normal rounding precision
166 */
167 if ((signed long long) rm < 0) {
168 rm =
169 (rm >> (64 - (DP_MBITS + 1 + 3))) |
170 ((rm << (DP_MBITS + 1 + 3)) != 0);
171 re++;
172 } else {
173 rm =
174 (rm >> (64 - (DP_MBITS + 1 + 3 + 1))) |
175 ((rm << (DP_MBITS + 1 + 3 + 1)) != 0);
176 }
177 assert(rm & (DP_HIDDEN_BIT << 3));
178 DPNORMRET2(rs, re, rm, "mul", x, y);
179 }
180 }
181