Rev 829 | Go to most recent revision | Details | Compare with Previous | Last modification | View Log | RSS feed
Rev | Author | Line No. | Line |
---|---|---|---|
731 | cejka | 1 | /* |
2 | * Copyright (C) 2005 Josef Cejka |
||
3 | * All rights reserved. |
||
4 | * |
||
5 | * Redistribution and use in source and binary forms, with or without |
||
6 | * modification, are permitted provided that the following conditions |
||
7 | * are met: |
||
8 | * |
||
9 | * - Redistributions of source code must retain the above copyright |
||
10 | * notice, this list of conditions and the following disclaimer. |
||
11 | * - Redistributions in binary form must reproduce the above copyright |
||
12 | * notice, this list of conditions and the following disclaimer in the |
||
13 | * documentation and/or other materials provided with the distribution. |
||
14 | * - The name of the author may not be used to endorse or promote products |
||
15 | * derived from this software without specific prior written permission. |
||
16 | * |
||
17 | * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR |
||
18 | * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES |
||
19 | * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. |
||
20 | * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, |
||
21 | * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT |
||
22 | * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, |
||
23 | * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY |
||
24 | * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
||
25 | * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF |
||
26 | * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
||
27 | */ |
||
28 | |||
29 | #include<sftypes.h> |
||
30 | #include<mul.h> |
||
31 | #include<comparison.h> |
||
829 | cejka | 32 | #include<common.h> |
731 | cejka | 33 | |
34 | /** Multiply two 32 bit float numbers |
||
35 | * |
||
36 | */ |
||
37 | float32 mulFloat32(float32 a, float32 b) |
||
38 | { |
||
39 | float32 result; |
||
1031 | cejka | 40 | uint64_t frac1, frac2; |
41 | int32_t exp; |
||
731 | cejka | 42 | |
43 | result.parts.sign = a.parts.sign ^ b.parts.sign; |
||
44 | |||
737 | cejka | 45 | if (isFloat32NaN(a) || isFloat32NaN(b) ) { |
731 | cejka | 46 | /* TODO: fix SigNaNs */ |
47 | if (isFloat32SigNaN(a)) { |
||
804 | cejka | 48 | result.parts.fraction = a.parts.fraction; |
731 | cejka | 49 | result.parts.exp = a.parts.exp; |
50 | return result; |
||
51 | }; |
||
52 | if (isFloat32SigNaN(b)) { /* TODO: fix SigNaN */ |
||
804 | cejka | 53 | result.parts.fraction = b.parts.fraction; |
731 | cejka | 54 | result.parts.exp = b.parts.exp; |
55 | return result; |
||
56 | }; |
||
57 | /* set NaN as result */ |
||
737 | cejka | 58 | result.binary = FLOAT32_NAN; |
731 | cejka | 59 | return result; |
60 | }; |
||
61 | |||
62 | if (isFloat32Infinity(a)) { |
||
63 | if (isFloat32Zero(b)) { |
||
64 | /* FIXME: zero * infinity */ |
||
737 | cejka | 65 | result.binary = FLOAT32_NAN; |
731 | cejka | 66 | return result; |
67 | } |
||
804 | cejka | 68 | result.parts.fraction = a.parts.fraction; |
731 | cejka | 69 | result.parts.exp = a.parts.exp; |
70 | return result; |
||
71 | } |
||
72 | |||
73 | if (isFloat32Infinity(b)) { |
||
74 | if (isFloat32Zero(a)) { |
||
75 | /* FIXME: zero * infinity */ |
||
737 | cejka | 76 | result.binary = FLOAT32_NAN; |
731 | cejka | 77 | return result; |
78 | } |
||
804 | cejka | 79 | result.parts.fraction = b.parts.fraction; |
731 | cejka | 80 | result.parts.exp = b.parts.exp; |
81 | return result; |
||
82 | } |
||
83 | |||
84 | /* exp is signed so we can easy detect underflow */ |
||
85 | exp = a.parts.exp + b.parts.exp; |
||
86 | exp -= FLOAT32_BIAS; |
||
87 | |||
737 | cejka | 88 | if (exp >= FLOAT32_MAX_EXPONENT) { |
731 | cejka | 89 | /* FIXME: overflow */ |
90 | /* set infinity as result */ |
||
737 | cejka | 91 | result.binary = FLOAT32_INF; |
92 | result.parts.sign = a.parts.sign ^ b.parts.sign; |
||
731 | cejka | 93 | return result; |
94 | }; |
||
95 | |||
96 | if (exp < 0) { |
||
97 | /* FIXME: underflow */ |
||
98 | /* return signed zero */ |
||
804 | cejka | 99 | result.parts.fraction = 0x0; |
731 | cejka | 100 | result.parts.exp = 0x0; |
101 | return result; |
||
102 | }; |
||
103 | |||
804 | cejka | 104 | frac1 = a.parts.fraction; |
737 | cejka | 105 | if (a.parts.exp > 0) { |
804 | cejka | 106 | frac1 |= FLOAT32_HIDDEN_BIT_MASK; |
731 | cejka | 107 | } else { |
108 | ++exp; |
||
109 | }; |
||
110 | |||
804 | cejka | 111 | frac2 = b.parts.fraction; |
737 | cejka | 112 | |
113 | if (b.parts.exp > 0) { |
||
804 | cejka | 114 | frac2 |= FLOAT32_HIDDEN_BIT_MASK; |
731 | cejka | 115 | } else { |
116 | ++exp; |
||
117 | }; |
||
118 | |||
804 | cejka | 119 | frac1 <<= 1; /* one bit space for rounding */ |
731 | cejka | 120 | |
804 | cejka | 121 | frac1 = frac1 * frac2; |
731 | cejka | 122 | /* round and return */ |
123 | |||
804 | cejka | 124 | while ((exp < FLOAT32_MAX_EXPONENT) && (frac1 >= ( 1 << (FLOAT32_FRACTION_SIZE + 2)))) { |
125 | /* 23 bits of fraction + one more for hidden bit (all shifted 1 bit left)*/ |
||
731 | cejka | 126 | ++exp; |
804 | cejka | 127 | frac1 >>= 1; |
731 | cejka | 128 | }; |
129 | |||
130 | /* rounding */ |
||
804 | cejka | 131 | /* ++frac1; FIXME: not works - without it is ok */ |
132 | frac1 >>= 1; /* shift off rounding space */ |
||
731 | cejka | 133 | |
804 | cejka | 134 | if ((exp < FLOAT32_MAX_EXPONENT) && (frac1 >= (1 << (FLOAT32_FRACTION_SIZE + 1)))) { |
731 | cejka | 135 | ++exp; |
804 | cejka | 136 | frac1 >>= 1; |
731 | cejka | 137 | }; |
138 | |||
737 | cejka | 139 | if (exp >= FLOAT32_MAX_EXPONENT ) { |
731 | cejka | 140 | /* TODO: fix overflow */ |
141 | /* return infinity*/ |
||
737 | cejka | 142 | result.parts.exp = FLOAT32_MAX_EXPONENT; |
804 | cejka | 143 | result.parts.fraction = 0x0; |
731 | cejka | 144 | return result; |
145 | } |
||
146 | |||
804 | cejka | 147 | exp -= FLOAT32_FRACTION_SIZE; |
731 | cejka | 148 | |
804 | cejka | 149 | if (exp <= FLOAT32_FRACTION_SIZE) { |
731 | cejka | 150 | /* denormalized number */ |
804 | cejka | 151 | frac1 >>= 1; /* denormalize */ |
152 | while ((frac1 > 0) && (exp < 0)) { |
||
153 | frac1 >>= 1; |
||
731 | cejka | 154 | ++exp; |
155 | }; |
||
804 | cejka | 156 | if (frac1 == 0) { |
731 | cejka | 157 | /* FIXME : underflow */ |
158 | result.parts.exp = 0; |
||
804 | cejka | 159 | result.parts.fraction = 0; |
731 | cejka | 160 | return result; |
161 | }; |
||
162 | }; |
||
163 | result.parts.exp = exp; |
||
804 | cejka | 164 | result.parts.fraction = frac1 & ( (1 << FLOAT32_FRACTION_SIZE) - 1); |
731 | cejka | 165 | |
166 | return result; |
||
167 | |||
168 | } |
||
169 | |||
737 | cejka | 170 | /** Multiply two 64 bit float numbers |
171 | * |
||
172 | */ |
||
173 | float64 mulFloat64(float64 a, float64 b) |
||
174 | { |
||
175 | float64 result; |
||
1031 | cejka | 176 | uint64_t frac1, frac2; |
177 | int32_t exp; |
||
731 | cejka | 178 | |
737 | cejka | 179 | result.parts.sign = a.parts.sign ^ b.parts.sign; |
180 | |||
181 | if (isFloat64NaN(a) || isFloat64NaN(b) ) { |
||
182 | /* TODO: fix SigNaNs */ |
||
183 | if (isFloat64SigNaN(a)) { |
||
804 | cejka | 184 | result.parts.fraction = a.parts.fraction; |
737 | cejka | 185 | result.parts.exp = a.parts.exp; |
186 | return result; |
||
187 | }; |
||
188 | if (isFloat64SigNaN(b)) { /* TODO: fix SigNaN */ |
||
804 | cejka | 189 | result.parts.fraction = b.parts.fraction; |
737 | cejka | 190 | result.parts.exp = b.parts.exp; |
191 | return result; |
||
192 | }; |
||
193 | /* set NaN as result */ |
||
194 | result.binary = FLOAT64_NAN; |
||
195 | return result; |
||
196 | }; |
||
197 | |||
198 | if (isFloat64Infinity(a)) { |
||
199 | if (isFloat64Zero(b)) { |
||
200 | /* FIXME: zero * infinity */ |
||
201 | result.binary = FLOAT64_NAN; |
||
202 | return result; |
||
203 | } |
||
804 | cejka | 204 | result.parts.fraction = a.parts.fraction; |
737 | cejka | 205 | result.parts.exp = a.parts.exp; |
206 | return result; |
||
207 | } |
||
731 | cejka | 208 | |
737 | cejka | 209 | if (isFloat64Infinity(b)) { |
210 | if (isFloat64Zero(a)) { |
||
211 | /* FIXME: zero * infinity */ |
||
212 | result.binary = FLOAT64_NAN; |
||
213 | return result; |
||
214 | } |
||
804 | cejka | 215 | result.parts.fraction = b.parts.fraction; |
737 | cejka | 216 | result.parts.exp = b.parts.exp; |
217 | return result; |
||
218 | } |
||
219 | |||
220 | /* exp is signed so we can easy detect underflow */ |
||
829 | cejka | 221 | exp = a.parts.exp + b.parts.exp - FLOAT64_BIAS; |
737 | cejka | 222 | |
804 | cejka | 223 | frac1 = a.parts.fraction; |
829 | cejka | 224 | |
737 | cejka | 225 | if (a.parts.exp > 0) { |
804 | cejka | 226 | frac1 |= FLOAT64_HIDDEN_BIT_MASK; |
737 | cejka | 227 | } else { |
228 | ++exp; |
||
229 | }; |
||
230 | |||
804 | cejka | 231 | frac2 = b.parts.fraction; |
737 | cejka | 232 | |
233 | if (b.parts.exp > 0) { |
||
804 | cejka | 234 | frac2 |= FLOAT64_HIDDEN_BIT_MASK; |
737 | cejka | 235 | } else { |
236 | ++exp; |
||
237 | }; |
||
238 | |||
829 | cejka | 239 | frac1 <<= (64 - FLOAT64_FRACTION_SIZE - 1); |
240 | frac2 <<= (64 - FLOAT64_FRACTION_SIZE - 2); |
||
737 | cejka | 241 | |
804 | cejka | 242 | mul64integers(frac1, frac2, &frac1, &frac2); |
737 | cejka | 243 | |
829 | cejka | 244 | frac2 |= (frac1 != 0); |
245 | if (frac2 & (0x1ll << 62)) { |
||
246 | frac2 <<= 1; |
||
247 | exp--; |
||
737 | cejka | 248 | } |
249 | |||
829 | cejka | 250 | result = finishFloat64(exp, frac2, result.parts.sign); |
251 | return result; |
||
737 | cejka | 252 | } |
253 | |||
254 | /** Multiply two 64 bit numbers and return result in two parts |
||
255 | * @param a first operand |
||
256 | * @param b second operand |
||
257 | * @param lo lower part from result |
||
258 | * @param hi higher part of result |
||
259 | */ |
||
1031 | cejka | 260 | void mul64integers(uint64_t a,uint64_t b, uint64_t *lo, uint64_t *hi) |
737 | cejka | 261 | { |
1031 | cejka | 262 | uint64_t low, high, middle1, middle2; |
263 | uint32_t alow, blow; |
||
829 | cejka | 264 | |
737 | cejka | 265 | alow = a & 0xFFFFFFFF; |
266 | blow = b & 0xFFFFFFFF; |
||
267 | |||
828 | cejka | 268 | a >>= 32; |
269 | b >>= 32; |
||
737 | cejka | 270 | |
1031 | cejka | 271 | low = ((uint64_t)alow) * blow; |
737 | cejka | 272 | middle1 = a * blow; |
273 | middle2 = alow * b; |
||
274 | high = a * b; |
||
275 | |||
276 | middle1 += middle2; |
||
1031 | cejka | 277 | high += (((uint64_t)(middle1 < middle2)) << 32) + (middle1 >> 32); |
804 | cejka | 278 | middle1 <<= 32; |
737 | cejka | 279 | low += middle1; |
280 | high += (low < middle1); |
||
281 | *lo = low; |
||
282 | *hi = high; |
||
828 | cejka | 283 | |
737 | cejka | 284 | return; |
285 | } |
||
286 | |||
287 |