Rev 1657 | 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 | |||
1740 | jermar | 29 | /** @addtogroup softfloat |
1657 | cejka | 30 | * @{ |
31 | */ |
||
32 | /** @file |
||
33 | */ |
||
34 | |||
731 | cejka | 35 | #include<sftypes.h> |
36 | #include<add.h> |
||
37 | #include<comparison.h> |
||
38 | |||
39 | /** Add two Float32 numbers with same signs |
||
40 | */ |
||
41 | float32 addFloat32(float32 a, float32 b) |
||
42 | { |
||
43 | int expdiff; |
||
1031 | cejka | 44 | uint32_t exp1, exp2,frac1, frac2; |
731 | cejka | 45 | |
732 | cejka | 46 | expdiff = a.parts.exp - b.parts.exp; |
47 | if (expdiff < 0) { |
||
731 | cejka | 48 | if (isFloat32NaN(b)) { |
804 | cejka | 49 | /* TODO: fix SigNaN */ |
731 | cejka | 50 | if (isFloat32SigNaN(b)) { |
51 | }; |
||
52 | |||
53 | return b; |
||
54 | }; |
||
55 | |||
732 | cejka | 56 | if (b.parts.exp == FLOAT32_MAX_EXPONENT) { |
731 | cejka | 57 | return b; |
58 | } |
||
59 | |||
804 | cejka | 60 | frac1 = b.parts.fraction; |
732 | cejka | 61 | exp1 = b.parts.exp; |
804 | cejka | 62 | frac2 = a.parts.fraction; |
732 | cejka | 63 | exp2 = a.parts.exp; |
64 | expdiff *= -1; |
||
731 | cejka | 65 | } else { |
804 | cejka | 66 | if ((isFloat32NaN(a)) || (isFloat32NaN(b))) { |
67 | /* TODO: fix SigNaN */ |
||
732 | cejka | 68 | if (isFloat32SigNaN(a) || isFloat32SigNaN(b)) { |
731 | cejka | 69 | }; |
804 | cejka | 70 | return (isFloat32NaN(a)?a:b); |
731 | cejka | 71 | }; |
72 | |||
732 | cejka | 73 | if (a.parts.exp == FLOAT32_MAX_EXPONENT) { |
731 | cejka | 74 | return a; |
75 | } |
||
76 | |||
804 | cejka | 77 | frac1 = a.parts.fraction; |
732 | cejka | 78 | exp1 = a.parts.exp; |
804 | cejka | 79 | frac2 = b.parts.fraction; |
732 | cejka | 80 | exp2 = b.parts.exp; |
731 | cejka | 81 | }; |
82 | |||
83 | if (exp1 == 0) { |
||
84 | /* both are denormalized */ |
||
804 | cejka | 85 | frac1 += frac2; |
86 | if (frac1 & FLOAT32_HIDDEN_BIT_MASK ) { |
||
731 | cejka | 87 | /* result is not denormalized */ |
88 | a.parts.exp = 1; |
||
89 | }; |
||
804 | cejka | 90 | a.parts.fraction = frac1; |
731 | cejka | 91 | return a; |
92 | }; |
||
93 | |||
804 | cejka | 94 | frac1 |= FLOAT32_HIDDEN_BIT_MASK; /* add hidden bit */ |
731 | cejka | 95 | |
96 | if (exp2 == 0) { |
||
97 | /* second operand is denormalized */ |
||
828 | cejka | 98 | --expdiff; |
731 | cejka | 99 | } else { |
100 | /* add hidden bit to second operand */ |
||
804 | cejka | 101 | frac2 |= FLOAT32_HIDDEN_BIT_MASK; |
731 | cejka | 102 | }; |
103 | |||
104 | /* create some space for rounding */ |
||
804 | cejka | 105 | frac1 <<= 6; |
106 | frac2 <<= 6; |
||
731 | cejka | 107 | |
804 | cejka | 108 | if (expdiff < (FLOAT32_FRACTION_SIZE + 2) ) { |
109 | frac2 >>= expdiff; |
||
110 | frac1 += frac2; |
||
834 | cejka | 111 | } else { |
112 | a.parts.exp = exp1; |
||
113 | a.parts.fraction = (frac1 >> 6) & (~(FLOAT32_HIDDEN_BIT_MASK)); |
||
114 | return a; |
||
115 | } |
||
731 | cejka | 116 | |
804 | cejka | 117 | if (frac1 & (FLOAT32_HIDDEN_BIT_MASK << 7) ) { |
731 | cejka | 118 | ++exp1; |
804 | cejka | 119 | frac1 >>= 1; |
731 | cejka | 120 | }; |
121 | |||
804 | cejka | 122 | /* rounding - if first bit after fraction is set then round up */ |
123 | frac1 += (0x1 << 5); |
||
731 | cejka | 124 | |
804 | cejka | 125 | if (frac1 & (FLOAT32_HIDDEN_BIT_MASK << 7)) { |
126 | /* rounding overflow */ |
||
731 | cejka | 127 | ++exp1; |
804 | cejka | 128 | frac1 >>= 1; |
731 | cejka | 129 | }; |
130 | |||
828 | cejka | 131 | |
132 | if ((exp1 == FLOAT32_MAX_EXPONENT ) || (exp2 > exp1)) { |
||
804 | cejka | 133 | /* overflow - set infinity as result */ |
134 | a.parts.exp = FLOAT32_MAX_EXPONENT; |
||
135 | a.parts.fraction = 0; |
||
136 | return a; |
||
137 | } |
||
138 | |||
731 | cejka | 139 | a.parts.exp = exp1; |
140 | |||
141 | /*Clear hidden bit and shift */ |
||
804 | cejka | 142 | a.parts.fraction = ((frac1 >> 6) & (~FLOAT32_HIDDEN_BIT_MASK)) ; |
731 | cejka | 143 | return a; |
144 | } |
||
145 | |||
146 | |||
147 | /** Add two Float64 numbers with same signs |
||
148 | */ |
||
149 | float64 addFloat64(float64 a, float64 b) |
||
150 | { |
||
151 | int expdiff; |
||
1031 | cejka | 152 | uint32_t exp1, exp2; |
153 | uint64_t frac1, frac2; |
||
731 | cejka | 154 | |
834 | cejka | 155 | expdiff = ((int )a.parts.exp) - b.parts.exp; |
732 | cejka | 156 | if (expdiff < 0) { |
731 | cejka | 157 | if (isFloat64NaN(b)) { |
804 | cejka | 158 | /* TODO: fix SigNaN */ |
731 | cejka | 159 | if (isFloat64SigNaN(b)) { |
160 | }; |
||
161 | |||
162 | return b; |
||
163 | }; |
||
164 | |||
165 | /* b is infinity and a not */ |
||
732 | cejka | 166 | if (b.parts.exp == FLOAT64_MAX_EXPONENT ) { |
731 | cejka | 167 | return b; |
168 | } |
||
169 | |||
804 | cejka | 170 | frac1 = b.parts.fraction; |
732 | cejka | 171 | exp1 = b.parts.exp; |
804 | cejka | 172 | frac2 = a.parts.fraction; |
732 | cejka | 173 | exp2 = a.parts.exp; |
174 | expdiff *= -1; |
||
731 | cejka | 175 | } else { |
176 | if (isFloat64NaN(a)) { |
||
804 | cejka | 177 | /* TODO: fix SigNaN */ |
732 | cejka | 178 | if (isFloat64SigNaN(a) || isFloat64SigNaN(b)) { |
731 | cejka | 179 | }; |
180 | return a; |
||
181 | }; |
||
182 | |||
183 | /* a is infinity and b not */ |
||
732 | cejka | 184 | if (a.parts.exp == FLOAT64_MAX_EXPONENT ) { |
731 | cejka | 185 | return a; |
186 | } |
||
187 | |||
804 | cejka | 188 | frac1 = a.parts.fraction; |
731 | cejka | 189 | exp1 = a.parts.exp; |
804 | cejka | 190 | frac2 = b.parts.fraction; |
731 | cejka | 191 | exp2 = b.parts.exp; |
192 | }; |
||
193 | |||
194 | if (exp1 == 0) { |
||
195 | /* both are denormalized */ |
||
804 | cejka | 196 | frac1 += frac2; |
197 | if (frac1 & FLOAT64_HIDDEN_BIT_MASK) { |
||
731 | cejka | 198 | /* result is not denormalized */ |
199 | a.parts.exp = 1; |
||
200 | }; |
||
804 | cejka | 201 | a.parts.fraction = frac1; |
731 | cejka | 202 | return a; |
203 | }; |
||
204 | |||
804 | cejka | 205 | /* add hidden bit - frac1 is sure not denormalized */ |
206 | frac1 |= FLOAT64_HIDDEN_BIT_MASK; |
||
731 | cejka | 207 | |
208 | /* second operand ... */ |
||
209 | if (exp2 == 0) { |
||
210 | /* ... is denormalized */ |
||
211 | --expdiff; |
||
212 | } else { |
||
213 | /* is not denormalized */ |
||
804 | cejka | 214 | frac2 |= FLOAT64_HIDDEN_BIT_MASK; |
731 | cejka | 215 | }; |
216 | |||
217 | /* create some space for rounding */ |
||
804 | cejka | 218 | frac1 <<= 6; |
219 | frac2 <<= 6; |
||
731 | cejka | 220 | |
804 | cejka | 221 | if (expdiff < (FLOAT64_FRACTION_SIZE + 2) ) { |
222 | frac2 >>= expdiff; |
||
223 | frac1 += frac2; |
||
834 | cejka | 224 | } else { |
225 | a.parts.exp = exp1; |
||
226 | a.parts.fraction = (frac1 >> 6) & (~(FLOAT64_HIDDEN_BIT_MASK)); |
||
227 | return a; |
||
228 | } |
||
731 | cejka | 229 | |
804 | cejka | 230 | if (frac1 & (FLOAT64_HIDDEN_BIT_MASK << 7) ) { |
731 | cejka | 231 | ++exp1; |
804 | cejka | 232 | frac1 >>= 1; |
731 | cejka | 233 | }; |
234 | |||
804 | cejka | 235 | /* rounding - if first bit after fraction is set then round up */ |
236 | frac1 += (0x1 << 5); |
||
731 | cejka | 237 | |
804 | cejka | 238 | if (frac1 & (FLOAT64_HIDDEN_BIT_MASK << 7)) { |
239 | /* rounding overflow */ |
||
731 | cejka | 240 | ++exp1; |
804 | cejka | 241 | frac1 >>= 1; |
731 | cejka | 242 | }; |
243 | |||
834 | cejka | 244 | if ((exp1 == FLOAT64_MAX_EXPONENT ) || (exp2 > exp1)) { |
804 | cejka | 245 | /* overflow - set infinity as result */ |
246 | a.parts.exp = FLOAT64_MAX_EXPONENT; |
||
247 | a.parts.fraction = 0; |
||
248 | return a; |
||
249 | } |
||
250 | |||
731 | cejka | 251 | a.parts.exp = exp1; |
252 | /*Clear hidden bit and shift */ |
||
804 | cejka | 253 | a.parts.fraction = ( (frac1 >> 6 ) & (~FLOAT64_HIDDEN_BIT_MASK)); |
834 | cejka | 254 | |
731 | cejka | 255 | return a; |
256 | } |
||
257 | |||
1740 | jermar | 258 | /** @} |
1657 | cejka | 259 | */ |