Rev 804 | Rev 1031 | 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<add.h> |
||
| 31 | #include<comparison.h> |
||
| 32 | |||
| 33 | /** Add two Float32 numbers with same signs |
||
| 34 | */ |
||
| 35 | float32 addFloat32(float32 a, float32 b) |
||
| 36 | { |
||
| 37 | int expdiff; |
||
| 804 | cejka | 38 | __u32 exp1, exp2,frac1, frac2; |
| 731 | cejka | 39 | |
| 732 | cejka | 40 | expdiff = a.parts.exp - b.parts.exp; |
| 41 | if (expdiff < 0) { |
||
| 731 | cejka | 42 | if (isFloat32NaN(b)) { |
| 804 | cejka | 43 | /* TODO: fix SigNaN */ |
| 731 | cejka | 44 | if (isFloat32SigNaN(b)) { |
| 45 | }; |
||
| 46 | |||
| 47 | return b; |
||
| 48 | }; |
||
| 49 | |||
| 732 | cejka | 50 | if (b.parts.exp == FLOAT32_MAX_EXPONENT) { |
| 731 | cejka | 51 | return b; |
| 52 | } |
||
| 53 | |||
| 804 | cejka | 54 | frac1 = b.parts.fraction; |
| 732 | cejka | 55 | exp1 = b.parts.exp; |
| 804 | cejka | 56 | frac2 = a.parts.fraction; |
| 732 | cejka | 57 | exp2 = a.parts.exp; |
| 58 | expdiff *= -1; |
||
| 731 | cejka | 59 | } else { |
| 804 | cejka | 60 | if ((isFloat32NaN(a)) || (isFloat32NaN(b))) { |
| 61 | /* TODO: fix SigNaN */ |
||
| 732 | cejka | 62 | if (isFloat32SigNaN(a) || isFloat32SigNaN(b)) { |
| 731 | cejka | 63 | }; |
| 804 | cejka | 64 | return (isFloat32NaN(a)?a:b); |
| 731 | cejka | 65 | }; |
| 66 | |||
| 732 | cejka | 67 | if (a.parts.exp == FLOAT32_MAX_EXPONENT) { |
| 731 | cejka | 68 | return a; |
| 69 | } |
||
| 70 | |||
| 804 | cejka | 71 | frac1 = a.parts.fraction; |
| 732 | cejka | 72 | exp1 = a.parts.exp; |
| 804 | cejka | 73 | frac2 = b.parts.fraction; |
| 732 | cejka | 74 | exp2 = b.parts.exp; |
| 731 | cejka | 75 | }; |
| 76 | |||
| 77 | if (exp1 == 0) { |
||
| 78 | /* both are denormalized */ |
||
| 804 | cejka | 79 | frac1 += frac2; |
| 80 | if (frac1 & FLOAT32_HIDDEN_BIT_MASK ) { |
||
| 731 | cejka | 81 | /* result is not denormalized */ |
| 82 | a.parts.exp = 1; |
||
| 83 | }; |
||
| 804 | cejka | 84 | a.parts.fraction = frac1; |
| 731 | cejka | 85 | return a; |
| 86 | }; |
||
| 87 | |||
| 804 | cejka | 88 | frac1 |= FLOAT32_HIDDEN_BIT_MASK; /* add hidden bit */ |
| 731 | cejka | 89 | |
| 90 | if (exp2 == 0) { |
||
| 91 | /* second operand is denormalized */ |
||
| 828 | cejka | 92 | --expdiff; |
| 731 | cejka | 93 | } else { |
| 94 | /* add hidden bit to second operand */ |
||
| 804 | cejka | 95 | frac2 |= FLOAT32_HIDDEN_BIT_MASK; |
| 731 | cejka | 96 | }; |
| 97 | |||
| 98 | /* create some space for rounding */ |
||
| 804 | cejka | 99 | frac1 <<= 6; |
| 100 | frac2 <<= 6; |
||
| 731 | cejka | 101 | |
| 804 | cejka | 102 | if (expdiff < (FLOAT32_FRACTION_SIZE + 2) ) { |
| 103 | frac2 >>= expdiff; |
||
| 104 | frac1 += frac2; |
||
| 105 | }; |
||
| 731 | cejka | 106 | |
| 804 | cejka | 107 | if (frac1 & (FLOAT32_HIDDEN_BIT_MASK << 7) ) { |
| 731 | cejka | 108 | ++exp1; |
| 804 | cejka | 109 | frac1 >>= 1; |
| 731 | cejka | 110 | }; |
| 111 | |||
| 804 | cejka | 112 | /* rounding - if first bit after fraction is set then round up */ |
| 113 | frac1 += (0x1 << 5); |
||
| 731 | cejka | 114 | |
| 804 | cejka | 115 | if (frac1 & (FLOAT32_HIDDEN_BIT_MASK << 7)) { |
| 116 | /* rounding overflow */ |
||
| 731 | cejka | 117 | ++exp1; |
| 804 | cejka | 118 | frac1 >>= 1; |
| 731 | cejka | 119 | }; |
| 120 | |||
| 828 | cejka | 121 | |
| 122 | if ((exp1 == FLOAT32_MAX_EXPONENT ) || (exp2 > exp1)) { |
||
| 804 | cejka | 123 | /* overflow - set infinity as result */ |
| 124 | a.parts.exp = FLOAT32_MAX_EXPONENT; |
||
| 125 | a.parts.fraction = 0; |
||
| 126 | return a; |
||
| 127 | } |
||
| 128 | |||
| 731 | cejka | 129 | a.parts.exp = exp1; |
| 130 | |||
| 131 | /*Clear hidden bit and shift */ |
||
| 804 | cejka | 132 | a.parts.fraction = ((frac1 >> 6) & (~FLOAT32_HIDDEN_BIT_MASK)) ; |
| 731 | cejka | 133 | return a; |
| 134 | } |
||
| 135 | |||
| 136 | |||
| 137 | /** Add two Float64 numbers with same signs |
||
| 138 | */ |
||
| 139 | float64 addFloat64(float64 a, float64 b) |
||
| 140 | { |
||
| 141 | int expdiff; |
||
| 734 | cejka | 142 | __u32 exp1, exp2; |
| 804 | cejka | 143 | __u64 frac1, frac2; |
| 731 | cejka | 144 | |
| 732 | cejka | 145 | expdiff = a.parts.exp - b.parts.exp; |
| 146 | if (expdiff < 0) { |
||
| 731 | cejka | 147 | if (isFloat64NaN(b)) { |
| 804 | cejka | 148 | /* TODO: fix SigNaN */ |
| 731 | cejka | 149 | if (isFloat64SigNaN(b)) { |
| 150 | }; |
||
| 151 | |||
| 152 | return b; |
||
| 153 | }; |
||
| 154 | |||
| 155 | /* b is infinity and a not */ |
||
| 732 | cejka | 156 | if (b.parts.exp == FLOAT64_MAX_EXPONENT ) { |
| 731 | cejka | 157 | return b; |
| 158 | } |
||
| 159 | |||
| 804 | cejka | 160 | frac1 = b.parts.fraction; |
| 732 | cejka | 161 | exp1 = b.parts.exp; |
| 804 | cejka | 162 | frac2 = a.parts.fraction; |
| 732 | cejka | 163 | exp2 = a.parts.exp; |
| 164 | expdiff *= -1; |
||
| 731 | cejka | 165 | } else { |
| 166 | if (isFloat64NaN(a)) { |
||
| 804 | cejka | 167 | /* TODO: fix SigNaN */ |
| 732 | cejka | 168 | if (isFloat64SigNaN(a) || isFloat64SigNaN(b)) { |
| 731 | cejka | 169 | }; |
| 170 | return a; |
||
| 171 | }; |
||
| 172 | |||
| 173 | /* a is infinity and b not */ |
||
| 732 | cejka | 174 | if (a.parts.exp == FLOAT64_MAX_EXPONENT ) { |
| 731 | cejka | 175 | return a; |
| 176 | } |
||
| 177 | |||
| 804 | cejka | 178 | frac1 = a.parts.fraction; |
| 731 | cejka | 179 | exp1 = a.parts.exp; |
| 804 | cejka | 180 | frac2 = b.parts.fraction; |
| 731 | cejka | 181 | exp2 = b.parts.exp; |
| 182 | }; |
||
| 183 | |||
| 184 | if (exp1 == 0) { |
||
| 185 | /* both are denormalized */ |
||
| 804 | cejka | 186 | frac1 += frac2; |
| 187 | if (frac1 & FLOAT64_HIDDEN_BIT_MASK) { |
||
| 731 | cejka | 188 | /* result is not denormalized */ |
| 189 | a.parts.exp = 1; |
||
| 190 | }; |
||
| 804 | cejka | 191 | a.parts.fraction = frac1; |
| 731 | cejka | 192 | return a; |
| 193 | }; |
||
| 194 | |||
| 804 | cejka | 195 | /* add hidden bit - frac1 is sure not denormalized */ |
| 196 | frac1 |= FLOAT64_HIDDEN_BIT_MASK; |
||
| 731 | cejka | 197 | |
| 198 | /* second operand ... */ |
||
| 199 | if (exp2 == 0) { |
||
| 200 | /* ... is denormalized */ |
||
| 201 | --expdiff; |
||
| 202 | } else { |
||
| 203 | /* is not denormalized */ |
||
| 804 | cejka | 204 | frac2 |= FLOAT64_HIDDEN_BIT_MASK; |
| 731 | cejka | 205 | }; |
| 206 | |||
| 207 | /* create some space for rounding */ |
||
| 804 | cejka | 208 | frac1 <<= 6; |
| 209 | frac2 <<= 6; |
||
| 731 | cejka | 210 | |
| 804 | cejka | 211 | if (expdiff < (FLOAT64_FRACTION_SIZE + 2) ) { |
| 212 | frac2 >>= expdiff; |
||
| 213 | frac1 += frac2; |
||
| 214 | }; |
||
| 731 | cejka | 215 | |
| 804 | cejka | 216 | if (frac1 & (FLOAT64_HIDDEN_BIT_MASK << 7) ) { |
| 731 | cejka | 217 | ++exp1; |
| 804 | cejka | 218 | frac1 >>= 1; |
| 731 | cejka | 219 | }; |
| 220 | |||
| 804 | cejka | 221 | /* rounding - if first bit after fraction is set then round up */ |
| 222 | frac1 += (0x1 << 5); |
||
| 731 | cejka | 223 | |
| 804 | cejka | 224 | if (frac1 & (FLOAT64_HIDDEN_BIT_MASK << 7)) { |
| 225 | /* rounding overflow */ |
||
| 731 | cejka | 226 | ++exp1; |
| 804 | cejka | 227 | frac1 >>= 1; |
| 731 | cejka | 228 | }; |
| 229 | |||
| 804 | cejka | 230 | if ((a.parts.exp == FLOAT64_MAX_EXPONENT ) || (a.parts.exp < exp1)) { |
| 231 | /* overflow - set infinity as result */ |
||
| 232 | a.parts.exp = FLOAT64_MAX_EXPONENT; |
||
| 233 | a.parts.fraction = 0; |
||
| 234 | return a; |
||
| 235 | } |
||
| 236 | |||
| 731 | cejka | 237 | a.parts.exp = exp1; |
| 238 | /*Clear hidden bit and shift */ |
||
| 804 | cejka | 239 | a.parts.fraction = ( (frac1 >> 6 ) & (~FLOAT64_HIDDEN_BIT_MASK)); |
| 731 | cejka | 240 | return a; |
| 241 | } |
||
| 242 | |||
| 243 |