Rev 874 | Rev 876 | Go to most recent revision | Details | Compare with Previous | Last modification | View Log | RSS feed
| Rev | Author | Line No. | Line |
|---|---|---|---|
| 647 | 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 | |||
| 697 | cejka | 29 | #include "sftypes.h" |
| 30 | #include "conversion.h" |
||
| 857 | cejka | 31 | #include "comparison.h" |
| 874 | cejka | 32 | #include "common.h" |
| 697 | cejka | 33 | |
| 34 | float64 convertFloat32ToFloat64(float32 a) |
||
| 35 | { |
||
| 36 | float64 result; |
||
| 804 | cejka | 37 | __u64 frac; |
| 697 | cejka | 38 | |
| 39 | result.parts.sign = a.parts.sign; |
||
| 804 | cejka | 40 | result.parts.fraction = a.parts.fraction; |
| 41 | result.parts.fraction <<= (FLOAT64_FRACTION_SIZE - FLOAT32_FRACTION_SIZE ); |
||
| 697 | cejka | 42 | |
| 43 | if ((isFloat32Infinity(a))||(isFloat32NaN(a))) { |
||
| 44 | result.parts.exp = 0x7FF; |
||
| 45 | /* TODO; check if its correct for SigNaNs*/ |
||
| 46 | return result; |
||
| 47 | }; |
||
| 48 | |||
| 49 | result.parts.exp = a.parts.exp + ( (int)FLOAT64_BIAS - FLOAT32_BIAS ); |
||
| 50 | if (a.parts.exp == 0) { |
||
| 51 | /* normalize denormalized numbers */ |
||
| 52 | |||
| 804 | cejka | 53 | if (result.parts.fraction == 0ll) { /* fix zero */ |
| 697 | cejka | 54 | result.parts.exp = 0ll; |
| 55 | return result; |
||
| 56 | } |
||
| 57 | |||
| 804 | cejka | 58 | frac = result.parts.fraction; |
| 697 | cejka | 59 | |
| 804 | cejka | 60 | while (!(frac & (0x10000000000000ll))) { |
| 61 | frac <<= 1; |
||
| 697 | cejka | 62 | --result.parts.exp; |
| 63 | }; |
||
| 698 | cejka | 64 | |
| 65 | ++result.parts.exp; |
||
| 804 | cejka | 66 | result.parts.fraction = frac; |
| 697 | cejka | 67 | }; |
| 68 | |||
| 69 | return result; |
||
| 70 | |||
| 857 | cejka | 71 | } |
| 697 | cejka | 72 | |
| 73 | float32 convertFloat64ToFloat32(float64 a) |
||
| 74 | { |
||
| 75 | float32 result; |
||
| 76 | __s32 exp; |
||
| 804 | cejka | 77 | __u64 frac; |
| 697 | cejka | 78 | |
| 79 | result.parts.sign = a.parts.sign; |
||
| 80 | |||
| 81 | if (isFloat64NaN(a)) { |
||
| 82 | |||
| 83 | result.parts.exp = 0xFF; |
||
| 84 | |||
| 85 | if (isFloat64SigNaN(a)) { |
||
| 804 | cejka | 86 | result.parts.fraction = 0x800000; /* set first bit of fraction nonzero */ |
| 697 | cejka | 87 | return result; |
| 88 | } |
||
| 89 | |||
| 804 | cejka | 90 | result.parts.fraction = 0x1; /* fraction nonzero but its first bit is zero */ |
| 697 | cejka | 91 | return result; |
| 92 | }; |
||
| 93 | |||
| 94 | if (isFloat64Infinity(a)) { |
||
| 804 | cejka | 95 | result.parts.fraction = 0; |
| 697 | cejka | 96 | result.parts.exp = 0xFF; |
| 97 | return result; |
||
| 98 | }; |
||
| 99 | |||
| 100 | exp = (int)a.parts.exp - FLOAT64_BIAS + FLOAT32_BIAS; |
||
| 101 | |||
| 102 | if (exp >= 0xFF) { |
||
| 103 | /*FIXME: overflow*/ |
||
| 804 | cejka | 104 | result.parts.fraction = 0; |
| 697 | cejka | 105 | result.parts.exp = 0xFF; |
| 106 | return result; |
||
| 107 | |||
| 108 | } else if (exp <= 0 ) { |
||
| 109 | |||
| 110 | /* underflow or denormalized */ |
||
| 111 | |||
| 112 | result.parts.exp = 0; |
||
| 113 | |||
| 114 | exp *= -1; |
||
| 804 | cejka | 115 | if (exp > FLOAT32_FRACTION_SIZE ) { |
| 697 | cejka | 116 | /* FIXME: underflow */ |
| 804 | cejka | 117 | result.parts.fraction = 0; |
| 697 | cejka | 118 | return result; |
| 119 | }; |
||
| 120 | |||
| 121 | /* denormalized */ |
||
| 122 | |||
| 804 | cejka | 123 | frac = a.parts.fraction; |
| 124 | frac |= 0x10000000000000ll; /* denormalize and set hidden bit */ |
||
| 697 | cejka | 125 | |
| 804 | cejka | 126 | frac >>= (FLOAT64_FRACTION_SIZE - FLOAT32_FRACTION_SIZE + 1); |
| 698 | cejka | 127 | |
| 697 | cejka | 128 | while (exp > 0) { |
| 129 | --exp; |
||
| 804 | cejka | 130 | frac >>= 1; |
| 697 | cejka | 131 | }; |
| 804 | cejka | 132 | result.parts.fraction = frac; |
| 697 | cejka | 133 | |
| 134 | return result; |
||
| 135 | }; |
||
| 136 | |||
| 137 | result.parts.exp = exp; |
||
| 804 | cejka | 138 | result.parts.fraction = a.parts.fraction >> (FLOAT64_FRACTION_SIZE - FLOAT32_FRACTION_SIZE); |
| 697 | cejka | 139 | return result; |
| 857 | cejka | 140 | } |
| 697 | cejka | 141 | |
| 857 | cejka | 142 | |
| 143 | /** Helping procedure for converting float32 to uint32 |
||
| 144 | * @param a floating point number in normalized form (no NaNs or Inf are checked ) |
||
| 145 | * @return unsigned integer |
||
| 146 | */ |
||
| 147 | static __u32 _float32_to_uint32_helper(float32 a) |
||
| 148 | { |
||
| 149 | __u32 frac; |
||
| 150 | |||
| 151 | if (a.parts.exp < FLOAT32_BIAS) { |
||
| 152 | /*TODO: rounding*/ |
||
| 153 | return 0; |
||
| 154 | } |
||
| 155 | |||
| 156 | frac = a.parts.fraction; |
||
| 157 | |||
| 158 | frac |= FLOAT32_HIDDEN_BIT_MASK; |
||
| 159 | /* shift fraction to left so hidden bit will be the most significant bit */ |
||
| 160 | frac <<= 32 - FLOAT32_FRACTION_SIZE - 1; |
||
| 161 | |||
| 162 | frac >>= 32 - (a.parts.exp - FLOAT32_BIAS) - 1; |
||
| 163 | if ((a.parts.sign == 1) && (frac != 0)) { |
||
| 164 | frac = ~frac; |
||
| 165 | ++frac; |
||
| 166 | } |
||
| 167 | |||
| 168 | return frac; |
||
| 169 | } |
||
| 170 | |||
| 171 | /* Convert float to unsigned int32 |
||
| 172 | * FIXME: Im not sure what to return if overflow/underflow happens |
||
| 173 | * - now its the biggest or the smallest int |
||
| 174 | */ |
||
| 175 | __u32 float32_to_uint32(float32 a) |
||
| 176 | { |
||
| 177 | if (isFloat32NaN(a)) { |
||
| 178 | return MAX_UINT32; |
||
| 179 | } |
||
| 180 | |||
| 181 | if (isFloat32Infinity(a) || (a.parts.exp >= (32 + FLOAT32_BIAS))) { |
||
| 182 | if (a.parts.sign) { |
||
| 183 | return MIN_UINT32; |
||
| 184 | } |
||
| 185 | return MAX_UINT32; |
||
| 186 | } |
||
| 187 | |||
| 188 | return _float32_to_uint32_helper(a); |
||
| 189 | } |
||
| 190 | |||
| 191 | /* Convert float to signed int32 |
||
| 192 | * FIXME: Im not sure what to return if overflow/underflow happens |
||
| 193 | * - now its the biggest or the smallest int |
||
| 194 | */ |
||
| 195 | __s32 float32_to_int32(float32 a) |
||
| 196 | { |
||
| 197 | if (isFloat32NaN(a)) { |
||
| 198 | return MAX_INT32; |
||
| 199 | } |
||
| 200 | |||
| 201 | if (isFloat32Infinity(a) || (a.parts.exp >= (32 + FLOAT32_BIAS))) { |
||
| 202 | if (a.parts.sign) { |
||
| 203 | return MIN_INT32; |
||
| 204 | } |
||
| 205 | return MAX_INT32; |
||
| 206 | } |
||
| 207 | return _float32_to_uint32_helper(a); |
||
| 208 | } |
||
| 209 | |||
| 210 | |||
| 865 | cejka | 211 | /** Helping procedure for converting float64 to uint64 |
| 212 | * @param a floating point number in normalized form (no NaNs or Inf are checked ) |
||
| 213 | * @return unsigned integer |
||
| 214 | */ |
||
| 215 | static __u64 _float64_to_uint64_helper(float64 a) |
||
| 216 | { |
||
| 217 | __u64 frac; |
||
| 218 | |||
| 219 | if (a.parts.exp < FLOAT64_BIAS) { |
||
| 220 | /*TODO: rounding*/ |
||
| 221 | return 0; |
||
| 222 | } |
||
| 223 | |||
| 224 | frac = a.parts.fraction; |
||
| 225 | |||
| 226 | frac |= FLOAT64_HIDDEN_BIT_MASK; |
||
| 227 | /* shift fraction to left so hidden bit will be the most significant bit */ |
||
| 228 | frac <<= 64 - FLOAT64_FRACTION_SIZE - 1; |
||
| 857 | cejka | 229 | |
| 865 | cejka | 230 | frac >>= 64 - (a.parts.exp - FLOAT64_BIAS) - 1; |
| 231 | if ((a.parts.sign == 1) && (frac != 0)) { |
||
| 232 | frac = ~frac; |
||
| 233 | ++frac; |
||
| 234 | } |
||
| 235 | |||
| 236 | return frac; |
||
| 237 | } |
||
| 238 | |||
| 239 | /* Convert float to unsigned int64 |
||
| 240 | * FIXME: Im not sure what to return if overflow/underflow happens |
||
| 241 | * - now its the biggest or the smallest int |
||
| 242 | */ |
||
| 243 | __u64 float64_to_uint64(float64 a) |
||
| 244 | { |
||
| 245 | if (isFloat64NaN(a)) { |
||
| 246 | return MAX_UINT64; |
||
| 247 | } |
||
| 248 | |||
| 249 | if (isFloat64Infinity(a) || (a.parts.exp >= (64 + FLOAT64_BIAS))) { |
||
| 250 | if (a.parts.sign) { |
||
| 251 | return MIN_UINT64; |
||
| 252 | } |
||
| 253 | return MAX_UINT64; |
||
| 254 | } |
||
| 255 | |||
| 256 | return _float64_to_uint64_helper(a); |
||
| 257 | } |
||
| 258 | |||
| 259 | /* Convert float to signed int64 |
||
| 260 | * FIXME: Im not sure what to return if overflow/underflow happens |
||
| 261 | * - now its the biggest or the smallest int |
||
| 262 | */ |
||
| 263 | __s64 float64_to_int64(float64 a) |
||
| 264 | { |
||
| 265 | if (isFloat64NaN(a)) { |
||
| 266 | return MAX_INT64; |
||
| 267 | } |
||
| 268 | |||
| 269 | if (isFloat64Infinity(a) || (a.parts.exp >= (64 + FLOAT64_BIAS))) { |
||
| 270 | if (a.parts.sign) { |
||
| 271 | return MIN_INT64; |
||
| 272 | } |
||
| 273 | return MAX_INT64; |
||
| 274 | } |
||
| 275 | return _float64_to_uint64_helper(a); |
||
| 276 | } |
||
| 277 | |||
| 278 | |||
| 279 | |||
| 280 | |||
| 281 | |||
| 282 | /** Helping procedure for converting float32 to uint64 |
||
| 283 | * @param a floating point number in normalized form (no NaNs or Inf are checked ) |
||
| 284 | * @return unsigned integer |
||
| 285 | */ |
||
| 286 | static __u64 _float32_to_uint64_helper(float32 a) |
||
| 287 | { |
||
| 288 | __u64 frac; |
||
| 289 | |||
| 290 | if (a.parts.exp < FLOAT32_BIAS) { |
||
| 291 | /*TODO: rounding*/ |
||
| 292 | return 0; |
||
| 293 | } |
||
| 294 | |||
| 295 | frac = a.parts.fraction; |
||
| 296 | |||
| 297 | frac |= FLOAT32_HIDDEN_BIT_MASK; |
||
| 298 | /* shift fraction to left so hidden bit will be the most significant bit */ |
||
| 299 | frac <<= 64 - FLOAT32_FRACTION_SIZE - 1; |
||
| 300 | |||
| 301 | frac >>= 64 - (a.parts.exp - FLOAT32_BIAS) - 1; |
||
| 302 | if ((a.parts.sign == 1) && (frac != 0)) { |
||
| 303 | frac = ~frac; |
||
| 304 | ++frac; |
||
| 305 | } |
||
| 306 | |||
| 307 | return frac; |
||
| 308 | } |
||
| 309 | |||
| 310 | /* Convert float to unsigned int64 |
||
| 311 | * FIXME: Im not sure what to return if overflow/underflow happens |
||
| 312 | * - now its the biggest or the smallest int |
||
| 313 | */ |
||
| 314 | __u64 float32_to_uint64(float32 a) |
||
| 315 | { |
||
| 316 | if (isFloat32NaN(a)) { |
||
| 317 | return MAX_UINT64; |
||
| 318 | } |
||
| 319 | |||
| 320 | if (isFloat32Infinity(a) || (a.parts.exp >= (64 + FLOAT32_BIAS))) { |
||
| 321 | if (a.parts.sign) { |
||
| 322 | return MIN_UINT64; |
||
| 323 | } |
||
| 324 | return MAX_UINT64; |
||
| 325 | } |
||
| 326 | |||
| 327 | return _float32_to_uint64_helper(a); |
||
| 328 | } |
||
| 329 | |||
| 330 | /* Convert float to signed int64 |
||
| 331 | * FIXME: Im not sure what to return if overflow/underflow happens |
||
| 332 | * - now its the biggest or the smallest int |
||
| 333 | */ |
||
| 334 | __s64 float32_to_int64(float32 a) |
||
| 335 | { |
||
| 336 | if (isFloat32NaN(a)) { |
||
| 337 | return MAX_INT64; |
||
| 338 | } |
||
| 339 | |||
| 340 | if (isFloat32Infinity(a) || (a.parts.exp >= (64 + FLOAT32_BIAS))) { |
||
| 341 | if (a.parts.sign) { |
||
| 342 | return (MIN_INT64); |
||
| 343 | } |
||
| 344 | return MAX_INT64; |
||
| 345 | } |
||
| 346 | return _float32_to_uint64_helper(a); |
||
| 347 | } |
||
| 348 | |||
| 349 | |||
| 350 | /* Convert float64 to unsigned int32 |
||
| 351 | * FIXME: Im not sure what to return if overflow/underflow happens |
||
| 352 | * - now its the biggest or the smallest int |
||
| 353 | */ |
||
| 354 | __u32 float64_to_uint32(float64 a) |
||
| 355 | { |
||
| 356 | if (isFloat64NaN(a)) { |
||
| 357 | return MAX_UINT32; |
||
| 358 | } |
||
| 359 | |||
| 360 | if (isFloat64Infinity(a) || (a.parts.exp >= (32 + FLOAT64_BIAS))) { |
||
| 361 | if (a.parts.sign) { |
||
| 362 | return MIN_UINT32; |
||
| 363 | } |
||
| 364 | return MAX_UINT32; |
||
| 365 | } |
||
| 366 | |||
| 367 | return (__u32)_float64_to_uint64_helper(a); |
||
| 368 | } |
||
| 369 | |||
| 370 | /* Convert float64 to signed int32 |
||
| 371 | * FIXME: Im not sure what to return if overflow/underflow happens |
||
| 372 | * - now its the biggest or the smallest int |
||
| 373 | */ |
||
| 374 | __s32 float64_to_int32(float64 a) |
||
| 375 | { |
||
| 376 | if (isFloat64NaN(a)) { |
||
| 377 | return MAX_INT32; |
||
| 378 | } |
||
| 379 | |||
| 380 | if (isFloat64Infinity(a) || (a.parts.exp >= (32 + FLOAT64_BIAS))) { |
||
| 381 | if (a.parts.sign) { |
||
| 382 | return MIN_INT32; |
||
| 383 | } |
||
| 384 | return MAX_INT32; |
||
| 385 | } |
||
| 386 | return (__s32)_float64_to_uint64_helper(a); |
||
| 387 | } |
||
| 388 | |||
| 874 | cejka | 389 | |
| 390 | /** Convert unsigned integer to float32 |
||
| 391 | * |
||
| 392 | * |
||
| 393 | */ |
||
| 394 | float32 uint32_to_float32(__u32 i) |
||
| 395 | { |
||
| 396 | int counter; |
||
| 397 | __s32 exp; |
||
| 398 | float32 result; |
||
| 399 | |||
| 400 | result.parts.sign = 0; |
||
| 401 | result.parts.fraction = 0; |
||
| 865 | cejka | 402 | |
| 874 | cejka | 403 | counter = countZeroes32(i); |
| 404 | |||
| 405 | exp = FLOAT32_BIAS + 32 - counter - 1; |
||
| 406 | |||
| 407 | if (counter == 32) { |
||
| 408 | result.binary = 0; |
||
| 409 | return result; |
||
| 410 | } |
||
| 411 | |||
| 412 | if (counter > 0) { |
||
| 413 | i <<= counter - 1; |
||
| 414 | } else { |
||
| 415 | i >>= 1; |
||
| 416 | } |
||
| 417 | |||
| 418 | roundFloat32(&exp, &i); |
||
| 419 | |||
| 420 | result.parts.fraction = i >> 7; |
||
| 421 | result.parts.exp = exp; |
||
| 422 | |||
| 423 | return result; |
||
| 424 | } |
||
| 425 | |||
| 426 | float32 int32_to_float32(__s32 i) |
||
| 427 | { |
||
| 428 | float32 result; |
||
| 429 | |||
| 430 | if (i < 0) { |
||
| 431 | result = uint32_to_float32((__u32)(-i)); |
||
| 432 | } else { |
||
| 433 | result = uint32_to_float32((__u32)i); |
||
| 434 | } |
||
| 435 | |||
| 436 | result.parts.sign = i < 0; |
||
| 437 | |||
| 438 | return result; |
||
| 439 | } |
||
| 440 | |||
| 441 | |||
| 442 | float32 uint64_to_float32(__u64 i) |
||
| 443 | { |
||
| 875 | cejka | 444 | int counter; |
| 445 | __s32 exp; |
||
| 446 | float32 result; |
||
| 447 | |||
| 448 | result.parts.sign = 0; |
||
| 449 | result.parts.fraction = 0; |
||
| 450 | |||
| 451 | counter = countZeroes64(i); |
||
| 452 | |||
| 453 | exp = FLOAT32_BIAS + 64 - counter - 1; |
||
| 454 | |||
| 455 | if (counter == 64) { |
||
| 456 | result.binary = 0; |
||
| 457 | return result; |
||
| 458 | } |
||
| 459 | |||
| 460 | /* Shift all to the first 31 bits (31. will be hidden 1)*/ |
||
| 461 | if (counter > 33) { |
||
| 462 | i <<= counter - 1 - 32; |
||
| 463 | } else { |
||
| 464 | i >>= 1 + 32 - counter; |
||
| 465 | } |
||
| 466 | |||
| 467 | roundFloat32(&exp, &i); |
||
| 468 | |||
| 469 | result.parts.fraction = i >> 7; |
||
| 470 | result.parts.exp = exp; |
||
| 471 | return result; |
||
| 874 | cejka | 472 | } |
| 473 | |||
| 474 | float32 int64_to_float32(__s64 i) |
||
| 475 | { |
||
| 476 | float32 result; |
||
| 477 | |||
| 478 | if (i < 0) { |
||
| 479 | result = uint64_to_float32((__u64)(-i)); |
||
| 480 | } else { |
||
| 481 | result = uint64_to_float32((__u64)i); |
||
| 482 | } |
||
| 483 | |||
| 484 | result.parts.sign = i < 0; |
||
| 485 | |||
| 486 | return result; |
||
| 487 | } |