1 /* 2 * Licensed to the Apache Software Foundation (ASF) under one or more 3 * contributor license agreements. See the NOTICE file distributed with 4 * this work for additional information regarding copyright ownership. 5 * The ASF licenses this file to You under the Apache License, Version 2.0 6 * (the "License"); you may not use this file except in compliance with 7 * the License. You may obtain a copy of the License at 8 * 9 * http://www.apache.org/licenses/LICENSE-2.0 10 * 11 * Unless required by applicable law or agreed to in writing, software 12 * distributed under the License is distributed on an "AS IS" BASIS, 13 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 * See the License for the specific language governing permissions and 15 * limitations under the License. 16 */ 17 18 #include <string.h> 19 #include "cbigint.h" 20 21 #if defined(__linux__) || defined(FREEBSD) || defined(ZOS) || defined(MACOSX) || defined(AIX) 22 #define USE_LL 23 #endif 24 25 #if __BYTE_ORDER == __LITTLE_ENDIAN 26 #define at(i) (i) 27 #else 28 #define at(i) ((i)^1) 29 /* the sequence for halfAt is -1, 2, 1, 4, 3, 6, 5, 8... */ 30 /* and it should correspond to 0, 1, 2, 3, 4, 5, 6, 7... */ 31 #define halfAt(i) (-((-(i)) ^ 1)) 32 #endif 33 34 #define HIGH_IN_U64(u64) ((u64) >> 32) 35 #if defined(USE_LL) 36 #define LOW_IN_U64(u64) ((u64) & 0x00000000FFFFFFFFLL) 37 #else 38 #if defined(USE_L) 39 #define LOW_IN_U64(u64) ((u64) & 0x00000000FFFFFFFFL) 40 #else 41 #define LOW_IN_U64(u64) ((u64) & 0x00000000FFFFFFFF) 42 #endif /* USE_L */ 43 #endif /* USE_LL */ 44 45 #if defined(USE_LL) 46 #define TEN_E1 (0xALL) 47 #define TEN_E2 (0x64LL) 48 #define TEN_E3 (0x3E8LL) 49 #define TEN_E4 (0x2710LL) 50 #define TEN_E5 (0x186A0LL) 51 #define TEN_E6 (0xF4240LL) 52 #define TEN_E7 (0x989680LL) 53 #define TEN_E8 (0x5F5E100LL) 54 #define TEN_E9 (0x3B9ACA00LL) 55 #define TEN_E19 (0x8AC7230489E80000LL) 56 #else 57 #if defined(USE_L) 58 #define TEN_E1 (0xAL) 59 #define TEN_E2 (0x64L) 60 #define TEN_E3 (0x3E8L) 61 #define TEN_E4 (0x2710L) 62 #define TEN_E5 (0x186A0L) 63 #define TEN_E6 (0xF4240L) 64 #define TEN_E7 (0x989680L) 65 #define TEN_E8 (0x5F5E100L) 66 #define TEN_E9 (0x3B9ACA00L) 67 #define TEN_E19 (0x8AC7230489E80000L) 68 #else 69 #define TEN_E1 (0xA) 70 #define TEN_E2 (0x64) 71 #define TEN_E3 (0x3E8) 72 #define TEN_E4 (0x2710) 73 #define TEN_E5 (0x186A0) 74 #define TEN_E6 (0xF4240) 75 #define TEN_E7 (0x989680) 76 #define TEN_E8 (0x5F5E100) 77 #define TEN_E9 (0x3B9ACA00) 78 #define TEN_E19 (0x8AC7230489E80000) 79 #endif /* USE_L */ 80 #endif /* USE_LL */ 81 82 #define TIMES_TEN(x) (((x) << 3) + ((x) << 1)) 83 #define bitSection(x, mask, shift) (((x) & (mask)) >> (shift)) 84 #define CREATE_DOUBLE_BITS(normalizedM, e) (((normalizedM) & MANTISSA_MASK) | ((static_cast<uint64_t>((e) + E_OFFSET)) << 52)) 85 86 #if defined(USE_LL) 87 #define MANTISSA_MASK (0x000FFFFFFFFFFFFFLL) 88 #define EXPONENT_MASK (0x7FF0000000000000LL) 89 #define NORMAL_MASK (0x0010000000000000LL) 90 #define SIGN_MASK (0x8000000000000000LL) 91 #else 92 #if defined(USE_L) 93 #define MANTISSA_MASK (0x000FFFFFFFFFFFFFL) 94 #define EXPONENT_MASK (0x7FF0000000000000L) 95 #define NORMAL_MASK (0x0010000000000000L) 96 #define SIGN_MASK (0x8000000000000000L) 97 #else 98 #define MANTISSA_MASK (0x000FFFFFFFFFFFFF) 99 #define EXPONENT_MASK (0x7FF0000000000000) 100 #define NORMAL_MASK (0x0010000000000000) 101 #define SIGN_MASK (0x8000000000000000) 102 #endif /* USE_L */ 103 #endif /* USE_LL */ 104 105 #define E_OFFSET (1075) 106 107 #define FLOAT_MANTISSA_MASK (0x007FFFFF) 108 #define FLOAT_EXPONENT_MASK (0x7F800000) 109 #define FLOAT_NORMAL_MASK (0x00800000) 110 #define FLOAT_E_OFFSET (150) 111 112 int32_t 113 simpleAddHighPrecision (uint64_t * arg1, int32_t length, uint64_t arg2) 114 { 115 /* assumes length > 0 */ 116 int32_t index = 1; 117 118 *arg1 += arg2; 119 if (arg2 <= *arg1) 120 return 0; 121 else if (length == 1) 122 return 1; 123 124 while (++arg1[index] == 0 && ++index < length); 125 126 return index == length; 127 } 128 129 int32_t 130 addHighPrecision (uint64_t * arg1, int32_t length1, uint64_t * arg2, int32_t length2) 131 { 132 /* addition is limited by length of arg1 as it this function is 133 * storing the result in arg1 */ 134 /* fix for cc (GCC) 3.2 20020903 (Red Hat Linux 8.0 3.2-7): code generated does not 135 * do the temp1 + temp2 + carry addition correct. carry is 64 bit because gcc has 136 * subtle issues when you mix 64 / 32 bit maths. */ 137 uint64_t temp1, temp2, temp3; /* temporary variables to help the SH-4, and gcc */ 138 uint64_t carry; 139 int32_t index; 140 141 if (length1 == 0 || length2 == 0) 142 { 143 return 0; 144 } 145 else if (length1 < length2) 146 { 147 length2 = length1; 148 } 149 150 carry = 0; 151 index = 0; 152 do 153 { 154 temp1 = arg1[index]; 155 temp2 = arg2[index]; 156 temp3 = temp1 + temp2; 157 arg1[index] = temp3 + carry; 158 if (arg2[index] < arg1[index]) 159 carry = 0; 160 else if (arg2[index] != arg1[index]) 161 carry = 1; 162 } 163 while (++index < length2); 164 if (!carry) 165 return 0; 166 else if (index == length1) 167 return 1; 168 169 while (++arg1[index] == 0 && ++index < length1); 170 171 return index == length1; 172 } 173 174 void 175 subtractHighPrecision (uint64_t * arg1, int32_t length1, uint64_t * arg2, int32_t length2) 176 { 177 /* assumes arg1 > arg2 */ 178 int32_t index; 179 for (index = 0; index < length1; ++index) 180 arg1[index] = ~arg1[index]; 181 simpleAddHighPrecision (arg1, length1, 1); 182 183 while (length2 > 0 && arg2[length2 - 1] == 0) 184 --length2; 185 186 addHighPrecision (arg1, length1, arg2, length2); 187 188 for (index = 0; index < length1; ++index) 189 arg1[index] = ~arg1[index]; 190 simpleAddHighPrecision (arg1, length1, 1); 191 } 192 193 static uint32_t simpleMultiplyHighPrecision(uint64_t* arg1, int32_t length, uint64_t arg2) { 194 /* assumes arg2 only holds 32 bits of information */ 195 uint64_t product; 196 int32_t index; 197 198 index = 0; 199 product = 0; 200 201 do 202 { 203 product = 204 HIGH_IN_U64 (product) + arg2 * LOW_U32_FROM_PTR (arg1 + index); 205 LOW_U32_FROM_PTR (arg1 + index) = LOW_U32_FROM_VAR (product); 206 product = 207 HIGH_IN_U64 (product) + arg2 * HIGH_U32_FROM_PTR (arg1 + index); 208 HIGH_U32_FROM_PTR (arg1 + index) = LOW_U32_FROM_VAR (product); 209 } 210 while (++index < length); 211 212 return HIGH_U32_FROM_VAR (product); 213 } 214 215 static void 216 simpleMultiplyAddHighPrecision (uint64_t * arg1, int32_t length, uint64_t arg2, 217 uint32_t * result) 218 { 219 /* Assumes result can hold the product and arg2 only holds 32 bits 220 of information */ 221 uint64_t product; 222 int32_t index, resultIndex; 223 224 index = resultIndex = 0; 225 product = 0; 226 227 do 228 { 229 product = 230 HIGH_IN_U64 (product) + result[at (resultIndex)] + 231 arg2 * LOW_U32_FROM_PTR (arg1 + index); 232 result[at (resultIndex)] = LOW_U32_FROM_VAR (product); 233 ++resultIndex; 234 product = 235 HIGH_IN_U64 (product) + result[at (resultIndex)] + 236 arg2 * HIGH_U32_FROM_PTR (arg1 + index); 237 result[at (resultIndex)] = LOW_U32_FROM_VAR (product); 238 ++resultIndex; 239 } 240 while (++index < length); 241 242 result[at (resultIndex)] += HIGH_U32_FROM_VAR (product); 243 if (result[at (resultIndex)] < HIGH_U32_FROM_VAR (product)) 244 { 245 /* must be careful with ++ operator and macro expansion */ 246 ++resultIndex; 247 while (++result[at (resultIndex)] == 0) 248 ++resultIndex; 249 } 250 } 251 252 #if __BYTE_ORDER != __LITTLE_ENDIAN 253 void simpleMultiplyAddHighPrecisionBigEndianFix(uint64_t* arg1, int32_t length, uint64_t arg2, uint32_t* result) { 254 /* Assumes result can hold the product and arg2 only holds 32 bits of information */ 255 int32_t index = 0; 256 int32_t resultIndex = 0; 257 uint64_t product = 0; 258 259 do { 260 product = HIGH_IN_U64(product) + result[halfAt(resultIndex)] + arg2 * LOW_U32_FROM_PTR(arg1 + index); 261 result[halfAt(resultIndex)] = LOW_U32_FROM_VAR(product); 262 ++resultIndex; 263 product = HIGH_IN_U64(product) + result[halfAt(resultIndex)] + arg2 * HIGH_U32_FROM_PTR(arg1 + index); 264 result[halfAt(resultIndex)] = LOW_U32_FROM_VAR(product); 265 ++resultIndex; 266 } while (++index < length); 267 268 result[halfAt(resultIndex)] += HIGH_U32_FROM_VAR(product); 269 if (result[halfAt(resultIndex)] < HIGH_U32_FROM_VAR(product)) { 270 /* must be careful with ++ operator and macro expansion */ 271 ++resultIndex; 272 while (++result[halfAt(resultIndex)] == 0) ++resultIndex; 273 } 274 } 275 #endif 276 277 void 278 multiplyHighPrecision (uint64_t * arg1, int32_t length1, uint64_t * arg2, int32_t length2, 279 uint64_t * result, int32_t length) 280 { 281 /* assumes result is large enough to hold product */ 282 uint64_t* temp; 283 uint32_t* resultIn32; 284 int32_t count, index; 285 286 if (length1 < length2) 287 { 288 temp = arg1; 289 arg1 = arg2; 290 arg2 = temp; 291 count = length1; 292 length1 = length2; 293 length2 = count; 294 } 295 296 memset (result, 0, sizeof (uint64_t) * length); 297 298 /* length1 > length2 */ 299 resultIn32 = reinterpret_cast<uint32_t*>(result); 300 index = -1; 301 for (count = 0; count < length2; ++count) 302 { 303 simpleMultiplyAddHighPrecision (arg1, length1, LOW_IN_U64 (arg2[count]), 304 resultIn32 + (++index)); 305 #if __BYTE_ORDER == __LITTLE_ENDIAN 306 simpleMultiplyAddHighPrecision(arg1, length1, HIGH_IN_U64(arg2[count]), resultIn32 + (++index)); 307 #else 308 simpleMultiplyAddHighPrecisionBigEndianFix(arg1, length1, HIGH_IN_U64(arg2[count]), resultIn32 + (++index)); 309 #endif 310 } 311 } 312 313 uint32_t 314 simpleAppendDecimalDigitHighPrecision (uint64_t * arg1, int32_t length, uint64_t digit) 315 { 316 /* assumes digit is less than 32 bits */ 317 uint64_t arg; 318 int32_t index = 0; 319 320 digit <<= 32; 321 do 322 { 323 arg = LOW_IN_U64 (arg1[index]); 324 digit = HIGH_IN_U64 (digit) + TIMES_TEN (arg); 325 LOW_U32_FROM_PTR (arg1 + index) = LOW_U32_FROM_VAR (digit); 326 327 arg = HIGH_IN_U64 (arg1[index]); 328 digit = HIGH_IN_U64 (digit) + TIMES_TEN (arg); 329 HIGH_U32_FROM_PTR (arg1 + index) = LOW_U32_FROM_VAR (digit); 330 } 331 while (++index < length); 332 333 return HIGH_U32_FROM_VAR (digit); 334 } 335 336 void 337 simpleShiftLeftHighPrecision (uint64_t * arg1, int32_t length, int32_t arg2) 338 { 339 /* assumes length > 0 */ 340 int32_t index, offset; 341 if (arg2 >= 64) 342 { 343 offset = arg2 >> 6; 344 index = length; 345 346 while (--index - offset >= 0) 347 arg1[index] = arg1[index - offset]; 348 do 349 { 350 arg1[index] = 0; 351 } 352 while (--index >= 0); 353 354 arg2 &= 0x3F; 355 } 356 357 if (arg2 == 0) 358 return; 359 while (--length > 0) 360 { 361 arg1[length] = arg1[length] << arg2 | arg1[length - 1] >> (64 - arg2); 362 } 363 *arg1 <<= arg2; 364 } 365 366 int32_t 367 highestSetBit (uint64_t * y) 368 { 369 uint32_t x; 370 int32_t result; 371 372 if (*y == 0) 373 return 0; 374 375 #if defined(USE_LL) 376 if (*y & 0xFFFFFFFF00000000LL) 377 { 378 x = HIGH_U32_FROM_PTR (y); 379 result = 32; 380 } 381 else 382 { 383 x = LOW_U32_FROM_PTR (y); 384 result = 0; 385 } 386 #else 387 #if defined(USE_L) 388 if (*y & 0xFFFFFFFF00000000L) 389 { 390 x = HIGH_U32_FROM_PTR (y); 391 result = 32; 392 } 393 else 394 { 395 x = LOW_U32_FROM_PTR (y); 396 result = 0; 397 } 398 #else 399 if (*y & 0xFFFFFFFF00000000) 400 { 401 x = HIGH_U32_FROM_PTR (y); 402 result = 32; 403 } 404 else 405 { 406 x = LOW_U32_FROM_PTR (y); 407 result = 0; 408 } 409 #endif /* USE_L */ 410 #endif /* USE_LL */ 411 412 if (x & 0xFFFF0000) 413 { 414 x = bitSection (x, 0xFFFF0000, 16); 415 result += 16; 416 } 417 if (x & 0xFF00) 418 { 419 x = bitSection (x, 0xFF00, 8); 420 result += 8; 421 } 422 if (x & 0xF0) 423 { 424 x = bitSection (x, 0xF0, 4); 425 result += 4; 426 } 427 if (x > 0x7) 428 return result + 4; 429 else if (x > 0x3) 430 return result + 3; 431 else if (x > 0x1) 432 return result + 2; 433 else 434 return result + 1; 435 } 436 437 int32_t 438 lowestSetBit (uint64_t * y) 439 { 440 uint32_t x; 441 int32_t result; 442 443 if (*y == 0) 444 return 0; 445 446 #if defined(USE_LL) 447 if (*y & 0x00000000FFFFFFFFLL) 448 { 449 x = LOW_U32_FROM_PTR (y); 450 result = 0; 451 } 452 else 453 { 454 x = HIGH_U32_FROM_PTR (y); 455 result = 32; 456 } 457 #else 458 #if defined(USE_L) 459 if (*y & 0x00000000FFFFFFFFL) 460 { 461 x = LOW_U32_FROM_PTR (y); 462 result = 0; 463 } 464 else 465 { 466 x = HIGH_U32_FROM_PTR (y); 467 result = 32; 468 } 469 #else 470 if (*y & 0x00000000FFFFFFFF) 471 { 472 x = LOW_U32_FROM_PTR (y); 473 result = 0; 474 } 475 else 476 { 477 x = HIGH_U32_FROM_PTR (y); 478 result = 32; 479 } 480 #endif /* USE_L */ 481 #endif /* USE_LL */ 482 483 if (!(x & 0xFFFF)) 484 { 485 x = bitSection (x, 0xFFFF0000, 16); 486 result += 16; 487 } 488 if (!(x & 0xFF)) 489 { 490 x = bitSection (x, 0xFF00, 8); 491 result += 8; 492 } 493 if (!(x & 0xF)) 494 { 495 x = bitSection (x, 0xF0, 4); 496 result += 4; 497 } 498 499 if (x & 0x1) 500 return result + 1; 501 else if (x & 0x2) 502 return result + 2; 503 else if (x & 0x4) 504 return result + 3; 505 else 506 return result + 4; 507 } 508 509 int32_t 510 highestSetBitHighPrecision (uint64_t * arg, int32_t length) 511 { 512 int32_t highBit; 513 514 while (--length >= 0) 515 { 516 highBit = highestSetBit (arg + length); 517 if (highBit) 518 return highBit + 64 * length; 519 } 520 521 return 0; 522 } 523 524 int32_t 525 lowestSetBitHighPrecision (uint64_t * arg, int32_t length) 526 { 527 int32_t lowBit, index = -1; 528 529 while (++index < length) 530 { 531 lowBit = lowestSetBit (arg + index); 532 if (lowBit) 533 return lowBit + 64 * index; 534 } 535 536 return 0; 537 } 538 539 int32_t 540 compareHighPrecision (uint64_t * arg1, int32_t length1, uint64_t * arg2, int32_t length2) 541 { 542 while (--length1 >= 0 && arg1[length1] == 0); 543 while (--length2 >= 0 && arg2[length2] == 0); 544 545 if (length1 > length2) 546 return 1; 547 else if (length1 < length2) 548 return -1; 549 else if (length1 > -1) 550 { 551 do 552 { 553 if (arg1[length1] > arg2[length1]) 554 return 1; 555 else if (arg1[length1] < arg2[length1]) 556 return -1; 557 } 558 while (--length1 >= 0); 559 } 560 561 return 0; 562 } 563 564 jdouble 565 toDoubleHighPrecision (uint64_t * arg, int32_t length) 566 { 567 int32_t highBit; 568 uint64_t mantissa, test64; 569 uint32_t test; 570 jdouble result; 571 572 while (length > 0 && arg[length - 1] == 0) 573 --length; 574 575 if (length == 0) 576 result = 0.0; 577 else if (length > 16) 578 { 579 DOUBLE_TO_LONGBITS (result) = EXPONENT_MASK; 580 } 581 else if (length == 1) 582 { 583 highBit = highestSetBit (arg); 584 if (highBit <= 53) 585 { 586 highBit = 53 - highBit; 587 mantissa = *arg << highBit; 588 DOUBLE_TO_LONGBITS (result) = 589 CREATE_DOUBLE_BITS (mantissa, -highBit); 590 } 591 else 592 { 593 highBit -= 53; 594 mantissa = *arg >> highBit; 595 DOUBLE_TO_LONGBITS (result) = 596 CREATE_DOUBLE_BITS (mantissa, highBit); 597 598 /* perform rounding, round to even in case of tie */ 599 test = (LOW_U32_FROM_PTR (arg) << (11 - highBit)) & 0x7FF; 600 if (test > 0x400 || ((test == 0x400) && (mantissa & 1))) 601 DOUBLE_TO_LONGBITS (result) = DOUBLE_TO_LONGBITS (result) + 1; 602 } 603 } 604 else 605 { 606 highBit = highestSetBit (arg + (--length)); 607 if (highBit <= 53) 608 { 609 highBit = 53 - highBit; 610 if (highBit > 0) 611 { 612 mantissa = 613 (arg[length] << highBit) | (arg[length - 1] >> 614 (64 - highBit)); 615 } 616 else 617 { 618 mantissa = arg[length]; 619 } 620 DOUBLE_TO_LONGBITS (result) = 621 CREATE_DOUBLE_BITS (mantissa, length * 64 - highBit); 622 623 /* perform rounding, round to even in case of tie */ 624 test64 = arg[--length] << highBit; 625 if (test64 > SIGN_MASK || ((test64 == SIGN_MASK) && (mantissa & 1))) 626 DOUBLE_TO_LONGBITS (result) = DOUBLE_TO_LONGBITS (result) + 1; 627 else if (test64 == SIGN_MASK) 628 { 629 while (--length >= 0) 630 { 631 if (arg[length] != 0) 632 { 633 DOUBLE_TO_LONGBITS (result) = 634 DOUBLE_TO_LONGBITS (result) + 1; 635 break; 636 } 637 } 638 } 639 } 640 else 641 { 642 highBit -= 53; 643 mantissa = arg[length] >> highBit; 644 DOUBLE_TO_LONGBITS (result) = 645 CREATE_DOUBLE_BITS (mantissa, length * 64 + highBit); 646 647 /* perform rounding, round to even in case of tie */ 648 test = (LOW_U32_FROM_PTR (arg + length) << (11 - highBit)) & 0x7FF; 649 if (test > 0x400 || ((test == 0x400) && (mantissa & 1))) 650 DOUBLE_TO_LONGBITS (result) = DOUBLE_TO_LONGBITS (result) + 1; 651 else if (test == 0x400) 652 { 653 do 654 { 655 if (arg[--length] != 0) 656 { 657 DOUBLE_TO_LONGBITS (result) = 658 DOUBLE_TO_LONGBITS (result) + 1; 659 break; 660 } 661 } 662 while (length > 0); 663 } 664 } 665 } 666 667 return result; 668 } 669 670 static uint64_t simpleMultiplyHighPrecision64(uint64_t* arg1, int32_t length, uint64_t arg2); 671 672 int32_t 673 timesTenToTheEHighPrecision (uint64_t * result, int32_t length, jint e) 674 { 675 /* assumes result can hold value */ 676 uint64_t overflow; 677 int exp10 = e; 678 679 if (e == 0) 680 return length; 681 682 /* bad O(n) way of doing it, but simple */ 683 /* 684 do { 685 overflow = simpleAppendDecimalDigitHighPrecision(result, length, 0); 686 if (overflow) 687 result[length++] = overflow; 688 } while (--e); 689 */ 690 /* Replace the current implementaion which performs a 691 * "multiplication" by 10 e number of times with an actual 692 * multiplication. 10e19 is the largest exponent to the power of ten 693 * that will fit in a 64-bit integer, and 10e9 is the largest exponent to 694 * the power of ten that will fit in a 64-bit integer. Not sure where the 695 * break-even point is between an actual multiplication and a 696 * simpleAappendDecimalDigit() so just pick 10e3 as that point for 697 * now. 698 */ 699 while (exp10 >= 19) 700 { 701 overflow = simpleMultiplyHighPrecision64 (result, length, TEN_E19); 702 if (overflow) 703 result[length++] = overflow; 704 exp10 -= 19; 705 } 706 while (exp10 >= 9) 707 { 708 overflow = simpleMultiplyHighPrecision (result, length, TEN_E9); 709 if (overflow) 710 result[length++] = overflow; 711 exp10 -= 9; 712 } 713 if (exp10 == 0) 714 return length; 715 else if (exp10 == 1) 716 { 717 overflow = simpleAppendDecimalDigitHighPrecision (result, length, 0); 718 if (overflow) 719 result[length++] = overflow; 720 } 721 else if (exp10 == 2) 722 { 723 overflow = simpleAppendDecimalDigitHighPrecision (result, length, 0); 724 if (overflow) 725 result[length++] = overflow; 726 overflow = simpleAppendDecimalDigitHighPrecision (result, length, 0); 727 if (overflow) 728 result[length++] = overflow; 729 } 730 else if (exp10 == 3) 731 { 732 overflow = simpleMultiplyHighPrecision (result, length, TEN_E3); 733 if (overflow) 734 result[length++] = overflow; 735 } 736 else if (exp10 == 4) 737 { 738 overflow = simpleMultiplyHighPrecision (result, length, TEN_E4); 739 if (overflow) 740 result[length++] = overflow; 741 } 742 else if (exp10 == 5) 743 { 744 overflow = simpleMultiplyHighPrecision (result, length, TEN_E5); 745 if (overflow) 746 result[length++] = overflow; 747 } 748 else if (exp10 == 6) 749 { 750 overflow = simpleMultiplyHighPrecision (result, length, TEN_E6); 751 if (overflow) 752 result[length++] = overflow; 753 } 754 else if (exp10 == 7) 755 { 756 overflow = simpleMultiplyHighPrecision (result, length, TEN_E7); 757 if (overflow) 758 result[length++] = overflow; 759 } 760 else if (exp10 == 8) 761 { 762 overflow = simpleMultiplyHighPrecision (result, length, TEN_E8); 763 if (overflow) 764 result[length++] = overflow; 765 } 766 return length; 767 } 768 769 uint64_t 770 doubleMantissa (jdouble z) 771 { 772 uint64_t m = DOUBLE_TO_LONGBITS (z); 773 774 if ((m & EXPONENT_MASK) != 0) 775 m = (m & MANTISSA_MASK) | NORMAL_MASK; 776 else 777 m = (m & MANTISSA_MASK); 778 779 return m; 780 } 781 782 int32_t 783 doubleExponent (jdouble z) 784 { 785 /* assumes positive double */ 786 int32_t k = HIGH_U32_FROM_VAR (z) >> 20; 787 788 if (k) 789 k -= E_OFFSET; 790 else 791 k = 1 - E_OFFSET; 792 793 return k; 794 } 795 796 uint32_t floatMantissa(jfloat z) { 797 uint32_t m = FLOAT_TO_INTBITS (z); 798 799 if ((m & FLOAT_EXPONENT_MASK) != 0) 800 m = (m & FLOAT_MANTISSA_MASK) | FLOAT_NORMAL_MASK; 801 else 802 m = (m & FLOAT_MANTISSA_MASK); 803 804 return m; 805 } 806 807 int32_t 808 floatExponent (jfloat z) 809 { 810 /* assumes positive float */ 811 int32_t k = FLOAT_TO_INTBITS (z) >> 23; 812 if (k) 813 k -= FLOAT_E_OFFSET; 814 else 815 k = 1 - FLOAT_E_OFFSET; 816 817 return k; 818 } 819 820 /* Allow a 64-bit value in arg2 */ 821 uint64_t 822 simpleMultiplyHighPrecision64 (uint64_t * arg1, int32_t length, uint64_t arg2) 823 { 824 uint64_t intermediate, carry1, carry2, prod1, prod2, sum; 825 uint64_t* pArg1; 826 int32_t index; 827 uint32_t buf32; 828 829 index = 0; 830 intermediate = 0; 831 pArg1 = arg1 + index; 832 carry1 = carry2 = 0; 833 834 do 835 { 836 if ((*pArg1 != 0) || (intermediate != 0)) 837 { 838 prod1 = 839 static_cast<uint64_t>(LOW_U32_FROM_VAR (arg2)) * static_cast<uint64_t>(LOW_U32_FROM_PTR (pArg1)); 840 sum = intermediate + prod1; 841 if ((sum < prod1) || (sum < intermediate)) 842 { 843 carry1 = 1; 844 } 845 else 846 { 847 carry1 = 0; 848 } 849 prod1 = 850 static_cast<uint64_t>(LOW_U32_FROM_VAR (arg2)) * static_cast<uint64_t>(HIGH_U32_FROM_PTR (pArg1)); 851 prod2 = 852 static_cast<uint64_t>(HIGH_U32_FROM_VAR (arg2)) * static_cast<uint64_t>(LOW_U32_FROM_PTR (pArg1)); 853 intermediate = carry2 + HIGH_IN_U64 (sum) + prod1 + prod2; 854 if ((intermediate < prod1) || (intermediate < prod2)) 855 { 856 carry2 = 1; 857 } 858 else 859 { 860 carry2 = 0; 861 } 862 LOW_U32_FROM_PTR (pArg1) = LOW_U32_FROM_VAR (sum); 863 buf32 = HIGH_U32_FROM_PTR (pArg1); 864 HIGH_U32_FROM_PTR (pArg1) = LOW_U32_FROM_VAR (intermediate); 865 intermediate = carry1 + HIGH_IN_U64 (intermediate) 866 + static_cast<uint64_t>(HIGH_U32_FROM_VAR (arg2)) * static_cast<uint64_t>(buf32); 867 } 868 pArg1++; 869 } 870 while (++index < length); 871 return intermediate; 872 } 873