1 /* 2 * QEMU float support macros 3 * 4 * The code in this source file is derived from release 2a of the SoftFloat 5 * IEC/IEEE Floating-point Arithmetic Package. Those parts of the code (and 6 * some later contributions) are provided under that license, as detailed below. 7 * It has subsequently been modified by contributors to the QEMU Project, 8 * so some portions are provided under: 9 * the SoftFloat-2a license 10 * the BSD license 11 * GPL-v2-or-later 12 * 13 * Any future contributions to this file after December 1st 2014 will be 14 * taken to be licensed under the Softfloat-2a license unless specifically 15 * indicated otherwise. 16 */ 17 18 /* 19 =============================================================================== 20 This C source fragment is part of the SoftFloat IEC/IEEE Floating-point 21 Arithmetic Package, Release 2a. 22 23 Written by John R. Hauser. This work was made possible in part by the 24 International Computer Science Institute, located at Suite 600, 1947 Center 25 Street, Berkeley, California 94704. Funding was partially provided by the 26 National Science Foundation under grant MIP-9311980. The original version 27 of this code was written as part of a project to build a fixed-point vector 28 processor in collaboration with the University of California at Berkeley, 29 overseen by Profs. Nelson Morgan and John Wawrzynek. More information 30 is available through the Web page `http://HTTP.CS.Berkeley.EDU/~jhauser/ 31 arithmetic/SoftFloat.html'. 32 33 THIS SOFTWARE IS DISTRIBUTED AS IS, FOR FREE. Although reasonable effort 34 has been made to avoid it, THIS SOFTWARE MAY CONTAIN FAULTS THAT WILL AT 35 TIMES RESULT IN INCORRECT BEHAVIOR. USE OF THIS SOFTWARE IS RESTRICTED TO 36 PERSONS AND ORGANIZATIONS WHO CAN AND WILL TAKE FULL RESPONSIBILITY FOR ANY 37 AND ALL LOSSES, COSTS, OR OTHER PROBLEMS ARISING FROM ITS USE. 38 39 Derivative works are acceptable, even for commercial purposes, so long as 40 (1) they include prominent notice that the work is derivative, and (2) they 41 include prominent notice akin to these four paragraphs for those parts of 42 this code that are retained. 43 44 =============================================================================== 45 */ 46 47 /* BSD licensing: 48 * Copyright (c) 2006, Fabrice Bellard 49 * All rights reserved. 50 * 51 * Redistribution and use in source and binary forms, with or without 52 * modification, are permitted provided that the following conditions are met: 53 * 54 * 1. Redistributions of source code must retain the above copyright notice, 55 * this list of conditions and the following disclaimer. 56 * 57 * 2. Redistributions in binary form must reproduce the above copyright notice, 58 * this list of conditions and the following disclaimer in the documentation 59 * and/or other materials provided with the distribution. 60 * 61 * 3. Neither the name of the copyright holder nor the names of its contributors 62 * may be used to endorse or promote products derived from this software without 63 * specific prior written permission. 64 * 65 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 66 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 67 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 68 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE 69 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 70 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 71 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 72 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 73 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 74 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF 75 * THE POSSIBILITY OF SUCH DAMAGE. 76 */ 77 78 /* Portions of this work are licensed under the terms of the GNU GPL, 79 * version 2 or later. See the COPYING file in the top-level directory. 80 */ 81 82 /*---------------------------------------------------------------------------- 83 | Shifts `a' right by the number of bits given in `count'. If any nonzero 84 | bits are shifted off, they are ``jammed'' into the least significant bit of 85 | the result by setting the least significant bit to 1. The value of `count' 86 | can be arbitrarily large; in particular, if `count' is greater than 32, the 87 | result will be either 0 or 1, depending on whether `a' is zero or nonzero. 88 | The result is stored in the location pointed to by `zPtr'. 89 *----------------------------------------------------------------------------*/ 90 91 static inline void shift32RightJamming(uint32_t a, int count, uint32_t *zPtr) 92 { 93 uint32_t z; 94 95 if ( count == 0 ) { 96 z = a; 97 } 98 else if ( count < 32 ) { 99 z = ( a>>count ) | ( ( a<<( ( - count ) & 31 ) ) != 0 ); 100 } 101 else { 102 z = ( a != 0 ); 103 } 104 *zPtr = z; 105 106 } 107 108 /*---------------------------------------------------------------------------- 109 | Shifts `a' right by the number of bits given in `count'. If any nonzero 110 | bits are shifted off, they are ``jammed'' into the least significant bit of 111 | the result by setting the least significant bit to 1. The value of `count' 112 | can be arbitrarily large; in particular, if `count' is greater than 64, the 113 | result will be either 0 or 1, depending on whether `a' is zero or nonzero. 114 | The result is stored in the location pointed to by `zPtr'. 115 *----------------------------------------------------------------------------*/ 116 117 static inline void shift64RightJamming(uint64_t a, int count, uint64_t *zPtr) 118 { 119 uint64_t z; 120 121 if ( count == 0 ) { 122 z = a; 123 } 124 else if ( count < 64 ) { 125 z = ( a>>count ) | ( ( a<<( ( - count ) & 63 ) ) != 0 ); 126 } 127 else { 128 z = ( a != 0 ); 129 } 130 *zPtr = z; 131 132 } 133 134 /*---------------------------------------------------------------------------- 135 | Shifts the 128-bit value formed by concatenating `a0' and `a1' right by 64 136 | _plus_ the number of bits given in `count'. The shifted result is at most 137 | 64 nonzero bits; this is stored at the location pointed to by `z0Ptr'. The 138 | bits shifted off form a second 64-bit result as follows: The _last_ bit 139 | shifted off is the most-significant bit of the extra result, and the other 140 | 63 bits of the extra result are all zero if and only if _all_but_the_last_ 141 | bits shifted off were all zero. This extra result is stored in the location 142 | pointed to by `z1Ptr'. The value of `count' can be arbitrarily large. 143 | (This routine makes more sense if `a0' and `a1' are considered to form a 144 | fixed-point value with binary point between `a0' and `a1'. This fixed-point 145 | value is shifted right by the number of bits given in `count', and the 146 | integer part of the result is returned at the location pointed to by 147 | `z0Ptr'. The fractional part of the result may be slightly corrupted as 148 | described above, and is returned at the location pointed to by `z1Ptr'.) 149 *----------------------------------------------------------------------------*/ 150 151 static inline void 152 shift64ExtraRightJamming( 153 uint64_t a0, uint64_t a1, int count, uint64_t *z0Ptr, uint64_t *z1Ptr) 154 { 155 uint64_t z0, z1; 156 int8_t negCount = ( - count ) & 63; 157 158 if ( count == 0 ) { 159 z1 = a1; 160 z0 = a0; 161 } 162 else if ( count < 64 ) { 163 z1 = ( a0<<negCount ) | ( a1 != 0 ); 164 z0 = a0>>count; 165 } 166 else { 167 if ( count == 64 ) { 168 z1 = a0 | ( a1 != 0 ); 169 } 170 else { 171 z1 = ( ( a0 | a1 ) != 0 ); 172 } 173 z0 = 0; 174 } 175 *z1Ptr = z1; 176 *z0Ptr = z0; 177 178 } 179 180 /*---------------------------------------------------------------------------- 181 | Shifts the 128-bit value formed by concatenating `a0' and `a1' right by the 182 | number of bits given in `count'. Any bits shifted off are lost. The value 183 | of `count' can be arbitrarily large; in particular, if `count' is greater 184 | than 128, the result will be 0. The result is broken into two 64-bit pieces 185 | which are stored at the locations pointed to by `z0Ptr' and `z1Ptr'. 186 *----------------------------------------------------------------------------*/ 187 188 static inline void 189 shift128Right( 190 uint64_t a0, uint64_t a1, int count, uint64_t *z0Ptr, uint64_t *z1Ptr) 191 { 192 uint64_t z0, z1; 193 int8_t negCount = ( - count ) & 63; 194 195 if ( count == 0 ) { 196 z1 = a1; 197 z0 = a0; 198 } 199 else if ( count < 64 ) { 200 z1 = ( a0<<negCount ) | ( a1>>count ); 201 z0 = a0>>count; 202 } 203 else { 204 z1 = (count < 128) ? (a0 >> (count & 63)) : 0; 205 z0 = 0; 206 } 207 *z1Ptr = z1; 208 *z0Ptr = z0; 209 210 } 211 212 /*---------------------------------------------------------------------------- 213 | Shifts the 128-bit value formed by concatenating `a0' and `a1' right by the 214 | number of bits given in `count'. If any nonzero bits are shifted off, they 215 | are ``jammed'' into the least significant bit of the result by setting the 216 | least significant bit to 1. The value of `count' can be arbitrarily large; 217 | in particular, if `count' is greater than 128, the result will be either 218 | 0 or 1, depending on whether the concatenation of `a0' and `a1' is zero or 219 | nonzero. The result is broken into two 64-bit pieces which are stored at 220 | the locations pointed to by `z0Ptr' and `z1Ptr'. 221 *----------------------------------------------------------------------------*/ 222 223 static inline void 224 shift128RightJamming( 225 uint64_t a0, uint64_t a1, int count, uint64_t *z0Ptr, uint64_t *z1Ptr) 226 { 227 uint64_t z0, z1; 228 int8_t negCount = ( - count ) & 63; 229 230 if ( count == 0 ) { 231 z1 = a1; 232 z0 = a0; 233 } 234 else if ( count < 64 ) { 235 z1 = ( a0<<negCount ) | ( a1>>count ) | ( ( a1<<negCount ) != 0 ); 236 z0 = a0>>count; 237 } 238 else { 239 if ( count == 64 ) { 240 z1 = a0 | ( a1 != 0 ); 241 } 242 else if ( count < 128 ) { 243 z1 = ( a0>>( count & 63 ) ) | ( ( ( a0<<negCount ) | a1 ) != 0 ); 244 } 245 else { 246 z1 = ( ( a0 | a1 ) != 0 ); 247 } 248 z0 = 0; 249 } 250 *z1Ptr = z1; 251 *z0Ptr = z0; 252 253 } 254 255 /*---------------------------------------------------------------------------- 256 | Shifts the 192-bit value formed by concatenating `a0', `a1', and `a2' right 257 | by 64 _plus_ the number of bits given in `count'. The shifted result is 258 | at most 128 nonzero bits; these are broken into two 64-bit pieces which are 259 | stored at the locations pointed to by `z0Ptr' and `z1Ptr'. The bits shifted 260 | off form a third 64-bit result as follows: The _last_ bit shifted off is 261 | the most-significant bit of the extra result, and the other 63 bits of the 262 | extra result are all zero if and only if _all_but_the_last_ bits shifted off 263 | were all zero. This extra result is stored in the location pointed to by 264 | `z2Ptr'. The value of `count' can be arbitrarily large. 265 | (This routine makes more sense if `a0', `a1', and `a2' are considered 266 | to form a fixed-point value with binary point between `a1' and `a2'. This 267 | fixed-point value is shifted right by the number of bits given in `count', 268 | and the integer part of the result is returned at the locations pointed to 269 | by `z0Ptr' and `z1Ptr'. The fractional part of the result may be slightly 270 | corrupted as described above, and is returned at the location pointed to by 271 | `z2Ptr'.) 272 *----------------------------------------------------------------------------*/ 273 274 static inline void 275 shift128ExtraRightJamming( 276 uint64_t a0, 277 uint64_t a1, 278 uint64_t a2, 279 int count, 280 uint64_t *z0Ptr, 281 uint64_t *z1Ptr, 282 uint64_t *z2Ptr 283 ) 284 { 285 uint64_t z0, z1, z2; 286 int8_t negCount = ( - count ) & 63; 287 288 if ( count == 0 ) { 289 z2 = a2; 290 z1 = a1; 291 z0 = a0; 292 } 293 else { 294 if ( count < 64 ) { 295 z2 = a1<<negCount; 296 z1 = ( a0<<negCount ) | ( a1>>count ); 297 z0 = a0>>count; 298 } 299 else { 300 if ( count == 64 ) { 301 z2 = a1; 302 z1 = a0; 303 } 304 else { 305 a2 |= a1; 306 if ( count < 128 ) { 307 z2 = a0<<negCount; 308 z1 = a0>>( count & 63 ); 309 } 310 else { 311 z2 = ( count == 128 ) ? a0 : ( a0 != 0 ); 312 z1 = 0; 313 } 314 } 315 z0 = 0; 316 } 317 z2 |= ( a2 != 0 ); 318 } 319 *z2Ptr = z2; 320 *z1Ptr = z1; 321 *z0Ptr = z0; 322 323 } 324 325 /*---------------------------------------------------------------------------- 326 | Shifts the 128-bit value formed by concatenating `a0' and `a1' left by the 327 | number of bits given in `count'. Any bits shifted off are lost. The value 328 | of `count' must be less than 64. The result is broken into two 64-bit 329 | pieces which are stored at the locations pointed to by `z0Ptr' and `z1Ptr'. 330 *----------------------------------------------------------------------------*/ 331 332 static inline void shortShift128Left(uint64_t a0, uint64_t a1, int count, 333 uint64_t *z0Ptr, uint64_t *z1Ptr) 334 { 335 *z1Ptr = a1 << count; 336 *z0Ptr = count == 0 ? a0 : (a0 << count) | (a1 >> (-count & 63)); 337 } 338 339 /*---------------------------------------------------------------------------- 340 | Shifts the 128-bit value formed by concatenating `a0' and `a1' left by the 341 | number of bits given in `count'. Any bits shifted off are lost. The value 342 | of `count' may be greater than 64. The result is broken into two 64-bit 343 | pieces which are stored at the locations pointed to by `z0Ptr' and `z1Ptr'. 344 *----------------------------------------------------------------------------*/ 345 346 static inline void shift128Left(uint64_t a0, uint64_t a1, int count, 347 uint64_t *z0Ptr, uint64_t *z1Ptr) 348 { 349 if (count < 64) { 350 *z1Ptr = a1 << count; 351 *z0Ptr = count == 0 ? a0 : (a0 << count) | (a1 >> (-count & 63)); 352 } else { 353 *z1Ptr = 0; 354 *z0Ptr = a1 << (count - 64); 355 } 356 } 357 358 /*---------------------------------------------------------------------------- 359 | Shifts the 192-bit value formed by concatenating `a0', `a1', and `a2' left 360 | by the number of bits given in `count'. Any bits shifted off are lost. 361 | The value of `count' must be less than 64. The result is broken into three 362 | 64-bit pieces which are stored at the locations pointed to by `z0Ptr', 363 | `z1Ptr', and `z2Ptr'. 364 *----------------------------------------------------------------------------*/ 365 366 static inline void 367 shortShift192Left( 368 uint64_t a0, 369 uint64_t a1, 370 uint64_t a2, 371 int count, 372 uint64_t *z0Ptr, 373 uint64_t *z1Ptr, 374 uint64_t *z2Ptr 375 ) 376 { 377 uint64_t z0, z1, z2; 378 int8_t negCount; 379 380 z2 = a2<<count; 381 z1 = a1<<count; 382 z0 = a0<<count; 383 if ( 0 < count ) { 384 negCount = ( ( - count ) & 63 ); 385 z1 |= a2>>negCount; 386 z0 |= a1>>negCount; 387 } 388 *z2Ptr = z2; 389 *z1Ptr = z1; 390 *z0Ptr = z0; 391 392 } 393 394 /*---------------------------------------------------------------------------- 395 | Adds the 128-bit value formed by concatenating `a0' and `a1' to the 128-bit 396 | value formed by concatenating `b0' and `b1'. Addition is modulo 2^128, so 397 | any carry out is lost. The result is broken into two 64-bit pieces which 398 | are stored at the locations pointed to by `z0Ptr' and `z1Ptr'. 399 *----------------------------------------------------------------------------*/ 400 401 static inline void 402 add128( 403 uint64_t a0, uint64_t a1, uint64_t b0, uint64_t b1, uint64_t *z0Ptr, uint64_t *z1Ptr ) 404 { 405 uint64_t z1; 406 407 z1 = a1 + b1; 408 *z1Ptr = z1; 409 *z0Ptr = a0 + b0 + ( z1 < a1 ); 410 411 } 412 413 /*---------------------------------------------------------------------------- 414 | Adds the 192-bit value formed by concatenating `a0', `a1', and `a2' to the 415 | 192-bit value formed by concatenating `b0', `b1', and `b2'. Addition is 416 | modulo 2^192, so any carry out is lost. The result is broken into three 417 | 64-bit pieces which are stored at the locations pointed to by `z0Ptr', 418 | `z1Ptr', and `z2Ptr'. 419 *----------------------------------------------------------------------------*/ 420 421 static inline void 422 add192( 423 uint64_t a0, 424 uint64_t a1, 425 uint64_t a2, 426 uint64_t b0, 427 uint64_t b1, 428 uint64_t b2, 429 uint64_t *z0Ptr, 430 uint64_t *z1Ptr, 431 uint64_t *z2Ptr 432 ) 433 { 434 uint64_t z0, z1, z2; 435 int8_t carry0, carry1; 436 437 z2 = a2 + b2; 438 carry1 = ( z2 < a2 ); 439 z1 = a1 + b1; 440 carry0 = ( z1 < a1 ); 441 z0 = a0 + b0; 442 z1 += carry1; 443 z0 += ( z1 < carry1 ); 444 z0 += carry0; 445 *z2Ptr = z2; 446 *z1Ptr = z1; 447 *z0Ptr = z0; 448 449 } 450 451 /*---------------------------------------------------------------------------- 452 | Subtracts the 128-bit value formed by concatenating `b0' and `b1' from the 453 | 128-bit value formed by concatenating `a0' and `a1'. Subtraction is modulo 454 | 2^128, so any borrow out (carry out) is lost. The result is broken into two 455 | 64-bit pieces which are stored at the locations pointed to by `z0Ptr' and 456 | `z1Ptr'. 457 *----------------------------------------------------------------------------*/ 458 459 static inline void 460 sub128( 461 uint64_t a0, uint64_t a1, uint64_t b0, uint64_t b1, uint64_t *z0Ptr, uint64_t *z1Ptr ) 462 { 463 464 *z1Ptr = a1 - b1; 465 *z0Ptr = a0 - b0 - ( a1 < b1 ); 466 467 } 468 469 /*---------------------------------------------------------------------------- 470 | Subtracts the 192-bit value formed by concatenating `b0', `b1', and `b2' 471 | from the 192-bit value formed by concatenating `a0', `a1', and `a2'. 472 | Subtraction is modulo 2^192, so any borrow out (carry out) is lost. The 473 | result is broken into three 64-bit pieces which are stored at the locations 474 | pointed to by `z0Ptr', `z1Ptr', and `z2Ptr'. 475 *----------------------------------------------------------------------------*/ 476 477 static inline void 478 sub192( 479 uint64_t a0, 480 uint64_t a1, 481 uint64_t a2, 482 uint64_t b0, 483 uint64_t b1, 484 uint64_t b2, 485 uint64_t *z0Ptr, 486 uint64_t *z1Ptr, 487 uint64_t *z2Ptr 488 ) 489 { 490 uint64_t z0, z1, z2; 491 int8_t borrow0, borrow1; 492 493 z2 = a2 - b2; 494 borrow1 = ( a2 < b2 ); 495 z1 = a1 - b1; 496 borrow0 = ( a1 < b1 ); 497 z0 = a0 - b0; 498 z0 -= ( z1 < borrow1 ); 499 z1 -= borrow1; 500 z0 -= borrow0; 501 *z2Ptr = z2; 502 *z1Ptr = z1; 503 *z0Ptr = z0; 504 505 } 506 507 /*---------------------------------------------------------------------------- 508 | Multiplies `a' by `b' to obtain a 128-bit product. The product is broken 509 | into two 64-bit pieces which are stored at the locations pointed to by 510 | `z0Ptr' and `z1Ptr'. 511 *----------------------------------------------------------------------------*/ 512 513 static inline void mul64To128( uint64_t a, uint64_t b, uint64_t *z0Ptr, uint64_t *z1Ptr ) 514 { 515 uint32_t aHigh, aLow, bHigh, bLow; 516 uint64_t z0, zMiddleA, zMiddleB, z1; 517 518 aLow = a; 519 aHigh = a>>32; 520 bLow = b; 521 bHigh = b>>32; 522 z1 = ( (uint64_t) aLow ) * bLow; 523 zMiddleA = ( (uint64_t) aLow ) * bHigh; 524 zMiddleB = ( (uint64_t) aHigh ) * bLow; 525 z0 = ( (uint64_t) aHigh ) * bHigh; 526 zMiddleA += zMiddleB; 527 z0 += ( ( (uint64_t) ( zMiddleA < zMiddleB ) )<<32 ) + ( zMiddleA>>32 ); 528 zMiddleA <<= 32; 529 z1 += zMiddleA; 530 z0 += ( z1 < zMiddleA ); 531 *z1Ptr = z1; 532 *z0Ptr = z0; 533 534 } 535 536 /*---------------------------------------------------------------------------- 537 | Multiplies the 128-bit value formed by concatenating `a0' and `a1' by 538 | `b' to obtain a 192-bit product. The product is broken into three 64-bit 539 | pieces which are stored at the locations pointed to by `z0Ptr', `z1Ptr', and 540 | `z2Ptr'. 541 *----------------------------------------------------------------------------*/ 542 543 static inline void 544 mul128By64To192( 545 uint64_t a0, 546 uint64_t a1, 547 uint64_t b, 548 uint64_t *z0Ptr, 549 uint64_t *z1Ptr, 550 uint64_t *z2Ptr 551 ) 552 { 553 uint64_t z0, z1, z2, more1; 554 555 mul64To128( a1, b, &z1, &z2 ); 556 mul64To128( a0, b, &z0, &more1 ); 557 add128( z0, more1, 0, z1, &z0, &z1 ); 558 *z2Ptr = z2; 559 *z1Ptr = z1; 560 *z0Ptr = z0; 561 562 } 563 564 /*---------------------------------------------------------------------------- 565 | Multiplies the 128-bit value formed by concatenating `a0' and `a1' to the 566 | 128-bit value formed by concatenating `b0' and `b1' to obtain a 256-bit 567 | product. The product is broken into four 64-bit pieces which are stored at 568 | the locations pointed to by `z0Ptr', `z1Ptr', `z2Ptr', and `z3Ptr'. 569 *----------------------------------------------------------------------------*/ 570 571 static inline void 572 mul128To256( 573 uint64_t a0, 574 uint64_t a1, 575 uint64_t b0, 576 uint64_t b1, 577 uint64_t *z0Ptr, 578 uint64_t *z1Ptr, 579 uint64_t *z2Ptr, 580 uint64_t *z3Ptr 581 ) 582 { 583 uint64_t z0, z1, z2, z3; 584 uint64_t more1, more2; 585 586 mul64To128( a1, b1, &z2, &z3 ); 587 mul64To128( a1, b0, &z1, &more2 ); 588 add128( z1, more2, 0, z2, &z1, &z2 ); 589 mul64To128( a0, b0, &z0, &more1 ); 590 add128( z0, more1, 0, z1, &z0, &z1 ); 591 mul64To128( a0, b1, &more1, &more2 ); 592 add128( more1, more2, 0, z2, &more1, &z2 ); 593 add128( z0, z1, 0, more1, &z0, &z1 ); 594 *z3Ptr = z3; 595 *z2Ptr = z2; 596 *z1Ptr = z1; 597 *z0Ptr = z0; 598 599 } 600 601 /*---------------------------------------------------------------------------- 602 | Returns an approximation to the 64-bit integer quotient obtained by dividing 603 | `b' into the 128-bit value formed by concatenating `a0' and `a1'. The 604 | divisor `b' must be at least 2^63. If q is the exact quotient truncated 605 | toward zero, the approximation returned lies between q and q + 2 inclusive. 606 | If the exact quotient q is larger than 64 bits, the maximum positive 64-bit 607 | unsigned integer is returned. 608 *----------------------------------------------------------------------------*/ 609 610 static inline uint64_t estimateDiv128To64(uint64_t a0, uint64_t a1, uint64_t b) 611 { 612 uint64_t b0, b1; 613 uint64_t rem0, rem1, term0, term1; 614 uint64_t z; 615 616 if ( b <= a0 ) return LIT64( 0xFFFFFFFFFFFFFFFF ); 617 b0 = b>>32; 618 z = ( b0<<32 <= a0 ) ? LIT64( 0xFFFFFFFF00000000 ) : ( a0 / b0 )<<32; 619 mul64To128( b, z, &term0, &term1 ); 620 sub128( a0, a1, term0, term1, &rem0, &rem1 ); 621 while ( ( (int64_t) rem0 ) < 0 ) { 622 z -= LIT64( 0x100000000 ); 623 b1 = b<<32; 624 add128( rem0, rem1, b0, b1, &rem0, &rem1 ); 625 } 626 rem0 = ( rem0<<32 ) | ( rem1>>32 ); 627 z |= ( b0<<32 <= rem0 ) ? 0xFFFFFFFF : rem0 / b0; 628 return z; 629 630 } 631 632 /* From the GNU Multi Precision Library - longlong.h __udiv_qrnnd 633 * (https://gmplib.org/repo/gmp/file/tip/longlong.h) 634 * 635 * Licensed under the GPLv2/LGPLv3 636 */ 637 static inline uint64_t udiv_qrnnd(uint64_t *r, uint64_t n1, 638 uint64_t n0, uint64_t d) 639 { 640 #if defined(__x86_64__) 641 uint64_t q; 642 asm("divq %4" : "=a"(q), "=d"(*r) : "0"(n0), "1"(n1), "rm"(d)); 643 return q; 644 #elif defined(__s390x__) && !defined(__clang__) 645 /* Need to use a TImode type to get an even register pair for DLGR. */ 646 unsigned __int128 n = (unsigned __int128)n1 << 64 | n0; 647 asm("dlgr %0, %1" : "+r"(n) : "r"(d)); 648 *r = n >> 64; 649 return n; 650 #elif defined(_ARCH_PPC64) && defined(_ARCH_PWR7) 651 /* From Power ISA 2.06, programming note for divdeu. */ 652 uint64_t q1, q2, Q, r1, r2, R; 653 asm("divdeu %0,%2,%4; divdu %1,%3,%4" 654 : "=&r"(q1), "=r"(q2) 655 : "r"(n1), "r"(n0), "r"(d)); 656 r1 = -(q1 * d); /* low part of (n1<<64) - (q1 * d) */ 657 r2 = n0 - (q2 * d); 658 Q = q1 + q2; 659 R = r1 + r2; 660 if (R >= d || R < r2) { /* overflow implies R > d */ 661 Q += 1; 662 R -= d; 663 } 664 *r = R; 665 return Q; 666 #else 667 uint64_t d0, d1, q0, q1, r1, r0, m; 668 669 d0 = (uint32_t)d; 670 d1 = d >> 32; 671 672 r1 = n1 % d1; 673 q1 = n1 / d1; 674 m = q1 * d0; 675 r1 = (r1 << 32) | (n0 >> 32); 676 if (r1 < m) { 677 q1 -= 1; 678 r1 += d; 679 if (r1 >= d) { 680 if (r1 < m) { 681 q1 -= 1; 682 r1 += d; 683 } 684 } 685 } 686 r1 -= m; 687 688 r0 = r1 % d1; 689 q0 = r1 / d1; 690 m = q0 * d0; 691 r0 = (r0 << 32) | (uint32_t)n0; 692 if (r0 < m) { 693 q0 -= 1; 694 r0 += d; 695 if (r0 >= d) { 696 if (r0 < m) { 697 q0 -= 1; 698 r0 += d; 699 } 700 } 701 } 702 r0 -= m; 703 704 *r = r0; 705 return (q1 << 32) | q0; 706 #endif 707 } 708 709 /*---------------------------------------------------------------------------- 710 | Returns an approximation to the square root of the 32-bit significand given 711 | by `a'. Considered as an integer, `a' must be at least 2^31. If bit 0 of 712 | `aExp' (the least significant bit) is 1, the integer returned approximates 713 | 2^31*sqrt(`a'/2^31), where `a' is considered an integer. If bit 0 of `aExp' 714 | is 0, the integer returned approximates 2^31*sqrt(`a'/2^30). In either 715 | case, the approximation returned lies strictly within +/-2 of the exact 716 | value. 717 *----------------------------------------------------------------------------*/ 718 719 static inline uint32_t estimateSqrt32(int aExp, uint32_t a) 720 { 721 static const uint16_t sqrtOddAdjustments[] = { 722 0x0004, 0x0022, 0x005D, 0x00B1, 0x011D, 0x019F, 0x0236, 0x02E0, 723 0x039C, 0x0468, 0x0545, 0x0631, 0x072B, 0x0832, 0x0946, 0x0A67 724 }; 725 static const uint16_t sqrtEvenAdjustments[] = { 726 0x0A2D, 0x08AF, 0x075A, 0x0629, 0x051A, 0x0429, 0x0356, 0x029E, 727 0x0200, 0x0179, 0x0109, 0x00AF, 0x0068, 0x0034, 0x0012, 0x0002 728 }; 729 int8_t index; 730 uint32_t z; 731 732 index = ( a>>27 ) & 15; 733 if ( aExp & 1 ) { 734 z = 0x4000 + ( a>>17 ) - sqrtOddAdjustments[ (int)index ]; 735 z = ( ( a / z )<<14 ) + ( z<<15 ); 736 a >>= 1; 737 } 738 else { 739 z = 0x8000 + ( a>>17 ) - sqrtEvenAdjustments[ (int)index ]; 740 z = a / z + z; 741 z = ( 0x20000 <= z ) ? 0xFFFF8000 : ( z<<15 ); 742 if ( z <= a ) return (uint32_t) ( ( (int32_t) a )>>1 ); 743 } 744 return ( (uint32_t) ( ( ( (uint64_t) a )<<31 ) / z ) ) + ( z>>1 ); 745 746 } 747 748 /*---------------------------------------------------------------------------- 749 | Returns 1 if the 128-bit value formed by concatenating `a0' and `a1' 750 | is equal to the 128-bit value formed by concatenating `b0' and `b1'. 751 | Otherwise, returns 0. 752 *----------------------------------------------------------------------------*/ 753 754 static inline flag eq128( uint64_t a0, uint64_t a1, uint64_t b0, uint64_t b1 ) 755 { 756 757 return ( a0 == b0 ) && ( a1 == b1 ); 758 759 } 760 761 /*---------------------------------------------------------------------------- 762 | Returns 1 if the 128-bit value formed by concatenating `a0' and `a1' is less 763 | than or equal to the 128-bit value formed by concatenating `b0' and `b1'. 764 | Otherwise, returns 0. 765 *----------------------------------------------------------------------------*/ 766 767 static inline flag le128( uint64_t a0, uint64_t a1, uint64_t b0, uint64_t b1 ) 768 { 769 770 return ( a0 < b0 ) || ( ( a0 == b0 ) && ( a1 <= b1 ) ); 771 772 } 773 774 /*---------------------------------------------------------------------------- 775 | Returns 1 if the 128-bit value formed by concatenating `a0' and `a1' is less 776 | than the 128-bit value formed by concatenating `b0' and `b1'. Otherwise, 777 | returns 0. 778 *----------------------------------------------------------------------------*/ 779 780 static inline flag lt128( uint64_t a0, uint64_t a1, uint64_t b0, uint64_t b1 ) 781 { 782 783 return ( a0 < b0 ) || ( ( a0 == b0 ) && ( a1 < b1 ) ); 784 785 } 786 787 /*---------------------------------------------------------------------------- 788 | Returns 1 if the 128-bit value formed by concatenating `a0' and `a1' is 789 | not equal to the 128-bit value formed by concatenating `b0' and `b1'. 790 | Otherwise, returns 0. 791 *----------------------------------------------------------------------------*/ 792 793 static inline flag ne128( uint64_t a0, uint64_t a1, uint64_t b0, uint64_t b1 ) 794 { 795 796 return ( a0 != b0 ) || ( a1 != b1 ); 797 798 } 799