Annotation of embedaddon/php/Zend/zend_strtod.c, revision 1.1
1.1 ! misho 1: /****************************************************************
! 2: *
! 3: * The author of this software is David M. Gay.
! 4: *
! 5: * Copyright (c) 1991 by AT&T.
! 6: *
! 7: * Permission to use, copy, modify, and distribute this software for any
! 8: * purpose without fee is hereby granted, provided that this entire notice
! 9: * is included in all copies of any software which is or includes a copy
! 10: * or modification of this software and in all copies of the supporting
! 11: * documentation for such software.
! 12: *
! 13: * THIS SOFTWARE IS BEING PROVIDED "AS IS", WITHOUT ANY EXPRESS OR IMPLIED
! 14: * WARRANTY. IN PARTICULAR, NEITHER THE AUTHOR NOR AT&T MAKES ANY
! 15: * REPRESENTATION OR WARRANTY OF ANY KIND CONCERNING THE MERCHANTABILITY
! 16: * OF THIS SOFTWARE OR ITS FITNESS FOR ANY PARTICULAR PURPOSE.
! 17: *
! 18: ***************************************************************/
! 19:
! 20: /* Please send bug reports to
! 21: David M. Gay
! 22: AT&T Bell Laboratories, Room 2C-463
! 23: 600 Mountain Avenue
! 24: Murray Hill, NJ 07974-2070
! 25: U.S.A.
! 26: dmg@research.att.com or research!dmg
! 27: */
! 28:
! 29: /* strtod for IEEE-, VAX-, and IBM-arithmetic machines.
! 30: *
! 31: * This strtod returns a nearest machine number to the input decimal
! 32: * string (or sets errno to ERANGE). With IEEE arithmetic, ties are
! 33: * broken by the IEEE round-even rule. Otherwise ties are broken by
! 34: * biased rounding (add half and chop).
! 35: *
! 36: * Inspired loosely by William D. Clinger's paper "How to Read Floating
! 37: * Point Numbers Accurately" [Proc. ACM SIGPLAN '90, pp. 92-101].
! 38: *
! 39: * Modifications:
! 40: *
! 41: * 1. We only require IEEE, IBM, or VAX double-precision
! 42: * arithmetic (not IEEE double-extended).
! 43: * 2. We get by with floating-point arithmetic in a case that
! 44: * Clinger missed -- when we're computing d * 10^n
! 45: * for a small integer d and the integer n is not too
! 46: * much larger than 22 (the maximum integer k for which
! 47: * we can represent 10^k exactly), we may be able to
! 48: * compute (d*10^k) * 10^(e-k) with just one roundoff.
! 49: * 3. Rather than a bit-at-a-time adjustment of the binary
! 50: * result in the hard case, we use floating-point
! 51: * arithmetic to determine the adjustment to within
! 52: * one bit; only in really hard cases do we need to
! 53: * compute a second residual.
! 54: * 4. Because of 3., we don't need a large table of powers of 10
! 55: * for ten-to-e (just some small tables, e.g. of 10^k
! 56: * for 0 <= k <= 22).
! 57: */
! 58:
! 59: /*
! 60: * #define IEEE_LITTLE_ENDIAN for IEEE-arithmetic machines where the least
! 61: * significant byte has the lowest address.
! 62: * #define IEEE_BIG_ENDIAN for IEEE-arithmetic machines where the most
! 63: * significant byte has the lowest address.
! 64: * #define Long int on machines with 32-bit ints and 64-bit longs.
! 65: * #define Sudden_Underflow for IEEE-format machines without gradual
! 66: * underflow (i.e., that flush to zero on underflow).
! 67: * #define IBM for IBM mainframe-style floating-point arithmetic.
! 68: * #define VAX for VAX-style floating-point arithmetic.
! 69: * #define Unsigned_Shifts if >> does treats its left operand as unsigned.
! 70: * #define No_leftright to omit left-right logic in fast floating-point
! 71: * computation of dtoa.
! 72: * #define Check_FLT_ROUNDS if FLT_ROUNDS can assume the values 2 or 3.
! 73: * #define RND_PRODQUOT to use rnd_prod and rnd_quot (assembly routines
! 74: * that use extended-precision instructions to compute rounded
! 75: * products and quotients) with IBM.
! 76: * #define ROUND_BIASED for IEEE-format with biased rounding.
! 77: * #define Inaccurate_Divide for IEEE-format with correctly rounded
! 78: * products but inaccurate quotients, e.g., for Intel i860.
! 79: * #define Just_16 to store 16 bits per 32-bit Long when doing high-precision
! 80: * integer arithmetic. Whether this speeds things up or slows things
! 81: * down depends on the machine and the number being converted.
! 82: * #define KR_headers for old-style C function headers.
! 83: * #define Bad_float_h if your system lacks a float.h or if it does not
! 84: * define some or all of DBL_DIG, DBL_MAX_10_EXP, DBL_MAX_EXP,
! 85: * FLT_RADIX, FLT_ROUNDS, and DBL_MAX.
! 86: * #define MALLOC your_malloc, where your_malloc(n) acts like malloc(n)
! 87: * if memory is available and otherwise does something you deem
! 88: * appropriate. If MALLOC is undefined, malloc will be invoked
! 89: * directly -- and assumed always to succeed.
! 90: */
! 91:
! 92: /* $Id: zend_strtod.c 316591 2011-09-13 07:07:06Z dmitry $ */
! 93:
! 94: #include <zend_operators.h>
! 95: #include <zend_strtod.h>
! 96:
! 97: #ifdef ZTS
! 98: #include <TSRM.h>
! 99: #endif
! 100:
! 101: #include <stddef.h>
! 102: #include <stdio.h>
! 103: #include <ctype.h>
! 104: #include <stdarg.h>
! 105: #include <string.h>
! 106: #include <stdlib.h>
! 107: #include <math.h>
! 108:
! 109: #ifdef HAVE_LOCALE_H
! 110: #include <locale.h>
! 111: #endif
! 112:
! 113: #ifdef HAVE_SYS_TYPES_H
! 114: #include <sys/types.h>
! 115: #endif
! 116:
! 117: #if defined(HAVE_INTTYPES_H)
! 118: #include <inttypes.h>
! 119: #elif defined(HAVE_STDINT_H)
! 120: #include <stdint.h>
! 121: #endif
! 122:
! 123: #ifndef HAVE_INT32_T
! 124: # if SIZEOF_INT == 4
! 125: typedef int int32_t;
! 126: # elif SIZEOF_LONG == 4
! 127: typedef long int int32_t;
! 128: # endif
! 129: #endif
! 130:
! 131: #ifndef HAVE_UINT32_T
! 132: # if SIZEOF_INT == 4
! 133: typedef unsigned int uint32_t;
! 134: # elif SIZEOF_LONG == 4
! 135: typedef unsigned long int uint32_t;
! 136: # endif
! 137: #endif
! 138:
! 139: #if (defined(__APPLE__) || defined(__APPLE_CC__)) && (defined(__BIG_ENDIAN__) || defined(__LITTLE_ENDIAN__))
! 140: # if defined(__LITTLE_ENDIAN__)
! 141: # undef WORDS_BIGENDIAN
! 142: # else
! 143: # if defined(__BIG_ENDIAN__)
! 144: # define WORDS_BIGENDIAN
! 145: # endif
! 146: # endif
! 147: #endif
! 148:
! 149: #ifdef WORDS_BIGENDIAN
! 150: #define IEEE_BIG_ENDIAN
! 151: #else
! 152: #define IEEE_LITTLE_ENDIAN
! 153: #endif
! 154:
! 155: #if defined(__arm__) && !defined(__VFP_FP__)
! 156: /*
! 157: * * Although the CPU is little endian the FP has different
! 158: * * byte and word endianness. The byte order is still little endian
! 159: * * but the word order is big endian.
! 160: * */
! 161: #define IEEE_BIG_ENDIAN
! 162: #undef IEEE_LITTLE_ENDIAN
! 163: #endif
! 164:
! 165: #ifdef __vax__
! 166: #define VAX
! 167: #undef IEEE_LITTLE_ENDIAN
! 168: #endif
! 169:
! 170: #if defined(_MSC_VER)
! 171: #define int32_t __int32
! 172: #define uint32_t unsigned __int32
! 173: #define IEEE_LITTLE_ENDIAN
! 174: #endif
! 175:
! 176: #define Long int32_t
! 177: #define ULong uint32_t
! 178:
! 179: #ifdef __cplusplus
! 180: #include "malloc.h"
! 181: #include "memory.h"
! 182: #else
! 183: #ifndef KR_headers
! 184: #include "stdlib.h"
! 185: #include "string.h"
! 186: #include "locale.h"
! 187: #else
! 188: #include "malloc.h"
! 189: #include "memory.h"
! 190: #endif
! 191: #endif
! 192:
! 193: #ifdef MALLOC
! 194: #ifdef KR_headers
! 195: extern char *MALLOC();
! 196: #else
! 197: extern void *MALLOC(size_t);
! 198: #endif
! 199: #else
! 200: #define MALLOC malloc
! 201: #endif
! 202:
! 203: #include "ctype.h"
! 204: #include "errno.h"
! 205:
! 206: #ifdef Bad_float_h
! 207: #ifdef IEEE_BIG_ENDIAN
! 208: #define IEEE_ARITHMETIC
! 209: #endif
! 210: #ifdef IEEE_LITTLE_ENDIAN
! 211: #define IEEE_ARITHMETIC
! 212: #endif
! 213:
! 214: #ifdef IEEE_ARITHMETIC
! 215: #define DBL_DIG 15
! 216: #define DBL_MAX_10_EXP 308
! 217: #define DBL_MAX_EXP 1024
! 218: #define FLT_RADIX 2
! 219: #define FLT_ROUNDS 1
! 220: #define DBL_MAX 1.7976931348623157e+308
! 221: #endif
! 222:
! 223: #ifdef IBM
! 224: #define DBL_DIG 16
! 225: #define DBL_MAX_10_EXP 75
! 226: #define DBL_MAX_EXP 63
! 227: #define FLT_RADIX 16
! 228: #define FLT_ROUNDS 0
! 229: #define DBL_MAX 7.2370055773322621e+75
! 230: #endif
! 231:
! 232: #ifdef VAX
! 233: #define DBL_DIG 16
! 234: #define DBL_MAX_10_EXP 38
! 235: #define DBL_MAX_EXP 127
! 236: #define FLT_RADIX 2
! 237: #define FLT_ROUNDS 1
! 238: #define DBL_MAX 1.7014118346046923e+38
! 239: #endif
! 240:
! 241:
! 242: #ifndef LONG_MAX
! 243: #define LONG_MAX 2147483647
! 244: #endif
! 245: #else
! 246: #include "float.h"
! 247: #endif
! 248: #ifndef __MATH_H__
! 249: #include "math.h"
! 250: #endif
! 251:
! 252: BEGIN_EXTERN_C()
! 253:
! 254: #ifndef CONST
! 255: #ifdef KR_headers
! 256: #define CONST /* blank */
! 257: #else
! 258: #define CONST const
! 259: #endif
! 260: #endif
! 261:
! 262: #ifdef Unsigned_Shifts
! 263: #define Sign_Extend(a,b) if (b < 0) a |= 0xffff0000;
! 264: #else
! 265: #define Sign_Extend(a,b) /*no-op*/
! 266: #endif
! 267:
! 268: #if defined(IEEE_LITTLE_ENDIAN) + defined(IEEE_BIG_ENDIAN) + defined(VAX) + \
! 269: defined(IBM) != 1
! 270: Exactly one of IEEE_LITTLE_ENDIAN IEEE_BIG_ENDIAN, VAX, or
! 271: IBM should be defined.
! 272: #endif
! 273:
! 274: typedef union {
! 275: double d;
! 276: ULong ul[2];
! 277: } _double;
! 278: #define value(x) ((x).d)
! 279: #ifdef IEEE_LITTLE_ENDIAN
! 280: #define word0(x) ((x).ul[1])
! 281: #define word1(x) ((x).ul[0])
! 282: #else
! 283: #define word0(x) ((x).ul[0])
! 284: #define word1(x) ((x).ul[1])
! 285: #endif
! 286:
! 287: /* The following definition of Storeinc is appropriate for MIPS processors.
! 288: * An alternative that might be better on some machines is
! 289: * #define Storeinc(a,b,c) (*a++ = b << 16 | c & 0xffff)
! 290: */
! 291: #if defined(IEEE_LITTLE_ENDIAN) + defined(VAX) + defined(__arm__)
! 292: #define Storeinc(a,b,c) (((unsigned short *)a)[1] = (unsigned short)b, \
! 293: ((unsigned short *)a)[0] = (unsigned short)c, a++)
! 294: #else
! 295: #define Storeinc(a,b,c) (((unsigned short *)a)[0] = (unsigned short)b, \
! 296: ((unsigned short *)a)[1] = (unsigned short)c, a++)
! 297: #endif
! 298:
! 299: /* #define P DBL_MANT_DIG */
! 300: /* Ten_pmax = floor(P*log(2)/log(5)) */
! 301: /* Bletch = (highest power of 2 < DBL_MAX_10_EXP) / 16 */
! 302: /* Quick_max = floor((P-1)*log(FLT_RADIX)/log(10) - 1) */
! 303: /* Int_max = floor(P*log(FLT_RADIX)/log(10) - 1) */
! 304:
! 305: #if defined(IEEE_LITTLE_ENDIAN) + defined(IEEE_BIG_ENDIAN)
! 306: #define Exp_shift 20
! 307: #define Exp_shift1 20
! 308: #define Exp_msk1 0x100000
! 309: #define Exp_msk11 0x100000
! 310: #define Exp_mask 0x7ff00000
! 311: #define P 53
! 312: #define Bias 1023
! 313: #define IEEE_Arith
! 314: #define Emin (-1022)
! 315: #define Exp_1 0x3ff00000
! 316: #define Exp_11 0x3ff00000
! 317: #define Ebits 11
! 318: #define Frac_mask 0xfffff
! 319: #define Frac_mask1 0xfffff
! 320: #define Ten_pmax 22
! 321: #define Bletch 0x10
! 322: #define Bndry_mask 0xfffff
! 323: #define Bndry_mask1 0xfffff
! 324: #define LSB 1
! 325: #define Sign_bit 0x80000000
! 326: #define Log2P 1
! 327: #define Tiny0 0
! 328: #define Tiny1 1
! 329: #define Quick_max 14
! 330: #define Int_max 14
! 331: #define Infinite(x) (word0(x) == 0x7ff00000) /* sufficient test for here */
! 332: #else
! 333: #undef Sudden_Underflow
! 334: #define Sudden_Underflow
! 335: #ifdef IBM
! 336: #define Exp_shift 24
! 337: #define Exp_shift1 24
! 338: #define Exp_msk1 0x1000000
! 339: #define Exp_msk11 0x1000000
! 340: #define Exp_mask 0x7f000000
! 341: #define P 14
! 342: #define Bias 65
! 343: #define Exp_1 0x41000000
! 344: #define Exp_11 0x41000000
! 345: #define Ebits 8 /* exponent has 7 bits, but 8 is the right value in b2d */
! 346: #define Frac_mask 0xffffff
! 347: #define Frac_mask1 0xffffff
! 348: #define Bletch 4
! 349: #define Ten_pmax 22
! 350: #define Bndry_mask 0xefffff
! 351: #define Bndry_mask1 0xffffff
! 352: #define LSB 1
! 353: #define Sign_bit 0x80000000
! 354: #define Log2P 4
! 355: #define Tiny0 0x100000
! 356: #define Tiny1 0
! 357: #define Quick_max 14
! 358: #define Int_max 15
! 359: #else /* VAX */
! 360: #define Exp_shift 23
! 361: #define Exp_shift1 7
! 362: #define Exp_msk1 0x80
! 363: #define Exp_msk11 0x800000
! 364: #define Exp_mask 0x7f80
! 365: #define P 56
! 366: #define Bias 129
! 367: #define Exp_1 0x40800000
! 368: #define Exp_11 0x4080
! 369: #define Ebits 8
! 370: #define Frac_mask 0x7fffff
! 371: #define Frac_mask1 0xffff007f
! 372: #define Ten_pmax 24
! 373: #define Bletch 2
! 374: #define Bndry_mask 0xffff007f
! 375: #define Bndry_mask1 0xffff007f
! 376: #define LSB 0x10000
! 377: #define Sign_bit 0x8000
! 378: #define Log2P 1
! 379: #define Tiny0 0x80
! 380: #define Tiny1 0
! 381: #define Quick_max 15
! 382: #define Int_max 15
! 383: #endif
! 384: #endif
! 385:
! 386: #ifndef IEEE_Arith
! 387: #define ROUND_BIASED
! 388: #endif
! 389:
! 390: #ifdef RND_PRODQUOT
! 391: #define rounded_product(a,b) a = rnd_prod(a, b)
! 392: #define rounded_quotient(a,b) a = rnd_quot(a, b)
! 393: #ifdef KR_headers
! 394: extern double rnd_prod(), rnd_quot();
! 395: #else
! 396: extern double rnd_prod(double, double), rnd_quot(double, double);
! 397: #endif
! 398: #else
! 399: #define rounded_product(a,b) a *= b
! 400: #define rounded_quotient(a,b) a /= b
! 401: #endif
! 402:
! 403: #define Big0 (Frac_mask1 | Exp_msk1*(DBL_MAX_EXP+Bias-1))
! 404: #define Big1 0xffffffff
! 405:
! 406: #ifndef Just_16
! 407: /* When Pack_32 is not defined, we store 16 bits per 32-bit Long.
! 408: * * This makes some inner loops simpler and sometimes saves work
! 409: * * during multiplications, but it often seems to make things slightly
! 410: * * slower. Hence the default is now to store 32 bits per Long.
! 411: * */
! 412: #ifndef Pack_32
! 413: #define Pack_32
! 414: #endif
! 415: #endif
! 416:
! 417: #define Kmax 15
! 418:
! 419: struct Bigint {
! 420: struct Bigint *next;
! 421: int k, maxwds, sign, wds;
! 422: ULong x[1];
! 423: };
! 424:
! 425: typedef struct Bigint Bigint;
! 426:
! 427: /* static variables, multithreading fun! */
! 428: static Bigint *freelist[Kmax+1];
! 429: static Bigint *p5s;
! 430:
! 431: static void destroy_freelist(void);
! 432:
! 433: #ifdef ZTS
! 434:
! 435: static MUTEX_T dtoa_mutex;
! 436: static MUTEX_T pow5mult_mutex;
! 437:
! 438: #define _THREAD_PRIVATE_MUTEX_LOCK(x) tsrm_mutex_lock(x);
! 439: #define _THREAD_PRIVATE_MUTEX_UNLOCK(x) tsrm_mutex_unlock(x);
! 440:
! 441: #else
! 442:
! 443: #define _THREAD_PRIVATE_MUTEX_LOCK(x)
! 444: #define _THREAD_PRIVATE_MUTEX_UNLOCK(x)
! 445:
! 446: #endif /* ZTS */
! 447:
! 448: #ifdef DEBUG
! 449: static void Bug(const char *message) {
! 450: fprintf(stderr, "%s\n", message);
! 451: }
! 452: #endif
! 453:
! 454: ZEND_API int zend_startup_strtod(void) /* {{{ */
! 455: {
! 456: #ifdef ZTS
! 457: dtoa_mutex = tsrm_mutex_alloc();
! 458: pow5mult_mutex = tsrm_mutex_alloc();
! 459: #endif
! 460: return 1;
! 461: }
! 462: /* }}} */
! 463: ZEND_API int zend_shutdown_strtod(void) /* {{{ */
! 464: {
! 465: destroy_freelist();
! 466: #ifdef ZTS
! 467: tsrm_mutex_free(dtoa_mutex);
! 468: dtoa_mutex = NULL;
! 469:
! 470: tsrm_mutex_free(pow5mult_mutex);
! 471: pow5mult_mutex = NULL;
! 472: #endif
! 473: return 1;
! 474: }
! 475: /* }}} */
! 476:
! 477: static Bigint * Balloc(int k)
! 478: {
! 479: int x;
! 480: Bigint *rv;
! 481:
! 482: if (k > Kmax) {
! 483: zend_error(E_ERROR, "Balloc() allocation exceeds list boundary");
! 484: }
! 485:
! 486: _THREAD_PRIVATE_MUTEX_LOCK(dtoa_mutex);
! 487: if ((rv = freelist[k])) {
! 488: freelist[k] = rv->next;
! 489: } else {
! 490: x = 1 << k;
! 491: rv = (Bigint *)MALLOC(sizeof(Bigint) + (x-1)*sizeof(Long));
! 492: if (!rv) {
! 493: _THREAD_PRIVATE_MUTEX_UNLOCK(dtoa_mutex);
! 494: zend_error(E_ERROR, "Balloc() failed to allocate memory");
! 495: }
! 496: rv->k = k;
! 497: rv->maxwds = x;
! 498: }
! 499: _THREAD_PRIVATE_MUTEX_UNLOCK(dtoa_mutex);
! 500: rv->sign = rv->wds = 0;
! 501: return rv;
! 502: }
! 503:
! 504: static void Bfree(Bigint *v)
! 505: {
! 506: if (v) {
! 507: _THREAD_PRIVATE_MUTEX_LOCK(dtoa_mutex);
! 508: v->next = freelist[v->k];
! 509: freelist[v->k] = v;
! 510: _THREAD_PRIVATE_MUTEX_UNLOCK(dtoa_mutex);
! 511: }
! 512: }
! 513:
! 514: #define Bcopy(x,y) memcpy((char *)&x->sign, (char *)&y->sign, \
! 515: y->wds*sizeof(Long) + 2*sizeof(int))
! 516:
! 517: /* return value is only used as a simple string, so mis-aligned parts
! 518: * inside the Bigint are not at risk on strict align architectures
! 519: */
! 520: static char * rv_alloc(int i) {
! 521: int j, k, *r;
! 522:
! 523: j = sizeof(ULong);
! 524: for(k = 0;
! 525: sizeof(Bigint) - sizeof(ULong) - sizeof(int) + j <= i;
! 526: j <<= 1) {
! 527: k++;
! 528: }
! 529: r = (int*)Balloc(k);
! 530: *r = k;
! 531: return (char *)(r+1);
! 532: }
! 533:
! 534:
! 535: static char * nrv_alloc(char *s, char **rve, int n)
! 536: {
! 537: char *rv, *t;
! 538:
! 539: t = rv = rv_alloc(n);
! 540: while((*t = *s++) !=0) {
! 541: t++;
! 542: }
! 543: if (rve) {
! 544: *rve = t;
! 545: }
! 546: return rv;
! 547: }
! 548:
! 549: static Bigint * multadd(Bigint *b, int m, int a) /* multiply by m and add a */
! 550: {
! 551: int i, wds;
! 552: ULong *x, y;
! 553: #ifdef Pack_32
! 554: ULong xi, z;
! 555: #endif
! 556: Bigint *b1;
! 557:
! 558: wds = b->wds;
! 559: x = b->x;
! 560: i = 0;
! 561: do {
! 562: #ifdef Pack_32
! 563: xi = *x;
! 564: y = (xi & 0xffff) * m + a;
! 565: z = (xi >> 16) * m + (y >> 16);
! 566: a = (int)(z >> 16);
! 567: *x++ = (z << 16) + (y & 0xffff);
! 568: #else
! 569: y = *x * m + a;
! 570: a = (int)(y >> 16);
! 571: *x++ = y & 0xffff;
! 572: #endif
! 573: }
! 574: while(++i < wds);
! 575: if (a) {
! 576: if (wds >= b->maxwds) {
! 577: b1 = Balloc(b->k+1);
! 578: Bcopy(b1, b);
! 579: Bfree(b);
! 580: b = b1;
! 581: }
! 582: b->x[wds++] = a;
! 583: b->wds = wds;
! 584: }
! 585: return b;
! 586: }
! 587:
! 588: static int hi0bits(ULong x)
! 589: {
! 590: int k = 0;
! 591:
! 592: if (!(x & 0xffff0000)) {
! 593: k = 16;
! 594: x <<= 16;
! 595: }
! 596: if (!(x & 0xff000000)) {
! 597: k += 8;
! 598: x <<= 8;
! 599: }
! 600: if (!(x & 0xf0000000)) {
! 601: k += 4;
! 602: x <<= 4;
! 603: }
! 604: if (!(x & 0xc0000000)) {
! 605: k += 2;
! 606: x <<= 2;
! 607: }
! 608: if (!(x & 0x80000000)) {
! 609: k++;
! 610: if (!(x & 0x40000000)) {
! 611: return 32;
! 612: }
! 613: }
! 614: return k;
! 615: }
! 616:
! 617: static int lo0bits(ULong *y)
! 618: {
! 619: int k;
! 620: ULong x = *y;
! 621:
! 622: if (x & 7) {
! 623: if (x & 1) {
! 624: return 0;
! 625: }
! 626: if (x & 2) {
! 627: *y = x >> 1;
! 628: return 1;
! 629: }
! 630: *y = x >> 2;
! 631: return 2;
! 632: }
! 633: k = 0;
! 634: if (!(x & 0xffff)) {
! 635: k = 16;
! 636: x >>= 16;
! 637: }
! 638: if (!(x & 0xff)) {
! 639: k += 8;
! 640: x >>= 8;
! 641: }
! 642: if (!(x & 0xf)) {
! 643: k += 4;
! 644: x >>= 4;
! 645: }
! 646: if (!(x & 0x3)) {
! 647: k += 2;
! 648: x >>= 2;
! 649: }
! 650: if (!(x & 1)) {
! 651: k++;
! 652: x >>= 1;
! 653: if (!(x & 1)) {
! 654: return 32;
! 655: }
! 656: }
! 657: *y = x;
! 658: return k;
! 659: }
! 660:
! 661: static Bigint * i2b(int i)
! 662: {
! 663: Bigint *b;
! 664:
! 665: b = Balloc(1);
! 666: b->x[0] = i;
! 667: b->wds = 1;
! 668: return b;
! 669: }
! 670:
! 671: static Bigint * mult(Bigint *a, Bigint *b)
! 672: {
! 673: Bigint *c;
! 674: int k, wa, wb, wc;
! 675: ULong carry, y, z;
! 676: ULong *x, *xa, *xae, *xb, *xbe, *xc, *xc0;
! 677: #ifdef Pack_32
! 678: ULong z2;
! 679: #endif
! 680:
! 681: if (a->wds < b->wds) {
! 682: c = a;
! 683: a = b;
! 684: b = c;
! 685: }
! 686: k = a->k;
! 687: wa = a->wds;
! 688: wb = b->wds;
! 689: wc = wa + wb;
! 690: if (wc > a->maxwds) {
! 691: k++;
! 692: }
! 693: c = Balloc(k);
! 694: for(x = c->x, xa = x + wc; x < xa; x++) {
! 695: *x = 0;
! 696: }
! 697: xa = a->x;
! 698: xae = xa + wa;
! 699: xb = b->x;
! 700: xbe = xb + wb;
! 701: xc0 = c->x;
! 702: #ifdef Pack_32
! 703: for(; xb < xbe; xb++, xc0++) {
! 704: if ((y = *xb & 0xffff)) {
! 705: x = xa;
! 706: xc = xc0;
! 707: carry = 0;
! 708: do {
! 709: z = (*x & 0xffff) * y + (*xc & 0xffff) + carry;
! 710: carry = z >> 16;
! 711: z2 = (*x++ >> 16) * y + (*xc >> 16) + carry;
! 712: carry = z2 >> 16;
! 713: Storeinc(xc, z2, z);
! 714: }
! 715: while(x < xae);
! 716: *xc = carry;
! 717: }
! 718: if ((y = *xb >> 16)) {
! 719: x = xa;
! 720: xc = xc0;
! 721: carry = 0;
! 722: z2 = *xc;
! 723: do {
! 724: z = (*x & 0xffff) * y + (*xc >> 16) + carry;
! 725: carry = z >> 16;
! 726: Storeinc(xc, z, z2);
! 727: z2 = (*x++ >> 16) * y + (*xc & 0xffff) + carry;
! 728: carry = z2 >> 16;
! 729: }
! 730: while(x < xae);
! 731: *xc = z2;
! 732: }
! 733: }
! 734: #else
! 735: for(; xb < xbe; xc0++) {
! 736: if (y = *xb++) {
! 737: x = xa;
! 738: xc = xc0;
! 739: carry = 0;
! 740: do {
! 741: z = *x++ * y + *xc + carry;
! 742: carry = z >> 16;
! 743: *xc++ = z & 0xffff;
! 744: }
! 745: while(x < xae);
! 746: *xc = carry;
! 747: }
! 748: }
! 749: #endif
! 750: for(xc0 = c->x, xc = xc0 + wc; wc > 0 && !*--xc; --wc) ;
! 751: c->wds = wc;
! 752: return c;
! 753: }
! 754:
! 755: static Bigint * s2b (CONST char *s, int nd0, int nd, ULong y9)
! 756: {
! 757: Bigint *b;
! 758: int i, k;
! 759: Long x, y;
! 760:
! 761: x = (nd + 8) / 9;
! 762: for(k = 0, y = 1; x > y; y <<= 1, k++) ;
! 763: #ifdef Pack_32
! 764: b = Balloc(k);
! 765: b->x[0] = y9;
! 766: b->wds = 1;
! 767: #else
! 768: b = Balloc(k+1);
! 769: b->x[0] = y9 & 0xffff;
! 770: b->wds = (b->x[1] = y9 >> 16) ? 2 : 1;
! 771: #endif
! 772:
! 773: i = 9;
! 774: if (9 < nd0) {
! 775: s += 9;
! 776: do b = multadd(b, 10, *s++ - '0');
! 777: while(++i < nd0);
! 778: s++;
! 779: } else {
! 780: s += 10;
! 781: }
! 782: for(; i < nd; i++) {
! 783: b = multadd(b, 10, *s++ - '0');
! 784: }
! 785: return b;
! 786: }
! 787:
! 788: static Bigint * pow5mult(Bigint *b, int k)
! 789: {
! 790: Bigint *b1, *p5, *p51;
! 791: int i;
! 792: static int p05[3] = { 5, 25, 125 };
! 793:
! 794: _THREAD_PRIVATE_MUTEX_LOCK(pow5mult_mutex);
! 795: if ((i = k & 3)) {
! 796: b = multadd(b, p05[i-1], 0);
! 797: }
! 798:
! 799: if (!(k >>= 2)) {
! 800: _THREAD_PRIVATE_MUTEX_UNLOCK(pow5mult_mutex);
! 801: return b;
! 802: }
! 803: if (!(p5 = p5s)) {
! 804: /* first time */
! 805: p5 = p5s = i2b(625);
! 806: p5->next = 0;
! 807: }
! 808: for(;;) {
! 809: if (k & 1) {
! 810: b1 = mult(b, p5);
! 811: Bfree(b);
! 812: b = b1;
! 813: }
! 814: if (!(k >>= 1)) {
! 815: break;
! 816: }
! 817: if (!(p51 = p5->next)) {
! 818: if (!(p51 = p5->next)) {
! 819: p51 = p5->next = mult(p5,p5);
! 820: p51->next = 0;
! 821: }
! 822: }
! 823: p5 = p51;
! 824: }
! 825: _THREAD_PRIVATE_MUTEX_UNLOCK(pow5mult_mutex);
! 826: return b;
! 827: }
! 828:
! 829:
! 830: static Bigint *lshift(Bigint *b, int k)
! 831: {
! 832: int i, k1, n, n1;
! 833: Bigint *b1;
! 834: ULong *x, *x1, *xe, z;
! 835:
! 836: #ifdef Pack_32
! 837: n = k >> 5;
! 838: #else
! 839: n = k >> 4;
! 840: #endif
! 841: k1 = b->k;
! 842: n1 = n + b->wds + 1;
! 843: for(i = b->maxwds; n1 > i; i <<= 1) {
! 844: k1++;
! 845: }
! 846: b1 = Balloc(k1);
! 847: x1 = b1->x;
! 848: for(i = 0; i < n; i++) {
! 849: *x1++ = 0;
! 850: }
! 851: x = b->x;
! 852: xe = x + b->wds;
! 853: #ifdef Pack_32
! 854: if (k &= 0x1f) {
! 855: k1 = 32 - k;
! 856: z = 0;
! 857: do {
! 858: *x1++ = *x << k | z;
! 859: z = *x++ >> k1;
! 860: }
! 861: while(x < xe);
! 862: if ((*x1 = z)) {
! 863: ++n1;
! 864: }
! 865: }
! 866: #else
! 867: if (k &= 0xf) {
! 868: k1 = 16 - k;
! 869: z = 0;
! 870: do {
! 871: *x1++ = *x << k & 0xffff | z;
! 872: z = *x++ >> k1;
! 873: }
! 874: while(x < xe);
! 875: if (*x1 = z) {
! 876: ++n1;
! 877: }
! 878: }
! 879: #endif
! 880: else do
! 881: *x1++ = *x++;
! 882: while(x < xe);
! 883: b1->wds = n1 - 1;
! 884: Bfree(b);
! 885: return b1;
! 886: }
! 887:
! 888: static int cmp(Bigint *a, Bigint *b)
! 889: {
! 890: ULong *xa, *xa0, *xb, *xb0;
! 891: int i, j;
! 892:
! 893: i = a->wds;
! 894: j = b->wds;
! 895: #ifdef DEBUG
! 896: if (i > 1 && !a->x[i-1])
! 897: Bug("cmp called with a->x[a->wds-1] == 0");
! 898: if (j > 1 && !b->x[j-1])
! 899: Bug("cmp called with b->x[b->wds-1] == 0");
! 900: #endif
! 901: if (i -= j)
! 902: return i;
! 903: xa0 = a->x;
! 904: xa = xa0 + j;
! 905: xb0 = b->x;
! 906: xb = xb0 + j;
! 907: for(;;) {
! 908: if (*--xa != *--xb)
! 909: return *xa < *xb ? -1 : 1;
! 910: if (xa <= xa0)
! 911: break;
! 912: }
! 913: return 0;
! 914: }
! 915:
! 916:
! 917: static Bigint * diff(Bigint *a, Bigint *b)
! 918: {
! 919: Bigint *c;
! 920: int i, wa, wb;
! 921: Long borrow, y; /* We need signed shifts here. */
! 922: ULong *xa, *xae, *xb, *xbe, *xc;
! 923: #ifdef Pack_32
! 924: Long z;
! 925: #endif
! 926:
! 927: i = cmp(a,b);
! 928: if (!i) {
! 929: c = Balloc(0);
! 930: c->wds = 1;
! 931: c->x[0] = 0;
! 932: return c;
! 933: }
! 934: if (i < 0) {
! 935: c = a;
! 936: a = b;
! 937: b = c;
! 938: i = 1;
! 939: } else {
! 940: i = 0;
! 941: }
! 942: c = Balloc(a->k);
! 943: c->sign = i;
! 944: wa = a->wds;
! 945: xa = a->x;
! 946: xae = xa + wa;
! 947: wb = b->wds;
! 948: xb = b->x;
! 949: xbe = xb + wb;
! 950: xc = c->x;
! 951: borrow = 0;
! 952: #ifdef Pack_32
! 953: do {
! 954: y = (*xa & 0xffff) - (*xb & 0xffff) + borrow;
! 955: borrow = y >> 16;
! 956: Sign_Extend(borrow, y);
! 957: z = (*xa++ >> 16) - (*xb++ >> 16) + borrow;
! 958: borrow = z >> 16;
! 959: Sign_Extend(borrow, z);
! 960: Storeinc(xc, z, y);
! 961: } while(xb < xbe);
! 962: while(xa < xae) {
! 963: y = (*xa & 0xffff) + borrow;
! 964: borrow = y >> 16;
! 965: Sign_Extend(borrow, y);
! 966: z = (*xa++ >> 16) + borrow;
! 967: borrow = z >> 16;
! 968: Sign_Extend(borrow, z);
! 969: Storeinc(xc, z, y);
! 970: }
! 971: #else
! 972: do {
! 973: y = *xa++ - *xb++ + borrow;
! 974: borrow = y >> 16;
! 975: Sign_Extend(borrow, y);
! 976: *xc++ = y & 0xffff;
! 977: } while(xb < xbe);
! 978: while(xa < xae) {
! 979: y = *xa++ + borrow;
! 980: borrow = y >> 16;
! 981: Sign_Extend(borrow, y);
! 982: *xc++ = y & 0xffff;
! 983: }
! 984: #endif
! 985: while(!*--xc) {
! 986: wa--;
! 987: }
! 988: c->wds = wa;
! 989: return c;
! 990: }
! 991:
! 992: static double ulp (double _x)
! 993: {
! 994: volatile _double x;
! 995: register Long L;
! 996: volatile _double a;
! 997:
! 998: value(x) = _x;
! 999: L = (word0(x) & Exp_mask) - (P-1)*Exp_msk1;
! 1000: #ifndef Sudden_Underflow
! 1001: if (L > 0) {
! 1002: #endif
! 1003: #ifdef IBM
! 1004: L |= Exp_msk1 >> 4;
! 1005: #endif
! 1006: word0(a) = L;
! 1007: word1(a) = 0;
! 1008: #ifndef Sudden_Underflow
! 1009: }
! 1010: else {
! 1011: L = -L >> Exp_shift;
! 1012: if (L < Exp_shift) {
! 1013: word0(a) = 0x80000 >> L;
! 1014: word1(a) = 0;
! 1015: }
! 1016: else {
! 1017: word0(a) = 0;
! 1018: L -= Exp_shift;
! 1019: word1(a) = L >= 31 ? 1 : 1 << (31 - L);
! 1020: }
! 1021: }
! 1022: #endif
! 1023: return value(a);
! 1024: }
! 1025:
! 1026: static double
! 1027: b2d
! 1028: #ifdef KR_headers
! 1029: (a, e) Bigint *a; int *e;
! 1030: #else
! 1031: (Bigint *a, int *e)
! 1032: #endif
! 1033: {
! 1034: ULong *xa, *xa0, w, y, z;
! 1035: int k;
! 1036: volatile _double d;
! 1037: #ifdef VAX
! 1038: ULong d0, d1;
! 1039: #else
! 1040: #define d0 word0(d)
! 1041: #define d1 word1(d)
! 1042: #endif
! 1043:
! 1044: xa0 = a->x;
! 1045: xa = xa0 + a->wds;
! 1046: y = *--xa;
! 1047: #ifdef DEBUG
! 1048: if (!y) Bug("zero y in b2d");
! 1049: #endif
! 1050: k = hi0bits(y);
! 1051: *e = 32 - k;
! 1052: #ifdef Pack_32
! 1053: if (k < Ebits) {
! 1054: d0 = Exp_1 | y >> (Ebits - k);
! 1055: w = xa > xa0 ? *--xa : 0;
! 1056: d1 = y << ((32-Ebits) + k) | w >> (Ebits - k);
! 1057: goto ret_d;
! 1058: }
! 1059: z = xa > xa0 ? *--xa : 0;
! 1060: if (k -= Ebits) {
! 1061: d0 = Exp_1 | y << k | z >> (32 - k);
! 1062: y = xa > xa0 ? *--xa : 0;
! 1063: d1 = z << k | y >> (32 - k);
! 1064: }
! 1065: else {
! 1066: d0 = Exp_1 | y;
! 1067: d1 = z;
! 1068: }
! 1069: #else
! 1070: if (k < Ebits + 16) {
! 1071: z = xa > xa0 ? *--xa : 0;
! 1072: d0 = Exp_1 | y << k - Ebits | z >> Ebits + 16 - k;
! 1073: w = xa > xa0 ? *--xa : 0;
! 1074: y = xa > xa0 ? *--xa : 0;
! 1075: d1 = z << k + 16 - Ebits | w << k - Ebits | y >> 16 + Ebits - k;
! 1076: goto ret_d;
! 1077: }
! 1078: z = xa > xa0 ? *--xa : 0;
! 1079: w = xa > xa0 ? *--xa : 0;
! 1080: k -= Ebits + 16;
! 1081: d0 = Exp_1 | y << k + 16 | z << k | w >> 16 - k;
! 1082: y = xa > xa0 ? *--xa : 0;
! 1083: d1 = w << k + 16 | y << k;
! 1084: #endif
! 1085: ret_d:
! 1086: #ifdef VAX
! 1087: word0(d) = d0 >> 16 | d0 << 16;
! 1088: word1(d) = d1 >> 16 | d1 << 16;
! 1089: #else
! 1090: #undef d0
! 1091: #undef d1
! 1092: #endif
! 1093: return value(d);
! 1094: }
! 1095:
! 1096:
! 1097: static Bigint * d2b(double _d, int *e, int *bits)
! 1098: {
! 1099: Bigint *b;
! 1100: int de, i, k;
! 1101: ULong *x, y, z;
! 1102: volatile _double d;
! 1103: #ifdef VAX
! 1104: ULong d0, d1;
! 1105: #endif
! 1106:
! 1107: value(d) = _d;
! 1108: #ifdef VAX
! 1109: d0 = word0(d) >> 16 | word0(d) << 16;
! 1110: d1 = word1(d) >> 16 | word1(d) << 16;
! 1111: #else
! 1112: #define d0 word0(d)
! 1113: #define d1 word1(d)
! 1114: #endif
! 1115:
! 1116: #ifdef Pack_32
! 1117: b = Balloc(1);
! 1118: #else
! 1119: b = Balloc(2);
! 1120: #endif
! 1121: x = b->x;
! 1122:
! 1123: z = d0 & Frac_mask;
! 1124: d0 &= 0x7fffffff; /* clear sign bit, which we ignore */
! 1125: #ifdef Sudden_Underflow
! 1126: de = (int)(d0 >> Exp_shift);
! 1127: #ifndef IBM
! 1128: z |= Exp_msk11;
! 1129: #endif
! 1130: #else
! 1131: if ((de = (int)(d0 >> Exp_shift)))
! 1132: z |= Exp_msk1;
! 1133: #endif
! 1134: #ifdef Pack_32
! 1135: if ((y = d1)) {
! 1136: if ((k = lo0bits(&y))) {
! 1137: x[0] = y | (z << (32 - k));
! 1138: z >>= k;
! 1139: } else {
! 1140: x[0] = y;
! 1141: }
! 1142: i = b->wds = (x[1] = z) ? 2 : 1;
! 1143: } else {
! 1144: #ifdef DEBUG
! 1145: if (!z)
! 1146: Bug("Zero passed to d2b");
! 1147: #endif
! 1148: k = lo0bits(&z);
! 1149: x[0] = z;
! 1150: i = b->wds = 1;
! 1151: k += 32;
! 1152: }
! 1153: #else
! 1154: if (y = d1) {
! 1155: if (k = lo0bits(&y)) {
! 1156: if (k >= 16) {
! 1157: x[0] = y | z << 32 - k & 0xffff;
! 1158: x[1] = z >> k - 16 & 0xffff;
! 1159: x[2] = z >> k;
! 1160: i = 2;
! 1161: } else {
! 1162: x[0] = y & 0xffff;
! 1163: x[1] = y >> 16 | z << 16 - k & 0xffff;
! 1164: x[2] = z >> k & 0xffff;
! 1165: x[3] = z >> k+16;
! 1166: i = 3;
! 1167: }
! 1168: } else {
! 1169: x[0] = y & 0xffff;
! 1170: x[1] = y >> 16;
! 1171: x[2] = z & 0xffff;
! 1172: x[3] = z >> 16;
! 1173: i = 3;
! 1174: }
! 1175: } else {
! 1176: #ifdef DEBUG
! 1177: if (!z)
! 1178: Bug("Zero passed to d2b");
! 1179: #endif
! 1180: k = lo0bits(&z);
! 1181: if (k >= 16) {
! 1182: x[0] = z;
! 1183: i = 0;
! 1184: } else {
! 1185: x[0] = z & 0xffff;
! 1186: x[1] = z >> 16;
! 1187: i = 1;
! 1188: }
! 1189: k += 32;
! 1190: }
! 1191: while(!x[i])
! 1192: --i;
! 1193: b->wds = i + 1;
! 1194: #endif
! 1195: #ifndef Sudden_Underflow
! 1196: if (de) {
! 1197: #endif
! 1198: #ifdef IBM
! 1199: *e = (de - Bias - (P-1) << 2) + k;
! 1200: *bits = 4*P + 8 - k - hi0bits(word0(d) & Frac_mask);
! 1201: #else
! 1202: *e = de - Bias - (P-1) + k;
! 1203: *bits = P - k;
! 1204: #endif
! 1205: #ifndef Sudden_Underflow
! 1206: } else {
! 1207: *e = de - Bias - (P-1) + 1 + k;
! 1208: #ifdef Pack_32
! 1209: *bits = 32*i - hi0bits(x[i-1]);
! 1210: #else
! 1211: *bits = (i+2)*16 - hi0bits(x[i]);
! 1212: #endif
! 1213: }
! 1214: #endif
! 1215: return b;
! 1216: }
! 1217: #undef d0
! 1218: #undef d1
! 1219:
! 1220:
! 1221: static double ratio (Bigint *a, Bigint *b)
! 1222: {
! 1223: volatile _double da, db;
! 1224: int k, ka, kb;
! 1225:
! 1226: value(da) = b2d(a, &ka);
! 1227: value(db) = b2d(b, &kb);
! 1228: #ifdef Pack_32
! 1229: k = ka - kb + 32*(a->wds - b->wds);
! 1230: #else
! 1231: k = ka - kb + 16*(a->wds - b->wds);
! 1232: #endif
! 1233: #ifdef IBM
! 1234: if (k > 0) {
! 1235: word0(da) += (k >> 2)*Exp_msk1;
! 1236: if (k &= 3) {
! 1237: da *= 1 << k;
! 1238: }
! 1239: } else {
! 1240: k = -k;
! 1241: word0(db) += (k >> 2)*Exp_msk1;
! 1242: if (k &= 3)
! 1243: db *= 1 << k;
! 1244: }
! 1245: #else
! 1246: if (k > 0) {
! 1247: word0(da) += k*Exp_msk1;
! 1248: } else {
! 1249: k = -k;
! 1250: word0(db) += k*Exp_msk1;
! 1251: }
! 1252: #endif
! 1253: return value(da) / value(db);
! 1254: }
! 1255:
! 1256: static CONST double
! 1257: tens[] = {
! 1258: 1e0, 1e1, 1e2, 1e3, 1e4, 1e5, 1e6, 1e7, 1e8, 1e9,
! 1259: 1e10, 1e11, 1e12, 1e13, 1e14, 1e15, 1e16, 1e17, 1e18, 1e19,
! 1260: 1e20, 1e21, 1e22
! 1261: #ifdef VAX
! 1262: , 1e23, 1e24
! 1263: #endif
! 1264: };
! 1265:
! 1266: #ifdef IEEE_Arith
! 1267: static CONST double bigtens[] = { 1e16, 1e32, 1e64, 1e128, 1e256 };
! 1268: static CONST double tinytens[] = { 1e-16, 1e-32, 1e-64, 1e-128, 1e-256 };
! 1269: #define n_bigtens 5
! 1270: #else
! 1271: #ifdef IBM
! 1272: static CONST double bigtens[] = { 1e16, 1e32, 1e64 };
! 1273: static CONST double tinytens[] = { 1e-16, 1e-32, 1e-64 };
! 1274: #define n_bigtens 3
! 1275: #else
! 1276: static CONST double bigtens[] = { 1e16, 1e32 };
! 1277: static CONST double tinytens[] = { 1e-16, 1e-32 };
! 1278: #define n_bigtens 2
! 1279: #endif
! 1280: #endif
! 1281:
! 1282:
! 1283: static int quorem(Bigint *b, Bigint *S)
! 1284: {
! 1285: int n;
! 1286: Long borrow, y;
! 1287: ULong carry, q, ys;
! 1288: ULong *bx, *bxe, *sx, *sxe;
! 1289: #ifdef Pack_32
! 1290: Long z;
! 1291: ULong si, zs;
! 1292: #endif
! 1293:
! 1294: n = S->wds;
! 1295: #ifdef DEBUG
! 1296: /*debug*/ if (b->wds > n)
! 1297: /*debug*/ Bug("oversize b in quorem");
! 1298: #endif
! 1299: if (b->wds < n)
! 1300: return 0;
! 1301: sx = S->x;
! 1302: sxe = sx + --n;
! 1303: bx = b->x;
! 1304: bxe = bx + n;
! 1305: q = *bxe / (*sxe + 1); /* ensure q <= true quotient */
! 1306: #ifdef DEBUG
! 1307: /*debug*/ if (q > 9)
! 1308: /*debug*/ Bug("oversized quotient in quorem");
! 1309: #endif
! 1310: if (q) {
! 1311: borrow = 0;
! 1312: carry = 0;
! 1313: do {
! 1314: #ifdef Pack_32
! 1315: si = *sx++;
! 1316: ys = (si & 0xffff) * q + carry;
! 1317: zs = (si >> 16) * q + (ys >> 16);
! 1318: carry = zs >> 16;
! 1319: y = (*bx & 0xffff) - (ys & 0xffff) + borrow;
! 1320: borrow = y >> 16;
! 1321: Sign_Extend(borrow, y);
! 1322: z = (*bx >> 16) - (zs & 0xffff) + borrow;
! 1323: borrow = z >> 16;
! 1324: Sign_Extend(borrow, z);
! 1325: Storeinc(bx, z, y);
! 1326: #else
! 1327: ys = *sx++ * q + carry;
! 1328: carry = ys >> 16;
! 1329: y = *bx - (ys & 0xffff) + borrow;
! 1330: borrow = y >> 16;
! 1331: Sign_Extend(borrow, y);
! 1332: *bx++ = y & 0xffff;
! 1333: #endif
! 1334: }
! 1335: while(sx <= sxe);
! 1336: if (!*bxe) {
! 1337: bx = b->x;
! 1338: while(--bxe > bx && !*bxe)
! 1339: --n;
! 1340: b->wds = n;
! 1341: }
! 1342: }
! 1343: if (cmp(b, S) >= 0) {
! 1344: q++;
! 1345: borrow = 0;
! 1346: carry = 0;
! 1347: bx = b->x;
! 1348: sx = S->x;
! 1349: do {
! 1350: #ifdef Pack_32
! 1351: si = *sx++;
! 1352: ys = (si & 0xffff) + carry;
! 1353: zs = (si >> 16) + (ys >> 16);
! 1354: carry = zs >> 16;
! 1355: y = (*bx & 0xffff) - (ys & 0xffff) + borrow;
! 1356: borrow = y >> 16;
! 1357: Sign_Extend(borrow, y);
! 1358: z = (*bx >> 16) - (zs & 0xffff) + borrow;
! 1359: borrow = z >> 16;
! 1360: Sign_Extend(borrow, z);
! 1361: Storeinc(bx, z, y);
! 1362: #else
! 1363: ys = *sx++ + carry;
! 1364: carry = ys >> 16;
! 1365: y = *bx - (ys & 0xffff) + borrow;
! 1366: borrow = y >> 16;
! 1367: Sign_Extend(borrow, y);
! 1368: *bx++ = y & 0xffff;
! 1369: #endif
! 1370: }
! 1371: while(sx <= sxe);
! 1372: bx = b->x;
! 1373: bxe = bx + n;
! 1374: if (!*bxe) {
! 1375: while(--bxe > bx && !*bxe)
! 1376: --n;
! 1377: b->wds = n;
! 1378: }
! 1379: }
! 1380: return q;
! 1381: }
! 1382:
! 1383: static void destroy_freelist(void)
! 1384: {
! 1385: int i;
! 1386: Bigint *tmp;
! 1387:
! 1388: _THREAD_PRIVATE_MUTEX_LOCK(dtoa_mutex);
! 1389: for (i = 0; i <= Kmax; i++) {
! 1390: Bigint **listp = &freelist[i];
! 1391: while ((tmp = *listp) != NULL) {
! 1392: *listp = tmp->next;
! 1393: free(tmp);
! 1394: }
! 1395: freelist[i] = NULL;
! 1396: }
! 1397: _THREAD_PRIVATE_MUTEX_UNLOCK(dtoa_mutex);
! 1398:
! 1399: }
! 1400:
! 1401:
! 1402: ZEND_API void zend_freedtoa(char *s)
! 1403: {
! 1404: Bigint *b = (Bigint *)((int *)s - 1);
! 1405: b->maxwds = 1 << (b->k = *(int*)b);
! 1406: Bfree(b);
! 1407: }
! 1408:
! 1409: /* dtoa for IEEE arithmetic (dmg): convert double to ASCII string.
! 1410: *
! 1411: * Inspired by "How to Print Floating-Point Numbers Accurately" by
! 1412: * Guy L. Steele, Jr. and Jon L. White [Proc. ACM SIGPLAN '90, pp. 92-101].
! 1413: *
! 1414: * Modifications:
! 1415: * 1. Rather than iterating, we use a simple numeric overestimate
! 1416: * to determine k = floor(log10(d)). We scale relevant
! 1417: * quantities using O(log2(k)) rather than O(k) multiplications.
! 1418: * 2. For some modes > 2 (corresponding to ecvt and fcvt), we don't
! 1419: * try to generate digits strictly left to right. Instead, we
! 1420: * compute with fewer bits and propagate the carry if necessary
! 1421: * when rounding the final digit up. This is often faster.
! 1422: * 3. Under the assumption that input will be rounded nearest,
! 1423: * mode 0 renders 1e23 as 1e23 rather than 9.999999999999999e22.
! 1424: * That is, we allow equality in stopping tests when the
! 1425: * round-nearest rule will give the same floating-point value
! 1426: * as would satisfaction of the stopping test with strict
! 1427: * inequality.
! 1428: * 4. We remove common factors of powers of 2 from relevant
! 1429: * quantities.
! 1430: * 5. When converting floating-point integers less than 1e16,
! 1431: * we use floating-point arithmetic rather than resorting
! 1432: * to multiple-precision integers.
! 1433: * 6. When asked to produce fewer than 15 digits, we first try
! 1434: * to get by with floating-point arithmetic; we resort to
! 1435: * multiple-precision integer arithmetic only if we cannot
! 1436: * guarantee that the floating-point calculation has given
! 1437: * the correctly rounded result. For k requested digits and
! 1438: * "uniformly" distributed input, the probability is
! 1439: * something like 10^(k-15) that we must resort to the Long
! 1440: * calculation.
! 1441: */
! 1442:
! 1443: ZEND_API char * zend_dtoa(double _d, int mode, int ndigits, int *decpt, int *sign, char **rve)
! 1444: {
! 1445: /* Arguments ndigits, decpt, sign are similar to those
! 1446: of ecvt and fcvt; trailing zeros are suppressed from
! 1447: the returned string. If not null, *rve is set to point
! 1448: to the end of the return value. If d is +-Infinity or NaN,
! 1449: then *decpt is set to 9999.
! 1450:
! 1451: mode:
! 1452: 0 ==> shortest string that yields d when read in
! 1453: and rounded to nearest.
! 1454: 1 ==> like 0, but with Steele & White stopping rule;
! 1455: e.g. with IEEE P754 arithmetic , mode 0 gives
! 1456: 1e23 whereas mode 1 gives 9.999999999999999e22.
! 1457: 2 ==> max(1,ndigits) significant digits. This gives a
! 1458: return value similar to that of ecvt, except
! 1459: that trailing zeros are suppressed.
! 1460: 3 ==> through ndigits past the decimal point. This
! 1461: gives a return value similar to that from fcvt,
! 1462: except that trailing zeros are suppressed, and
! 1463: ndigits can be negative.
! 1464: 4-9 should give the same return values as 2-3, i.e.,
! 1465: 4 <= mode <= 9 ==> same return as mode
! 1466: 2 + (mode & 1). These modes are mainly for
! 1467: debugging; often they run slower but sometimes
! 1468: faster than modes 2-3.
! 1469: 4,5,8,9 ==> left-to-right digit generation.
! 1470: 6-9 ==> don't try fast floating-point estimate
! 1471: (if applicable).
! 1472:
! 1473: Values of mode other than 0-9 are treated as mode 0.
! 1474:
! 1475: Sufficient space is allocated to the return value
! 1476: to hold the suppressed trailing zeros.
! 1477: */
! 1478:
! 1479: int bbits, b2, b5, be, dig, i, ieps, ilim = 0, ilim0, ilim1,
! 1480: j, j1, k, k0, k_check, leftright, m2, m5, s2, s5,
! 1481: spec_case = 0, try_quick;
! 1482: Long L;
! 1483: #ifndef Sudden_Underflow
! 1484: int denorm;
! 1485: ULong x;
! 1486: #endif
! 1487: Bigint *b, *b1, *delta, *mlo, *mhi, *S, *tmp;
! 1488: double ds;
! 1489: char *s, *s0;
! 1490: volatile _double d, d2, eps;
! 1491:
! 1492: value(d) = _d;
! 1493:
! 1494: if (word0(d) & Sign_bit) {
! 1495: /* set sign for everything, including 0's and NaNs */
! 1496: *sign = 1;
! 1497: word0(d) &= ~Sign_bit; /* clear sign bit */
! 1498: }
! 1499: else
! 1500: *sign = 0;
! 1501:
! 1502: #if defined(IEEE_Arith) + defined(VAX)
! 1503: #ifdef IEEE_Arith
! 1504: if ((word0(d) & Exp_mask) == Exp_mask)
! 1505: #else
! 1506: if (word0(d) == 0x8000)
! 1507: #endif
! 1508: {
! 1509: /* Infinity or NaN */
! 1510: *decpt = 9999;
! 1511: #ifdef IEEE_Arith
! 1512: if (!word1(d) && !(word0(d) & 0xfffff))
! 1513: return nrv_alloc("Infinity", rve, 8);
! 1514: #endif
! 1515: return nrv_alloc("NaN", rve, 3);
! 1516: }
! 1517: #endif
! 1518: #ifdef IBM
! 1519: value(d) += 0; /* normalize */
! 1520: #endif
! 1521: if (!value(d)) {
! 1522: *decpt = 1;
! 1523: return nrv_alloc("0", rve, 1);
! 1524: }
! 1525:
! 1526: b = d2b(value(d), &be, &bbits);
! 1527: #ifdef Sudden_Underflow
! 1528: i = (int)(word0(d) >> Exp_shift1 & (Exp_mask>>Exp_shift1));
! 1529: #else
! 1530: if ((i = (int)(word0(d) >> Exp_shift1 & (Exp_mask>>Exp_shift1)))) {
! 1531: #endif
! 1532: value(d2) = value(d);
! 1533: word0(d2) &= Frac_mask1;
! 1534: word0(d2) |= Exp_11;
! 1535: #ifdef IBM
! 1536: if (j = 11 - hi0bits(word0(d2) & Frac_mask))
! 1537: value(d2) /= 1 << j;
! 1538: #endif
! 1539:
! 1540: /* log(x) ~=~ log(1.5) + (x-1.5)/1.5
! 1541: * log10(x) = log(x) / log(10)
! 1542: * ~=~ log(1.5)/log(10) + (x-1.5)/(1.5*log(10))
! 1543: * log10(d) = (i-Bias)*log(2)/log(10) + log10(d2)
! 1544: *
! 1545: * This suggests computing an approximation k to log10(d) by
! 1546: *
! 1547: * k = (i - Bias)*0.301029995663981
! 1548: * + ( (d2-1.5)*0.289529654602168 + 0.176091259055681 );
! 1549: *
! 1550: * We want k to be too large rather than too small.
! 1551: * The error in the first-order Taylor series approximation
! 1552: * is in our favor, so we just round up the constant enough
! 1553: * to compensate for any error in the multiplication of
! 1554: * (i - Bias) by 0.301029995663981; since |i - Bias| <= 1077,
! 1555: * and 1077 * 0.30103 * 2^-52 ~=~ 7.2e-14,
! 1556: * adding 1e-13 to the constant term more than suffices.
! 1557: * Hence we adjust the constant term to 0.1760912590558.
! 1558: * (We could get a more accurate k by invoking log10,
! 1559: * but this is probably not worthwhile.)
! 1560: */
! 1561:
! 1562: i -= Bias;
! 1563: #ifdef IBM
! 1564: i <<= 2;
! 1565: i += j;
! 1566: #endif
! 1567: #ifndef Sudden_Underflow
! 1568: denorm = 0;
! 1569: }
! 1570: else {
! 1571: /* d is denormalized */
! 1572:
! 1573: i = bbits + be + (Bias + (P-1) - 1);
! 1574: x = i > 32 ? (word0(d) << (64 - i)) | (word1(d) >> (i - 32))
! 1575: : (word1(d) << (32 - i));
! 1576: value(d2) = x;
! 1577: word0(d2) -= 31*Exp_msk1; /* adjust exponent */
! 1578: i -= (Bias + (P-1) - 1) + 1;
! 1579: denorm = 1;
! 1580: }
! 1581: #endif
! 1582: ds = (value(d2)-1.5)*0.289529654602168 + 0.1760912590558 + i*0.301029995663981;
! 1583: k = (int)ds;
! 1584: if (ds < 0. && ds != k)
! 1585: k--; /* want k = floor(ds) */
! 1586: k_check = 1;
! 1587: if (k >= 0 && k <= Ten_pmax) {
! 1588: if (value(d) < tens[k])
! 1589: k--;
! 1590: k_check = 0;
! 1591: }
! 1592: j = bbits - i - 1;
! 1593: if (j >= 0) {
! 1594: b2 = 0;
! 1595: s2 = j;
! 1596: }
! 1597: else {
! 1598: b2 = -j;
! 1599: s2 = 0;
! 1600: }
! 1601: if (k >= 0) {
! 1602: b5 = 0;
! 1603: s5 = k;
! 1604: s2 += k;
! 1605: }
! 1606: else {
! 1607: b2 -= k;
! 1608: b5 = -k;
! 1609: s5 = 0;
! 1610: }
! 1611: if (mode < 0 || mode > 9)
! 1612: mode = 0;
! 1613: try_quick = 1;
! 1614: if (mode > 5) {
! 1615: mode -= 4;
! 1616: try_quick = 0;
! 1617: }
! 1618: leftright = 1;
! 1619: switch(mode) {
! 1620: case 0:
! 1621: case 1:
! 1622: ilim = ilim1 = -1;
! 1623: i = 18;
! 1624: ndigits = 0;
! 1625: break;
! 1626: case 2:
! 1627: leftright = 0;
! 1628: /* no break */
! 1629: case 4:
! 1630: if (ndigits <= 0)
! 1631: ndigits = 1;
! 1632: ilim = ilim1 = i = ndigits;
! 1633: break;
! 1634: case 3:
! 1635: leftright = 0;
! 1636: /* no break */
! 1637: case 5:
! 1638: i = ndigits + k + 1;
! 1639: ilim = i;
! 1640: ilim1 = i - 1;
! 1641: if (i <= 0)
! 1642: i = 1;
! 1643: }
! 1644: s = s0 = rv_alloc(i);
! 1645:
! 1646: if (ilim >= 0 && ilim <= Quick_max && try_quick) {
! 1647:
! 1648: /* Try to get by with floating-point arithmetic. */
! 1649:
! 1650: i = 0;
! 1651: value(d2) = value(d);
! 1652: k0 = k;
! 1653: ilim0 = ilim;
! 1654: ieps = 2; /* conservative */
! 1655: if (k > 0) {
! 1656: ds = tens[k&0xf];
! 1657: j = k >> 4;
! 1658: if (j & Bletch) {
! 1659: /* prevent overflows */
! 1660: j &= Bletch - 1;
! 1661: value(d) /= bigtens[n_bigtens-1];
! 1662: ieps++;
! 1663: }
! 1664: for(; j; j >>= 1, i++)
! 1665: if (j & 1) {
! 1666: ieps++;
! 1667: ds *= bigtens[i];
! 1668: }
! 1669: value(d) /= ds;
! 1670: }
! 1671: else if ((j1 = -k)) {
! 1672: value(d) *= tens[j1 & 0xf];
! 1673: for(j = j1 >> 4; j; j >>= 1, i++)
! 1674: if (j & 1) {
! 1675: ieps++;
! 1676: value(d) *= bigtens[i];
! 1677: }
! 1678: }
! 1679: if (k_check && value(d) < 1. && ilim > 0) {
! 1680: if (ilim1 <= 0)
! 1681: goto fast_failed;
! 1682: ilim = ilim1;
! 1683: k--;
! 1684: value(d) *= 10.;
! 1685: ieps++;
! 1686: }
! 1687: value(eps) = ieps*value(d) + 7.;
! 1688: word0(eps) -= (P-1)*Exp_msk1;
! 1689: if (ilim == 0) {
! 1690: S = mhi = 0;
! 1691: value(d) -= 5.;
! 1692: if (value(d) > value(eps))
! 1693: goto one_digit;
! 1694: if (value(d) < -value(eps))
! 1695: goto no_digits;
! 1696: goto fast_failed;
! 1697: }
! 1698: #ifndef No_leftright
! 1699: if (leftright) {
! 1700: /* Use Steele & White method of only
! 1701: * generating digits needed.
! 1702: */
! 1703: value(eps) = 0.5/tens[ilim-1] - value(eps);
! 1704: for(i = 0;;) {
! 1705: L = value(d);
! 1706: value(d) -= L;
! 1707: *s++ = '0' + (int)L;
! 1708: if (value(d) < value(eps))
! 1709: goto ret1;
! 1710: if (1. - value(d) < value(eps))
! 1711: goto bump_up;
! 1712: if (++i >= ilim)
! 1713: break;
! 1714: value(eps) *= 10.;
! 1715: value(d) *= 10.;
! 1716: }
! 1717: }
! 1718: else {
! 1719: #endif
! 1720: /* Generate ilim digits, then fix them up. */
! 1721: value(eps) *= tens[ilim-1];
! 1722: for(i = 1;; i++, value(d) *= 10.) {
! 1723: L = value(d);
! 1724: value(d) -= L;
! 1725: *s++ = '0' + (int)L;
! 1726: if (i == ilim) {
! 1727: if (value(d) > 0.5 + value(eps))
! 1728: goto bump_up;
! 1729: else if (value(d) < 0.5 - value(eps)) {
! 1730: while(*--s == '0');
! 1731: s++;
! 1732: goto ret1;
! 1733: }
! 1734: break;
! 1735: }
! 1736: }
! 1737: #ifndef No_leftright
! 1738: }
! 1739: #endif
! 1740: fast_failed:
! 1741: s = s0;
! 1742: value(d) = value(d2);
! 1743: k = k0;
! 1744: ilim = ilim0;
! 1745: }
! 1746:
! 1747: /* Do we have a "small" integer? */
! 1748:
! 1749: if (be >= 0 && k <= Int_max) {
! 1750: /* Yes. */
! 1751: ds = tens[k];
! 1752: if (ndigits < 0 && ilim <= 0) {
! 1753: S = mhi = 0;
! 1754: if (ilim < 0 || value(d) <= 5*ds)
! 1755: goto no_digits;
! 1756: goto one_digit;
! 1757: }
! 1758: for(i = 1;; i++) {
! 1759: L = value(d) / ds;
! 1760: value(d) -= L*ds;
! 1761: #ifdef Check_FLT_ROUNDS
! 1762: /* If FLT_ROUNDS == 2, L will usually be high by 1 */
! 1763: if (value(d) < 0) {
! 1764: L--;
! 1765: value(d) += ds;
! 1766: }
! 1767: #endif
! 1768: *s++ = '0' + (int)L;
! 1769: if (i == ilim) {
! 1770: value(d) += value(d);
! 1771: if (value(d) > ds || (value(d) == ds && (L & 1))) {
! 1772: bump_up:
! 1773: while(*--s == '9')
! 1774: if (s == s0) {
! 1775: k++;
! 1776: *s = '0';
! 1777: break;
! 1778: }
! 1779: ++*s++;
! 1780: }
! 1781: break;
! 1782: }
! 1783: if (!(value(d) *= 10.))
! 1784: break;
! 1785: }
! 1786: goto ret1;
! 1787: }
! 1788:
! 1789: m2 = b2;
! 1790: m5 = b5;
! 1791: mhi = mlo = 0;
! 1792: if (leftright) {
! 1793: if (mode < 2) {
! 1794: i =
! 1795: #ifndef Sudden_Underflow
! 1796: denorm ? be + (Bias + (P-1) - 1 + 1) :
! 1797: #endif
! 1798: #ifdef IBM
! 1799: 1 + 4*P - 3 - bbits + ((bbits + be - 1) & 3);
! 1800: #else
! 1801: 1 + P - bbits;
! 1802: #endif
! 1803: }
! 1804: else {
! 1805: j = ilim - 1;
! 1806: if (m5 >= j)
! 1807: m5 -= j;
! 1808: else {
! 1809: s5 += j -= m5;
! 1810: b5 += j;
! 1811: m5 = 0;
! 1812: }
! 1813: if ((i = ilim) < 0) {
! 1814: m2 -= i;
! 1815: i = 0;
! 1816: }
! 1817: }
! 1818: b2 += i;
! 1819: s2 += i;
! 1820: mhi = i2b(1);
! 1821: }
! 1822: if (m2 > 0 && s2 > 0) {
! 1823: i = m2 < s2 ? m2 : s2;
! 1824: b2 -= i;
! 1825: m2 -= i;
! 1826: s2 -= i;
! 1827: }
! 1828: if (b5 > 0) {
! 1829: if (leftright) {
! 1830: if (m5 > 0) {
! 1831: mhi = pow5mult(mhi, m5);
! 1832: b1 = mult(mhi, b);
! 1833: Bfree(b);
! 1834: b = b1;
! 1835: }
! 1836: if ((j = b5 - m5)) {
! 1837: b = pow5mult(b, j);
! 1838: }
! 1839: } else {
! 1840: b = pow5mult(b, b5);
! 1841: }
! 1842: }
! 1843: S = i2b(1);
! 1844: if (s5 > 0)
! 1845: S = pow5mult(S, s5);
! 1846: /* Check for special case that d is a normalized power of 2. */
! 1847:
! 1848: if (mode < 2) {
! 1849: if (!word1(d) && !(word0(d) & Bndry_mask)
! 1850: #ifndef Sudden_Underflow
! 1851: && word0(d) & Exp_mask
! 1852: #endif
! 1853: ) {
! 1854: /* The special case */
! 1855: b2 += Log2P;
! 1856: s2 += Log2P;
! 1857: spec_case = 1;
! 1858: } else {
! 1859: spec_case = 0;
! 1860: }
! 1861: }
! 1862:
! 1863: /* Arrange for convenient computation of quotients:
! 1864: * shift left if necessary so divisor has 4 leading 0 bits.
! 1865: *
! 1866: * Perhaps we should just compute leading 28 bits of S once
! 1867: * and for all and pass them and a shift to quorem, so it
! 1868: * can do shifts and ors to compute the numerator for q.
! 1869: */
! 1870: #ifdef Pack_32
! 1871: if ((i = ((s5 ? 32 - hi0bits(S->x[S->wds-1]) : 1) + s2) & 0x1f))
! 1872: i = 32 - i;
! 1873: #else
! 1874: if ((i = ((s5 ? 32 - hi0bits(S->x[S->wds-1]) : 1) + s2) & 0xf))
! 1875: i = 16 - i;
! 1876: #endif
! 1877: if (i > 4) {
! 1878: i -= 4;
! 1879: b2 += i;
! 1880: m2 += i;
! 1881: s2 += i;
! 1882: }
! 1883: else if (i < 4) {
! 1884: i += 28;
! 1885: b2 += i;
! 1886: m2 += i;
! 1887: s2 += i;
! 1888: }
! 1889: if (b2 > 0)
! 1890: b = lshift(b, b2);
! 1891: if (s2 > 0)
! 1892: S = lshift(S, s2);
! 1893: if (k_check) {
! 1894: if (cmp(b,S) < 0) {
! 1895: k--;
! 1896: b = multadd(b, 10, 0); /* we botched the k estimate */
! 1897: if (leftright)
! 1898: mhi = multadd(mhi, 10, 0);
! 1899: ilim = ilim1;
! 1900: }
! 1901: }
! 1902: if (ilim <= 0 && mode > 2) {
! 1903: if (ilim < 0 || cmp(b,S = multadd(S,5,0)) <= 0) {
! 1904: /* no digits, fcvt style */
! 1905: no_digits:
! 1906: k = -1 - ndigits;
! 1907: goto ret;
! 1908: }
! 1909: one_digit:
! 1910: *s++ = '1';
! 1911: k++;
! 1912: goto ret;
! 1913: }
! 1914: if (leftright) {
! 1915: if (m2 > 0)
! 1916: mhi = lshift(mhi, m2);
! 1917:
! 1918: /* Compute mlo -- check for special case
! 1919: * that d is a normalized power of 2.
! 1920: */
! 1921:
! 1922: mlo = mhi;
! 1923: if (spec_case) {
! 1924: mhi = Balloc(mhi->k);
! 1925: Bcopy(mhi, mlo);
! 1926: mhi = lshift(mhi, Log2P);
! 1927: }
! 1928:
! 1929: for(i = 1;;i++) {
! 1930: dig = quorem(b,S) + '0';
! 1931: /* Do we yet have the shortest decimal string
! 1932: * that will round to d?
! 1933: */
! 1934: j = cmp(b, mlo);
! 1935: delta = diff(S, mhi);
! 1936: j1 = delta->sign ? 1 : cmp(b, delta);
! 1937: Bfree(delta);
! 1938: #ifndef ROUND_BIASED
! 1939: if (j1 == 0 && !mode && !(word1(d) & 1)) {
! 1940: if (dig == '9')
! 1941: goto round_9_up;
! 1942: if (j > 0)
! 1943: dig++;
! 1944: *s++ = dig;
! 1945: goto ret;
! 1946: }
! 1947: #endif
! 1948: if (j < 0 || (j == 0 && !mode
! 1949: #ifndef ROUND_BIASED
! 1950: && !(word1(d) & 1)
! 1951: #endif
! 1952: )) {
! 1953: if (j1 > 0) {
! 1954: b = lshift(b, 1);
! 1955: j1 = cmp(b, S);
! 1956: if ((j1 > 0 || (j1 == 0 && (dig & 1)))
! 1957: && dig++ == '9')
! 1958: goto round_9_up;
! 1959: }
! 1960: *s++ = dig;
! 1961: goto ret;
! 1962: }
! 1963: if (j1 > 0) {
! 1964: if (dig == '9') { /* possible if i == 1 */
! 1965: round_9_up:
! 1966: *s++ = '9';
! 1967: goto roundoff;
! 1968: }
! 1969: *s++ = dig + 1;
! 1970: goto ret;
! 1971: }
! 1972: *s++ = dig;
! 1973: if (i == ilim)
! 1974: break;
! 1975: b = multadd(b, 10, 0);
! 1976: if (mlo == mhi)
! 1977: mlo = mhi = multadd(mhi, 10, 0);
! 1978: else {
! 1979: mlo = multadd(mlo, 10, 0);
! 1980: mhi = multadd(mhi, 10, 0);
! 1981: }
! 1982: }
! 1983: }
! 1984: else
! 1985: for(i = 1;; i++) {
! 1986: *s++ = dig = quorem(b,S) + '0';
! 1987: if (i >= ilim)
! 1988: break;
! 1989: b = multadd(b, 10, 0);
! 1990: }
! 1991:
! 1992: /* Round off last digit */
! 1993:
! 1994: b = lshift(b, 1);
! 1995: j = cmp(b, S);
! 1996: if (j > 0 || (j == 0 && (dig & 1))) {
! 1997: roundoff:
! 1998: while(*--s == '9')
! 1999: if (s == s0) {
! 2000: k++;
! 2001: *s++ = '1';
! 2002: goto ret;
! 2003: }
! 2004: ++*s++;
! 2005: }
! 2006: else {
! 2007: while(*--s == '0');
! 2008: s++;
! 2009: }
! 2010: ret:
! 2011: Bfree(S);
! 2012: if (mhi) {
! 2013: if (mlo && mlo != mhi)
! 2014: Bfree(mlo);
! 2015: Bfree(mhi);
! 2016: }
! 2017: ret1:
! 2018:
! 2019: _THREAD_PRIVATE_MUTEX_LOCK(pow5mult_mutex);
! 2020: while (p5s) {
! 2021: tmp = p5s;
! 2022: p5s = p5s->next;
! 2023: free(tmp);
! 2024: }
! 2025: _THREAD_PRIVATE_MUTEX_UNLOCK(pow5mult_mutex);
! 2026:
! 2027: Bfree(b);
! 2028:
! 2029: if (s == s0) { /* don't return empty string */
! 2030: *s++ = '0';
! 2031: k = 0;
! 2032: }
! 2033: *s = 0;
! 2034: *decpt = k + 1;
! 2035: if (rve)
! 2036: *rve = s;
! 2037: return s0;
! 2038: }
! 2039:
! 2040: /* F* VC6 */
! 2041: #if _MSC_VER <= 1300
! 2042: # pragma optimize( "", off )
! 2043: #endif
! 2044: ZEND_API double zend_strtod (CONST char *s00, char **se)
! 2045: {
! 2046: int bb2, bb5, bbe, bd2, bd5, bbbits, bs2, c, dsign,
! 2047: e, e1, esign, i, j, k, nd, nd0, nf, nz, nz0, sign;
! 2048: CONST char *s, *s0, *s1;
! 2049: volatile double aadj, aadj1, adj;
! 2050: volatile _double rv, rv0;
! 2051: Long L;
! 2052: ULong y, z;
! 2053: Bigint *bb, *bb1, *bd, *bd0, *bs, *delta, *tmp;
! 2054: double result;
! 2055:
! 2056: CONST char decimal_point = '.';
! 2057:
! 2058: sign = nz0 = nz = 0;
! 2059: value(rv) = 0.;
! 2060:
! 2061:
! 2062: for(s = s00; isspace((unsigned char) *s); s++)
! 2063: ;
! 2064:
! 2065: if (*s == '-') {
! 2066: sign = 1;
! 2067: s++;
! 2068: } else if (*s == '+') {
! 2069: s++;
! 2070: }
! 2071:
! 2072: if (*s == '\0') {
! 2073: s = s00;
! 2074: goto ret;
! 2075: }
! 2076:
! 2077: if (*s == '0') {
! 2078: nz0 = 1;
! 2079: while(*++s == '0') ;
! 2080: if (!*s)
! 2081: goto ret;
! 2082: }
! 2083: s0 = s;
! 2084: y = z = 0;
! 2085: for(nd = nf = 0; (c = *s) >= '0' && c <= '9'; nd++, s++)
! 2086: if (nd < 9)
! 2087: y = 10*y + c - '0';
! 2088: else if (nd < 16)
! 2089: z = 10*z + c - '0';
! 2090: nd0 = nd;
! 2091: if (c == decimal_point) {
! 2092: c = *++s;
! 2093: if (!nd) {
! 2094: for(; c == '0'; c = *++s)
! 2095: nz++;
! 2096: if (c > '0' && c <= '9') {
! 2097: s0 = s;
! 2098: nf += nz;
! 2099: nz = 0;
! 2100: goto have_dig;
! 2101: }
! 2102: goto dig_done;
! 2103: }
! 2104: for(; c >= '0' && c <= '9'; c = *++s) {
! 2105: have_dig:
! 2106: nz++;
! 2107: if (c -= '0') {
! 2108: nf += nz;
! 2109: for(i = 1; i < nz; i++)
! 2110: if (nd++ < 9)
! 2111: y *= 10;
! 2112: else if (nd <= DBL_DIG + 1)
! 2113: z *= 10;
! 2114: if (nd++ < 9)
! 2115: y = 10*y + c;
! 2116: else if (nd <= DBL_DIG + 1)
! 2117: z = 10*z + c;
! 2118: nz = 0;
! 2119: }
! 2120: }
! 2121: }
! 2122: dig_done:
! 2123: e = 0;
! 2124: if (c == 'e' || c == 'E') {
! 2125: if (!nd && !nz && !nz0) {
! 2126: s = s00;
! 2127: goto ret;
! 2128: }
! 2129: s00 = s;
! 2130: esign = 0;
! 2131: switch(c = *++s) {
! 2132: case '-':
! 2133: esign = 1;
! 2134: case '+':
! 2135: c = *++s;
! 2136: }
! 2137: if (c >= '0' && c <= '9') {
! 2138: while(c == '0')
! 2139: c = *++s;
! 2140: if (c > '0' && c <= '9') {
! 2141: L = c - '0';
! 2142: s1 = s;
! 2143: while((c = *++s) >= '0' && c <= '9')
! 2144: L = 10*L + c - '0';
! 2145: if (s - s1 > 8 || L > 19999)
! 2146: /* Avoid confusion from exponents
! 2147: * so large that e might overflow.
! 2148: */
! 2149: e = 19999; /* safe for 16 bit ints */
! 2150: else
! 2151: e = (int)L;
! 2152: if (esign)
! 2153: e = -e;
! 2154: }
! 2155: else
! 2156: e = 0;
! 2157: }
! 2158: else
! 2159: s = s00;
! 2160: }
! 2161: if (!nd) {
! 2162: if (!nz && !nz0)
! 2163: s = s00;
! 2164: goto ret;
! 2165: }
! 2166: e1 = e -= nf;
! 2167:
! 2168: /* Now we have nd0 digits, starting at s0, followed by a
! 2169: * decimal point, followed by nd-nd0 digits. The number we're
! 2170: * after is the integer represented by those digits times
! 2171: * 10**e */
! 2172:
! 2173: if (!nd0)
! 2174: nd0 = nd;
! 2175: k = nd < DBL_DIG + 1 ? nd : DBL_DIG + 1;
! 2176: value(rv) = y;
! 2177: if (k > 9)
! 2178: value(rv) = tens[k - 9] * value(rv) + z;
! 2179: bd0 = 0;
! 2180: if (nd <= DBL_DIG
! 2181: #ifndef RND_PRODQUOT
! 2182: && FLT_ROUNDS == 1
! 2183: #endif
! 2184: ) {
! 2185: if (!e)
! 2186: goto ret;
! 2187: if (e > 0) {
! 2188: if (e <= Ten_pmax) {
! 2189: #ifdef VAX
! 2190: goto vax_ovfl_check;
! 2191: #else
! 2192: /* value(rv) = */ rounded_product(value(rv),
! 2193: tens[e]);
! 2194: goto ret;
! 2195: #endif
! 2196: }
! 2197: i = DBL_DIG - nd;
! 2198: if (e <= Ten_pmax + i) {
! 2199: /* A fancier test would sometimes let us do
! 2200: * this for larger i values.
! 2201: */
! 2202: e -= i;
! 2203: value(rv) *= tens[i];
! 2204: #ifdef VAX
! 2205: /* VAX exponent range is so narrow we must
! 2206: * worry about overflow here...
! 2207: */
! 2208: vax_ovfl_check:
! 2209: word0(rv) -= P*Exp_msk1;
! 2210: /* value(rv) = */ rounded_product(value(rv),
! 2211: tens[e]);
! 2212: if ((word0(rv) & Exp_mask)
! 2213: > Exp_msk1*(DBL_MAX_EXP+Bias-1-P))
! 2214: goto ovfl;
! 2215: word0(rv) += P*Exp_msk1;
! 2216: #else
! 2217: /* value(rv) = */ rounded_product(value(rv),
! 2218: tens[e]);
! 2219: #endif
! 2220: goto ret;
! 2221: }
! 2222: }
! 2223: #ifndef Inaccurate_Divide
! 2224: else if (e >= -Ten_pmax) {
! 2225: /* value(rv) = */ rounded_quotient(value(rv),
! 2226: tens[-e]);
! 2227: goto ret;
! 2228: }
! 2229: #endif
! 2230: }
! 2231: e1 += nd - k;
! 2232:
! 2233: /* Get starting approximation = rv * 10**e1 */
! 2234:
! 2235: if (e1 > 0) {
! 2236: if ((i = e1 & 15))
! 2237: value(rv) *= tens[i];
! 2238: if (e1 &= ~15) {
! 2239: if (e1 > DBL_MAX_10_EXP) {
! 2240: ovfl:
! 2241: errno = ERANGE;
! 2242: #ifndef Bad_float_h
! 2243: value(rv) = HUGE_VAL;
! 2244: #else
! 2245: /* Can't trust HUGE_VAL */
! 2246: #ifdef IEEE_Arith
! 2247: word0(rv) = Exp_mask;
! 2248: word1(rv) = 0;
! 2249: #else
! 2250: word0(rv) = Big0;
! 2251: word1(rv) = Big1;
! 2252: #endif
! 2253: #endif
! 2254: if (bd0)
! 2255: goto retfree;
! 2256: goto ret;
! 2257: }
! 2258: if (e1 >>= 4) {
! 2259: for(j = 0; e1 > 1; j++, e1 >>= 1)
! 2260: if (e1 & 1)
! 2261: value(rv) *= bigtens[j];
! 2262: /* The last multiplication could overflow. */
! 2263: word0(rv) -= P*Exp_msk1;
! 2264: value(rv) *= bigtens[j];
! 2265: if ((z = word0(rv) & Exp_mask)
! 2266: > Exp_msk1*(DBL_MAX_EXP+Bias-P))
! 2267: goto ovfl;
! 2268: if (z > Exp_msk1*(DBL_MAX_EXP+Bias-1-P)) {
! 2269: /* set to largest number */
! 2270: /* (Can't trust DBL_MAX) */
! 2271: word0(rv) = Big0;
! 2272: word1(rv) = Big1;
! 2273: }
! 2274: else
! 2275: word0(rv) += P*Exp_msk1;
! 2276: }
! 2277:
! 2278: }
! 2279: }
! 2280: else if (e1 < 0) {
! 2281: e1 = -e1;
! 2282: if ((i = e1 & 15))
! 2283: value(rv) /= tens[i];
! 2284: if (e1 &= ~15) {
! 2285: e1 >>= 4;
! 2286: if (e1 >= 1 << n_bigtens)
! 2287: goto undfl;
! 2288: for(j = 0; e1 > 1; j++, e1 >>= 1)
! 2289: if (e1 & 1)
! 2290: value(rv) *= tinytens[j];
! 2291: /* The last multiplication could underflow. */
! 2292: value(rv0) = value(rv);
! 2293: value(rv) *= tinytens[j];
! 2294: if (!value(rv)) {
! 2295: value(rv) = 2.*value(rv0);
! 2296: value(rv) *= tinytens[j];
! 2297: if (!value(rv)) {
! 2298: undfl:
! 2299: value(rv) = 0.;
! 2300: errno = ERANGE;
! 2301: if (bd0)
! 2302: goto retfree;
! 2303: goto ret;
! 2304: }
! 2305: word0(rv) = Tiny0;
! 2306: word1(rv) = Tiny1;
! 2307: /* The refinement below will clean
! 2308: * this approximation up.
! 2309: */
! 2310: }
! 2311: }
! 2312: }
! 2313:
! 2314: /* Now the hard part -- adjusting rv to the correct value.*/
! 2315:
! 2316: /* Put digits into bd: true value = bd * 10^e */
! 2317:
! 2318: bd0 = s2b(s0, nd0, nd, y);
! 2319:
! 2320: for(;;) {
! 2321: bd = Balloc(bd0->k);
! 2322: Bcopy(bd, bd0);
! 2323: bb = d2b(value(rv), &bbe, &bbbits); /* rv = bb * 2^bbe */
! 2324: bs = i2b(1);
! 2325:
! 2326: if (e >= 0) {
! 2327: bb2 = bb5 = 0;
! 2328: bd2 = bd5 = e;
! 2329: }
! 2330: else {
! 2331: bb2 = bb5 = -e;
! 2332: bd2 = bd5 = 0;
! 2333: }
! 2334: if (bbe >= 0)
! 2335: bb2 += bbe;
! 2336: else
! 2337: bd2 -= bbe;
! 2338: bs2 = bb2;
! 2339: #ifdef Sudden_Underflow
! 2340: #ifdef IBM
! 2341: j = 1 + 4*P - 3 - bbbits + ((bbe + bbbits - 1) & 3);
! 2342: #else
! 2343: j = P + 1 - bbbits;
! 2344: #endif
! 2345: #else
! 2346: i = bbe + bbbits - 1; /* logb(rv) */
! 2347: if (i < Emin) /* denormal */
! 2348: j = bbe + (P-Emin);
! 2349: else
! 2350: j = P + 1 - bbbits;
! 2351: #endif
! 2352: bb2 += j;
! 2353: bd2 += j;
! 2354: i = bb2 < bd2 ? bb2 : bd2;
! 2355: if (i > bs2)
! 2356: i = bs2;
! 2357: if (i > 0) {
! 2358: bb2 -= i;
! 2359: bd2 -= i;
! 2360: bs2 -= i;
! 2361: }
! 2362: if (bb5 > 0) {
! 2363: bs = pow5mult(bs, bb5);
! 2364: bb1 = mult(bs, bb);
! 2365: Bfree(bb);
! 2366: bb = bb1;
! 2367: }
! 2368: if (bb2 > 0)
! 2369: bb = lshift(bb, bb2);
! 2370: if (bd5 > 0)
! 2371: bd = pow5mult(bd, bd5);
! 2372: if (bd2 > 0)
! 2373: bd = lshift(bd, bd2);
! 2374: if (bs2 > 0)
! 2375: bs = lshift(bs, bs2);
! 2376: delta = diff(bb, bd);
! 2377: dsign = delta->sign;
! 2378: delta->sign = 0;
! 2379: i = cmp(delta, bs);
! 2380: if (i < 0) {
! 2381: /* Error is less than half an ulp -- check for
! 2382: * special case of mantissa a power of two.
! 2383: */
! 2384: if (dsign || word1(rv) || word0(rv) & Bndry_mask)
! 2385: break;
! 2386: delta = lshift(delta,Log2P);
! 2387: if (cmp(delta, bs) > 0)
! 2388: goto drop_down;
! 2389: break;
! 2390: }
! 2391: if (i == 0) {
! 2392: /* exactly half-way between */
! 2393: if (dsign) {
! 2394: if ((word0(rv) & Bndry_mask1) == Bndry_mask1
! 2395: && word1(rv) == 0xffffffff) {
! 2396: /*boundary case -- increment exponent*/
! 2397: word0(rv) = (word0(rv) & Exp_mask)
! 2398: + Exp_msk1
! 2399: #ifdef IBM
! 2400: | Exp_msk1 >> 4
! 2401: #endif
! 2402: ;
! 2403: word1(rv) = 0;
! 2404: break;
! 2405: }
! 2406: }
! 2407: else if (!(word0(rv) & Bndry_mask) && !word1(rv)) {
! 2408: drop_down:
! 2409: /* boundary case -- decrement exponent */
! 2410: #ifdef Sudden_Underflow
! 2411: L = word0(rv) & Exp_mask;
! 2412: #ifdef IBM
! 2413: if (L < Exp_msk1)
! 2414: #else
! 2415: if (L <= Exp_msk1)
! 2416: #endif
! 2417: goto undfl;
! 2418: L -= Exp_msk1;
! 2419: #else
! 2420: L = (word0(rv) & Exp_mask) - Exp_msk1;
! 2421: #endif
! 2422: word0(rv) = L | Bndry_mask1;
! 2423: word1(rv) = 0xffffffff;
! 2424: #ifdef IBM
! 2425: goto cont;
! 2426: #else
! 2427: break;
! 2428: #endif
! 2429: }
! 2430: #ifndef ROUND_BIASED
! 2431: if (!(word1(rv) & LSB))
! 2432: break;
! 2433: #endif
! 2434: if (dsign)
! 2435: value(rv) += ulp(value(rv));
! 2436: #ifndef ROUND_BIASED
! 2437: else {
! 2438: value(rv) -= ulp(value(rv));
! 2439: #ifndef Sudden_Underflow
! 2440: if (!value(rv))
! 2441: goto undfl;
! 2442: #endif
! 2443: }
! 2444: #endif
! 2445: break;
! 2446: }
! 2447: if ((aadj = ratio(delta, bs)) <= 2.) {
! 2448: if (dsign)
! 2449: aadj = aadj1 = 1.;
! 2450: else if (word1(rv) || word0(rv) & Bndry_mask) {
! 2451: #ifndef Sudden_Underflow
! 2452: if (word1(rv) == Tiny1 && !word0(rv))
! 2453: goto undfl;
! 2454: #endif
! 2455: aadj = 1.;
! 2456: aadj1 = -1.;
! 2457: }
! 2458: else {
! 2459: /* special case -- power of FLT_RADIX to be */
! 2460: /* rounded down... */
! 2461:
! 2462: if (aadj < 2./FLT_RADIX)
! 2463: aadj = 1./FLT_RADIX;
! 2464: else
! 2465: aadj *= 0.5;
! 2466: aadj1 = -aadj;
! 2467: }
! 2468: }
! 2469: else {
! 2470: aadj *= 0.5;
! 2471: aadj1 = dsign ? aadj : -aadj;
! 2472: #ifdef Check_FLT_ROUNDS
! 2473: switch(FLT_ROUNDS) {
! 2474: case 2: /* towards +infinity */
! 2475: aadj1 -= 0.5;
! 2476: break;
! 2477: case 0: /* towards 0 */
! 2478: case 3: /* towards -infinity */
! 2479: aadj1 += 0.5;
! 2480: }
! 2481: #else
! 2482: if (FLT_ROUNDS == 0)
! 2483: aadj1 += 0.5;
! 2484: #endif
! 2485: }
! 2486: y = word0(rv) & Exp_mask;
! 2487:
! 2488: /* Check for overflow */
! 2489:
! 2490: if (y == Exp_msk1*(DBL_MAX_EXP+Bias-1)) {
! 2491: value(rv0) = value(rv);
! 2492: word0(rv) -= P*Exp_msk1;
! 2493: adj = aadj1 * ulp(value(rv));
! 2494: value(rv) += adj;
! 2495: if ((word0(rv) & Exp_mask) >=
! 2496: Exp_msk1*(DBL_MAX_EXP+Bias-P)) {
! 2497: if (word0(rv0) == Big0 && word1(rv0) == Big1)
! 2498: goto ovfl;
! 2499: word0(rv) = Big0;
! 2500: word1(rv) = Big1;
! 2501: goto cont;
! 2502: }
! 2503: else
! 2504: word0(rv) += P*Exp_msk1;
! 2505: }
! 2506: else {
! 2507: #ifdef Sudden_Underflow
! 2508: if ((word0(rv) & Exp_mask) <= P*Exp_msk1) {
! 2509: value(rv0) = value(rv);
! 2510: word0(rv) += P*Exp_msk1;
! 2511: adj = aadj1 * ulp(value(rv));
! 2512: value(rv) += adj;
! 2513: #ifdef IBM
! 2514: if ((word0(rv) & Exp_mask) < P*Exp_msk1)
! 2515: #else
! 2516: if ((word0(rv) & Exp_mask) <= P*Exp_msk1)
! 2517: #endif
! 2518: {
! 2519: if (word0(rv0) == Tiny0
! 2520: && word1(rv0) == Tiny1)
! 2521: goto undfl;
! 2522: word0(rv) = Tiny0;
! 2523: word1(rv) = Tiny1;
! 2524: goto cont;
! 2525: }
! 2526: else
! 2527: word0(rv) -= P*Exp_msk1;
! 2528: }
! 2529: else {
! 2530: adj = aadj1 * ulp(value(rv));
! 2531: value(rv) += adj;
! 2532: }
! 2533: #else
! 2534: /* Compute adj so that the IEEE rounding rules will
! 2535: * correctly round rv + adj in some half-way cases.
! 2536: * If rv * ulp(rv) is denormalized (i.e.,
! 2537: * y <= (P-1)*Exp_msk1), we must adjust aadj to avoid
! 2538: * trouble from bits lost to denormalization;
! 2539: * example: 1.2e-307 .
! 2540: */
! 2541: if (y <= (P-1)*Exp_msk1 && aadj >= 1.) {
! 2542: aadj1 = (double)(int)(aadj + 0.5);
! 2543: if (!dsign)
! 2544: aadj1 = -aadj1;
! 2545: }
! 2546: adj = aadj1 * ulp(value(rv));
! 2547: value(rv) += adj;
! 2548: #endif
! 2549: }
! 2550: z = word0(rv) & Exp_mask;
! 2551: if (y == z) {
! 2552: /* Can we stop now? */
! 2553: L = aadj;
! 2554: aadj -= L;
! 2555: /* The tolerances below are conservative. */
! 2556: if (dsign || word1(rv) || word0(rv) & Bndry_mask) {
! 2557: if (aadj < .4999999 || aadj > .5000001)
! 2558: break;
! 2559: }
! 2560: else if (aadj < .4999999/FLT_RADIX)
! 2561: break;
! 2562: }
! 2563: cont:
! 2564: Bfree(bb);
! 2565: Bfree(bd);
! 2566: Bfree(bs);
! 2567: Bfree(delta);
! 2568: }
! 2569: retfree:
! 2570: Bfree(bb);
! 2571: Bfree(bd);
! 2572: Bfree(bs);
! 2573: Bfree(bd0);
! 2574: Bfree(delta);
! 2575: ret:
! 2576: if (se)
! 2577: *se = (char *)s;
! 2578: result = sign ? -value(rv) : value(rv);
! 2579:
! 2580: _THREAD_PRIVATE_MUTEX_LOCK(pow5mult_mutex);
! 2581: while (p5s) {
! 2582: tmp = p5s;
! 2583: p5s = p5s->next;
! 2584: free(tmp);
! 2585: }
! 2586: _THREAD_PRIVATE_MUTEX_UNLOCK(pow5mult_mutex);
! 2587:
! 2588: return result;
! 2589: }
! 2590:
! 2591: ZEND_API double zend_hex_strtod(const char *str, char **endptr)
! 2592: {
! 2593: const char *s = str;
! 2594: char c;
! 2595: int any = 0;
! 2596: double value = 0;
! 2597:
! 2598: if (*s == '0' && (s[1] == 'x' || s[1] == 'X')) {
! 2599: s += 2;
! 2600: }
! 2601:
! 2602: while ((c = *s++)) {
! 2603: if (c >= '0' && c <= '9') {
! 2604: c -= '0';
! 2605: } else if (c >= 'A' && c <= 'F') {
! 2606: c -= 'A' - 10;
! 2607: } else if (c >= 'a' && c <= 'f') {
! 2608: c -= 'a' - 10;
! 2609: } else {
! 2610: break;
! 2611: }
! 2612:
! 2613: any = 1;
! 2614: value = value * 16 + c;
! 2615: }
! 2616:
! 2617: if (endptr != NULL) {
! 2618: *endptr = (char *)(any ? s - 1 : str);
! 2619: }
! 2620:
! 2621: return value;
! 2622: }
! 2623:
! 2624: ZEND_API double zend_oct_strtod(const char *str, char **endptr)
! 2625: {
! 2626: const char *s = str;
! 2627: char c;
! 2628: double value = 0;
! 2629: int any = 0;
! 2630:
! 2631: /* skip leading zero */
! 2632: s++;
! 2633:
! 2634: while ((c = *s++)) {
! 2635: if (c < '0' || c > '7') {
! 2636: /* break and return the current value if the number is not well-formed
! 2637: * that's what Linux strtol() does
! 2638: */
! 2639: break;
! 2640: }
! 2641: value = value * 8 + c - '0';
! 2642: any = 1;
! 2643: }
! 2644:
! 2645: if (endptr != NULL) {
! 2646: *endptr = (char *)(any ? s - 1 : str);
! 2647: }
! 2648:
! 2649: return value;
! 2650: }
! 2651:
! 2652: /*
! 2653: * Local variables:
! 2654: * tab-width: 4
! 2655: * c-basic-offset: 4
! 2656: * End:
! 2657: * vim600: sw=4 ts=4 fdm=marker
! 2658: * vim<600: sw=4 ts=4
! 2659: */
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>
![[BACK]](/icons/cvsweb/back.gif){kind=icon}
Return to zend_strtod.c CVS log ![[TXT]](/icons/cvsweb/text.gif){kind=icon}
![[DIR]](/icons/cvsweb/dir.gif){kind=icon}
Up to [ELWIX - Embedded LightWeight unIX -] / embedaddon / php / Zend