2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 1991-2000, University of Groningen, The Netherlands.
5 * Copyright (c) 2001-2004, The GROMACS development team.
6 * Copyright (c) 2013,2014, by the GROMACS development team, led by
7 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
8 * and including many others, as listed in the AUTHORS file in the
9 * top-level source directory and at http://www.gromacs.org.
11 * GROMACS is free software; you can redistribute it and/or
12 * modify it under the terms of the GNU Lesser General Public License
13 * as published by the Free Software Foundation; either version 2.1
14 * of the License, or (at your option) any later version.
16 * GROMACS is distributed in the hope that it will be useful,
17 * but WITHOUT ANY WARRANTY; without even the implied warranty of
18 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
19 * Lesser General Public License for more details.
21 * You should have received a copy of the GNU Lesser General Public
22 * License along with GROMACS; if not, see
23 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
24 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
26 * If you want to redistribute modifications to GROMACS, please
27 * consider that scientific software is very special. Version
28 * control is crucial - bugs must be traceable. We will be happy to
29 * consider code for inclusion in the official distribution, but
30 * derived work must not be called official GROMACS. Details are found
31 * in the README & COPYING files - if they are missing, get the
32 * official version at http://www.gromacs.org.
34 * To help us fund GROMACS development, we humbly ask that you cite
35 * the research papers on the package. Check out http://www.gromacs.org.
39 #include "utilities.h"
56 result = (a < 0.) ? ((int)(a - half)) : ((int)(a + half));
64 return (-pow(-x, 1.0/3.0));
68 return (pow(x, 1.0/3.0));
72 real sign(real x, real y)
84 /* Double and single precision erf() and erfc() from
85 * the Sun Freely Distributable Math Library FDLIBM.
86 * See http://www.netlib.org/fdlibm
87 * Specific file used: s_erf.c, version 1.3 95/01/18
90 * ====================================================
91 * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
93 * Developed at SunSoft, a Sun Microsystems, Inc. business.
94 * Permission to use, copy, modify, and distribute this
95 * software is freely granted, provided that this notice
97 * ====================================================
102 half = 5.00000000000000000000e-01, /* 0x3FE00000, 0x00000000 */
103 one = 1.00000000000000000000e+00, /* 0x3FF00000, 0x00000000 */
104 two = 2.00000000000000000000e+00, /* 0x40000000, 0x00000000 */
105 /* c = (float)0.84506291151 */
106 erx = 8.45062911510467529297e-01, /* 0x3FEB0AC1, 0x60000000 */
108 * Coefficients for approximation to erf on [0,0.84375]
110 efx = 1.28379167095512586316e-01, /* 0x3FC06EBA, 0x8214DB69 */
111 efx8 = 1.02703333676410069053e+00, /* 0x3FF06EBA, 0x8214DB69 */
112 pp0 = 1.28379167095512558561e-01, /* 0x3FC06EBA, 0x8214DB68 */
113 pp1 = -3.25042107247001499370e-01, /* 0xBFD4CD7D, 0x691CB913 */
114 pp2 = -2.84817495755985104766e-02, /* 0xBF9D2A51, 0xDBD7194F */
115 pp3 = -5.77027029648944159157e-03, /* 0xBF77A291, 0x236668E4 */
116 pp4 = -2.37630166566501626084e-05, /* 0xBEF8EAD6, 0x120016AC */
117 qq1 = 3.97917223959155352819e-01, /* 0x3FD97779, 0xCDDADC09 */
118 qq2 = 6.50222499887672944485e-02, /* 0x3FB0A54C, 0x5536CEBA */
119 qq3 = 5.08130628187576562776e-03, /* 0x3F74D022, 0xC4D36B0F */
120 qq4 = 1.32494738004321644526e-04, /* 0x3F215DC9, 0x221C1A10 */
121 qq5 = -3.96022827877536812320e-06, /* 0xBED09C43, 0x42A26120 */
123 * Coefficients for approximation to erf in [0.84375,1.25]
125 pa0 = -2.36211856075265944077e-03, /* 0xBF6359B8, 0xBEF77538 */
126 pa1 = 4.14856118683748331666e-01, /* 0x3FDA8D00, 0xAD92B34D */
127 pa2 = -3.72207876035701323847e-01, /* 0xBFD7D240, 0xFBB8C3F1 */
128 pa3 = 3.18346619901161753674e-01, /* 0x3FD45FCA, 0x805120E4 */
129 pa4 = -1.10894694282396677476e-01, /* 0xBFBC6398, 0x3D3E28EC */
130 pa5 = 3.54783043256182359371e-02, /* 0x3FA22A36, 0x599795EB */
131 pa6 = -2.16637559486879084300e-03, /* 0xBF61BF38, 0x0A96073F */
132 qa1 = 1.06420880400844228286e-01, /* 0x3FBB3E66, 0x18EEE323 */
133 qa2 = 5.40397917702171048937e-01, /* 0x3FE14AF0, 0x92EB6F33 */
134 qa3 = 7.18286544141962662868e-02, /* 0x3FB2635C, 0xD99FE9A7 */
135 qa4 = 1.26171219808761642112e-01, /* 0x3FC02660, 0xE763351F */
136 qa5 = 1.36370839120290507362e-02, /* 0x3F8BEDC2, 0x6B51DD1C */
137 qa6 = 1.19844998467991074170e-02, /* 0x3F888B54, 0x5735151D */
139 * Coefficients for approximation to erfc in [1.25,1/0.35]
141 ra0 = -9.86494403484714822705e-03, /* 0xBF843412, 0x600D6435 */
142 ra1 = -6.93858572707181764372e-01, /* 0xBFE63416, 0xE4BA7360 */
143 ra2 = -1.05586262253232909814e+01, /* 0xC0251E04, 0x41B0E726 */
144 ra3 = -6.23753324503260060396e+01, /* 0xC04F300A, 0xE4CBA38D */
145 ra4 = -1.62396669462573470355e+02, /* 0xC0644CB1, 0x84282266 */
146 ra5 = -1.84605092906711035994e+02, /* 0xC067135C, 0xEBCCABB2 */
147 ra6 = -8.12874355063065934246e+01, /* 0xC0545265, 0x57E4D2F2 */
148 ra7 = -9.81432934416914548592e+00, /* 0xC023A0EF, 0xC69AC25C */
149 sa1 = 1.96512716674392571292e+01, /* 0x4033A6B9, 0xBD707687 */
150 sa2 = 1.37657754143519042600e+02, /* 0x4061350C, 0x526AE721 */
151 sa3 = 4.34565877475229228821e+02, /* 0x407B290D, 0xD58A1A71 */
152 sa4 = 6.45387271733267880336e+02, /* 0x40842B19, 0x21EC2868 */
153 sa5 = 4.29008140027567833386e+02, /* 0x407AD021, 0x57700314 */
154 sa6 = 1.08635005541779435134e+02, /* 0x405B28A3, 0xEE48AE2C */
155 sa7 = 6.57024977031928170135e+00, /* 0x401A47EF, 0x8E484A93 */
156 sa8 = -6.04244152148580987438e-02, /* 0xBFAEEFF2, 0xEE749A62 */
158 * Coefficients for approximation to erfc in [1/.35,28]
160 rb0 = -9.86494292470009928597e-03, /* 0xBF843412, 0x39E86F4A */
161 rb1 = -7.99283237680523006574e-01, /* 0xBFE993BA, 0x70C285DE */
162 rb2 = -1.77579549177547519889e+01, /* 0xC031C209, 0x555F995A */
163 rb3 = -1.60636384855821916062e+02, /* 0xC064145D, 0x43C5ED98 */
164 rb4 = -6.37566443368389627722e+02, /* 0xC083EC88, 0x1375F228 */
165 rb5 = -1.02509513161107724954e+03, /* 0xC0900461, 0x6A2E5992 */
166 rb6 = -4.83519191608651397019e+02, /* 0xC07E384E, 0x9BDC383F */
167 sb1 = 3.03380607434824582924e+01, /* 0x403E568B, 0x261D5190 */
168 sb2 = 3.25792512996573918826e+02, /* 0x40745CAE, 0x221B9F0A */
169 sb3 = 1.53672958608443695994e+03, /* 0x409802EB, 0x189D5118 */
170 sb4 = 3.19985821950859553908e+03, /* 0x40A8FFB7, 0x688C246A */
171 sb5 = 2.55305040643316442583e+03, /* 0x40A3F219, 0xCEDF3BE6 */
172 sb6 = 4.74528541206955367215e+02, /* 0x407DA874, 0xE79FE763 */
173 sb7 = -2.24409524465858183362e+01; /* 0xC03670E2, 0x42712D62 */
175 double gmx_erfd(double x)
177 #ifdef GMX_FLOAT_FORMAT_IEEE754
178 gmx_int32_t hx, ix, i;
179 double R, S, P, Q, s, y, z, r;
190 #ifdef GMX_IEEE754_BIG_ENDIAN_WORD_ORDER
197 if (ix >= 0x7ff00000)
200 i = ((gmx_uint32_t)hx>>31)<<1;
201 return (double)(1-i)+one/x; /* erf(+-inf)=+-1 */
212 return 0.125*(8.0*x+efx8*x); /*avoid underflow */
217 r = pp0+z*(pp1+z*(pp2+z*(pp3+z*pp4)));
218 s = one+z*(qq1+z*(qq2+z*(qq3+z*(qq4+z*qq5))));
224 /* 0.84375 <= |x| < 1.25 */
226 P = pa0+s*(pa1+s*(pa2+s*(pa3+s*(pa4+s*(pa5+s*pa6)))));
227 Q = one+s*(qa1+s*(qa2+s*(qa3+s*(qa4+s*(qa5+s*qa6)))));
237 if (ix >= 0x40180000)
254 R = ra0+s*(ra1+s*(ra2+s*(ra3+s*(ra4+s*(ra5+s*(ra6+s*ra7))))));
255 S = one+s*(sa1+s*(sa2+s*(sa3+s*(sa4+s*(sa5+s*(sa6+s*(sa7+s*sa8)))))));
260 R = rb0+s*(rb1+s*(rb2+s*(rb3+s*(rb4+s*(rb5+s*rb6)))));
261 S = one+s*(sb1+s*(sb2+s*(sb3+s*(sb4+s*(sb5+s*(sb6+s*sb7))))));
266 #ifdef GMX_IEEE754_BIG_ENDIAN_WORD_ORDER
274 r = exp(-z*z-0.5625)*exp((z-x)*(z+x)+R/S);
284 /* No IEEE754 information. We need to trust that the OS provides erf(). */
290 double gmx_erfcd(double x)
292 #ifdef GMX_FLOAT_FORMAT_IEEE754
294 double R, S, P, Q, s, y, z, r;
305 #ifdef GMX_IEEE754_BIG_ENDIAN_WORD_ORDER
312 if (ix >= 0x7ff00000)
315 /* erfc(+-inf)=0,2 */
316 return (double)(((gmx_uint32_t)hx>>31)<<1)+one/x;
322 double r1, r2, s1, s2, s3, z2, z4;
323 if (ix < 0x3c700000) /* |x|<2**-56 */
328 r = pp0+z*(pp1+z*(pp2+z*(pp3+z*pp4)));
329 s = one+z*(qq1+z*(qq2+z*(qq3+z*(qq4+z*qq5))));
346 /* 0.84375 <= |x| < 1.25 */
348 P = pa0+s*(pa1+s*(pa2+s*(pa3+s*(pa4+s*(pa5+s*pa6)))));
349 Q = one+s*(qa1+s*(qa2+s*(qa3+s*(qa4+s*(qa5+s*qa6)))));
352 z = one-erx; return z - P/Q;
356 z = erx+P/Q; return one+z;
366 /* |x| < 1/.35 ~ 2.857143*/
367 R = ra0+s*(ra1+s*(ra2+s*(ra3+s*(ra4+s*(ra5+s*(ra6+s*ra7))))));
368 S = one+s*(sa1+s*(sa2+s*(sa3+s*(sa4+s*(sa5+s*(sa6+s*(sa7+s*sa8)))))));
372 /* |x| >= 1/.35 ~ 2.857143 */
373 if (hx < 0 && ix >= 0x40180000)
375 return two-tiny; /* x < -6 */
377 R = rb0+s*(rb1+s*(rb2+s*(rb3+s*(rb4+s*(rb5+s*rb6)))));
378 S = one+s*(sb1+s*(sb2+s*(sb3+s*(sb4+s*(sb5+s*(sb6+s*sb7))))));
383 #ifdef GMX_IEEE754_BIG_ENDIAN_WORD_ORDER
391 r = exp(-z*z-0.5625)*exp((z-x)*(z+x)+R/S);
414 /* No IEEE754 information. We need to trust that the OS provides erfc(). */
422 halff = 5.0000000000e-01, /* 0x3F000000 */
423 onef = 1.0000000000e+00, /* 0x3F800000 */
424 twof = 2.0000000000e+00, /* 0x40000000 */
425 /* c = (subfloat)0.84506291151 */
426 erxf = 8.4506291151e-01, /* 0x3f58560b */
428 * Coefficients for approximation to erf on [0,0.84375]
430 efxf = 1.2837916613e-01, /* 0x3e0375d4 */
431 efx8f = 1.0270333290e+00, /* 0x3f8375d4 */
432 pp0f = 1.2837916613e-01, /* 0x3e0375d4 */
433 pp1f = -3.2504209876e-01, /* 0xbea66beb */
434 pp2f = -2.8481749818e-02, /* 0xbce9528f */
435 pp3f = -5.7702702470e-03, /* 0xbbbd1489 */
436 pp4f = -2.3763017452e-05, /* 0xb7c756b1 */
437 qq1f = 3.9791721106e-01, /* 0x3ecbbbce */
438 qq2f = 6.5022252500e-02, /* 0x3d852a63 */
439 qq3f = 5.0813062117e-03, /* 0x3ba68116 */
440 qq4f = 1.3249473704e-04, /* 0x390aee49 */
441 qq5f = -3.9602282413e-06, /* 0xb684e21a */
443 * Coefficients for approximation to erf in [0.84375,1.25]
445 pa0f = -2.3621185683e-03, /* 0xbb1acdc6 */
446 pa1f = 4.1485610604e-01, /* 0x3ed46805 */
447 pa2f = -3.7220788002e-01, /* 0xbebe9208 */
448 pa3f = 3.1834661961e-01, /* 0x3ea2fe54 */
449 pa4f = -1.1089469492e-01, /* 0xbde31cc2 */
450 pa5f = 3.5478305072e-02, /* 0x3d1151b3 */
451 pa6f = -2.1663755178e-03, /* 0xbb0df9c0 */
452 qa1f = 1.0642088205e-01, /* 0x3dd9f331 */
453 qa2f = 5.4039794207e-01, /* 0x3f0a5785 */
454 qa3f = 7.1828655899e-02, /* 0x3d931ae7 */
455 qa4f = 1.2617121637e-01, /* 0x3e013307 */
456 qa5f = 1.3637083583e-02, /* 0x3c5f6e13 */
457 qa6f = 1.1984500103e-02, /* 0x3c445aa3 */
459 * Coefficients for approximation to erfc in [1.25,1/0.35]
461 ra0f = -9.8649440333e-03, /* 0xbc21a093 */
462 ra1f = -6.9385856390e-01, /* 0xbf31a0b7 */
463 ra2f = -1.0558626175e+01, /* 0xc128f022 */
464 ra3f = -6.2375331879e+01, /* 0xc2798057 */
465 ra4f = -1.6239666748e+02, /* 0xc322658c */
466 ra5f = -1.8460508728e+02, /* 0xc3389ae7 */
467 ra6f = -8.1287437439e+01, /* 0xc2a2932b */
468 ra7f = -9.8143291473e+00, /* 0xc11d077e */
469 sa1f = 1.9651271820e+01, /* 0x419d35ce */
470 sa2f = 1.3765776062e+02, /* 0x4309a863 */
471 sa3f = 4.3456588745e+02, /* 0x43d9486f */
472 sa4f = 6.4538726807e+02, /* 0x442158c9 */
473 sa5f = 4.2900814819e+02, /* 0x43d6810b */
474 sa6f = 1.0863500214e+02, /* 0x42d9451f */
475 sa7f = 6.5702495575e+00, /* 0x40d23f7c */
476 sa8f = -6.0424413532e-02, /* 0xbd777f97 */
478 * Coefficients for approximation to erfc in [1/.35,28]
480 rb0f = -9.8649431020e-03, /* 0xbc21a092 */
481 rb1f = -7.9928326607e-01, /* 0xbf4c9dd4 */
482 rb2f = -1.7757955551e+01, /* 0xc18e104b */
483 rb3f = -1.6063638306e+02, /* 0xc320a2ea */
484 rb4f = -6.3756646729e+02, /* 0xc41f6441 */
485 rb5f = -1.0250950928e+03, /* 0xc480230b */
486 rb6f = -4.8351919556e+02, /* 0xc3f1c275 */
487 sb1f = 3.0338060379e+01, /* 0x41f2b459 */
488 sb2f = 3.2579251099e+02, /* 0x43a2e571 */
489 sb3f = 1.5367296143e+03, /* 0x44c01759 */
490 sb4f = 3.1998581543e+03, /* 0x4547fdbb */
491 sb5f = 2.5530502930e+03, /* 0x451f90ce */
492 sb6f = 4.7452853394e+02, /* 0x43ed43a7 */
493 sb7f = -2.2440952301e+01; /* 0xc1b38712 */
500 } ieee_float_shape_type;
502 #define GET_FLOAT_WORD(i, d) \
504 ieee_float_shape_type gf_u; \
510 #define SET_FLOAT_WORD(d, i) \
512 ieee_float_shape_type sf_u; \
518 float gmx_erff(float x)
520 gmx_int32_t hx, ix, i;
521 float R, S, P, Q, s, y, z, r;
534 if (ix >= 0x7f800000)
537 i = ((gmx_uint32_t)hx>>31)<<1;
538 return (float)(1-i)+onef/x; /* erf(+-inf)=+-1 */
549 return (float)0.125*((float)8.0*x+efx8f*x); /*avoid underflow */
554 r = pp0f+z*(pp1f+z*(pp2f+z*(pp3f+z*pp4f)));
555 s = onef+z*(qq1f+z*(qq2f+z*(qq3f+z*(qq4f+z*qq5f))));
561 /* 0.84375 <= |x| < 1.25 */
563 P = pa0f+s*(pa1f+s*(pa2f+s*(pa3f+s*(pa4f+s*(pa5f+s*pa6f)))));
564 Q = onef+s*(qa1f+s*(qa2f+s*(qa3f+s*(qa4f+s*(qa5f+s*qa6f)))));
574 if (ix >= 0x40c00000)
591 R = ra0f+s*(ra1f+s*(ra2f+s*(ra3f+s*(ra4f+s*(ra5f+s*(ra6f+s*ra7f))))));
592 S = onef+s*(sa1f+s*(sa2f+s*(sa3f+s*(sa4f+s*(sa5f+s*(sa6f+s*(sa7f+s*sa8f)))))));
597 R = rb0f+s*(rb1f+s*(rb2f+s*(rb3f+s*(rb4f+s*(rb5f+s*rb6f)))));
598 S = onef+s*(sb1f+s*(sb2f+s*(sb3f+s*(sb4f+s*(sb5f+s*(sb6f+s*sb7f))))));
602 conv.i = conv.i & 0xfffff000;
605 r = exp(-z*z-(float)0.5625)*exp((z-x)*(z+x)+R/S);
616 float gmx_erfcf(float x)
619 float R, S, P, Q, s, y, z, r;
632 if (ix >= 0x7f800000)
635 /* erfc(+-inf)=0,2 */
636 return (float)(((gmx_uint32_t)hx>>31)<<1)+onef/x;
644 return onef-x; /* |x|<2**-56 */
647 r = pp0f+z*(pp1f+z*(pp2f+z*(pp3f+z*pp4f)));
648 s = onef+z*(qq1f+z*(qq2f+z*(qq3f+z*(qq4f+z*qq5f))));
664 /* 0.84375 <= |x| < 1.25 */
666 P = pa0f+s*(pa1f+s*(pa2f+s*(pa3f+s*(pa4f+s*(pa5f+s*pa6f)))));
667 Q = onef+s*(qa1f+s*(qa2f+s*(qa3f+s*(qa4f+s*(qa5f+s*qa6f)))));
670 z = onef-erxf; return z - P/Q;
674 z = erxf+P/Q; return onef+z;
684 /* |x| < 1/.35 ~ 2.857143*/
685 R = ra0f+s*(ra1f+s*(ra2f+s*(ra3f+s*(ra4f+s*(ra5f+s*(ra6f+s*ra7f))))));
686 S = onef+s*(sa1f+s*(sa2f+s*(sa3f+s*(sa4f+s*(sa5f+s*(sa6f+s*(sa7f+s*sa8f)))))));
690 /* |x| >= 1/.35 ~ 2.857143 */
691 if (hx < 0 && ix >= 0x40c00000)
693 return twof-tinyf; /* x < -6 */
695 R = rb0f+s*(rb1f+s*(rb2f+s*(rb3f+s*(rb4f+s*(rb5f+s*rb6f)))));
696 S = onef+s*(sb1f+s*(sb2f+s*(sb3f+s*(sb4f+s*(sb5f+s*(sb6f+s*sb7f))))));
700 conv.i = conv.i & 0xfffff000;
703 r = exp(-z*z-(float)0.5625)*exp((z-x)*(z+x)+R/S);
727 gmx_bool gmx_isfinite(real gmx_unused x)
732 returnval = _finite(x);
733 #elif defined HAVE_ISFINITE
734 returnval = isfinite(x);
735 #elif defined HAVE__ISFINITE
736 returnval = _isfinite(x);
738 /* If no suitable function was found, assume the value is
745 gmx_bool gmx_isnan(real x)
751 gmx_within_tol(double f1,
755 /* The or-equal is important - otherwise we return false if f1==f2==0 */
756 if (fabs(f1-f2) <= tol*0.5*(fabs(f1)+fabs(f2)) )
767 gmx_numzero(double a)
769 return gmx_within_tol(a, 0.0, GMX_REAL_MIN/GMX_REAL_EPS);
773 gmx_log2i(unsigned int n)
775 assert(n != 0); /* behavior differs for 0 */
776 #if defined(__INTEL_COMPILER)
777 return _bit_scan_reverse(n);
778 #elif defined(__GNUC__) && UINT_MAX == 4294967295U /*also for clang*/
779 return __builtin_clz(n) ^ 31U; /* xor gets optimized out */
780 #elif defined(_MSC_VER) && _MSC_VER >= 1400
783 _BitScanReverse(&i, n);
786 #elif defined(__xlC__)
787 return 31 - __cntlz4(n);
789 /* http://graphics.stanford.edu/~seander/bithacks.html#IntegerLogLookup */
790 #define LT(n) n, n, n, n, n, n, n, n, n, n, n, n, n, n, n, n
791 static const char LogTable256[256] = {
792 -1, 0, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3,
793 LT(4), LT(5), LT(5), LT(6), LT(6), LT(6), LT(6),
794 LT(7), LT(7), LT(7), LT(7), LT(7), LT(7), LT(7), LT(7)
798 unsigned int r; /* r will be lg(n) */
799 unsigned int t, tt; /* temporaries */
801 if ((tt = n >> 16) != 0)
803 r = ((t = tt >> 8) != 0) ? 24 + LogTable256[t] : 16 + LogTable256[tt];
807 r = ((t = n >> 8) != 0) ? 8 + LogTable256[t] : LogTable256[n];
814 check_int_multiply_for_overflow(gmx_int64_t a,
818 gmx_int64_t sign = 1;
819 if ((0 == a) || (0 == b))
834 if (GMX_INT64_MAX / b < a)
836 *result = (sign > 0) ? GMX_INT64_MAX : GMX_INT64_MIN;
839 *result = sign * a * b;
843 int gmx_greatest_common_divisor(int p, int q)