[BACK]Return to real.c CVS log [TXT][DIR] Up to [local] / OpenXM_contrib2 / asir2000 / engine

Annotation of OpenXM_contrib2/asir2000/engine/real.c, Revision 1.3

1.3     ! noro        1: /*
        !             2:  * Copyright (c) 1994-2000 FUJITSU LABORATORIES LIMITED
        !             3:  * All rights reserved.
        !             4:  *
        !             5:  * FUJITSU LABORATORIES LIMITED ("FLL") hereby grants you a limited,
        !             6:  * non-exclusive and royalty-free license to use, copy, modify and
        !             7:  * redistribute, solely for non-commercial and non-profit purposes, the
        !             8:  * computer program, "Risa/Asir" ("SOFTWARE"), subject to the terms and
        !             9:  * conditions of this Agreement. For the avoidance of doubt, you acquire
        !            10:  * only a limited right to use the SOFTWARE hereunder, and FLL or any
        !            11:  * third party developer retains all rights, including but not limited to
        !            12:  * copyrights, in and to the SOFTWARE.
        !            13:  *
        !            14:  * (1) FLL does not grant you a license in any way for commercial
        !            15:  * purposes. You may use the SOFTWARE only for non-commercial and
        !            16:  * non-profit purposes only, such as academic, research and internal
        !            17:  * business use.
        !            18:  * (2) The SOFTWARE is protected by the Copyright Law of Japan and
        !            19:  * international copyright treaties. If you make copies of the SOFTWARE,
        !            20:  * with or without modification, as permitted hereunder, you shall affix
        !            21:  * to all such copies of the SOFTWARE the above copyright notice.
        !            22:  * (3) An explicit reference to this SOFTWARE and its copyright owner
        !            23:  * shall be made on your publication or presentation in any form of the
        !            24:  * results obtained by use of the SOFTWARE.
        !            25:  * (4) In the event that you modify the SOFTWARE, you shall notify FLL by
        !            26:  * e-mail at risa-admin@flab.fujitsu.co.jp of the detailed specification
        !            27:  * for such modification or the source code of the modified part of the
        !            28:  * SOFTWARE.
        !            29:  *
        !            30:  * THE SOFTWARE IS PROVIDED AS IS WITHOUT ANY WARRANTY OF ANY KIND. FLL
        !            31:  * MAKES ABSOLUTELY NO WARRANTIES, EXPRESSED, IMPLIED OR STATUTORY, AND
        !            32:  * EXPRESSLY DISCLAIMS ANY IMPLIED WARRANTY OF MERCHANTABILITY, FITNESS
        !            33:  * FOR A PARTICULAR PURPOSE OR NONINFRINGEMENT OF THIRD PARTIES'
        !            34:  * RIGHTS. NO FLL DEALER, AGENT, EMPLOYEES IS AUTHORIZED TO MAKE ANY
        !            35:  * MODIFICATIONS, EXTENSIONS, OR ADDITIONS TO THIS WARRANTY.
        !            36:  * UNDER NO CIRCUMSTANCES AND UNDER NO LEGAL THEORY, TORT, CONTRACT,
        !            37:  * OR OTHERWISE, SHALL FLL BE LIABLE TO YOU OR ANY OTHER PERSON FOR ANY
        !            38:  * DIRECT, INDIRECT, SPECIAL, INCIDENTAL, PUNITIVE OR CONSEQUENTIAL
        !            39:  * DAMAGES OF ANY CHARACTER, INCLUDING, WITHOUT LIMITATION, DAMAGES
        !            40:  * ARISING OUT OF OR RELATING TO THE SOFTWARE OR THIS AGREEMENT, DAMAGES
        !            41:  * FOR LOSS OF GOODWILL, WORK STOPPAGE, OR LOSS OF DATA, OR FOR ANY
        !            42:  * DAMAGES, EVEN IF FLL SHALL HAVE BEEN INFORMED OF THE POSSIBILITY OF
        !            43:  * SUCH DAMAGES, OR FOR ANY CLAIM BY ANY OTHER PARTY. EVEN IF A PART
        !            44:  * OF THE SOFTWARE HAS BEEN DEVELOPED BY A THIRD PARTY, THE THIRD PARTY
        !            45:  * DEVELOPER SHALL HAVE NO LIABILITY IN CONNECTION WITH THE USE,
        !            46:  * PERFORMANCE OR NON-PERFORMANCE OF THE SOFTWARE.
        !            47:  *
        !            48:  * $OpenXM: OpenXM_contrib2/asir2000/engine/real.c,v 1.2 2000/02/04 09:27:31 noro Exp $
        !            49: */
1.1       noro       50: #include "ca.h"
                     51: #include "base.h"
                     52: #include <math.h>
                     53:
                     54: #if defined(THINK_C)
                     55: double RatnToReal(a)
                     56: Q a;
                     57: {
                     58:        double nm,dn,t;
                     59:        int enm,edn;
                     60:        char buf[BUFSIZ];
                     61:
                     62:        nm = NatToReal(NM(a),&enm);
                     63:        if ( INT(a) )
                     64:                if ( enm >= 1 )
                     65:                        error("RatnToReal : Overflow");
                     66:                else
                     67:                        return SGN(a)>0 ? nm : -nm;
                     68:        else {
                     69:                dn = NatToReal(DN(a),&edn);
                     70:                sprintf(buf,"1.0E%d",enm-edn);
                     71:                sscanf(buf,"%lf",&t);
                     72:                if ( SGN(a) < 0 )
                     73:                        t = -t;
                     74:                return nm/dn*t;
                     75:        }
                     76: }
                     77:
                     78: double NatToReal(a,expo)
                     79: N a;
                     80: int *expo;
                     81: {
                     82:        int *p;
                     83:        int l,all,i,j,s,tb,top,tail;
                     84:        unsigned int t,m[2];
                     85:
                     86:        p = BD(a); l = PL(a) - 1;
                     87:        for ( top = 0, t = p[l]; t; t >>= 1, top++ );
                     88:        all = top + BSH*l; tail = (53-top)%BSH; i = l-(53-top)/BSH-1;
                     89:        m[1] = i < 0 ? 0 : p[i]>>(BSH-tail);
                     90:        for ( j = 1, i++, tb = tail; i <= l; i++ ) {
                     91:                s = 32-tb; t = i < 0 ? 0 : p[i];
                     92:                if ( BSH > s ) {
                     93:                        m[j] |= ((t&((1<<s)-1))<<tb);
                     94:                        if ( !j )
                     95:                                break;
                     96:                        else {
                     97:                                j--; m[j] = t>>s; tb = BSH-s;
                     98:                        }
                     99:                } else {
                    100:                        m[j] |= (t<<tb); tb += BSH;
                    101:                }
                    102:        }
                    103:        s = (all-1)+1023;
                    104:        m[0] = (m[0]&((1<<20)-1))|(MIN(2046,s)<<20); *expo = MAX(s-2046,0);
                    105: #ifdef vax
                    106:        t = m[0]; m[0] = m[1]; m[1] = t; itod(m);
                    107: #endif
                    108: #if defined(MIPSEL)
                    109:        t = m[0]; m[0] = m[1]; m[1] = t;
                    110: #endif
                    111:        return (double)(*((short double *)m));
                    112: }
                    113: #else
                    114: double RatnToReal(a)
                    115: Q a;
                    116: {
                    117:        double nm,dn,t,man;
                    118:        int enm,edn,e;
                    119:        unsigned int *p,s;
                    120:
                    121:        nm = NatToReal(NM(a),&enm);
                    122:        if ( INT(a) )
                    123:                if ( enm >= 1 )
                    124:                        error("RatnToReal : Overflow");
                    125:                else
                    126:                        return SGN(a)>0 ? nm : -nm;
                    127:        else {
                    128:                dn = NatToReal(DN(a),&edn);
                    129:                man = nm/dn;
                    130:                if ( SGN(a) < 0 )
                    131:                        man = -man;
                    132:                if ( ((e = enm - edn) >= 1024) || (e <= -1023) )
                    133:                        error("RatnToReal : Overflow"); /* XXX */
                    134:                else if ( !e )
                    135:                        return man;
                    136:                else
                    137:                        return man*pow(2,e);
                    138:        }
                    139: }
                    140:
                    141: double NatToReal(a,expo)
                    142: N a;
                    143: int *expo;
                    144: {
                    145:        unsigned int *p;
                    146:        int l,all,i,j,s,tb,top,tail;
                    147:        unsigned int t,m[2];
                    148:
                    149:        p = BD(a); l = PL(a) - 1;
                    150:        for ( top = 0, t = p[l]; t; t >>= 1, top++ );
                    151:        all = top + BSH*l; tail = (53-top)%BSH; i = l-(53-top)/BSH-1;
                    152:        m[1] = i < 0 ? 0 : p[i]>>(BSH-tail);
                    153:        for ( j = 1, i++, tb = tail; i <= l; i++ ) {
                    154:                s = 32-tb; t = i < 0 ? 0 : p[i];
                    155:                if ( BSH > s ) {
                    156:                        m[j] |= ((t&((1<<s)-1))<<tb);
                    157:                        if ( !j )
                    158:                                break;
                    159:                        else {
                    160:                                j--; m[j] = t>>s; tb = BSH-s;
                    161:                        }
                    162:                } else {
                    163:                        m[j] |= (t<<tb); tb += BSH;
                    164:                }
                    165:        }
                    166:        s = (all-1)+1023;
                    167:        m[0] = (m[0]&((1<<20)-1))|(MIN(2046,s)<<20); *expo = MAX(s-2046,0);
                    168: #ifdef vax
                    169:        t = m[0]; m[0] = m[1]; m[1] = t; itod(m);
                    170: #endif
1.2       noro      171: #if defined(__i386__) || defined(MIPSEL) || defined(VISUAL) || defined(__alpha) || defined(__FreeBSD__) || defined(__NetBSD__)
1.1       noro      172:        t = m[0]; m[0] = m[1]; m[1] = t;
                    173: #endif
                    174:        return *((double *)m);
                    175: }
                    176: #endif
                    177:
                    178: void addreal(a,b,c)
                    179: Num a,b;
                    180: Real *c;
                    181: {
                    182:        double t;
                    183:
                    184:        t = ToReal(a)+ToReal(b); MKReal(t,*c);
                    185: }
                    186:
                    187: void subreal(a,b,c)
                    188: Num a,b;
                    189: Real *c;
                    190: {
                    191:        double t;
                    192:
                    193:        t = ToReal(a)-ToReal(b); MKReal(t,*c);
                    194: }
                    195:
                    196: void mulreal(a,b,c)
                    197: Num a,b;
                    198: Real *c;
                    199: {
                    200:        double t;
                    201:
                    202:        if ( !a || !b )
                    203:                *c = 0;
                    204:        else {
                    205:                t = ToReal(a)*ToReal(b);
                    206:                if ( !t )
                    207:                        error("mulreal : Underflow"); /* XXX */
                    208:                else
                    209:                        MKReal(t,*c);
                    210:        }
                    211: }
                    212:
                    213: void divreal(a,b,c)
                    214: Num a,b;
                    215: Real *c;
                    216: {
                    217:        double t;
                    218:
                    219:        if ( !a )
                    220:                *c = 0;
                    221:        else {
                    222:                t = ToReal(a)/ToReal(b);
                    223:                if ( !t )
                    224:                        error("divreal : Underflow"); /* XXX */
                    225:                else
                    226:                        MKReal(t,*c);
                    227:        }
                    228: }
                    229:
                    230: void chsgnreal(a,c)
                    231: Num a,*c;
                    232: {
                    233:        double t;
                    234:        Real s;
                    235:
                    236:        if ( !a )
                    237:                *c = 0;
                    238:        else if ( NID(a) == N_Q )
                    239:                chsgnq((Q)a,(Q *)c);
                    240:        else {
                    241:                t = -ToReal(a); MKReal(t,s); *c = (Num)s;
                    242:        }
                    243: }
                    244:
                    245: void pwrreal(a,b,c)
                    246: Num a,b;
                    247: Real *c;
                    248: {
                    249:        double t;
                    250:        double pwrreal0();
                    251:
                    252:        if ( !b )
                    253:                MKReal(1.0,*c);
                    254:        else if ( !a )
                    255:                *c = 0;
                    256:        else if ( !RATN(b) || !INT(b) || (PL(NM((Q)b)) > 1) ) {
                    257:                t = (double)pow((double)ToReal(a),(double)ToReal(b));
                    258:                if ( !t )
                    259:                        error("pwrreal : Underflow"); /* XXX */
                    260:                else
                    261:                        MKReal(t,*c);
                    262:        } else {
                    263:                t = pwrreal0(BDY((Real)a),BD(NM((Q)b))[0]);
                    264:                t = SGN((Q)b)>0?t:1/t;
                    265:                if ( !t )
                    266:                        error("pwrreal : Underflow"); /* XXX */
                    267:                else
                    268:                        MKReal(t,*c);
                    269:        }
                    270: }
                    271:
                    272: double pwrreal0(n,e)
                    273: double n;
                    274: int e;
                    275: {
                    276:        double t;
                    277:
                    278:        if ( e == 1 )
                    279:                return n;
                    280:        else {
                    281:                t = pwrreal0(n,e / 2);
                    282:                return e%2 ? t*t*n : t*t;
                    283:        }
                    284: }
                    285:
                    286: int cmpreal(a,b)
                    287: Real a,b;
                    288: {
                    289:        double t;
                    290:
                    291:        t = ToReal(a)-ToReal(b);
                    292:        return t>0.0 ? 1 : t<0.0?-1 : 0;
                    293: }

FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>