/* mpfr_set_d, mpfr_get_d -- convert a multiple precision floating-point number from/to a machine double precision float Copyright (C) 1999, 2001 Free Software Foundation, Inc. This file is part of the MPFR Library. The MPFR Library is free software; you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation; either version 2.1 of the License, or (at your option) any later version. The MPFR Library is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details. You should have received a copy of the GNU Lesser General Public License along with the MPFR Library; see the file COPYING.LIB. If not, write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */ #include "gmp.h" #include "mpfr.h" #include "mpfr-impl.h" #include "gmp-impl.h" #include "longlong.h" #if (BITS_PER_MP_LIMB==32) #define MPFR_LIMBS_PER_DOUBLE 2 #elif (BITS_PER_MP_LIMB >= 64) #define MPFR_LIMBS_PER_DOUBLE 1 #elif (BITS_PER_MP_LIMB == 16) #define MPFR_LIMBS_PER_DOUBLE 4 #endif static int __mpfr_extract_double _PROTO ((mp_ptr, double, int)); static double __mpfr_scale2 _PROTO ((double, int)); #define NaN (0./0.) /* ensures a machine-independent NaN */ #define Infp (1/0.) #define Infm (-1/0.) /* Included from gmp-2.0.2, patched to support denorms */ #ifdef XDEBUG #undef _GMP_IEEE_FLOATS #endif #ifndef _GMP_IEEE_FLOATS #define _GMP_IEEE_FLOATS 0 #endif static int __mpfr_extract_double (mp_ptr rp, double d, int e) /* e=0 iff BITS_PER_MP_LIMB=32 and rp has only one limb */ { long exp; mp_limb_t manl; #if BITS_PER_MP_LIMB == 32 mp_limb_t manh; #endif /* BUGS 1. Should handle Inf and NaN in IEEE specific code. 2. Handle Inf and NaN also in default code, to avoid hangs. 3. Generalize to handle all BITS_PER_MP_LIMB >= 32. 4. This lits is incomplete and misspelled. */ if (d == 0.0) { rp[0] = 0; #if BITS_PER_MP_LIMB == 32 if (e) rp[1] = 0; #endif return 0; } #if _GMP_IEEE_FLOATS { union ieee_double_extract x; x.d = d; exp = x.s.exp; if (exp) { #if BITS_PER_MP_LIMB == 64 manl = ((MP_LIMB_T_ONE << 63) | ((mp_limb_t) x.s.manh << 43) | ((mp_limb_t) x.s.manl << 11)); #else manh = (MP_LIMB_T_ONE << 31) | (x.s.manh << 11) | (x.s.manl >> 21); manl = x.s.manl << 11; #endif } else { #if BITS_PER_MP_LIMB == 64 manl = ((mp_limb_t) x.s.manh << 43) | ((mp_limb_t) x.s.manl << 11); #else manh = (x.s.manh << 11) | (x.s.manl >> 21); manl = x.s.manl << 11; #endif } } #else { /* Unknown (or known to be non-IEEE) double format. */ exp = 0; if (d >= 1.0) { if (d * 0.5 == d) abort (); while (d >= 32768.0) { d *= (1.0 / 65536.0); exp += 16; } while (d >= 1.0) { d *= 0.5; exp += 1; } } else if (d < 0.5) { while (d < (1.0 / 65536.0)) { d *= 65536.0; exp -= 16; } while (d < 0.5) { d *= 2.0; exp -= 1; } } d *= MP_BASE_AS_DOUBLE; #if BITS_PER_MP_LIMB == 64 manl = d; #else manh = d; manl = (d - manh) * MP_BASE_AS_DOUBLE; #endif exp += 1022; } #endif if (exp) exp = (unsigned) exp - 1022; else exp = -1021; #if BITS_PER_MP_LIMB == 64 rp[0] = manl; #else if (e) { rp[1] = manh; rp[0] = manl; } else { rp[0] = manh; } #endif return exp; } /* End of part included from gmp-2.0.2 */ /* Part included from gmp temporary releases */ static double __mpfr_scale2 (double d, int exp) { #if _GMP_IEEE_FLOATS { union ieee_double_extract x; x.d = d; exp += x.s.exp; x.s.exp = exp; if (exp >= 2047) { /* Return +-infinity */ x.s.exp = 2047; x.s.manl = x.s.manh = 0; } else if (exp < 1) { x.s.exp = 1; /* smallest exponent (biased) */ /* Divide result by 2 until we have scaled it to the right IEEE denormalized number, but stop if it becomes zero. */ while (exp < 1 && x.d != 0) { x.d *= 0.5; exp++; } } return x.d; } #else { double factor, r; factor = 2.0; if (exp < 0) { factor = 0.5; exp = -exp; } r = d; if (exp != 0) { if ((exp & 1) != 0) r *= factor; exp >>= 1; while (exp != 0) { factor *= factor; if ((exp & 1) != 0) r *= factor; exp >>= 1; } } return r; } #endif } /* End of part included from gmp */ int mpfr_set_d (mpfr_ptr r, double d, mp_rnd_t rnd_mode) { int signd, sizer, sizetmp, inexact; unsigned int cnt; mpfr_ptr tmp; TMP_DECL(marker); TMP_MARK(marker); MPFR_CLEAR_FLAGS(r); if (d == 0) { union ieee_double_extract x; MPFR_SET_ZERO(r); /* set correct sign */ x.d = d; if (((x.s.sig == 1) && (MPFR_SIGN(r) > 0)) || ((x.s.sig == 0) && (MPFR_SIGN(r) < 0))) MPFR_CHANGE_SIGN(r); return 0; /* 0 is exact */ } else if (DOUBLE_ISNAN(d)) { MPFR_SET_NAN(r); return 1; /* a NaN is always inexact */ } else if (DOUBLE_ISINF(d)) { MPFR_SET_INF(r); if ((d > 0 && (MPFR_SIGN(r) == -1)) || (d < 0 && (MPFR_SIGN(r) == 1))) MPFR_CHANGE_SIGN(r); return 0; /* infinity is exact */ } sizer = (MPFR_PREC(r) - 1) / BITS_PER_MP_LIMB + 1; /* warning: don't use tmp=r here, even if sizer >= MPFR_LIMBS_PER_DOUBLE, since PREC(r) may be different from PREC(tmp), and then both variables would have same precision in the mpfr_set4 call below. */ tmp = (mpfr_ptr) TMP_ALLOC(sizeof(mpfr_t)); MPFR_MANT(tmp) = TMP_ALLOC(MPFR_LIMBS_PER_DOUBLE * BYTES_PER_MP_LIMB); MPFR_PREC(tmp) = 53; MPFR_SIZE(tmp) = MPFR_LIMBS_PER_DOUBLE; sizetmp = MPFR_LIMBS_PER_DOUBLE; signd = (d < 0) ? -1 : 1; d = ABS (d); MPFR_EXP(tmp) = __mpfr_extract_double (MPFR_MANT(tmp), d, 1); count_leading_zeros(cnt, MPFR_MANT(tmp)[sizetmp - 1]); if (cnt) mpn_lshift (MPFR_MANT(tmp), MPFR_MANT(tmp), sizetmp, cnt); MPFR_EXP(tmp) -= cnt; /* tmp is exact since PREC(tmp)=53 */ inexact = mpfr_set4(r, tmp, rnd_mode, signd); TMP_FREE(marker); return inexact; } double mpfr_get_d2 (mpfr_srcptr src, long e) { double res; mp_size_t size, i, n_limbs_to_use; mp_ptr qp; int negative; if (MPFR_IS_NAN(src)) { #ifdef DEBUG printf("recognized NaN\n"); #endif return NaN; } if (MPFR_IS_INF(src)) { #ifdef DEBUG printf("Found Inf.\n"); #endif return (MPFR_SIGN(src) == 1 ? Infp : Infm); } if (MPFR_NOTZERO(src) == 0) return 0.0; size = 1+(MPFR_PREC(src)-1)/BITS_PER_MP_LIMB; qp = MPFR_MANT(src); negative = (MPFR_SIGN(src) < 0); /* Warning: don't compute the abs(res) and set the sign afterwards, otherwise the current machine rounding mode will not be taken correctly into account. */ /* res = (negative) ? -(double)qp[size - 1] : qp[size - 1]; */ res = 0.0; /* Warning: an arbitrary number of limbs may be required for an exact rounding. The following code is correct but not optimal since one may be able to decide without considering all limbs. */ /* n_limbs_to_use = MIN (MPFR_LIMBS_PER_DOUBLE, size); */ n_limbs_to_use = size; /* Accumulate the limbs from less significant to most significant otherwise due to rounding we may accumulate several ulps, especially in rounding towards -/+infinity. */ for (i = n_limbs_to_use; i>=1; i--) { #if (BITS_PER_MP_LIMB == 32) res = res / MP_BASE_AS_DOUBLE + ((negative) ? -(double)qp[size - i] : qp[size - i]); #else #if (BITS_PER_MP_LIMB == 64) mp_limb_t q; q = qp[size - i] & CNST_LIMB(0xFFFFFFFF); res = res / MP_BASE_AS_DOUBLE + ((negative) ? -(double)q : q); q = qp[size - i] - q; res = res + ((negative) ? -(double)q : q); #endif /* BITS_PER_MP_LIMB == 64 */ #endif /* BITS_PER_MP_LIMB == 32 */ } res = __mpfr_scale2 (res, e - BITS_PER_MP_LIMB); return res; } double mpfr_get_d (mpfr_srcptr src) { return mpfr_get_d2 (src, MPFR_EXP(src)); }