NAME

Math::LongDouble - perl interface to C's long double operations

DESCRIPTION

   use Math::LongDouble qw(:all);

   $arg = ~0; # largest UV
   $d1 = Math::LongDouble->new($arg); # Assign the UV ~0 to $d2.
   $d2 = UVtoLD($arg);                # Assign the UV ~0 to $d2.

   $arg = -21;
   $d1 = Math::LongDouble->new($arg); # Assign the IV -21 to $d2.
   $d2 = IVtoLD($arg);                # Assign the IV -21 to $d2.

   $arg = 32.1;
   $d1 = Math::LongDouble->new($arg); # Assign the NV 32.1 to $d2.
   $d2 = NVtoLD($arg);                # Assign the NV 32.1 to $d2.

   $arg = "32.1";
   $d1 = Math::LongDouble->new($arg); # Assign strtold("32.1") to $d2.
   $d2 = STRtoLD($arg);               # Assign strtold("32.1") to $d2.

   $d3 = Math::LongDouble->new($d1); # Assign the value of $d1 to $d3.
   $d4 = LDtoLD($d1);                # Assign the value of $d1 to $d4.
   $d5 = $d1;                        # Assign the value of $d1 to $d5.

   This behaviour has changed from 0.06 and earlier.

   A number of the functions below accept string arguments. These arguments
   having been handed to strtold() will be checked for the presence of
   non-numeric characters. If any such non-numeric characters are detected,
   then the global non-numeric flag (which is initially set to 0) will be
   incremented.
   Neither leading nor trailing whitespace is deemed non-numeric, but any
   other (ie internal) whitespace *is* regarded as non-numeric.
   You can query the value held by the global non-numeric flag by running
   Math::Float128::nnumflag() and you can manually alter the value of this
   global using Math::Float128::set_nnum and Math::Float128::clear_nnum.
   These functions are documented below.

   NOTE:
    Math::LongDouble->new(32.1) != Math::LongDouble->new('32.1')
    unless $Config{nvtype} reports 'long double'. The same holds
    for many (but not all) numeric values. In general, it's not
    always true (and is often untrue) that
    Math::LongDouble->new($n) == Math::LongDouble->new("$n")

OVERLOADING

   The following operations are overloaded:
    + - * / **
    += -= *= /= **=
    != == <= >= <=> < >
    ++ --
    ""
    abs bool ! int
    sqrt log exp
    sin cos atan2
    =

    NOTE: Making use of the '=' overloading is not recommended unless
          you understand its caveats. See 'perldoc overload' and
          read it thoroughly, including the documentation regarding
          'copy constructors'.

    In those situations where the overload subroutine operates on 2
    perl variables, then obviously one of those perl variables is
    a Math::LongDouble object. To determine the value of the other
    variable the subroutine works through the following steps (in
    order), using the first value it finds, or croaking if it gets
    to step 6:

    1. If the variable is a UV (unsigned integer value) then that
       value is used. The variable is considered to be a UV if
       (perl 5.8) the UOK flag is set or if (perl 5.6) SvIsUV()
       returns true.

    2. If the variable is an IV (signed integer value) then that
       value is used. The variable is considered to be an IV if the
       IOK flag is set.

    3. If the variable is an NV (floating point value) then that
       value is used. The variable is considered to be an NV if the
       NOK flag is set.

    4. If the variable is a string (ie the POK flag is set) then the
       value of that string is used.

    5. If the variable is a Math::LongDouble object then the value
       encapsulated in that object is used.

    6. If none of the above is true, then the second variable is
       deemed to be of an invalid type. The subroutine croaks with
       an appropriate error message.

ASSIGNMENT FUNCTIONS

   The following create and assign a new Math::LongDouble.

    $ld = Math::LongDouble->new($arg);
     Returns a Math::LongDouble object to which the numeric value of $arg
     has been assigned.
     If $arg is not provided then the value of $ld will be NaN.

    $ld = UVtoLD($arg);
     Returns a Math::LongDouble object to which the numeric (unsigned
     integer) value of $arg has been assigned.

    $ld = IVtoLD($arg);
     Returns a Math::LongDouble object to which the numeric (signed
     integer) value of $arg has been assigned.

    $ld = NVtoLD($arg);
     Returns a Math::LongDouble object to which the numeric (floating
     point) value of $arg has been assigned.

    $ld2 = LDtoLD($ld1);
     Returns a Math::LongDouble object that is a copy of the
     Math::LongDouble object provided as the argument.
     Courtesy of overloading, this is in effect no different to doing:
     $ld2 = $ld1;

    $ld = STRtoLD($str);
     Returns a Math::LongDouble object that has the value of the string
     $str.

ASSIGNMENT OF INF, NAN, UNITY and ZERO

   $ld = InfLD($sign);
    If $sign < 0, returns a Math::LongDouble object set to
    negative infinity; else returns a Math::LongDouble object set
    to positive infinity.

   $ld = NaNLD();
    If $sign < 0, returns a Math::longDouble object set to NaN.

   $ld = ZeroLD($sign);
    If $sign < 0, returns a Math::LongDouble object set to
    negative zero; else returns a Math::LongDouble object set to
    zero.

   $ld = UnityLD($sign);
    If $sign < 0, returns a Math::LongDouble object set to
    negative one; else returns a Math::LongDouble object set to
    one.

   ld_set_prec($precision);
    Sets the precision of stringified values to $precision decimal
    digits. Default precision is set in the XS global _DIGITS to
        1 + ceil(MANT_PREC * log(2) / log(10)
    where MANT_PREC is LDBL_MANT_DIG if float.h defines that symbol.
    Else MANT_PREC is DBL_MANT_DIG if float.h defines that symbol.
    Else MANT_PREC is 21 (which is the correct value for a 64-bit
    precision mantissa).


   $precision = ld_get_prec();
    Returns the precision (in decimal digits) that will be used
    when stringifying values (by printing them, or calling
    LDtoSTR).

ASSIGNMENT OF FLOAT.H & MATH.H CONSTANTS

   The following functions return their values as either normal
   perl scalar integer values ($iv) or Math::LongDouble objects
   ($ld), as appropriate.
   Those LD_DBL_* functions that return 'double' values could have been
   structured to return an NV, but they *do* return Math::LongDouble
   objects - mainly for consistency with their LD_LDBL_* counterparts.

    $iv = LD_DBL_DIG;
    $iv = LD_LDBL_DIG;
     Returns DBL_DIG/LDBL_DIG or croaks if DBL_DIG/LDBL_DIG is not
     defined.

    $ld = LD_DBL_MAX;
    $ld = LD_LDBL_MAX;
     Returns DBL_MAX/LDBL_MAX or croaks if DBL_MAX/LDBL_MAX is not defined.

    $ld = LD_DBL_MIN;
    $ld = LD_LDBL_MIN;
     Returns DBL_MIN/LDBL_MIN or croaks if DBL_MIN/LDBL_MIN is not defined.

    $ld = LD_DBL_EPSILON;
    $ld = LD_LDBL_EPSILON;
     Returns DBL_EPSILON/LDBL_EPSILON or croaks if
     DBL_EPSILON/LDBL_EPSILON is not defined.

    $ld = LD_DBL_DENORM_MIN;
    $ld = LD_LDBL_DENORM_MIN;
     Returns DBL_DENORM_MIN/LDBL_DENORM_MIN or croaks if
     DBL_DENORM_MIN/LDBL_DENORM_MIN is not defined.

    $iv = LD_DBL_MANT_DIG;
    $iv = LD_LDBL_MANT_DIG;
     Returns DBL_MANT_DIG/LDBL_MANT_DIG or croaks if
     DBL_MANT_DIG/LDBL_MANT_DIG is not defined.

    $iv = LD_DBL_MIN_EXP;
    $iv = LD_LDBL_MIN_EXP;
     Returns DBL_MIN_EXP/LDBL_MIN_EXP or croaks if
     DBL_MIN_EXP/LDBL_MIN_EXP is not defined.

    $iv = LD_DBL_MAX_EXP;
    $iv = LD_LDBL_MAX_EXP;
     Returns DBL_MAX_EXP/LDBL_MAX_EXP or croaks if
     DBL_MAX_EXP/LDBL_MAX_EXP is not defined.

    $iv = LD_DBL_MIN_10_EXP;
    $iv = LD_LDBL_MIN_10_EXP;
     Returns DBL_MIN_10_EXP/LDBL_MIN_10_EXP or croaks if
     DBL_MIN_10_EXP/LDBL_MIN_10_EXP is not defined.

    $iv = LD_DBL_MAX_10_EXP;
    $iv = LD_LDBL_MAX_10_EXP;
     Returns DBL_MAX_10_EXP/LDBL_MAX_10_EXP or croaks if
     DBL_MAX_10_EXP/LDBL_MAX_10_EXP is not defined.

    $ld = M_El;
     Returns M_El (e) or expl(1.0) if M_El is not defined.

    $ld = M_LOG2El;
     Returns M_LOG2El or log2l(expl(1.0)) if M_LOG2El is not
     defined.

    $ld = M_LOG10El;
     Returns M_LOG10El or log10l(expl(1.0)) if M_LOG10El is not
     defined.

    $ld = M_LN2l;
     Returns M_LN2l or logl(2) if M_LN2l is not defined.

    $ld = M_LN10l;
     Returns M_LN10l or logl(10) if M_LN10l is not defined.

    $ld = M_PIl;
     Returns M_PIl (pi) or 2 * asinl(1) if M_PIl is not defined.

    $ld = M_PI_2l;
     Returns M_PI_2l (pi/2) or asinl(1) if M_PI_2l is not defined.

    $ld = M_PI_4l;
     Returns M_PI_4l (pi/4) or asinl(1)/2 if M_PI_4l is not defined.

    $ld = M_1_PIl;
     Returns M_1_PIl (1/pi) or 0.5/asinl(1) if M_1_PIl is not
     defined.

    $ld = M_2_PIl;
     Returns M_2_PIl (2/pi) or 1/asinl(1) if M_2_PIl is not defined.

    $ld = M_2_SQRTPIl;
     Returns M_2_SQRTPIl (2/sqrt(pi)) or 2/sqrtl(pi) if M_2_SQRTPIl
     is not defined.

    $ld = M_SQRT2l;
     Returns M_SQRT2l or sqrtl(2)) if M_SQRT2l is not defined.

    $ld = M_SQRT1_2l;
     Returns M_SQRT1_2l or 1/sqrtl(2)) if M_SQRT1_2l is not defined.

RETRIEVAL FUNCTIONS

   The following functions provide ways of seeing the value of
   Math::LongDouble objects.

   $nv = LDtoNV($ld);
    This function returns the value of the Math::LongDouble object to
    a perl scalar (NV). It may not translate the value accurately.

   $string = LDtoSTR($ld);
    Returns the value of the Math::LongDouble object as a string.
    The returned string will contain the same as is displayed by
    "print $ld", except that print() will strip the trailing zeroes
    in the mantissa (significand) whereas LDtoSTR won't.
    By default, provides 21 decimal digits of precision for the
    typical 80-bit long double or 17 decimal digits if the long double
    is a double. The number of digits dispalayed can be altered by
    specifying the desired precision (in decimal digits) in a call to
    ld_set_prec.

   $string = LDtoSTRP($ld, $precision);
    Same as LDtoSTR, but takes an additional arg that specifies the
    precision (in decimal digits) of the stringified return value.

MATH LIBRARY FUNCTIONS

   With the following functions, "$rop" and "$op" are Math::LongDouble
   objects, and "$iv" is just a normal perl scalar that either
   holds a signed integer value, or to which a signed integer value
   will be returned.
   These are just interfaces to the standard math library functions.
   I'm assuming you already have access to their documentation.
   These functions do not check their argument types - if you get
   a segfault, check that you've supplied the correct argument type(s).

   acos_LD($rop, $op);
    acosl($op) is assigned to $rop.

   acosh_LD($rop, $op);
    acoshl($op) is assigned to $rop.

   asin_LD($rop, $op);
    asinl($op) is assigned to $rop.

   asinh_LD($rop, $op);
    asinhl($op) is assigned to $rop.

   atan_LD($rop, $op);
    atanl($op) is assigned to $rop.

   atanh_LD($rop, $op);
    atanhl($op) is assigned to $rop.

   atan2_LD($rop, $op1, $op2);
    atan2l($op1, $op2) is assigned to $rop.

   cbrt_LD($rop, $op);
    cbrtl($op) is assigned to $rop.

   ceil_LD($rop, $op);
    ceill($op) is assigned to $rop.

   copysign_LD($rop, $op1, $op2);
    copysignl($op1, $op2) is assigned to $rop.

   cosh_LD($rop, $op);
    coshl($op) is assigned to $rop.

   cos_LD($rop, $op);
    cosl($op) is assigned to $rop.

   erf_LD($rop, $op);
    erfl($op) is assigned to $rop.

   erfc_LD($rop, $op);
    erfcl($op) is assigned to $rop.

   exp_LD($rop, $op);
    expl($op) is assigned to $rop.

   expm1_LD($rop, $op);
    expm1l($op) is assigned to $rop.

   fabs_LD($rop, $op);
    fabsl($op) is assigned to $rop.

   fdim_LD($rop, $op1, $op2);
    fdiml($op1, $op2) is assigned to $rop.

   $iv = finite_LD($op);
    finite($op) is assigned to $iv.

   floor_LD($rop, $op);
    floorl($op) is assigned to $rop.

   fma_LD($rop, $op1, $op2, $op3);
    fmal($op1, $op2, $op3) is assigned to $rop.
    On mingw-w64 compilers, fmaq() crashes, so for those compilers
    we assign ($op1 * $op2)+$op3 to $rop.

   fmax_LD($rop, $op1, $op2);
    fmaxl($op1, $op2) is assigned to $rop.

   fmin_LD($rop, $op1, $op2);
    fmin($op1, $op2) is assigned to $rop.

   fmod_LD($rop, $op1, $op2);
    fmodl($op1, $op2) is assigned to $rop.

   frexp_LD($rop, $iv, $op);
    frexpl($op) is assigned to ($rop, $iv)

   hypot_LD($rop, $op1, $op2);
    hypotl($op1, $op2) is assigned to $rop.

   $iv = isinf_LD($op);
    isinf($op) is assigned to $iv.

   $iv = ilogb_LD($op);
    ilogbl($op) is assigned to $iv.

   $iv = isnan_LD($op);
    isnanl($op) is assigned to $iv.
    If Math::LOngDouble::_have_isnanl returns false, uses custom
    (_is_nan) XSub instead.

   ldexp_LD($rop, $op, $iv);
    ldexpl($op, $iv) is assigned to $rop.
    $iv should not contain a value that won't fit into a signed int

   lgamma_LD($rop, $op);
    lgammal($op) is assigned to $rop.

   $iv = llrint_LD($op);
    llrintl($op) is assigned to $iv.
    This requires that perl's IV is large enough to hold a longlong
    int. Otherwise attempts to use this function will result in a fatal
    error, accompanied by a message stating that the function is
    unimplemented.

   $iv = llround_LD($op);
    llroundl($op) is assigned to $rop.
    This requires that perl's IV is large enough to hold a longlong
    int. Otherwise attempts to use this function will result in a fatal
    error, accompanied by a message stating that the function is
    unimplemented.

   log_LD($rop, $op);
    logl($op) is assigned to $rop. # base e

   log10_LD($rop, $op);
    log10l($op) is assigned to $rop. # base 10

   log2_LD($rop, $op);
    log2l($op) is assigned to $rop. # base 2

   log1p_LD($rop, $op);
    log1pl($op) is assigned to $rop. # base e

   $iv = lrint_LD($op);
    lrintl($op) is assigned to $iv.
    This requires that perl's IV is large enough to hold a long int.
    Otherwise attempts to use this function will result in a fatal
    error, accompanied by a message stating that the function is
    unimplemented.

   $iv = lround_LD($op);
    lroundl($op) is assigned to $iv
    This requires that perl's IV is large enough to hold a long int.
    Otherwise attempts to use this function will result in a fatal
    error, accompanied by a message stating that the function is
    unimplemented.

   modf_LD($rop1, $rop2, $op);
    modfl($op) is assigned to ($rop1, $rop2).

   nan_LD($rop, $op);
    nanl($op) is assigned to $rop.
    If Math::LongDouble::_have_nanl returns false, uses custom
    (_get_nan) XSub instead.

   nearbyint_LD($rop, $op);
    nearbyintl($op) is assigned to $rop.

   nextafter_LD($rop, $op1, $op2);
    nextafterl($op1, $op2) is assigned to $rop.

   pow_LD($rop, $op1, $op2);
    pow($op1, $op2) is assigned to $rop.

   remainder_LD($rop, $op1, $op2);
    remainderl($op1, $op2) is assigned to $rop.

   remquo_LD($rop1, $rop2, $op1, $op2);
    remquol($op1, $op2) is assigned to ($rop1, $rop2).
    I find this function can return unexpected results with some
    compilers. Therefore, this function is not tested in the test suite.
    Use it at your own risk.

   $iv = rint_LD($op);
    rintl($op) is assigned to $rop.

   $iv = round_LD($op);
    roundl($op) is assigned to $iv.

   scalbln_LD($rop, $op, $iv);
    scalblnl($op, $iv) is assigned to $rop.
    $iv should not contain a value that won't fit into a signed
    long int.

   scalbn_LD($rop, $op, $iv);
    scalbnl($op, $iv) is assigned to $rop.
    $iv should not contain a value that won't fir into a signed int.

   $iv = signbit_LD($op);
    signbitl($op) is assigned to $iv.
    If Math::LongDouble::_have_signbitl returns false signbit() is
    used instead.

   sincos_LD($rop1, $rop2, $op);
    sinl($op) is assigned to $rop1.
    cosl($op) is assigned to $rop2.

   sinh_LD($rop, $op);
    sinhl($op) is assigned to $rop.

   sin_LD($rop, $op);
    sin($op) is assigned to $rop.

   sqrt_LD($rop, $op);
    sqrtl($op) is assigned to $rop.

   tan_LD($rop, $op);
    tanl($op) is assigned to $rop.

   tanh_LD($rop, $op);
    tanhl($op) is assigned to $rop.

   tgamma_LD($rop, $op);
    gammal($op) is assigned to $rop.

   trunc_LD($rop, $op);
    truncl($op) is assigned to $rop.

OTHER FUNCTIONS

   $iv = Math::LongDouble::nnumflag(); # not exported
    Returns the value of the non-numeric flag. This flag is
    initialized to zero, but incemented by 1 whenever a function
    is handed a string containing non-numeric characters. The
    value of the flag therefore tells us how many times functions
    have been handed such a string. The flag can be reset to 0 by
    running Math::LongDouble::clear_nnum().

   Math::LongDouble::set_nnum($iv); # not exported
    Resets the global non-numeric flag to the value specified by
    $iv.

   Math::LongDouble::clear_nnum(); # not exported
    Resets the global non-numeric flag to 0.(Essentially the same
    as running Math::LongDouble::set_nnum(0).)

   $bool = is_NaNLD($ld);
    Returns 1 if $ld is a Math::LongDouble NaN.
    Else returns 0

   $int = is_InfLD($ld)
    If the Math::LongDouble object $ld is -inf, returns -1.
    If it is +inf, returns 1.
    Otherwise returns 0.

   $int = is_ZeroLD($ld);
    If the Math::LongDouble object $ld is -0, returns -1.
    If it is zero, returns 1.
    Otherwise returns 0.

   $int = cmp_NV($ld, $nv);
    $nv can be any perl number - ie NV, UV or IV.
    If the Math::LongDouble object $ld < $nv returns -1.
    If it is > $nv, returns 1.
    Otherwise returns 0.

   $hex = ld_bytes($ld);
    Returns the  hex representation (in big-endian order) of the
    byte structure of $ld.

BASE CONVERSIONS

   $min_prec = ld_min_inter_prec($orig_base, $orig_prec, $to_base);

    NOTE: $min_prec can be (very rarely) off by one if $orig_prec is in
          the millions, or if either $orig_base or $to_base are
          outside of the range 2..64.

    Example 1:
    Let's say we have some base 10 integers comprising 16 base 10
    digits, and we want to represent those numbers in base 2 (binary).
    What is the minimum required number of bits, such that it can be
    guaranteed that converting the base 2 representations back to base
    10 will result in the original 16 digit representations ?

    We can calculate that minimum required precision with:
      $min_prec = ld_min_inter_prec($orig_base, $orig_prec, $to_base);

    In this example case that becomes:
    $min_prec = mpfr_min_inter_prec(10, 16, 2);
    which will set $min_prec to 55.

    That is, so long as our base 2 representations provide at least 55
    bits, we can pass 16-digit, base 10, integer values to them,
    and be assured of retrieving the original base 10 representation when
    we convert the base 2 representations back to base 10.
    Sure ... not all 16-digit values require 55 bits, but there are some
    that do ... and there are none that require more than  55 bits.

    Example 2:
    $min_prec = ld_min_inter_prec(2, 53, 10);
    $min_prec is set to 17.
    This tells us that a base 10 representation of a 53-bit integer needs
    to comprise at least 17 digits if we are to be assured that assigning
    that base 10 representation to a 53-bit integer will result in a
    53-bit integer that is identical to the first.
    Otherwise, there is no such assurance.


   $max_prec = ld_max_orig_prec ($orig_base, $to_base, $to_prec);

    NOTE: $max_prec can be (very rarely) off by one if $to_prec is in
          the millions, or if either $orig_base or $to_base are
          outside of the range 2..64.

    For example:
    To determine the maximum significant number of base 10 digits that
    can be specified, when assigning to a 53-bit double. We have:

    $max_len = ld_max_orig_len($orig_base, $to_base, $to_prec);

    For this example that becomes:

    $max_len = mpfr_max_orig_len(10, 2, 53);

    which will set $max_len to 15.
    That is, so long as our base 10 integer consists of no more than
    15 significant digits, we can assign it to a 53-bit double and be
    assured of retrieving the original value upon converting that
    double back to a 15-digit base 10 representation.
    Otherwise, there is no such assurance.

    It is to be expected that
     mpfr_max_orig_len(10, 2, 53) == DBL_DIG
     and
     mpfr_max_orig_len(10, 2, 113) == FLT128_DIG
     and
     mpfr_max_orig_len(10, 2, $long_double_prec) == LDBL_DIG

    (where $long_double_prec is the precision, in bits, of the
    the C 'long double' type - usually either 53 or 64 or 113.)

LICENSE

   This program is free software; you may redistribute it and/or modify
   it under the same terms as Perl itself.
   Copyright 2012-16, 2020, 2024 Sisyphus

AUTHOR

   Sisyphus <sisyphus at(@) cpan dot (.) org>