Note that this module is alpha software. It seems to work ok
  for me on MS Windows (compiling with gcc-4.6.3, gcc-4.7.0)
  but, at time of release, is untested elsewhere.

  Math::Decimal64 supports up to 16 decimal digits of significand
  (mantissa) and an exponent range of -383 to +384.
  The smallest expressable value is -9.999999999999999e384 (which
  is also equivalent to -9999999999999999e369).
  The largest expressable value is 9.999999999999999e384 (which
  also equivalent to 9999999999999999e369).
  The closest we can get to zero is (plus or minus) 1e-384
  (which is also equivalent to 1000000000000000e-399).

  This module allows decimal floating point arithmetic via
  operator overloading - see "OVERLOADING".

  In the documentation that follows, "$mantissa" is a perl scalar
  holding a string of up to 16 decimal digits:
   $mantissa = '1234';
   $mantissa = '1234567890123456';

  For many values, it normally shouldn't matter if $mantissa is
  assigned as a number:
   $mantissa = 1234;      # should work ok.

  But on some perls there are values that *need* to be assigned
  as a string. For example, on perls where nvtype is an 8 byte
  'double':
   $mantissa = '-9307199254740993'; # works fine
   $mantissa = -9307199254740993;   # will assign wrong value

  So ... where you see "$mantissa" in the following docs, think
  *string* of up to 16 decimal digits".

• Igor Vlasenko

• perl-Math-Decimal64
• perl-Math-Decimal64-debuginfo