If your perl's nvtype is 'long double' or your compiler's 'long double' is the same size as 'double', then I don't think there's any point in installing this module. It should, however, still build and pass its tests (unless you're using a Microsoft compiler, or are building on a perl that was built by a Microsoft compiler - in which case all bets are off). Build with: perl Makefile.PL make test make install It is assumed that your 'long double' has precision p = 64 (ie 64-bits, or 20 decimal digits of precision). If this is not the case, you can either: 1) Set $digits to x (in the Makefile.PL) or 2) Run 'perl Makefile DIG=x' where x is ceil(p * 0.30103) Note that this module has bugs on perls built with a Microsoft compiler (eg ActivePerl) - even if the binaries installed onto the MSVC-built perl were built using MinGW on a MinGW-built perl such as Strawberry Perl (where no such problem exists). By some means that is still unclear, in this situation the 'long double' precision is usually (apparently) reduced to 'double' precision whenever a Math::LongDouble object is raised to a power (or a square root taken) on MSVC-built perls. This bug manifests itself in causing test failures in t/cmp.t and t/pow.t.