Fixed issue with internal real representation, incorrect for non x86-32 architectures. Cleaned up CallExp::toElem, moved implementation to tocall.cpp providing a single procedure to call arbitrary D functions fairly easily.

// ifloat

module typeinfo1.ti_ifloat;

private import typeinfo1.ti_float;

class TypeInfo_o : TypeInfo_f
{
    char[] toString() { return "ifloat"; }
}