Semiautomatic characterization of Fortran 90 compilers

This article discusses a semiautomatic method of determining the floating-point and integer types available to a Fortran compiler on a given machine. These types are then characterized using a collection of intrinsic routines and other algorithms. Also, the possibility of unexpected truncation when assigning constants to variables is examined. The use of the algorithms discussed herein can lead to a better understanding of a particular compiler or machine and, more portable and accurate programs.