The gfortran, ifort, and pgf90 compilers can be instructed that variables declared as “real” should actually be stored and handled as double precision quantities. Details are at https://198.82.212.30/userguide/fortran_realsize/