Re: double precision g77

Richard Pitman wrote:

double precision vector(1:50)

This is not Fortran 77, nor within the small subset of f95 which g77 supports. Nor does g77 have a promote-to-double option.
g77 is not gfortran nor g95, in spite of several posters suggesting that it may have mysteriously adopted some traits of those compilers. As Richard repeated, read the fine print (e.g. 'info g77').
g77, with normal options, does support interchangeability of single and double precision function results (only on x86 CPUs).