Default Real and Integer Length

[Stéphane Guilloteau <Stephane dot Guilloteau at obs dot u-bordeaux1 dot fr>]

Dear gfortran contributors

        We have been developping an image processing software for radio 
interferometry over the last 20 or so years, most of it in Fortran. The 
package has been successfully ported to a wide variety of computers (from 
VAX to PCs and Macs, including 64-bits Alpha machines and more) and 
compilers (initially Fortran VAX, then HP, IBM xlf, g77, Lahey-Fujitsu 
lf95, Intel ifort, etc...),
as well as operating systems (VMS, Unix, Linux, Windows, Mac-OS) over the 
years. We had to abandon supporting a g77 version about two years ago 
because our applications needed Fortran-90 features, and manpower was scarce.

        We are currently facing a direct difficulty when compiling our 
code to "gfortran" under 64 bits machines (AMD Opteron or
Intel EMT64). This difficulty is due do the lack of
        -f-default-integer-4
        -f-default-real-4
or similar option which would allow unspecified integers / reals to be 
considered as 32 bit quantities.

Is there any intention to implement such a capability in future releases of 
"gfortran"?  This would greatly help porting 32-bit applications to the now 
available 64 bits machines.

        Best regards

                Stephane Guilloteau

cc: Dr.J.Pety,  manager of the GILDAS software


Dr. Stephane GUILLOTEAU
L3AB, Observatoire de Bordeaux
2 rue de l'Observatoire
        F-33270 FLOIRAC,  France

Tel: (33) 557 77 61 68
Mobile (33) 683 84 66 70
E-Mail: guilloteau@obs.u-bordeaux1.fr