compiling code for 64-bit: random # generater (rand) real*4 or real*8

[Dan_Fishman]

I am trying to run a program on Windows XP x64. The code already runs on ia32 (on a PC) and ia64 (on a SGI Altix Linux machine).

I am using Intel VF 11.1.060 64 bit compiler with the same compile options as I have used for the ia32 compilation.

The program compilies but doesn't run properly.

One issue I noticed is that the random # generator "rand()" only works if I define it as a single precision variable. (It ran okay when I defined it real*8 for the ia32 compiler).

Does anyone know the reason for this?

Are there specific compilation options that should be set to help the code adjust to the 64-bit compilation?

[TimP]

You should have
USE ifport
to maintain correct interface for ifport subroutines (which aren't standard Fortran).
Fortran standard intrinsic random_number should adjust automatically to your declared data type.
Only if you use a function in 32-bit compilation with /arch:IA32 (x87 code) can you gloss over the difference between single and double precision function return values.