Problem with code migration


I am now migration my fortran program from Digital Fortran of the Alpha station to PGF90 on linux. I can run the program but the results
are quite different (in the order of 1e-4 where magnitude the the variable is around 1000). Is there anyone had same problem before? And what would be the reason for this difference? Thanks a lot.

Also, I would like to ask if there’s any ways to set the PGDBG GUI on Fedora Linux back to the same interface as Redhat 8.0?

Many thanks.

We have a FAQ about precision issues here. If you have a SSE2 enabled processor, you can use the “-Mscalarsse” flag to use SSE floating point operations instead of x87. This flag is automatically included when you use “-fastsse”. SSE uses 64-bit floating point operations while x87 uses 80-bit operations. Using SSE instead of “-pc 64” or “-Kieee” will significantly improve your performance.

With the last release of PGDBG we added a Java interface so the look on Fedora and Redhat should be very similar (unless your Java versions are very different). You might have different versions of PGDBG which would also account for the different interfaces. We don’t support the older Motiff interface any longer so there isn’t a way to go back to it in the current release (5.2-4). If you prefer the new Java interface, then you’ll need to update your installation.

Hope this helps,