Win32 ftn95 binaries on Win64 - pitfalls?

A client is running our Win32 software on Win64 XP. Are there are any issues with floating point underflow? Our program has generated an underflow and crashed. Currently, we don't have Win64 for testing, so debugging this may take time. A quick work-around would be to remove the (deliberate) underflow - it comes from the DLAMCH routine in Lapack, but I would think that underflows were fairly common, so this would not be a full solution.

John,

I work on Win64 XP and have no issues with floating point underflow (or overflow or anything else for that matter), using an AMD Phenom quadcore processor. The 64bit OS appears to run 32bit stuff flawlessly (and fast!). However, I do not use any third party libraries like Lapack (do you compile this from source or use a DLL?), other than OpenGL.

cheers, John

It's ftn95-compiled from downloaded source obtained from Netlib. I guess, I'll have to persuade my management to stump up for Win64 box. Although it happens in real life, it's always bad style to get the users to debug your programs

I see FTN95 works on Vista 64 no problems (did not try debugger yet as I've heard there were some compaints about it), but today noticed that OpenGL does not work at all. Freezes program. Will look further, may be there exist some bug somewhere in my code.

No, seems there is some other reason for program crash. The original examples supplied with FTN95 also do not work. Could be opengl32.dll and glu32.dll as they are 32-bit builds causing these problems in Vita64. Or Vista itself which is still immature and has a lot of small problems which need fixing

I have no problems with OpenGL on XP64, my code only uses opengl32.dll (not glu32.dll) and all the original examples supplied with FTN95 also work.