Intel(R) Visual Fortran Compiler Professional Edition 10.

[ggslavcheva]

Hello,

I have been using Visual Fortran compiler for more than 3 years on a Dell Precision 690 workstation with two dual core Intel Xeon processors under Windows XP x64 Professional and haven't had any problems until recently when I changed the workstation currently based on Intel Core 2 quad Q6600 running under Windows Vista x64 Business.

I found out that the results of my parallel (and sequential) calculations are substantially different from the ones obtained on the former platform. The source code is absolutely identical on both platforms. I am using MS Visual Studio 2005 and the project settings are absolutely identical with the exception of the processors settings. In order to resolve the problem, I have tried all possible settings, e.g. Pentium 4 compatible with streaming etc., different libraries (dll, static), parallel and sequential code generation, but the results seem to be the same (wrong) despite the change of the settings. I cannot give you details of the fortran code that I am running since it is a very complicated mathematical simulation.

Would you give me any information concerning any patches, releases, updates or a way to solve the problem.

Thank you very much for your assistance.

Sincerely,

Gaby Slavcheva

[TimP]

This question would be more appropriate for the forum on Windows Fortran. You get updates at the site https://registrationcenter.intel.com It is always possible that some additional optimization brings out a problem; such problems are associated typically with undeclared or uninitialized variables, or subscript range violation. The /check options are meant to help diagnose these.

[Steven_L_Intel1]

In the cases I have seen before with such symptoms, the problem was that a function was not properly declared as double precision so the caller assumed it was single precision. On IA-32, this will "seem" to work, you'll get a return value that is close to the actual value, but on Intel 64, you will get a very different (and wrong) result.

Enabling the "Generate Interface Blocks" and "Check Generated Interfaces" options under Diagnostics can help detect such errors.

[ggslavcheva]

Just read on my 1st posting...

"I found out that the results of my parallel (and sequential) calculations are substantially different from the ones obtained on the former platform. The source code is absolutely identical on both platforms. I am using MS Visual Studio 2005 and the project settings are absolutely identical with the exception of the processors settings. In order to resolve the problem, I have tried all possible settings, e.g. Pentium 4 compatible with streaming etc., different libraries (dll, static), parallel and sequential code generation, but the results seem to be the same (wrong) despite the change of the settings."

The only significant difference between the platforms is the Processor - Intel Xeon 3.2GHz (where all the results are OK) and Intel Core 2 Quad 2.4GHz (where the results are quite different)....

Please HELP !

[Steven_L_Intel1]

Since we have no example, I'd suggest that you submit a report to Intel Premier Support and attach a test case that can be built and run to demonstrate the problem. Be sure to provide details such as all compiler options used to build.

If you take the SAME executable and run it on both systems, do the results differ?