Compilation of Fortran Code in VS 2022

[Jaensch]

Hello,

At the moment I am trying to compile an older .F file for a combustion model in GT Power.
In the next step, I want to make some changes to the older .F combustion model. But first, I want to find a solution to compile the Fortran code.

Using the procedure from the UserCode.pdf and Visual Studio, I can build a .dll file.

But if I use the same GT model for the older .dll and for the new .dll, I do not get the same results.
In both cases, however, I used the same .F code and the same GT model.

So I think the only difference is the compiler. Because now I use Visual Studio 2022, and the older .dll it was 2015. But should the compiler affect the results, or did I overlook something?

Here are the Setups:

My old setup in GT 2024:

• GT version: v2024
• Fortran code
• UserCodeReference folder used: %GTIHOME%\v2020\usercode\Fortran\UserCodeReference
• Microsoft Visual Studio: version 2015 à Intel Parallel Studio XE 2017
• Binary file: .dll generated from the above
• Results: stored after running the GT v2024 model with the .dll

My current setup:

• GT version: v2024
• Fortan code (unchanged)
• UserCodeReference folder used: %GTIHOME%\v2020\usercode\Fortran\UserCodeReference
• Microsoft Visual Studio: version 2022 à Intel® Fortran Compiler 2025.3.2 [Intel(R) 64]
• Binary file: .dll generated from the above
• Results: stored after running the GT v2024 model with the .dll
  

  I know it is difficult to present a solution with so little information, but maybe anybody have a few ideas about what it could be caused by.

  Thanks for any helpt!

Best regards,

Jaensch

[jimdempseyatthecove · Accepted solution]

This may be a floating point issue.

First try disabling optimizations.

That failing, in the project properties, try setting the

Properties | Fortran | Code Generation | Intel Processor-Specific Optimization | (pull down) ...

and select older ISA (QxSSE3, or QxAVX).

Jim Dempsey

[Jaensch]

Thank you very much!!!

Disabling optimizations works! Now it gives me the same results.

But why did that work? What exactly is the compiler doing differently now?

[witwald]

@Jaensch, from your description, it appears that the original code was compiled with optimizations off. Whoever did that was a defensive programmer. When performing optimizations, the compiler writes object code that it thinks will improve the speed of execution in comparison to the non-optimized object code. As you have encountered, there may be differences in the results that are produced by optimized and non-optimized code. With any program, particularly larger codebases, it may be prudent to initially compile with all optimizations turned off. In my experience, this seems to be a reasonable and robust approach to take.

[Jaensch]

Thank you for the explanation!