github-actions: cleanup and oneapi 2025.0 release

[marcfehling]

In your Makefile, I reverted back from mpiifx to mpiifort. Further following your suggestions in the oneapi workflows, I specified INTEL_IFORT as an environmental variable. I also made some smaller miscellaneous changes along the way.

With this patch, you will compile FDS with ifx on linux and windows on the latest oneapi release 2025.0. It turns out you really only need to specify INTEL_IFORT=ifx. It appears that Intel MPI then automatically calls mpiifort -fc=ifx.

Thank you @mcgratta @rmcdermo @cxp484 for the discussion.

[mcgratta]

I would prefer to keep mpiifx in the makefile because mpiifort is not supported anymore. We can still use the "classic" Fortran compiler ifort with mpiifx using the environment variable. Everything appears to be working fine now. Unless I am not understanding something, I recommend that we close this PR.

[marcfehling]

I would prefer to keep mpiifx in the makefile because mpiifort is not supported anymore.

Okay, understood. I can remove the first of my commits to just bring the CI up-to-date.

Everything appears to be working fine now. Unless I am not understanding something, I recommend that we close this PR.

The outcome of this PR is that we are updating oneapi on Windows to 2025.0. Previously we still use 2024.2.

[marcfehling]

Please note that you are currently using ifort and mpiifx by default.
https://github.com/firemodels/fds/blob/master/Build/makefile

fds/Build/makefile

Line 16 in d57658c

I_IFORT=ifort

fds/Build/makefile

Line 244 in d57658c

impi_intel_linux : FCOMPL = mpiifx

[mcgratta]

Yes, we are using mpiifix and ifort. We would like to switch over to ifx, but we have found that it slows FDS by approximately 20%. This is not unusual. We have read various un-line discussions that indicate that floating-point intensive "engineering" codes, like CFD, see a slow down with ifx for various reasons. For the moment, we will continue to use ifort until we decide either to switch to ifx or look for another compiler.

Randy, the changes look fine to me, but could you double-check.

[marcfehling]

I understand your reasoning to switch to mpiifx while keeping ifort. Thanks for clarifying!