Cannot find mpi module in VS 2022 oneAPI 2024.1 Fortran

[CarlSteefel]

Still having problems to get Visual Studio to find the mpi module.  Basically the "USE mpi" does not work.  I followed closely (exactly) the recommendations from Devorah at Intel, including updates that recognize that I_MPI_ONEAPI_ROOT has been phased out.  The syntax recommended is <Installation_Dir>, I tried this directly, but then also tried:

$(ONEAPI_ROOT)\mpi\latest plus the various sub-directories

(you name it, I tried them all)

In Fortran General:

$(ONEAPI_ROOT)\mpi\latest
$(ONEAPI_ROOT)\mpi\latest\lib

and in LINKER General:

C:\Program Files (x86)\Intel\oneAPI\mpi\latest\lib

and 

$(ONEAPI_ROOT)\mpi\latest\lib

$(ONEAPI_ROOT)\mpi\latest\include

in the INCLUDE library path. 

Also I added these to the ENVIRONMENT field in the DEBUGGING part of the Visual Studio Properties Page, e.g.:

PATH=$(ONEAPI_ROOT)\mpi\latest\bin;$(ONEAPI_ROOT)\mpi\latest\opt\mpi\libfabric\bin;$(PATH)

Error (one of many of similar ilk) is:

C:\Whatever\CrunchODiTi\Source\AddElectrode.F90(20): error #7002: Error in opening the compiled module file. Check INCLUDE paths. [MPI]

[Barbara_P_Intel]

What version of IMPI are you using on Windows?

I'll move this over to the MPI Forum. They'll be able to help you better.

[CarlSteefel]

Perhaps it is more complicated than I thought.  It seems to be a conflict between some of the PETSc libraries (specifically petscmat) and the MPI module.  Going back to older versions of PETSc allows it to build, so the problem may be in the petscmat routines.  All of the paths for MPI from Intel (2021.12) seem to be set properly.  However, when one builds on Linux instead of Windows and Visual Studio 2022, no such problem occurs.

[Sergey_K_Intel3]

I tried to reproduce this issue on my end, following the instructions here Configuring a Microsoft Visual Studio* Project (intel.com), using Intel oneAPI Base Kit 2024.1 and oneAPI HPC Kit 2024.1. I am able to build and run my MPI project without any issues.

If you believe there are other factors that are contributing to the issue (PETSc libraries?), please provide a sample code that I can use as a reproducer.