# include "mpi.h"
int main(int argc, char *argv[])
{
char mpi_lib_ver[MPI_MAX_LIBRARY_VERSION_STRING];
int len;
MPI_Get_library_version(mpi_lib_ver, &len);
printf("%s\n",mpi_lib_ver);
}
(1129) $ cat CMakeLists.txt
cmake_minimum_required(VERSION 3.14)
project(mpiversion VERSION 1.0.0 LANGUAGES C)
find_package(MPI REQUIRED)
message(STATUS "MPI_C_COMPILER: ${MPI_C_COMPILER}")
add_executable(mpiversion.exe mpiversion.c)
target_link_libraries(mpiversion.exe PRIVATE MPI::MPI_C)
...
(1132) $ cmake ..
-- The C compiler identification is GNU 8.3.0
-- Check for working C compiler: /usr/local/other/gcc/8.3.0/bin/gcc
-- Check for working C compiler: /usr/local/other/gcc/8.3.0/bin/gcc -- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Detecting C compile features
-- Detecting C compile features - done
-- Found MPI_C: /gpfsm/dulocal/sles12/intel/2019/compilers_and_libraries_2019.5.281/linux/mpi/intel64/lib/release/libmpi.so (found version "3.1")
-- Found MPI: TRUE (found version "3.1")
-- MPI_C_COMPILER: /usr/local/intel/2019/compilers_and_libraries_2019.5.281/linux/mpi/intel64/bin/mpigcc
-- Configuring done
-- Generating done
-- Build files have been written to: /home/mathomp4/MPIVersionCMake/C/build
(1133) $ make
Scanning dependencies of target mpiversion.exe
[50%] Building C object CMakeFiles/mpiversion.exe.dir/mpiversion.c.o
[100%] Linking C executable mpiversion.exe
[100%] Built target mpiversion.exe
(1134) $ ./mpiversion.exe
Intel(R) MPI Library 2019 Update 5 for Linux* OS
```
it even correctly used `mpigcc`. But as about 95% of my coding is in Fortran and using Fortran libraries, I’d like to get the gfortran working.
---
### Author:  [@brad.king](https://discourse.cmake.org/u/brad.king)
#### Post date: [December 16, 2019, 10:18pm UTC](https://discourse.cmake.org/t/help-with-cmake-intel-mpi-and-gnu-fortran/377/2 "2019-12-16T22:18:15Z")
Thanks for the detailed report. I think your `MPI_Fortran_COMPILER` setting is being used but there should not be any difference between `mpifc` and `mpif90` in this use case. The FindMPI module [sets environment variables](https://gitlab.kitware.com/cmake/cmake/blob/v3.16.1/Modules/FindMPI.cmake#L404-417) while using the MPI compiler `-show` option that configure the matching underlying Fortran compiler anyway. `mpifc` will just forward to `mpif90`.
FindMPI is flipping the include directory order from that reported by `mpif90 -showme:compile`. I tracked down the code doing that and opened [CMake Issue 20098](https://gitlab.kitware.com/cmake/cmake/issues/20098) for this bug. Please follow that issue for further updates.
---
### Author:  [@chuckatkins](https://discourse.cmake.org/u/chuckatkins)
#### Post date: [December 19, 2019, 5:23pm UTC](https://discourse.cmake.org/t/help-with-cmake-intel-mpi-and-gnu-fortran/377/3 "2019-12-19T17:23:15Z")
Please try [https://gitlab.kitware.com/cmake/cmake/merge\_requests/4153](https://gitlab.kitware.com/cmake/cmake/merge_requests/4153) , I believe it should fix the issue.
---
### Author:  [@mathomp4](https://discourse.cmake.org/u/mathomp4)
#### Post date: [January 6, 2020, 4:53pm UTC](https://discourse.cmake.org/t/help-with-cmake-intel-mpi-and-gnu-fortran/377/4 "2020-01-06T16:53:25Z")
@chuckatkins I copied your new `FindMPI.cmake` into a 3.16.2 download and built and all seems good! I can run my little tester with GCC 9.2 and Intel MPI and PGI 19.10 and Intel MPI and both seem happy. I even tried it with a more important code that doesn’t need `-DMPI_Fortran_MODULE_DIR` anymore to succeed.
Thank you, and I eagerly await CMake 3.17 (or 3.18 or whenever it gets in).
---
### Author:  [@Dom\_Heinzeller](https://discourse.cmake.org/u/Dom_Heinzeller)
#### Post date: [June 25, 2022, 1:25am UTC](https://discourse.cmake.org/t/help-with-cmake-intel-mpi-and-gnu-fortran/377/5 "2022-06-25T01:25:09Z")
Matt, I am seeing the same problem still, with cmake 3.22.1 but for Intel OneAPI MPI 2021.4.0 with GNU 10.1.0.
```auto
f951: Fatal Error: Reading module ‘/gpfsm/dulocal/sles12/intel/oneapi/2021/mpi/2021.4.0/include/mpi.mod’ at line 1 column 2: Unexpected EOF
```
It should be picking the gfortran module instead? Looks like it is getting the order wrong:
```auto
... -I/gpfsm/dulocal/sles12/intel/oneapi/2021/mpi/2021.4.0/include -I/gpfsm/dulocal/sles12/intel/oneapi/2021/mpi/2021.4.0/include/gfortran/10.2.0 -ffree-line-length-none ...
```
In `CMakeCache.txt`:
```auto
MPI_Fortran_COMPILER_INCLUDE_DIRS:STRING=/gpfsm/dulocal/sles12/intel/oneapi/2021/mpi/2021.4.0/include/gfortran/10.2.0;/gpfsm/dulocal/sles12/intel/oneapi/2021/mpi/2021.4.0/include
```