hello-sparse-blas

Demonstrate sparse matrix-vector multiplication with Intel MKL. Also has various structural components that are intended to be transferable across projects.

Use either pFUnit or test-drive for unit testing.

Make sure you have the Intel oneAPI HPC Toolkit or Intel Fortran Essentials.

Then, you can configure, build, and run tests for the project.

# test-drive builds faster than pfunit
./config/intel_ubuntu_test_drive

Alternatively, if on a system where you need to specify a different MKL library and MKL include path, you can do this with the flags -DMKLLIB and -DMKLINCLUDE during configuration. The below bash script does the necessary configuration, as an example, for DKRZ's Levante.

./config/intel_levante

You can build the documentation separately with

cmake --build build --target doxygen_docs 

References

• Test file based on gist.
• Cmake linking based on link line advisor
• Installing intel fortran compiler
• Installing mkl
• Set environment vars
• I did INCLUDE mkl_spblas.f90 above the PROGRAM statement and the use mkl_spblas in the program block since i mimicked the direct sparse solver example for f90 on pg. 1941 of the mkl manual
• Directory structure from fpm even though using cmake
• fftpack uses cmake for dependencies instead of git submodule and fetch content
• Doxygen real Fortran examples: NOAA-ECM/fv3atm