Coarray Fortran library and programs for running some iterative solvers in parallel.
The iterative solvers currently implemented are:
-
Chebyshev;
-
Conjugate Gradient (CG);
-
Preconditioned CG (PCG);
-
Flexible PCG.
The library and programs were compiled and tested with Intel Fortran 2017 (2018). The Intel MKL library (i.e., BLAS, Sparse BLAS, LAPACK, and Pardiso) was used. Sparse matrices are handled with the library libsparse (https://github.com/jvdp1/libsparse).
Earlier versions of the programs were compiled with the GNU compiler Collection (GCC) Fortran and OpenCoarrays.
The modulbe modcoarraysolver includes the different iterative solvers, as derived types (objects). The coefficient matrix and the preconditioner can be provided to the iterative solver under different derived types. Different examples are provided. For example, one example uses a coefficient matrix stored in a CRS format, while another example uses a matrix-free coefficient matrix.
For compilation of Fortran programs: make
For compilation of shared-memory Coarray Fortran programs: make COARRAYENABLE=shared
For compilation of distributed-memory Coarray Fortran programs: make COARRAYENABLE=dist
Compilation with debug options (-g -check all -traceback) is possible by adding the argument DEBUGENABLE=1.