This directory has a few examples demonstrating the work with matrices.

MKL

Define directory where mkl libraries are installed, e.g.
I have MKL in HOMEMKL=/opt/intel/mkl2020
and then
 source $HOMEMKL/bin/compilervars.csh intel64 -platform linux

If you have full intel parallel studio, this step is not needed after

source /opt/intel/bin/compilervars.csh intel64

check:

 echo ${MKLROOT}

cd src:

dgemm_example.f  has a commented example to
compute real matrix C=alpha*A*B+beta*C using Intel® MKL function dgemm

define F95ROOT as MKLROOT, e.g.
 export F95ROOT=$MKLROOT
or
 setenv F95ROOT $MKLROOT

gfortran -fdefault-integer-8 -I${F95ROOT}/include/intel64/ilp64 -m64 -I${MKLROOT}/include dgemm_example.f -o tdgemm.exe ${F95ROOT}/lib/intel64/libmkl_lapack95_ilp64.a -Wl,--start-group ${MKLROOT}/lib/intel64/libmkl_gf_ilp64.a ${MKLROOT}/lib/intel64/libmkl_sequential.a ${MKLROOT}/lib/intel64/libmkl_core.a -Wl,--end-group -lpthread -lm -ldl

run:

./tdgemm.exe


To find code examples that use Intel® MKL PARDISO routines to solve systems of linear equations,
unzip the archive file in the examples folder of the Intel® MKL installation directory.
Code examples will be in the examples/solverc/source folder.

E.g. I have MKL in HOMEMKL=/opt/intel/mkl2020

ls $HOMEMKL/compilers_and_libraries_2020.0.166/linux/mkl/examples

gives

 examples_cluster_c.tgz  examples_core_c.tgz  examples_f95.tgz
 examples_cluster_f.tgz  examples_core_f.tgz

core_f.txt  has the list of codes from examples_core_f.tgz
f95.txt has the list of codes from examples_f95.tgz


pardiso_unsym_f.f is an example program to show the use of the "PARDISO" routine
for nonsymmetric linear systems, define F95ROOT as described above

cd src:

   gfortran -fdefault-integer-8 -I${F95ROOT}/include/intel64/ilp64 -m64 -I${MKLROOT}/include pardiso_unsym_f.f -o tpardiso.exe ${F95ROOT}/lib/intel64/libmkl_lapack95_ilp64.a -Wl,--start-group ${MKLROOT}/lib/intel64/libmkl_gf_ilp64.a ${MKLROOT}/lib/intel64/libmkl_sequential.a ${MKLROOT}/lib/intel64/libmkl_core.a -Wl,--end-group -lpthread -lm -ldl

or try this:

   gfortran -m64  -w -fno-second-underscore -fcray-pointer -x f77-cpp-input -I${MKLROOT}/include \
   ./pardiso_unsym_f.f -x none -Wl,--start-group  "${MKLROOT}/lib/intel64/libmkl_gf_lp64.a" \
   "${MKLROOT}/lib/intel64/libmkl_gnu_thread.a" "${MKLROOT}/lib/intel64/libmkl_core.a" \
   -Wl,--end-group -L"${MKLROOT}/../compiler/lib/intel64" -liomp5 -lpthread -lm -ldl -o tpardiso.exe


run
   ./tpardiso.exe


Another example for 3-diagonal N=10 matrix:

gfortran -m64 -w -fno-second-underscore -fcray-pointer -x f77-cpp-input -I${MKLROOT}/include ./pardisoTridiag1.f -x none -Wl,--start-group "${MKLROOT}/lib/intel64/libmkl_gf_lp64.a" "${MKLROOT}/lib/intel64/libmkl_gnu_thread.a" "${MKLROOT}/lib/intel64/libmkl_core.a" -Wl,--end-group -L"${MKLROOT}/../compiler/lib/intel64" -liomp5 -lpthread -lm -ldl -o tpardisoTridiag1.exe
or
ifort -I${MKLROOT}/include ./pardisoTridiag1.f -Wl,--start-group "${MKLROOT}/lib/intel64/libmkl_gf_lp64.a" "${MKLROOT}/lib/intel64/libmkl_gnu_thread.a" "${MKLROOT}/lib/intel64/libmkl_core.a" -Wl,--end-group -L"${MKLROOT}/../compiler/lib/intel64" -liomp5 -lpthread -lm -ldl -o tpardisoTridiag1.exe

For arbitrary N
reading form files like SBlcsrTridiag{N}.dat, rename it to SBlcsrTridiagN.dat

gfortran -m64 -w -fno-second-underscore -fcray-pointer -x f77-cpp-input -I${MKLROOT}/include ./pardisoTridiagN.f -x none -Wl,--start-group "${MKLROOT}/lib/intel64/libmkl_gf_lp64.a" "${MKLROOT}/lib/intel64/libmkl_gnu_thread.a" "${MKLROOT}/lib/intel64/libmkl_core.a" -Wl,--end-group -L"${MKLROOT}/../compiler/lib/intel64" -liomp5 -lpthread -lm -ldl -o tpardisoTridiagN.exe
or
ifort -I${MKLROOT}/include ./pardisoTridiagN.f -Wl,--start-group "${MKLROOT}/lib/intel64/libmkl_gf_lp64.a" "${MKLROOT}/lib/intel64/libmkl_gnu_thread.a" "${MKLROOT}/lib/intel64/libmkl_core.a" -Wl,--end-group -L"${MKLROOT}/../compiler/lib/intel64" -liomp5 -lpthread -lm -ldl -o tpardisoTridiagN.exe

The code
 pardisoTridiagN.f
fails with segmentation fault due to allocation statements

pardisoTridiag3.f with fixed array dimensions for N=1000 works, but produces bad results like MUMPS and our m28y12.f:

gfortran -m64 -w -fno-second-underscore -fcray-pointer -x f77-cpp-input -I${MKLROOT}/include ./pardisoTridiag3.f -x none -Wl,--start-group "${MKLROOT}/lib/intel64/libmkl_gf_lp64.a" "${MKLROOT}/lib/intel64/libmkl_gnu_thread.a" "${MKLROOT}/lib/intel64/libmkl_core.a" -Wl,--end-group -L"${MKLROOT}/../compiler/lib/intel64" -liomp5 -lpthread -lm -ldl -o tpardisoTridiag3.exe
or
ifort -I${MKLROOT}/include ./pardisoTridiag3.f -Wl,--start-group "${MKLROOT}/lib/intel64/libmkl_gf_lp64.a" "${MKLROOT}/lib/intel64/libmkl_gnu_thread.a" "${MKLROOT}/lib/intel64/libmkl_core.a" -Wl,--end-group -L"${MKLROOT}/../compiler/lib/intel64" -liomp5 -lpthread -lm -ldl -o tpardisoTridiag3.exe

much better for diagonal-dominated case:

gfortran -m64 -w -fno-second-underscore -fcray-pointer -x f77-cpp-input -I${MKLROOT}/include ./pardisoTridiag3C1000.f -x none -Wl,--start-group "${MKLROOT}/lib/intel64/libmkl_gf_lp64.a" "${MKLROOT}/lib/intel64/libmkl_gnu_thread.a" "${MKLROOT}/lib/intel64/libmkl_core.a" -Wl,--end-group -L"${MKLROOT}/../compiler/lib/intel64" -liomp5 -lpthread -lm -ldl -o tpardisoTrid3C1000.exe
or
ifort -I${MKLROOT}/include ./pardisoTridiag3C1000.f -Wl,--start-group "${MKLROOT}/lib/intel64/libmkl_gf_lp64.a" "${MKLROOT}/lib/intel64/libmkl_gnu_thread.a" "${MKLROOT}/lib/intel64/libmkl_core.a" -Wl,--end-group -L"${MKLROOT}/../compiler/lib/intel64" -liomp5 -lpthread -lm -ldl -o tpardisoTrid3C1000.exe


Next is only in mkl2020 (mkl for 2013 does not work):

mkl2020 # aliased to e.g.
source /opt/intel/mkl2020/compilers_and_libraries_2020.3.279/linux/mkl/bin/mklvars.csh intel64

ifort -I${MKLROOT}/include mkl_spblas.f90 fgmres_no_precon_f.f90 -Wl,--start-group "${MKLROOT}/lib/intel64/libmkl_gf_lp64.a" "${MKLROOT}/lib/intel64/libmkl_gnu_thread.a" "${MKLROOT}/lib/intel64/libmkl_core.a" -Wl,--end-group -L"${MKLROOT}/../compiler/lib/intel64" -liomp5 -lpthread -lm -ldl -o gmres_no_precon_f.exe

ifort -I${MKLROOT}/include mkl_spblas.f90 fgmresTridiag1.f90 -Wl,--start-group "${MKLROOT}/lib/intel64/libmkl_gf_lp64.a" "${MKLROOT}/lib/intel64/libmkl_gnu_thread.a" "${MKLROOT}/lib/intel64/libmkl_core.a" -Wl,--end-group -L"${MKLROOT}/../compiler/lib/intel64" -liomp5 -lpthread -lm -ldl -o fgmresTridiag1.exe

-- the same for Tridiag3, but nor convergence with slight perturbations of expected solution