SMG2S - Interface

SMG2S install command will generate a shared library libsmg2s.so into ${INSTALL_DIRECTORY}/lib. It can be used to profit the C wrapper of SMG2S. The compile command for C codes using SMG2S :

Include header file:

#include <interface/C/c_wrapper.h>

create Nilpotency object :

struct NilpotencyInt *n;
n = newNilpotencyInt();
NilpType1(n, 2, 10);

After that, you need to create the parallel Sparse Matrix Object Mt like this :

struct parMatrixSparseDoubleInt *m;
m = newParMatrixSparseDoubleInt();

Generate by SMG2S :

smg2s(m, 10, n, 3 ," ",MPI_COMM_WORLD);

GRelease Nilpotency Object and parMatrixSparse Object :

ReleaseNilpotencyInt(&n);
ReleaseParMatrixSparseDoubleInt(&m);

Interface to Python

Generate the shared library and install the python module of smg2s

cd ./interface/python;
mpicxx -fpic -c smg2s_wrap.cxx  -I/apps/python/3/include/python3.5m -std=c++0x
mpicxx -shared smg2s_wrap.o -o _smg2s.so
python setup.py install

Before the utilisation, make sure that mpi4py installed.

A little example of usge :

from mpi4py import MPI
import smg2s
import sys

size = MPI.COMM_WORLD.Get_size()
rank = MPI.COMM_WORLD.Get_rank()
name = MPI.Get_processor_name()

sys.stdout.write(
    "Hello, World! I am process %d of %d on %s.\n"
    % (rank, size, name))

if rank == 0:
        print ('INFO ]> Starting ...')
        print("INFO ]> The MPI Comm World Size is %d" %size)

#bandwidth for the lower band of initial matrix
lbandwidth = 3

#create the nilpotent matrix
nilp = smg2s.NilpotencyInt()

#setup the nilpotent matrix: 2 = continous 1 nb, 10 = matrix size
nilp.NilpType1(2,10)

if rank == 0:
        print("Nilptency matrix continuous one nb = %d" %nilp.nbOne)

Mt = smg2s.parMatrixSparseDoubleInt()

#Generate Mt by SMG2S
#vector.txt is the file that stores the given spectral distribution in local filesystem.
Mt=smg2s.smg2sDoubleInt(10,nilp,lbandwidth,"vector.txt", MPI.COMM_WORLD)

Interface to PETSc

Include header file

#include <interface/PETSc/petsc_interface.h>

Create parMatrixSparse type matrix :

parMatrixSparse<std::complex<double>,int> *Mt;

Restore this matrix into CSR format :

Mt->Loc_ConvertToCSR();

Create PETSc MAT type :

MatCreate(PETSC_COMM_WORLD,&A);

Convert to PETSc MAT format :

A = ConvertToPETSCMat(Mt);

Interface to Trilinos

Include header file

#include <interface/Trilinos/trilinos_interface.hpp>

Create parMatrixSparse type matrix :

parMatrixSparse<std::complex<double>,int> *Mt;

Create Trilinos/Teptra MAT type :

RCP<CrsMatrix<ScalarType> > K ;

Convert to Trilinos MAT format :

K = ConvertToTrilinosMat(Mt);

Interface

Interface to C

Interface to Python

Interface to PETSc

Interface to Trilinos