Lis (linear algebra library)

Lis (Library of Iterative Solvers for linear systems) is a scalable parallel library for solving systems of linear equations and standard eigenvalue problems with real sparse matrices using iterative solvers.

Features

Lis provides facilities for:

• Automatic makefile generation
• NUMA aware hybrid implementation with MPI and OpenMP
• Exchangeable dense and sparse matrix storage formats
• Basic linear algebra operations for dense and sparse matrices
• Parallel iterative solvers for systems of linear equations and standard eigenvalue problems
• Parallel preconditioners for iterative solvers
• Hardware accelerated internal quadruple (double-double) precision floating point operations
• Performance analysis

Example

A C program to solve a system of linear equations Ax=b is written as follows:

#include <stdio.h>
#include "lis_config.h"
#include "lis.h"
 
int main(int argc, char* argv[])
{
  LIS_MATRIX A;
  LIS_VECTOR b,x;
  LIS_SOLVER solver;
  int iter;
  double times;
 
  lis_initialize(&argc, &argv);
  lis_matrix_create(LIS_COMM_WORLD,&A);
  lis_vector_create(LIS_COMM_WORLD,&b);
  lis_vector_create(LIS_COMM_WORLD,&x);
  lis_solver_create(&solver);
  lis_solver_set_optionC(solver);
  lis_input(A,b,x,argv[1]);
  lis_vector_set_all(1.0,b);
  lis_vector_duplicate(A,&x);
  lis_solve(A,b,x,solver);
  lis_solver_get_iters(solver,&iter);
  lis_solver_get_time(solver,&times);
  printf("iter= %d time = %e\n",iter,times);
  lis_solver_destroy(solver);
  lis_matrix_destroy(A);
  lis_vector_destroy(b);
  lis_vector_destroy(x);
  lis_finalize();
  return 0;
}

System requirements

Installation of Lis requires a C compiler. To use the Fortran interface, a FORTRAN 77 compiler is required. To use the algebraic multigrid preconditioner, a Fortran 90 compiler is required. OpenMP and MPI-1 are used in parallel computing environments. Both the Harwell-Boeing and Matrix Market formats are supported for the import and export of user data.

References

• Akira Nishida (2010). "Experience in Developing an Open Source Scalable Software Infrastructure in Japan". Computational Science and Its Applications – ICCSA 2010. Lecture Notes in Computer Science 6017. Springer. pp. 87–98. doi:10.1007/978-3-642-12165-4_36. ISBN 3642121640. 
• Hisashi Kotakemori, Hidehiko Hasegawa, Tamito Kajiyama, Akira Nukada, Reiji Suda, and Akira Nishida (2008). "Performance Evaluation of Parallel Sparse Matrix-Vector Products on SGI Altix 3700". OpenMP Shared Memory Parallel Programming. Lecture Notes in Computer Science 4315. Springer. pp. 153–163. doi:10.1007/978-3-540-68555-5_13. ISBN 3540685545. 
• Akihiro Fujii, Akira Nishida, and Yoshio Oyanagi (2005). "Evaluation of Parallel Aggregate Creation Orders : Smoothed Aggregation Algebraic Multigrid Method". High Performance Computational Science and Engineering. Springer. pp. 99–122. doi:10.1007/0-387-24049-7_6. ISBN 144193684X. 

See also

• List of numerical libraries

External links

• Official Lis web page
• Jack Dongarra's Linear Algebra Software Page
• RPM Free Download for Linux (Courtesy of Fedora Project)