Code project

• Solve Ax=b
• A matrix A could be a dense or sparse matrix
  • Symmetric or unsymmetric condition might be tested
  • Random matrix for unsymmetric conditions
• For a dense matrix of double-precision
  • LAPACK for CPU solver
    • dgesv for general/unsymmetric matrices
    • dsysv for symmetric matrices
  • CuSolver for GPU solver
    • cusolverDnDgetrf for general/unsymmetric matrices
    • cusolverDnDpotrf for symmetric matrices
• For a sparse matrix using MPI
  • HYPRE for distributed CPU solver
    • Flexible GMRES solver + AMG preconditioner
  • AmgX for distributed GPU solver
    • FGMRES + AMG preconditioner
• Objectives of the project
  • Practice LAPACK/HYPRE/CuSolver/AmgX
  • Practice of CMake with cuda
  • Practice of unit-test using CTest
  • Practice of C++11 or C++14 implementation