combiningmpiandopenmp(编辑修改稿)

combiningmpiandopenmp(编辑修改稿)内容摘要：

eads Copyright 169。 The McGrawHill Companies, Inc. Permission required for reproduction or display. Results of Benchmarking Copyright 169。 The McGrawHill Companies, Inc. Permission required for reproduction or display. Analysis of Results  C+MPI+OpenMP program slower on 2, 4 CPUs because C+MPI+OpenMP threads are sharing memory bandwidth, while C+MPI processes are not  C+MPI+OpenMP programs faster on 6, 8 CPUs because they have lower munication cost Copyright 169。 The McGrawHill Companies, Inc. Permission required for reproduction or display. Case Study: Jacobi Method  Begin with C+MPI program that uses Jacobi method to solve steady state heat distribution problem of Chapter 13  Program based on rowwise block striped deposition of twodimensional matrix containing finite difference mesh Copyright 169。 The McGrawHill Companies, Inc. Permission required for reproduction or display. Methodology  Profile execution of C+MPI program  Focus on adding OpenMP directives to most puteintensive function Copyright 169。 The McGrawHill Companies, Inc. Permission required for reproduction or display. Result of Profiling Function 1 CPU 8 CPUs initialize_mesh % % find_steady_state % % print_solution % % Copyright 169。 The McGrawHill Companies, Inc. Permission required for reproduction or display. Function find_steady_state (1/2) its = 0。 for (。 ) { if (id 0) MPI_Send (u[1], N, MPI_DOUBLE, id1, 0, MPI_COMM_WORLD)。 if (id p1) { MPI_Send (u[my_rows2], N, MPI_DOUBLE, id+1, 0, MPI_COMM_WORLD)。 MPI_Recv (u[my_rows1], N, MPI_DOUBLE, id+1, 0, MPI_COMM_WORLD, amp。 status)。 } if (id 0) MPI_Recv (u[0], N, MPI_DOUBLE, id1, 0, MPI_COMM_WORLD, amp。 status)。 Copy。