JSU Course 2005 Homework 4

Homework 4

High Performance Computing and Computational Science

  • Nearly all students had trouble with homework 3 and so this is a consolidation homework
  • Consider the code shown in http://www.old-npac.org/projects/cpsedu/summer98summary/examples/mpi-c/examples96/simpson-rule_with_mpi.c. Consider 2 cases -- replace #define N 32 by #define N 1000000 or #define N 10000000 
  • Time using UNIX command time ./executable
  • Run jobs with the two values of N and 1 2 or 4 or 8 MPI processes. Run the 8 MPI processor case on 4 real processors. Run other cases on a number of real processors equal to number of MPI proceeses
  • Record results and discuss speedup. Note N is number of points per processor so if you fix N and change number of processes, the size of problem changes proportional to number of processes. This is called sclaed speedup
  • Email Homework document or link to me at gcf@indiana.edu
  • Due Midnight Sunday March 27 2005