[SOLVED] EE451 Homework4-  Pass Message in a Ring

Examples

The “mpi examples” folder includes the source codes used in discussions and a pbs file ‘queue.pbs ’. To run an mpi program, for example, the ‘scatter.c’, follow the steps:

1. login hpc-login3.usc.edu
2. source /usr/usc/openmpi/default/setup.sh
3. Go to your working directory which has ‘job.sl’ and ‘scatter.c’.
4. mpicc -o go scatter.c
5. sbatch job.sl
6. After you get the email saying your job is completed, check ‘mpijob.out’ for output and ‘mpijob.err’ for any possible error.

For more information, you can visit https://hpcc.usc.edu/support/documentation/examples-of-mpi-programs/

1             Pass Message in a Ring

Write an MPI program that passes a value around a ring of 4 processes using the following steps.

1. Process 0 initializes Msg = 451 and prints value of Msg
2. Process 0 sends the value of Msg to Process 1
3. Process 1 receives the value of Msg, increases it by 1, prints the value and sends the current value of Msg to Process 2
4. Process 2 receives the value of Msg, increases it by 1, prints the value and sends the current value of Msg to Process 3
5. Process 3 receives the value of Msg, increases it by 1, prints the value and sends the current value of Msg to Process 0
6. Process 0 receives the value of Msg from Process 3 and prints the value

Name this program as ‘p1.c’. Figure 1 illustrates the steps. The output messages look like:

• Process 0: Initially Msg = 451
• Process 1: Msg = 452
• Process 2: Msg = 453
• Process 3: Msg = 454
• Process 0: Received Msg = 454. Done!

Figure 1: Example diagram

2             Add 64 numbers using 4 processes

In the “number.txt” file, you can find 64 numbers. Your task is to write an MPI program with 4 processes to compute the sum of these 64 numbers. There are 3 approaches:

1. Approach 1, name this program as p2 1.c:
   • Each process reads the entire array.
   • Do in parallel: Process 0 computes]; Process 1 computes]; Process 2 computes]; Process 3 computes
   • Process 1,2,3 send their partial sum to Process 0.
   • Process 0 computes the sum of all the partial sums and prints it out.
2. Approach 2, name this program as p2 2.c:
   • Process 0 reads the array
   • Process 0 broadcasts the entire array to every process
   • Do in parallel: Process 0 computes]; Process 1 computes]; Process 2 computes]; Process 3 computes
   • Process 0 uses MPI SUM reduction to sum these partial sum
   • Process 0 prints out the result.
3. Approach 3, name this program as p2 3.c:
   • Process 0 reads the array and scatters the entire array to every process using the scatter
   • Each process sums up the portion of the array it receives.
   • Process 0 uses the gather operation to gather these partial sums, computes the sum of all the partial sums and prints it out.