Hands-on Lab

  1. Scalability Test.

    • Set TRIES to 100,000 and measure the performance with different numbers of processes multiple times. Then, use Google Sheets to create a Scalability chart.
    #!/bin/bash
    #PBS -N pi-mt-Intel-OMPI
    #PBS -P ACD110018
    #PBS -q cf160
    #PBS -o out.log
    #PBS -e err.log
    #PBS -l mpiprocs=$(TRY_IT_YOURSELF!)
    
    # Please set TRIES to 100,000 and experiment with different numbers of processes (NUM_OF_PROCS).
    NUM_OF_PROCS=$(TRY_IT_YOURSELF!)
    TRIES=100000
    
    module purge
    module load gcc/9.3.0
    module load mpi/openmpi-4.0.5/gcc930
    
    make clean && make
    
    echo "Calculat pi by Monte Carlo method with 100000 tries!!"
    
    time mpirun --mca btl ^openib -np $NUM_OF_PROCS /home/tim911004/SC25-WinterCamp-Lab/mpi-monte-carlo $TRIES
    
  2. Error Test.

    • Keep the number of processes constant at 1000, and vary the number of TRIES. Then, create a graph to show how the error changes with different TRIES values.
    #!/bin/bash
    #PBS -N pi-mt-Intel-OMPI
    #PBS -P ACD110018
    #PBS -q cf160
    #PBS -o out.log
    #PBS -e err.log
    #PBS -l mpiprocs=1000
    
    # Please set NUM_OF_PROCS to 1000 and experiment with various values for TRIES.
    NUM_OF_PROCS=1000
    TRIES=$(TRY_IT_YOURSELF!)
    
    module purge
    module load gcc/9.3.0
    module load mpi/openmpi-4.0.5/gcc930
    
    make clean && make
    
    echo "Calculat pi by Monte Carlo method with 100000 tries!!"
    
    time mpirun --mca btl ^openib -np $NUM_OF_PROCS /home/tim911004/SC25-WinterCamp-Lab/mpi-monte-carlo $TRIES