Assignment 2: Cluster Computing
Background:
In this assignment you are going to work with a parallel code written in C using
the OpenMP library1. The necessary files that need to be used are indicated in the or
can be found in your home directory on Snellius2 ($HOME/linux-cluster-
computing/cluster/batch/)
Assignment:
Analyse of the computation of PI based on the Leibniz Series 3, 4
Given the previous introduction, perform the following tasks:
• Task1: Thread Scalability Study for Parallel Pi Computation
▪ Compile and run a sequence of pi computations with different number of
threads: from 8 threads, you should increase the number of threads in 4
for every successive simulation, until you reach 48 threads.
▪ Plot a graph to show the execution times of the computation (Y-axis)
with respect to the amount of threads (X-axis).
▪ Question: How would you explain the observed performance,
considering that you have requested 32 cores from a node on Snellius?
• Task2: Execution Time Analysis of Pi Computation
▪ Compile the code with 31250000, 62500000, 125000000, 250000000,
500000000, 1000000000 and 2000000000 numbers in the Leibniz
series (changing the value of "niter") and run the pi computation only
with 32 threads in all cases.
▪ Plot a graph to show the execution times of the computation (Y-axis)
with respect to the amount of iterations (X-axis).
▪ Question: How does the execution time scale with the number of
iterations in the Leibniz series?
Submission:
Submit a short report in Canvas before the deadline, including the graphs and answers
to the questions.
Note: Submitting after the deadline will result in losing points (1 point for every 30
minutes after the deadline)
1 The parallel code is provided, you don’t need to understand it, there will be a full workshop on
parallel programming (OpenMP and MPI)
2 [Link]
3 [Link]
4 [Link]
Code and Scripts:
▪ The code needs to be compiled with the following commands on Snellius:
#!/bin/bash
module load 2022
module load GCCcore-11.3.0
gcc -fopenmp -o pi pi.c -lm
▪ The basis script to submit the job would be as follows:
#!/bin/bash
#SBATCH --job-name="pi"
#SBATCH --nodes=1
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=32
#SBATCH --time=[Link]
#SBATCH --partition=normal
#SBATCH --output=pi_%[Link]
#SBATCH --error=pi_%[Link]
module purge
module load 2022
module load GCCcore-11.3.0
echo "OpenMP parallelism"
echo
for ncores in `seq 8 4 48`
do
export OMP_NUM_THREADS=$ncores
echo "CPUS: " $OMP_NUM_THREADS
echo "CPUS: " $OMP_NUM_THREADS >&2
./pi
echo "DONE "
done
▪ C PI program:
#include <omp.h>
#include <math.h>
#include <stdio.h>
#include <stdlib.h>
int main (int argc, char *argv[])
{
//initialize variables
int i;
double pi = 0;
int niter = 1000000000;
// Get timing
double start,end;
start=omp_get_wtime();
// Calculate PI using Leibnitz sum
/* Fork a team of threads */
#pragma omp parallel for reduction(+ : pi)
for(i = 0; i < niter; i++)
{
pi = pi + pow(-1, i) * (4 / (2*((double) i)+1));
}
/* Reduction operation is done.
All threads join master thread
and disband */
// Stop timing
end=omp_get_wtime();
// Print result
printf("Pi estimate: %.20f, obtained in %f seconds\n", pi,
end-start);
