High_Performance_Computing_CSE_4014

High Performance Computing - CSE4014

Setup

There are two main frameworks which are necessary for proceeding with these labs -- openmp and mpi. Programming Experience with C language is also recommended to follow through.

I have performed the installation on a Mac OSX environment, and thus the installation process on other Operating systems would differ.

1. Open MP :

• Installed using GCC Compiler

• I have installed gcc-10 using homebrew in mac.

• Compilation

gcc-10 -fopenmp <file_name.c>

• Execution

./a.out

• To set the number of threads, run this command in the terminal before compiling the program.

export OMP_NUM_THREADS=<number of threads to use> 

2. MPI :

• Installed open-mpi using

brew install openmpi

• To Compile -

mpicc -o <output_file_name> <file_name>.c

• To Run -

mpirun -np <number_of_processors> <output_file_name>

• Default Compiler Location : /usr/local/bin/mpicc

• To create more than 2 processors :

mpirun -np 3 --oversubscribe a.out

Index

1. Fibonacci Series

In this, we use the Open MP constructs to use multiple threads to compute and store the fibonacci series in a parallel manner.

2. Inter Process Communication - Introduction

A template for MPI constructs, and a basic program to get a hang of different constructs used with MPI.
We will also display the details of the processor in the system.

3. MPI - Scatter and Gather Constructs

A program to find the average of a few numbers using the Scatter and Gather constructs of MPI.

4. MPI - Reduction

A program to find the average of a few numbers using the Scatter and Reduce constructs of MPI.

5. MPI - Passing Integers

A program to send and recieve integers between two processes.

6. MPI - Passing Strings

A program to send and recieve string messages between two processes. Note : This program has not been written by me, but instead the author has been credited in the code.

7. MPI - Message Passing with Probing

A program to send and recieve string messages between two processes with probing using MPI_Probe and MPI_Get_count constructs.

8. MPI - Passing Arrays

A program to send and recieve arrays of integers between two processes.

9. MPI - Message Broadcast

A program to broadcast messages from one process to all other processes. This is mainly used for data synchronization among the processes and is achieved through the MPI_Bcast construct.

10. MPI - Ping Pong

A program which will cause a message to bounce back and forth between two processes till a condition is met.

11. MPI - Token Ring

A program which will arrange multiple processes in the form of a ring and make a message float through them. The message passes from say Process 1 through all other processes and returns back to Process 1.