michaelk4/parallel-implementation-of-submatrix-search-final-project

This project is aimed at developing a simplified recognition algorithm for identifying objects in each given picture with an appropriate matching value. The program takes a set of pictures and a set of objects, and searches for three different objects in each picture. The output includes the list of objects found in each picture.

Parallel-implementation-of-Submatrix-Search-Final-project

This project is aimed at developing a simplified recognition algorithm for identifying objects in each given picture with an appropriate matching value. The program takes a set of pictures and a set of objects and searches for three different objects in each picture. The output includes the list of objects found in each picture.

The program runs on the Linux operating system to use the program in the terminal do this

To compile the code write in the terminal: make all

To run the program write in the terminal: make run

To clear all: write in the terminal: make clean

Works for both the sequential solution and the parallel solution.

The parallelization

The parallelization uses MPI(Message Passing Interface) and OpenMP(Open Multi-Processing), with MPI dividing the work on the pictures between two processes and OpenMP was used to assign each thread to process a subset of pictures and to execute multiple iterations of a loop in parallel. Each process received two threads.

Overall the sequential solution under the Linux operating system with 4 cores runs on average about 32.5 seconds and the parallel solution for about 17.6 seconds.

parallel-implementation-of-submatrix-search-final-project's Issues

The program doesn't run on two different computers as described in the instructions.

The program starts running on both computers using each computer as a process but after starting a search with the master rank it goes on to finish the job with the worker rank, sends the results back to the master and gets stuck. To terminate this use ctrl+c in the terminal or don't use this option at all with make runOn2.

The MPI functionality doesn't change enough the runtime of the program

Bad load blancing