Parallel Mandelbrot Set Generator

This repository hosts the implementation of a parallel Mandelbrot set generator. The code in mandelbrot-parallel.cpp leverages multi-threading to compute the Mandelbrot set using either static or dynamic work allocation strategies. This project includes detailed performance analysis, comparing the efficiency of both work allocation methods.

Author

Saleh Elkaza

Features

• Implementation of Mandelbrot set generation using C++ and parallel programming techniques.
• Static and dynamic work allocation for thread management.
• Performance analysis with different thread counts on ALMA node, which has 16 cores with hyperthreading enabled.
• Atomic operations to prevent race conditions when counting pixels inside the set.

Performance Analysis

The run-test.sh script is included to automate the running of the program with various thread counts and work allocation strategies. Results are outputted to performance_results.csv for easy analysis.

Execution Time vs. Number of Threads

Mandelbrot Set Visualizations

Dynamic Worker Allocation

Static Worker Allocation

Building and Running

To build the program, make sure you have a C++ compiler and support for C++20 or later. You can compile the program with a command such as:

g++-11.2.0 -std=c++20 -O2 -lpthread -o s mandelbrot-parallel.cpp

To run:

srun --nodes=<nodes number>

Please replace nodes number with any number you wish and keep in mind that your system support this number of nodes.

To run the tests and generate the performance data:

chmod +x run-test.sh
./run-test.sh

Additional Information

The mandelbrot-helpers.hpp file provided by the parallel computing course contains utility functions and data structures used by the main program.