This project aims to analyze the performance of three different implementations (sequential, Pthreads, and OpenMP) of a video filtering algorithm. The algorithm applies high-pass filters to each frame of a video, and the performance is measured by the execution time and speedup achieved using parallel programming techniques.
- Sequential Implementation: A single-threaded approach that processes each frame of the video one after the other.
- Pthreads Implementation: A multi-threaded approach using the Pthreads library to process multiple frames in parallel.
- OpenMP Implementation: Another multi-threaded approach using the OpenMP library to parallelize the frame processing.
- g++
- OpenCV
- Python 3
- Matplotlib
- NumPy
The programs can be compiled using the provided Python script, which also manages the execution and analysis.
Execute the runner.py script to compile the programs, run the analysis on the provided video files, and generate the plots.
python3 runner.pyFor each video file, we measure the execution time and speedup for the sequential, Pthreads, and OpenMP implementations. The results are saved as individual plots in the plots directory.
A combined plot shows the execution time across different videos for each implementation.
An average plot shows the execution time and speedup across all videos for each implementation.
This plot shows the average speedup achieved by the Pthreads and OpenMP implementations as the number of threads varies (6, 8, 10, and 12).
The analysis demonstrates the effectiveness of parallel programming in improving the performance of video processing algorithms. The results indicate how the number of threads impacts the speedup and overall execution time, providing valuable insights for optimizing performance in multi-threaded environments.