Histogram Analysis with Parallel Computing

This repository contains a suite of programs for histogram analysis that demonstrate the use of parallel computing techniques in C++. The parallel implementations are derived from a sequential version provided by a parallel computing course and are designed to showcase different synchronization strategies.

Author

Saleh Elkaza

Course Project Overview

The original sequential version of the histogram (histogram.cpp) was supplied by the course. The parallel versions created from this base include atomic operations, mutexes, and a hybrid approach for thread-safe updates to a shared histogram during large dataset processing.

Implementations and Usage

The following versions are available in this repository:

• histogram-atomic.cpp
• histogram-best.cpp
• histogram-mutex-per-bucket.cpp
• histogram-mutex.cpp
• histogram.cpp (Sequential baseline provided by the course)

Running the Programs

To run any version of the histogram program, compile it using:

g++ -std=c++11 -pthread -o histogram <version>.cpp

And then execute it with appropriate arguments. For example:

./histogram --num-threads 32 --N 1000 --sample-size 30000000 --print-level 1

Automated Testing

The run-test.sh script facilitates automated testing across all versions for different histogram sizes. It measures and logs the performance of each version:

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

This will output the execution times to performance_for_all_versions_results.csv for further analysis.

Performance Comparison