This repository contains solutions to lab exercises for the Parallel and Distributed Programming course during the fifth semester at AGH University of Krakow. The exercises cover various parallel programming techniques and paradigms using C and Java. Each lab focuses on a specific topic.
Simulation of a multi-threaded coffee shop where 300 clients (threads) take a mug (limited resource), use one of 4 coffee machines (limited resource), drink their coffee (wait), and repeat the process multiple times before leaving. While not waiting for resources, threads perform 'work' (simulated with busy-waiting and addition operation).
- Clone and Fork: Process creation and management, focusing on understanding process isolation and resource sharing.
- Pthreads: Multithreading with POSIX threads, emphasizing synchronization mechanisms like mutexes, barriers, and condition variables to handle RAW (Read-After-Write), WAR (Write-After-Read), and WAW (Write-After-Write) dependencies.
- OpenMP: Parallel programming with OpenMP directives for shared memory systems, optimizing performance through loop parallelization, task scheduling, and data sharing.
- OpenMPI: Message Passing Interface (MPI) for distributed computing, focusing on inter-process communication, synchronization, and scalability.
- Thread and Runnable: Basic concepts of multithreading, including thread lifecycle management and synchronization.
- Executor and ThreadPool: Efficient thread pooling and task execution, ensuring optimal resource utilization and task scheduling.
- Callable and Future: Tasks that return results, with asynchronous retrieval using Future, enabling better control over task execution and results.
- RecursiveTask and RecursiveAction: Parallel programming with Java's Fork/Join framework for divide-and-conquer tasks, focusing on work-stealing algorithms and load balancing.
- Synchronization: Ensuring thread safety and avoiding race conditions through proper use of synchronization primitives like mutexes, locks, and barriers.
- Performance Optimization: Writing efficient parallel code by minimizing overhead, reducing contention, and maximizing resource utilization.
- Dependency Management: Understanding and resolving RAW, WAR, and WAW dependencies to ensure correct and predictable execution.
- Code Safety: Designing robust and maintainable parallel programs that are free from deadlocks, livelocks, and data corruption.
Some of the exercises in the repository involve modifying or extending code provided as part of the course. While I strive to write clean, well-structured, and maintainable code, the starting code for certain exercises may not reflect my personal coding style or standards. These files are included as-is to provide context for the solutions and to demonstrate how the exercises were approached.
- All C code was originally written on Linux and exercises using Clone, Fork and Pthreads will not work outside of it as they rely on POSIX-specific APIs.
- Each lab directory involving C code should have a working makefile, making use of GCC compiler (MPICC in case of OpenMPI).
- OpenMPI may require a separate installation of MPICC not provided in the repository.
- For Java, only source code is provided and can be compiled with any IDE or directly from the command line.