Parallel & Distributed Computing 2 - Distributed Memory Programming with MPI скачать в хорошем качестве

Parallel & Distributed Computing 2 - Distributed Memory Programming with MPI

📘 Video Overview This video provides a comprehensive guide to Distributed Memory Programming using MPI (Message-Passing Interface). We begin by setting up the MPI environment using C, covering the compilation process with mpicc and execution with mpiexec. The lesson explores the SPMD (Single-Program Multiple-Data) model, where we learn how to identify processes by rank and manage communication within communicators like MPI_COMM_WORLD. We then apply these concepts to parallelize the Trapezoidal Rule for integration, move into efficient collective communication techniques, and analyze matrix-vector multiplication. Finally, we discuss how to evaluate parallel performance using speedup and efficiency metrics, and review safety practices to prevent deadlocks. 🧠 Topics Covered • MPI Basics: Writing a "Hello World" program, initializing MPI, and understanding Communicators and Ranks. • Compilation & Execution: Using wrapper scripts (mpicc) and launching processes (mpiexec). • Point-to-Point Communication: Implementing MPI_Send and MPI_Recv, and understanding message matching via tags and source ranks. • The Trapezoidal Rule: Parallelizing numerical integration and handling I/O (input/output) limitations where only process 0 accesses stdin. • Collective Communication: optimizing global operations using MPI_Reduce, MPI_Bcast (Broadcast), MPI_Scatter, and MPI_Gather. • Data Distribution: Strategies for partitioning vectors (Block, Cyclic, Block-cyclic) and performing Matrix-Vector Multiplication. • Advanced Features: Creating MPI Derived Datatypes to send complex data structures and using MPI_Barrier for synchronization. • Performance Evaluation: Measuring elapsed time with MPI_Wtime, calculating Speedup and Efficiency, and defining Strong vs. Weak Scalability. • Parallel Sorting & Safety: Implementing the Odd-Even Transposition Sort and avoiding deadlocks using synchronous sends (MPI_Ssend) and MPI_Sendrecv 🎓 About CS Course Companion For every computer science course I took, I went to YouTube looking for clear, high-quality explanations—and almost never found what I needed. So I started creating my own visual study videos to better understand the material. CS Course Companion is a collection of concise, example-driven CS tutorials designed to help students learn faster and with less frustration. 👍 If this helped If you found this video useful: Like the video (it really helps the channel) Subscribe for more CS course walkthroughs Leave a comment if you want a topic explained next This video was created using NotebookLM.