14. Multithreading [HPC in Julia]

In this video we will introduce multithreading, which allows you to write code that executes in parallel. Multithreading is a shared memory parallel programming paradigm, and is the first paradigm we will cover in this series. Writing multithreaded code can be notoriously difficult, as it is very easy to introduce correctness bugs due to race conditions. We will see the common methods used to combat race conditions, including mutexes, semaphores and algorithm redesign. This module was designed as an MPAGS (Midlands Physics Alliance Graduate School) module and aimed at postgraduates and early career researchers. Simultaneous Multithreading (SMT) / Hyperthreading: https://en.wikipedia.org/wiki/Simulta... https://www.intel.com/content/www/us/... Timestamps: 00:00:00 Introduction 00:00:40 Concurrency 00:00:56 Multicore CPUs 00:02:33 Multithreading 00:05:11 Launching Julia with multiple threads 00:06:18 Task based parallelism 00:08:46 Parallel "for" loop 00:10:45 Example: Monte-Carlo calculation 00:11:55 Race conditions 00:15:32 Atomics 00:19:52 Mutexes 00:23:58 Semaphores 00:27:12 Mutexes vs Semaphores 00:28:04 Separate memory per thread 00:31:59 Benchmarks 00:35:21 Multithreading considerations Useful links: Git/GitHub for Researchers:    • MLiS1 Introduction to Git by Jamie Mair   Julia Documentation: https://docs.julialang.org/en/v1/ MPAGS Registration: https://warwick.ac.uk/fac/sci/physics... Course Notes: https://jamiemair.co.uk/courses/hpc