The Upcoming Multi-threading Infrastructure in Julia
summary
Claims a concrete multithreading implementation for Julia: a data-parallel for with broadcast and barrier synchronization, atomic support in the runtime, and API like k-thread splits. Benchmarks compare against MATLAB and GPU-accelerated code, showing parity on a PDE solver (≈2.96s vs 3.01s) and scaling up to 18 cores for some workloads though some kernels plateau at 5–6 threads. Results are preliminary but show the runtime and primitives actually work.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 0:48Why Multithreading
- 2:47Julia Data Algorithm
- 4:00Multithreading Infrastructure
- 5:00How Multithreading works
- 5:44Things to consider
- 6:24Per thread safety
- 7:05Julia workloads
- 8:38Julia benchmarks
- 9:15Final workflow
- 9:43Preliminary results
- 10:40Does it work
related talks
01h 37mTech Tutorial: Use Julia for your next scientific computing job!
01h 06mJulia: Fast Performance, Distributed Computing & Multiple Dispatch
01h 27mIntroduction to Julia for Scientific Computing and Data Science
01h 56mIntro to Julia
02h 00mIntroduction to Julia
30mJulia for Modern Data Analysis
27mJulia (workshop)
30mProductivity and performance in one programming language - A tour of Julia