A Tour of Julia
summary
Julia aims to be a fast dynamic language by design: a method-based JIT to LLVM compiles per-method native code, using 90% of dynamic behavior is static to specialize aggressively. Demonstrations show emitted LLVM and x86 assembly, examples where simple arithmetic and map chains become single-machine instructions, and claims of being within a few percentage points of Fortran or even beats Fortran by up to 3x in some cases. Notable users include Celeste processing 178 TB on 650,000 cores and a next‑generation climate model at MIT/JPL.
AI-written summary · about
chapters (30)
- 0:00Start
- 3:45What is Julia?
- 6:52Celeste Project
- 10:51What does it look like?
- 15:09Julia REPL
- 15:41Hello World
- 15:59Composite Types
- 18:38Field Access
- 20:14For Loops
- 21:29If Statement
- 24:15Creating a Dictionary
- 25:22Accessing Elements
- 26:13Anonymous Functions and Closures
- 28:05Partial Application
- 30:39Overhead from Classes
- 32:14Memory Fragmentation
- 33:19Fragmentation of Arrays in Memory
- 34:35Julia Version
- 36:45Types Change
- 37:54Is it Justin Time Compilation?
- 39:52Type Annotations
- 40:52Language Design
- 43:47Archers, Pikemen and Knights
- 44:59Making Archers, Pikeman and Knights Fight
- 46:18Archer vs Everybody Else
- 46:51Pikeman vs Everybody Else
- 47:25Single vs Multiple Dispatch
- 47:47Dynamic Single Dispatch
- 49:50Static Single Dispatch
- 55:30Simulate JIT in Julia
related talks
43mHow Julia Goes Fast
02h 00mIntroduction to Julia
01h 38mAn Introduction to Julia
25mThe Unreasonable Performance Of The Julia Language
43mJulia: A Fresh Approach to Machine Learning
43mJulia for Data Analysis features, interfaces and future directions
01h 56mIntro to Julia
16mEmpowering Innovation with the Julia Language