Advanced Scientific Computing using Julia
Información General
Julia es un lenguaje de programación moderno, diseñado para generar código rápido, extensible y reproducible, lo que lo convierte en una herramienta ideal para la computación científica. Sin embargo, aprender a utilizarlo correctamente por cuenta propia puede ser un desafío. En este taller de dos días, pusimos manos a la obra: comenzamos con un ejemplo práctico, creamos un primer paquete en Julia y optimizamos el código, descubriendo cómo lograr mejoras de rendimiento de hasta 1000 veces utilizando las herramientas del lenguaje.
Este evento fue organizado por LIBRE hub y se realizó de manera presencial en la Pontificia Universidad Católica de Chile, Campus San Joaquín, el viernes 10 y lunes 13 de enero de 2025.
Visita nuestra galería (Flickr) para ver las fotos del taller. NOTA: Si compartes alguna de las fotos en redes sociales, por favor, usa #LIBREhub en la descripción o etiquétanos en tus publicaciones.
Conocimientos previos que tuvieron los participantes:
- Experiencia con otros lenguajes como Python o MATLAB.
Materiales necesarios:
- Computador personal
- Julia v1.10 instalado (usando juliaup)
- Cuenta en GitHub
- VSCode instalado con extensión de Julia
- Lectura previa obligatoria: https://modernjuliaworkflows.org/writing/
Temario
- Part 1: Julia Basics
- Broadcasting (
f.(x)) - Structs: Concrete, parametric, and abstract types (About.jl)
- Multiple dispatch (functions vs. methods,
@which f(x),methodswith(MyType),methods(f))- Type stability (
similar,eltype,ones,zeros,randn)
- Type stability (
- Broadcasting (
- Part 2: Julia Packages
- Using packages (
usingvs.import) - Environments (
Project.toml) - Creating a basic package (
generate MyPkg,module MyPkg,publicandexported methods) - Extending functions from a different package (
MyPkg.fun(x) = ...)- How not to extend functions: Type piracy
- Package extensions
- Using packages (
- Part 3: Making Your Code Fast
- Basic benchmarks:
@timeand@profview - Run time vs. compile time
- Letting the compiler know how much memory you need
- Stack vs. heap: memory allocations
- In-place mutation (
f(x)vs.f!(x)) - Views (
@viewand@views)
- Letting the compiler optimize with type-stable code (Cthulhu.jl)
- Basic benchmarks:
- Part 4: Advanced Applications Showcase
- Automatic differentiation (DifferentiationInterface.jl and Enzyme.jl)
- Forward vs. reverse mode
- GPU programming
- Array-based programming (CUDA.jl, AMDGPU.jl, Metal.jl, and oneAPI.jl)
- Kernel-based programming (KernelAbstractions.jl)
- Deep learning (Flux.jl and Lux.jl)
- Automatic differentiation (DifferentiationInterface.jl and Enzyme.jl)
Agenda
Place: Sala de clases AP504, Edificio de Aulas Las Pataguas (Piso 5), Campus San Joaquín - Google Maps
| Start | End | Day 1 | Day 2 |
|---|---|---|---|
| 9:00 | 9:30 | 🟢 Welcome and Introduction to Julia | 🟡 Measuring performance |
| 9:30 | 10:00 | 🟢 Fundamentals | 🟡 Compiler |
| 10:00 | 10:30 | 🟢 Types and multiple dispatch | 🟡 Optimizing performance |
| 10:30 | 11:00 | 💻 Activity 1 - Babylonian square root | 💻 Activity 3 - Optimizing code |
| 11:00 | 11:30 | Break | Break |
| 11:30 | 12:00 | 🔵 Packages | 🔴 Parallel computing |
| 12:00 | 12:30 | 🔵 Environments | 🔴 Automatic differentiation |
| 12:30 | 13:00 | 🔵 Composability | 🔴 Deep Learning |
| 13:00 | 14:00 | Lunch | Lunch |
| 14:00 | 17:00 | 💻 Activity 2 - Free Induction Decay (FID) | 💻 Activity 4 - Optimizing FID code |
| 17:00 | 17:30 | Closing remarks and awards |
Legend: 🟢 Julia basics 🔵 Julia packages 🟡 Making your code fast 🔴 Advanced applications
Documentation
You will find the slides from Day 1 and Day 2 below. For more information, visit the GitHub repository of Julia Workshop 2025.
Code of Conduct
LIBRE hub, a Pontificia Universidad Católica de Chile initiative promoting Open Hardware, adheres to the UC Code of Conduct, the CZI Code of Conduct, and the GOSH Code of Conduct, which promote open access, collaboration, and transparency in science. By participating, you agree to abide by these standards.
Subscribe to our LIBRE hub mailing list to receive more related content. All our activities are recorded and published on our Youtube channel where they can be accessed with automatic subtitles in Spanish and Portuguese.