- When: Monday, 16 November 2026, 8:30am - 5:00pm CST, full day
- Where W185d
- Conference info = SC26 event webpage
- Handouts slides = Julia for HPC SC26 Handouts
Welcome to the Julia for High-performance Computing (HPC) Tutorial at SC26 in Chicago, IL! This full-day, 6-hour hands-on tutorial introduces participants to using the Julia programming language effectively in HPC environments, from first principles to end-to-end deployment on a leadership-class system. Registered participants: please look at the end of this file for access to the Oak Ridge National Laboratory Odo system (32 nodes of Frontier, each node has 8 AMD MI250X GPUs).
- Those familiar with a programming language that want to learn more about Julia.
- Those without previous HPC familiarity. In fact, Julia is a great introductory language for HPC.
- Those interested in how Julia can help them go from idea to deploy at large scale on HPC systems - all in one language.
- Introduction to the Julia language and ecosystem for high-productivity
- Vendor-neutral parallel computing in Julia, intra-node: KernelAbstractions.jl, JACC.jl, inter-node: MPI.jl, Dagger.jl
- Running and profiling a Julia application, the Gray-Scott diffusion-reaction partial differential equation solver on a HPC Cluster (CPU and GPU).
Julia for HPC Tutorial
- Date Monday, 16 November 2026
- Time 8:30am - 5:00pm CST
- Location W185d
8:30 AM - 12:00 PM: Intro to the language, ecosystem, and CPU/GPU programming in Julia
8:30 AM - 10:00 AM - Intro to Julia
- Welcome and setup up - 15 minutes
- Intro to the Julia language and ecosystem - 75 minutes
- Packaging, package, type system w/ exercises.
- Julia features: inspect low-level IR
@code_llvm, macro system@time, performance pitfalls@inbounds
10:00 AM - 10:30 AM Conference Morning Break
10:30 - 12:00 PM - Parallel code with Julia - Part I
- Intro to kernel programming in Julia - 15 minutes
- CPU/GPU portable programming: KernelAbstractions.jl - 30 minutes
- CPU/GPU portable programming: JACC.jl, Julia for Accelerators - 45 minutes
12:00 PM - 1:30 PM Conference Lunch Break
1:30 PM - 5:00 PM Julia on HPC systems
1:30 AM - 3:00 PM - Parallel code with Julia - Part II
- Recap of morning contents - 15 minutes
- Auto-tuning with KernelTuner.jl - 30 minutes
- Out-of-core parallel execution with Dagger.jl - 45 minutes
3:00 PM - 3:30 PM Conference Morning Break
3:30 PM - 5:00 PM - Deploying a Julia Application on HPC systems
- Running Gray-Scott on Odo, Frontier's training system - 75 minutes
- Deploying Gray-Scott: Julia packaging and HPC module system interaction
- CPU/GPU kernels with JACC.jl + KernelAbstractions.jl (interoperabily) + MPI.jl + parallel I/O code with ADIOS2.jl
- HPC performance measurements, weak and strong scaling exercises
- Visualization of the simulation results on OLCF JupyterHub-open
- Profiling GPU kernels with rocprof
- Gray-Scott code organization and description
The "Julia for HPC" Tutorial series has been an international multi-institutional and multi-disciplinary collaboration through the years:
Massachusetts Institute of Technology (MIT), JuliaLab:
- Evelyne Ringoot
- Felipe de Alcântara Tomé
- Julian Samaroo
- Raye Kimmerer, now a PhD Student at Princeton University
- Alan Edelman
King Abdullah University of Science and Technology (KAUST)
- Rabab Alomairy, also with MIT
Lawrence Berkeley National Laboratory, National Energy Research Scientific Computing Center
Oak Ridge National Laboratory, Advanced Computing Systems Research Section
Participants are encourage to sign up for access to the Oak Ridge National Laboratory (ORNL) Leadership Computing Facility (OLCF) Odo Training system (32 nodes of Frontier, 8 AMD MI250X nodes).
- Please use a verified institutional email (no gmail or other similar service)
- Please make sure you registered for the SC26 Tutorials
- Deadline to register: November 6, 2026
- To register: https://my.olcf.ornl.gov/, follow instructions and use ProjectID: TRN047, and account type: Open. NOTE: "Funding source" means simply "who pays you?" (exact precision is not necessary as this is a training account)
- Questions about Odo access? Please reach out to William F Godoy, godoywf@ornl.gov
Once you have access to Odo please follow these steps to setup access to Julia and this repo:
cd /gpfs/wolf2/olcf/trn047/scratch/$USER
git clone https://github.com/JuliaParallel/julia-hpc-tutorial-sc26.git
cd julia-hpc-tutorial-sc26
source setenv-odo.sh
Each instructor will provide content-specific steps after that.