Python code acceleration with Numba [online, PYT301]

Python is easy to use, but sometimes too slow for your intensive computations? Numba compiler automatically translates your Python code to make it faster, whether on CPU or GPU. This workshop teaches you the essential techniques to optimize your most time-consuming functions, all while keeping the simplicity of Python.

Prerequisites

• Know the basics of submitting jobs to a compute cluster or have completed the workshop “First steps on supercomputers (CIP101)”
• Know how to identify and correct inefficient jobs or have completed the workshop “Benefits of using computing resources effectively (CIP102)“
• Have a general understanding of GPUs and basic experience in running GPU tasks on compute clusters
• Have a good knowledge of Python programming or have completed the workshops “Introduction to programming with Python (PYT101)” and “Enhancing your Python programming skills (PYT102)”

Agenda

• Why use GPUs for computing?
• Understanding difference between CPU and GPU
• Numba: just-in-time compiling
• Parallel computing: distributing workload among CPU and GPU cores
• Very short intro to CUDA
• Numba + CUDA: porting a CPU code to GPU
• Numba + CUDA: more acceleration with shared memory, streams
• Practical Sessions: lots of examples and exercises

Instructor

Nikolaï Sergueev, analyst in advanced research computing at Calcul Québec.

Language

English

Registration

Regular rate: $250

Academic or Nonprofit rate: Free registration for anyone who studies, teaches, or works at a university, CEGEP, CCTT, or research institute affiliated with a Canadian university, or anyone working for a nonprofit organization in Canada

⚠ Even if you qualify for free registration, signing up still has a cost and uses up resources! Please register only once for the same event, and unregister in advance if something comes up.

Technical prerequisites

We will use the Zoom platform. Because this event is a practical workshop, it is very useful having a secondary screen where you would get the instructor window on one screen and your own window on your main screen.

We will use the Jupyter Lab interface. Make sure you have a modern Web browser like Google Chrome, Firefox, Edge or Safari.

Notes

• A certificate of participation will be send to each participant who attends at least 60% of the workshop.
• The workshop is not recorded.
• The workshop could be canceled if the number of registrations is too low.

Contact

For any question, please write us to training@calculquebec.ca

To stay up to date on our upcoming events, subscribe to our Eventbrite page!