Leads: Brian Stucky (SCINet Computational Biologist), Heather Savoy (SCINet Computational Biologist)
This session will introduce key concepts of GPU-based computing, including how they differ from CPUs and what kinds of computing tasks can benefit from GPUs. There will also be a hands-on tutorial showing how to use the GPUs on the Atlas cluster and how to evaluate the effect of using GPUs on your computation time. A primary goal of this session is to help participants build intuition about when GPUs might be useful in scientific computing and how to use them.
Session objectives
The goals of this session are to:
- Understand the key differences between CPUs and GPUs.
- Build intuition about when GPUs can be helpful in scientific computing.
- Provide a practical introduction to using GPUs for scientific computing with Python.
Agenda
This session will begin with a short presentation followed by an interactive tutorial.
- Presentation: An introduction to GPU-based computing
- Python tutorial:
- GPU-based computing with CuPy
- Creating and working with GPU-based multidimensional arrays
- Accelerated computing with 1D CuPy arrays (vectors)
- Accelerated computing with 2D CuPy arrays (matrices)
- GPU-based computing with RAPIDS
- GPU-based dataframes with RAPIDS
- GPU-based machine learning with RAPIDS
- GPU-based computing with CuPy
Tutorial instructions
Coming soon!