Python for High Performance Computing by MPCDF

Dec 3, 2025, 9:00 AM → Dec 4, 2025, 1:00 PM Europe/Berlin

Zoom

Sebastian Kehl (MPCDF), Sebastian Ohlmann (MPCDF), Klaus Reuter (MPCDF)

Python for High Performance Computing

Python is indispensable in modern scientific computing, but bridging the gap between a simple prototype and a high-performance production code can be challenging. This course teaches you how to maximize Python's efficiency in an HPC environment, moving beyond standard scripting to high-speed parallel computing.

Course Content

• HPC Fundamentals: An introduction to High Performance Computing.
• Performance & Compilation: Deep dives into NumPy, SciPy, Numba, and JAX, as well as interfacing with C and Fortran.
• Parallelism & Scaling: Profiling code, GPU acceleration, and parallel computing using `multiprocessing`, `concurrent.futures`, `mpi4py`, and Dask.
• Data & IO: Efficient Input/Output strategies using HDF5.
• Software Engineering: Best practices for packaging, publishing, testing, and generating documentation.

Format

• Timeline: Two consecutive days, Dec 3 - 4, 9 am – 1 pm (Online via Zoom).
• Materials: Instruction is based on Jupyter notebooks with reusable code and optional hands-on exercises.
• Support: Daily Q&A sessions and supplementary material on specific topics such as Cython and visualization are provided.

Wed, December 3

9:00 AM → 9:30 AM Introduction to HPC with Python

9:30 AM → 11:00 AM The Foundation of Scientific Computing with Python: NumPy, SciPy, and H5py

11:15 AM → 12:45 PM Software Engineering with Python: Packaging, Testing, Interfacing with C/C++ and Fortran, Profiling

12:45 PM → 1:00 PM Q&A

Thu, December 4

9:00 AM → 10:00 AM From CPU to GPU: High-Performance Python with Numba, JAX, and CuPy

10:00 AM → 11:00 AM Introduction to Parallel Computing; Shared Memory Parallelism with multiprocessing and concurrent.futures

11:15 AM → 12:45 PM Scaling Out: Distributed Computing with Dask and mpi4py

12:45 PM → 1:00 PM Q&A