Python for High Performance and Scientific Computing
summary
German space-agency teams use Python to wrap and drive existing C/C++ simulation libraries, keeping validated solvers and adding NumPy/SciPy front ends, plotting and postprocessing on the cluster. Parallel work uses MPI4Py and PyCUDA/OpenCL for distributing embarrassingly parallel jobs and GPU kernels; getting code running across ~100 machines typically takes hours to days. A modular C++ CFD stack with mesh, solver and postprocessing layers exists and operates in production; about 100 Python developers contribute.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 2:59High performance computing overview
- 5:53Importance of HPC
- 8:50HPC system architectures
- 11:31Python in research
- 15:15Python and legacy HPC
- 20:06NumPy and SciPy in HPC
- 23:59Parallel computing with Python
- 25:53CUDA and GPU computing
- 27:45HPC challenges
- 31:01Profiling and performance tools
- 34:24HPC simulations case studies
- 37:44HPC conferences and community
- 41:38Questions and closing remarks
related talks
31mPython in High Energy Physics
45mPython for Speed, Scale, and Science
55mAccelerated Python: The Community and Ecosystem
44mPython's Bright Future in Science
35mPython at the Speed of Light: A Complete Tour of the CUDA Ecosystem
01h 06mBodo: Automatic HPC Performance and Scaling for Data Processing in Python
31m"Optimizing Scientific Python with C++, CUDA, and Serverless Compute"
35mCUDA in Python- A New Era for GPU Acceleration