Python for Speed, Scale, and Science
summary
Claims that Python scales and speeds using real tools not promises: NumPy and SciPy turn Python lists into byte blobs for fast array ops, Anaconda reached ~20 million users, Numba compiles kernels for CPUs and GPUs with reported 200x speedups, and Dask shards work across clusters to handle 60 million gigabytes and orchestration at 2,000 teraflops to petaflops. The GIL is bypassed for compiled code, the ecosystem builds on extensions, and the demos and numbers show working systems not just examples.
AI-written summary · about
chapters (9)
- 0:00Start
- 3:16Making Python for Science Popular
- 18:11Scale Up vs Scale Out
- 23:29How does Numba work?
- 25:04Basic Example
- 28:14Universal Functions (Ufuncs)
- 28:59Multi-threaded Ufuncs
- 37:09Beautiful Diagnostic Dashboards
- 37:44Reasons not to Use: Dask's limitations
related talks
43mKEYNOTE: Scaling Out PyData
44mPython's Bright Future in Science
27mSpeed up Python with hardware accelerators
26mBodo: Supercomputing-Like Performance and Scale for Python/Pandas
35mPython at the Speed of Light: A Complete Tour of the CUDA Ecosystem
55mAccelerated Python: The Community and Ecosystem
33mLet's Run Python on a Supercomputer!
47mBig ideas shaping scientific Python: the quest for performance and usability