Seminarium fizyki atmosfery

PySDM: Pythonic particle-based cloud microphysics package

mgr. Piotr Bartman

Uniwersytet Jagielloński


30 października 2020 13:15, on-line via ZOOM

The package core is a Pythonic implementation of the Super-Droplet Method (SDM).  
The implementation introduces an extension to the SDM algorithm offering adaptive time stepping.  
PySDM architecture features separation of a number-crunching layer -- backend.
The developed backend implementations based on Numba (CPU computing, multi-threading) and ThrustRTC (GPU computing) leverage different Python acceleration techniques dubbed just-in-time and runtime compilation, respectively.

As a result, PySDM offers performance on par with compiled-language solutions with little-to-no trade-offs with respect to such advantages of the Python language as succinct and readable source code and portability (seamless interoperability between Windows, OSX and Linux).  

PySDM together with a set of bundled usage examples constitutes a tool for research on cloud microphysical processes, and for testing and development of novel modelling methods.  

The usage examples were developed embracing the Jupyter interactive platform allowing control of the simulations via a web browser.



PySDM: Pythonic particle-based cloud microphysics package - Slajdy do prezentacji

PySDM_IGF.pdf (11 MiB) Uploaded: 22 listopada 2020 21:51

Bieżące wydarzenia
Wydarzenia archiwalne