Modeling and Simulation Group Meeting

Some software for scientific programming

Speaker: Florian Wechsung, Ondrej Maxian, Huy Dinh, and Raul Perez

Location: Warren Weaver Hall 1314

Date: Thursday, February 25, 2021, 12:30 p.m.

Synopsis:

Florian Wechsung: pybind11 and xtensor

Ondrej Maxian: PyBind11 + CUDA

Huy Dinh: R & Shiny

Raul Perez: AWK + piping example for real-time visualization

Florian's example codes for pybind11+eigen+xtensor are at:

http://github.com/florianwechsung/pybind-demo

Ondrej's example used pybind11 together with thrust to interface to the GPU. Thrust is a powerful C++ library of templates (but not for the faint of heart):

https://thrust.github.io/

Shiny tutorials can be found at:

https://shiny.rstudio.com/gallery/

Some other things we mentioned are numba as a way to get numpy code to be compiled and parallelized using OpenMP and made more efficient auto-magically (works like magic when it does):

We also brought up

https://github.com/google/jax

Often used for machine learning / fast python on CPUs and especially GPUs. It’s a Google project that’s winning some of the speed tests against tensor flow, etc. It’s also really high-level and easy to write.