MSA Seminar: GPU libraries - Get More for Less

Modular Supercomputing Architectures Seminar

Speaker: Jan Meinke, JSC

GPU libraries come in many shapes and sizes. They make GPU programming easier, more portable, and less error prone. In this talk, I start with very general libraries such as CUB and Thrust, present some of the numerical libraries like cuBLAS and cuRand, and give examples of domain specific libraries such as OpenMM and QUDA. Short code snippets illustrate how to use GPU libraries from C++, Fortran, and Python.

This seminar is also offered via DFNconf:

Meeting Room Name: MSA-Seminar
Meeting Room Number: 97983314