The OpenMP API specification for parallel programming
Webinar: Large-Scale Materials Science Codes Porting Strategies on GPU Architectures, the BerkeleyGW Case Study
BerkeleyGW is a massively parallel software package employed to study the excited state properties of electrons in materials by using GW, Bethe-Salpeter Equation (BSE) methods, and more. The code effectively utilizes strong-scaling GPU architectures and can scale out to tens of thousands of nodes. [More about BerkeleyGW]
This webinar is presented by Mauro Del Ben, Research Scientist, Applied Mathematics and Computational Research Division, LBNL In this webinar we will discuss our experiences porting BerkeleyGW to three different GPU programming models (CUDA, OpenACC, and OpenMP Target) with various GPU vendor architectures. We’ll cover some of the challenges we encountered along the way that impede true performance portability. Special attention will be paid to code modernization practices that we found useful in the porting pipeline.
Who should attend?
This webinar will be of particular interest to those who are looking for the best strategies to port scientific software to GPU acceleration. Particular attention will be made between the tradeoff between implementation effort and final performance.
Date: This webinar was on Wednesday, April. 20, 2022