In order to inform and ensure a good collaboration between dScience Digital Resources and the research environment at UiO, we will invite you to regular meetings.
Topic
Scientific software is often characterized by a need to process large amounts of data in high precision. This processing can range from meteorological simulations over large swaths of the earth to simulating the tiniest molecules. One characteristic of these computations is the need for high throughput. One way of achieving this throughput is to utilize GPUs which are able to process large amounts of data in parallel.
In this lecture, we will present how to integrate GPU programming into new or existing scientific software. We will contrast how CPUs and GPUs differ and how this affects the programming model. By going through a few select examples, we will highlight the main differences and where one needs to pay special attention when using a GPU. We will also point out a few different technologies that can be used for GPU programming and how these affect the portability of your software.
Lecturer
Jørgen Nordmoen is a senior engineer at USIT and is the team lead for NRIS’ GPU team. His main focus is helping researchers get started with and optimize their use of GPUs on HPC hardware. He has a deep understanding of GPU programming and is well versed in technologies such as CUDA, HIP, SYCL, OpenMP offloading and OpenACC. He has a PhD in Evolutionary Robotics from the University of Oslo and has published on both software simulation and real world robotics experiments.
Prerequisite knowledge
Basic knowledge of C/C++ programming/syntax or related languages and basic understanding of OpenMP shared memory parallelization is necessary to attend this webinar.