Partnerships and collaborations¶
GPU Education Center (formerly CUDA Teaching Center)¶
Do your science; let us handle your computation.
Note: Our GPU Education Center (formerly CUDA Teaching Center) status ended on 31st of December 2016. We proudly stuck with the program until the very end. Thanks to NVIDIA for all the help and support! We will continue to teach Parallel Programming on Heterogeneous Systems, use CUDA and OpenCL (an open standard for parallel programming on heterogeneous systems consisting of CPUs, GPUs, FPGAs, and potentially other types of chips) in our research, and promote GPU computing in general. The following content is now considered to be legacy.
Visual computing landscape is evolving at an impressive pace.¶
Computing platforms are increasingly heterogeneous, with computing power dived across CPUs and GPUs. At the same time, computer programming and scientific computing courses and workshops are essentially taught the same way as two decades ago. Certainly, there is a room for improvement.
In year 2012 CUDA Teaching Center was established at University of Rijeka under a partnership with NVIDIA. Our goal here is to provide a platform for education in domain of GPU computing using CUDA. In 2015 the Center has been renamed to GPU Education Center.
Demand for compute power in many branches of science and engineering is ever-increasing.¶
Computational Physics, Computational Chemistry, Applied Mathematics, Bioinformatics, Medical Imaging, Fluid Simulation, Artificial Inteligence, Data Science. We could go on; each and every one of these and many others wouldn’t mind just a little more compute power. Our goal is to bring together scientists, engineers and high performance computing enthusiasts to work together on improving our understaning of nature and developing next-generation techologies.
GPU computing is going mainstream and finding applications in industry.¶
In 2007, just before CUDA appeared, GPU computing was mostly a playground. When it worked, it indeed offered nice speedups. However, programming the GPU was hard, and usually accessible only to niche applications that fit GPU programming model. These days, with many readily available libraries, GPU computing is usable with effort ranging from change includes and recompile to restructure entire application for more parallelization. And there are speedups to be gained at any step.
It is the role of the university to talk about advancements made possible by new technologies.¶
The world is a large and heterogenous place. With more people having access to the internet and quality education, new technologies appear more rapidly. It is not simple for any organization to adopt all the latest technological advances and provide core service in a stable and predictable way. On the other hand, computational scientist’s everyday work is to research, develop and share information about computation, to talk about technologies that work well in particular areas.
We teach Parallel Programming on Heterogeneous Systems offered at Department of Informatics and GPU Education Center Workshop organized on a yearly basis.
Parallel Programming on Heterogeneous Systems is an elective course offered in the third year undergraduate single major informatics, consting both of lectures and hands-on laboratory exercises.
GPU Education Center Workshop consists of hands-on laboratory exercises. It is organized on a yearly basis, and open to all university staff. Students can also apply; please check below for more details. Next workshop will take place in University of Rijeka Departments Building on XXth of YY 20ZZ (details to be added).
Research in GPU computing¶
Scientists and engineers with domain expertise can focus on solving problems from their domain, as we take care of computation. Which also happens to be what scientists and engineers prefer to do.
- E-Learning Activities Recommender System (ELARS) is an educational recommender system that predicts student’s preference of e-learning activities. Optimization of preference prediction is done using CUDA. Learn more.
- Prototype WDM Network Simulator (PWNS) is a network simulator based on ns-3 that provides models for optical network components. Optimization of routing and wavelength assignment is done using CUDA. To learn more, see Photonic WDM Network Simulator (PWNS).
- DNA methylation is a fundamental mechanism in functional organization of the human genome. Using CUDA-enabled NVIDIA GPU with open-source computational chemistry software, we have developed in-silico a set of potential inhibitors of human DNA methyltransferase Dnmt1. Learn more.
- Your project! Have a research project idea or an ongoing project? Think it could benefit from GPU computing? We would like to hear about it; get in touch with us.
GPU computing and CUDA has expanded its reach outside of scientific laboratories and made way into the industry applications. And this is just the beginning.
Are you considering investments in heterogeneous systems with CPUs and GPUs, but aren’t sure if they will perform well on your application? Or perhaps comparing CPU-only systems against GPUs in terms of performance, cost and power efficiency? We can help you decide.
We enjoy talking about computation to a wider audience. Organizing an event aimed at a wider audience related to computation? We might be interested in attending.
We are an interdisciplinary team consisting of professors, research and teaching assistants, postdocs and PhD students from University of Rijeka working in various domains of science and engineering.