Use OpenCL with R for HPC & OpenGL/Vulkan for interactive visualisation
This time around, we will have distinguished speaker Dr Josh Bowden from CSIRO, talking to us about developing OpenCL applications for High Performance Computing, as a package in R, and Daniel Filonik who will talk about his DataChopin novel interface for collaborative visualisation.
We will run this meet up at the ARC Centre of Excellence for Mathematical and Statistical Frontiers, QUT.
(We had to shift meeting to September due to personal request from the speakers)
Package development in R with OpenCL for high performance computing
The statistical programming language “R” has a firm place in the toolkit of scientific data analysis due to its accessibility, extensibility, plethora of powerful statistical tools and high quality graphics outputs suitable for publication. Its popularity has meant that high performance computing facilities have had to ensure it is part of available software on their systems. This has also meant that facilities have had to try to address R’s downsides, of which its design as a single threaded application is often an impediment. There have been many ways that this performance bottleneck has been addressed, and this talk goes into the use of OpenCL as another avenue for deriving performance from the R programming environment. I’ll present a couple of case studies that have benefited from use of the OpenCL and look at R package structure that allows OpenCL based code to be integrated with R.
Josh Bowden received a PhD in Materials Science from The University of South Australia in 2007 and has worked in a number of materials science based projects including cartilage 3D biochemistry and structure and high throughput wood phenomics data analysis. Many of his projects have required statistical programming and visualisation and he started developing code using OpenGL in aide of this during his PhD. An interest in the power of GPUs and the burgeoning need for higher performance computing led to his interest in OpenCL and he has been working at CSIRO for the past 7 years in the eResearch HPC area applying these skills to various projects.
DataChopin is a novel interface for co-located, collaborative visualisation composition. It uses large-scale display walls as a shared desktop, allowing multiple co-located users to work together simultaneously, and supports a flexible compositional model for incremental and piece-wise construction of visualisations. Composability was chosen as a guiding principle in the design, since it is essential to both, open-ended exploration as well as collaborative analysis. If data and visualisations are composable, they can split into independent parts and recombined during the analytical process, allowing analysts to seamlessly transition between closely- and loosely-coupled work. The research is developing a conceptual framework and software architecture for implementing such highly dynamic, composable visualisations, along with suitable interaction methods for manipulating these graphical representations on the fly.
Daniel Filonik (Urban Informatics & Institute for Future Environments, Queensland University of Technology, Brisbane, Australia) organizes the Brisbane GPU User Meetup and is a PhD candidate researching in area of Human-Computer Interaction, with a focus on Interaction Environments for Collaborative Data Composition and Visualization.