OpenCL Efficiency

Bohrium can be configured to produce OpenCL kernels that target an arbitrary number of threads. The project goal is to benchmark a set of applications on a set of hardware platforms with a varying number of threads. The outcome is intended for a journal publication that identifies the strengths and weaknesses of OpenCL as a portable programming model for high-performance computing. The successful student will be coauthor to such a paper.

Area: Project

Tags: Bohrium:hardware:benchmarking

Contact: Mads Kristensen madsbk@nbi.ku.dk or Brian Vinter vinter@nbi.ku.dk

Activities: Experiment design and execution, result analysis, extracting conclusions.