What is it about?
We show how a non-trivial image codec is accelerated with GPUs and multi-core CPUs using OpenCL. Among other things, the paper discusses the challenges encountered during this process and the choices made to obtain the best possible performance. A thorough performance analysis is also included.
Featured Image
Why is it important?
The paper discusses the fits heterogeneous implementation of a JPEG 2000 JP3D codec. It can serve as an example of how a complex program can be accelerated on a heterogeneous platform using available processing power of CPUs and GPUs and OpenCL. The description of this process together with the performance analysis may help others to achieve a similar task. Finally, the paper may also serve as inspiration for those that try to find better ways to achieve a similar task (e.g. by developing higher level language constructs, ...).
Perspectives
This paper will hopefully contribute to the knowledge of efficient implementation of HPC applications. The growing computational demands and the energy problem make this a very important topic in the foreseeable future. By sharing our experiences we can hopefully learn how to write better - more efficient - programs.
Mr. Jan G Cornelis
Vrije Universiteit Brussel
Read the Original
This page is a summary of: Heterogeneous acceleration of volumetric JPEG 2000 using OpenCL, The International Journal of High Performance Computing Applications, May 2016, SAGE Publications,
DOI: 10.1177/1094342016646438.
You can read the full text:
Contributors
The following have contributed to this page
