The Application of OpenCL to Precision Time and Frequency Algorithms

Andrew Dowd, William DeCook

Abstract:	The wide availability of GPUs in desktop and mobile computers is a potential resource for accelerating computationally intensive algorithms. The precision time and frequency industry uses many algorithms that would benefit from parallel processing. Like most industries, the wide acceptance of these techniques has been hindered by a lack of standards and vendor specific solutions. OpenCL has the potential to solve these shortcomings. This paper will examine how to apply OpenCL to several commonly used timing algorithms and look at the practical considerations of deploying OpenCL. OpenCL is a cross-platform, open source standard for implementing parallel programming algorithms. It represents the most successful effort to define a standard approach to heterogeneous computing. OpenCL has been implemented by most vendors of desktop GPUs and CPUs. It is also beginning to find support in the embedded community. While OpenCL is best known as a language for GPGPU computing (General Purpose Graphical Processing Unit), it can be used with standard multi-core CPUs. In fact, both processors can be used in parallel to solve the same problem, i.e. heterogeneous computing. This paper will describe OpenCL versions of common Sigma-Tau algorithms used in precision time applications. For large datasets, these algorithms can be very computationally intensive. We will show kernels that implement parallel versions of these algorithms and examine the optimization of memory. We will present benchmarks that compare the OpenCL implementations to conventional coding methods using optimized compilers.
Published in:	Proceedings of the 46th Annual Precise Time and Time Interval Systems and Applications Meeting December 1 - 4, 2014 Seaport Boston Hotel Boston, Massachusetts
Pages:	15 - 22
Cite this article:	Dowd, Andrew, DeCook,  William, "The Application of OpenCL to Precision Time and Frequency Algorithms," Proceedings of the 46th Annual Precise Time and Time Interval Systems and Applications Meeting, Boston, Massachusetts, December 2014, pp. 15-22.
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In