Computational & Technology Resources
an online resource for computational,
engineering & technology publications
Computational Science, Engineering & Technology Series
PATTERNS FOR PARALLEL PROGRAMMING ON GPUS
Edited by: F. Magoulès
Programming GPUs from High Level Data Flow Models
M. Barreteau, R. Barrère and E. Lenormand
Thales Research & Technology, Palaiseau, France
M. Barreteau, R. Barrère, E. Lenormand, "Programming GPUs from High Level Data Flow Models", in F. Magoulès, (Editor), "Patterns for Parallel Programming on GPUs", Saxe-Coburg Publications, Stirlingshire, UK, Chapter 4, pp 73-109, 2014. doi:10.4203/csets.34.4
Keywords: GPU, OPENCL , high level programming, design tool, data flow applications, efficient code generation.
This chapter shows how data flow applications can be programmed efficiently on GPUs from a unique high level capture. We will rely on a tooling approach (through the SPEAR Design Environment) to point out the underlying productivity gain with respect to performance. For efficient code generation purpose, several optimisations at different levels will be detailed. A representative Radar application will illustrate our results.
purchase the full-text of this chapter (price £20)