Thrust2D: A new design abstraction framework for structuredrngrid class of algorithms

摘要

An important goal of structured parallel programming has been to provide a design frameworkrnthat balancesbetweentheextentof abstractionbuiltover thehardwareandtheamountof controlrngiven to the programmer to leverage the hardware resource features. Towards this goal, NVIDIArnhas released an open-source design framework called Thrust based on C++ STL,where the developersrncan express the functionality in STL style, without having to know the architectural detailsrnof the underlying parallel infrastructure.While the framework is generic and portable, it does notrnsupport the right abstraction for two-dimensional data, which is heavily used in most of the popularrnparallel algorithms. In this paper, we proposed Thrust2D, an extension of Thrust to supportrnthe abstraction for two-dimensional data, targeted towards structured grid class of applications.rnWe took several structured grid examples from Rodinia benchmark, OpenCV framework, andrnNVIDIA samples and rewrote them using Thrust2D.We demonstrated that, in some cases, we getrnnearly 80% reduction in code complexity, and for 12 out of 17 applications we have tested, thernkernel performance of Thrust2D versions arewellwithin85%of the nativeCUDAversions.Whenrnwe consider the total execution time, 14 out of 17 Thrust2D versions performance are withinrn85% of the native CUDA versions. In some cases, the performance of the Thrust2D versions hasrnoutperformed the native versions.