Please use this identifier to cite or link to this item: http://hdl.handle.net/1822/43704

TitleExploring heterogeneous computing with advanced path tracing algorithms
Author(s)Oliveira, André
Perdigão, César Morais
Santos, Luís Paulo
Proença, Alberto José
KeywordsGlobal illumination
Path space based integration
Heterogeneous parallel computing
Issue date2016
PublisherIEEE
Abstract(s)The CG research community has a renewed interest on rendering algorithms based on path space integration, mainly due to new approaches to discover, generate and exploit relevant light paths while keeping the numerical integrator unbiased or, at the very least, consistent. Simultaneously, the current trend towards massive parallelism and heterogeneous environments, based on a mix of conventional computing units with accelerators, is playing a major role both in HPC and embedded platforms. To efficiently use the available resources in these and future systems, algorithms and software packages are being revisited and reevaluated to assess their adequateness to these environments. This paper assesses the performance and scalability of three different path based algorithms running on homogeneous servers (dual multicore Xeons) and heterogeneous systems (those multicore plus manycore Xeon and NVidia Kepler GPU devices). These algorithms include path tracing (PT), its bidirectional counterpart (BPT) and the more recent Vertex Connect and Merge (VCM). Experimental results with two conventional scenes (one mainly diffuse, the other exhibiting specular-diffuse-specular paths) show that all algorithms scale well across the different platforms, the actual scalability depending on whether shared data structures are accessed or not (PT vs. BPT vs. VCM).
TypeconferencePaper
URIhttp://hdl.handle.net/1822/43704
ISBN978-1-5090-5387
DOI10.1109/EPCGI.2016.7851185
Peer-Reviewedyes
AccessopenAccess
Appears in Collections:CAlg - Artigos em livros de atas/Papers in proceedings
DI/CCTC - Artigos (papers)

Files in This Item:
File Description SizeFormat 
EPCGI2016.pdf2,76 MBAdobe PDFView/Open

Partilhe no FacebookPartilhe no TwitterPartilhe no DeliciousPartilhe no LinkedInPartilhe no DiggAdicionar ao Google BookmarksPartilhe no MySpacePartilhe no Orkut
Exporte no formato BibTex mendeley Exporte no formato Endnote Adicione ao seu Currículo DeGóis