Services on Demand
Journal
Article
Indicators
- Cited by SciELO
- Access statistics
Related links
- Cited by Google
- Similars in SciELO
- Similars in Google
Share
Ingeniería y Desarrollo
Print version ISSN 0122-3461On-line version ISSN 2145-9371
Abstract
CARRILLO, Snaider; SIEGEL, Jakob and LI, Xiaoming. Impact analysis of conditional and loop statements for the NVIDIA G80 architecture. Ing. Desarro. [online]. 2010, n.27, pp.130-150. ISSN 0122-3461.
In this paper, we propose two novel techniques to transform control statements so they can be executed efficiently on the NVIDIA G80 architecture. Our techniques called loop splitting and branch splitting smartly increase code redundancy, which might be deemed as "de-optimization" for CPU; but for a GPU framework these techniques improve the occupancy of a program on the GPU device and therefore improve its performance. We demonstrate our optimizations on an artificial benchmark and the results show that these techniques are very efficient and, depending on the problem layout, can lead to an increase in occupancy and a drastic improvement in performance compared to non-split version of the same algorithm.
Keywords : Branch splitting; instruction level optimization; loop splitting; NVIDIA G80 architecture.