A Scalable and Energy Efficient GPU Thread Map for m-Simplex Domains

This work proposes a new GPU thread map for -simplex domains, that scales its speedup with dimension and is energy efficient compared to other state of the art approaches. The main contributions of this work are i) the formulation of the new block-space map for regular orthogonal simplex domains, which is analyzed in terms of resource usage, and ii) the experimental evaluation in terms of speedup over a bounding box approach and energy efficiency as elements per second per Watt. Results from the analysis show that has a potential speedup of up to and for and -simplices, respectively. Experimental evaluation shows that is competitive for -simplices, reaching of speedup for different tests, which is on par with the fastest state of the art approaches. For -simplices reaches up to of speedup making it the fastest of all. The extension of to higher dimensional -simplices is feasible and has a potential speedup that scales as given a proper selection of parameters which are the scaling and replication factors, respectively. In terms of energy consumption, although is among the highest in power consumption, it compensates by its short duration, making it one of the most energy efficient approaches. Lastly, further improvements with Tensor and Ray Tracing Cores are analyzed, giving insights to leverage each one of them. The results obtained in this work show that is a scalable and energy efficient map that can contribute to the efficiency of GPU applications when they need to process -simplex domains, such as Cellular Automata or PDE simulations.
View on arXiv