Abstract
Graphics processing unit (GPU) is a low-cost, low-power (watts per flop) and very high performance alternative to conventional microprocessors. In addition to their applications in graphics processing, researches are more and more attracted by their potential application in numerical computing due to their powerful ability of parallel computing and rapidly improved programmability. This paper makes a first try of applying GPU to accelerate computation for boundary face method (BFM). The element integration possessing high level of parallelism is a computationally intensive part of the BFM. As a primary step, we have implemented the parallelization of the regular integration and singular integration in CUDA programming environment. Comparative computations are made on both NVIDIA GTX 680 GPU and Intel(R) Core(TM) i7-3700 K CPU. Results show that, at the same level of accuracy, the speedup of regular integration and singular integration is up to 18.2 and 34.4, respectively.
Disclosure statement
No potential conflict of interest was reported by the authors.