强激光与粒子束, 2013, 25 (9): 2418, 网络出版: 2013-08-28
基于CUDA的多GPU加速SART迭代重建算法
Accelerating simultaneous algebraic reconstruction technique by multi CUDAenabled GPU
计算机断层成像 统一计算机架构 多GPU computer tomography computer unified device architecture SART simultaneous algebraic reconstruction technique multi GPU
摘要
为解决SART迭代重建算法计算耗时的问题,在单GPU基础上,利用多块GPU 的并行计算能力,提出了一种多GPU加速迭代重建算法。实验结果表明,与CPU重建相比,在不影响重建图像质量的情况下,采用GPU重建速度有明显提高,且增加GPU数量可以进一步提高重建速度。
Abstract
In order to solve the problem that the simultaneous algebraic reconstruction technique is very timeconsuming, this paper proposes a multiGPU parallel acceleration method on the basis of singleGPUaccelerated reconstruction. The experimental result shows that compared with the CPU reconstruction speed, the mutiGPU reconstruction speed improves significantly, while the precision of reconstructed images keeps the same, and the reconstruction speed can be further improved by increasing the number of GPUs. This method is of a higher value in engineering.
雷德川, 陈浩, 王远, 张成鑫, 陈云斌, 胡栋材. 基于CUDA的多GPU加速SART迭代重建算法[J]. 强激光与粒子束, 2013, 25(9): 2418. Lei Dechuan, Chen Hao, Wang Yuan, Zhang Chengxin, Chen Yunbin, Hu Dongcai. Accelerating simultaneous algebraic reconstruction technique by multi CUDAenabled GPU[J]. High Power Laser and Particle Beams, 2013, 25(9): 2418.