首页> 外文会议>International conference on mathematics, computational methods reactor physics;MC 2009 >STUDY OF THE ACCELERATION OF NUCLIDE BURNUP CALCULATION USING GPU WITH CUDA
【24h】

STUDY OF THE ACCELERATION OF NUCLIDE BURNUP CALCULATION USING GPU WITH CUDA

机译:使用CUDA和GPU加速核素燃尽计算的研究

获取原文

摘要

The computation costs of neutronics calculation code become higher as physics models and methods are complicated. The degree of them in neutronics calculation tends to be limited due to available computing power. In order to open a door to the new world, use of GPU for general purpose computing, called GPGPU, has been studied [1]. GPU has multi-threads computing mechanism enabled with multi-processors which realize mush higher performance than CPUs. NVIDIA recently released the CUDA language for general purpose computation which is a C-like programming language. It is relatively easy to learn compared to the conventional ones used for GPGPU, such as OpenGL or CG. Therefore application of GPU to the numerical calculation became much easier. In this paper, we tried to accelerate nuclide burnup calculation, which is important to predict nuclides time dependence in the core, using GPU with CUDA. We chose the 4th-order Runge-Kutta method to solve the nuclide burnup equation. The nuclide burnup calculation and the 4th-order Runge-Kutta method were suitable to the first step of introduction CUDA into numerical calculation because these consist of simple operations of matrices and vectors of single precision where actual codes were written in the C++ language. Our experimental results showed that nuclide burnup calculations with GPU have possibility of speedup by factor of 100 compared to that with CPU.
机译:随着物理模型和方法的复杂化,中子学计算代码的计算成本也越来越高。由于可用的计算能力,它们在中子学计算中的程度往往受到限制。为了向新世界敞开大门,已经研究了将GPU用于通用计算(称为GPGPU)[1]。 GPU具有支持多处理器的多线程计算机制,可实现比CPU更高的性能。 NVIDIA最近发布了用于通用计算的CUDA语言,这是一种类似于C的编程语言。与用于GPGPU的常规工具(例如OpenGL或CG)相比,它相对容易学习。因此,GPU在数值计算中的应用变得更加容易。在本文中,我们尝试使用带有CUDA的GPU加速核素燃耗计算,这对于预测核素中核素的时间依赖性非常重要。我们选择了四阶Runge-Kutta方法来求解核素燃耗方程。核素燃耗计算和四阶Runge-Kutta方法适合将CUDA引入数值计算的第一步,因为它们包括矩阵的简单运算和单精度矢量,其中实际代码以C ++语言编写。我们的实验结果表明,与CPU相比,使用GPU进行核素燃耗计算的速度提高了100倍。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号