cuda-by-example

进行相关运算时必须要是用的一些步骤,可以解决相关算法的一些问题(Related operations must be used in a number of steps can be solved a number of issues related to algorithm)

CUDA代码，主要是要实现一个具体ADD操作，并且可以高速在GPU上运行(An Algorithm based on UKF)

用此程序可以方便的查讯日期和星期几，还可以查当天和未来几天的．(This procedure can be used to facilitate inquiry date and day of the week, you can check the same day and the next few days.)

采用改进的三维Hasin准则，考虑基体和纤维损伤，以及复合材料分层损伤( U91C7 u7539 u659 u8F09 u7F4 u7R4 u7354 u03 U6599 u5206 u5C42 u635F u4F24)

格子Boltzmann方法 格子Boltzmann方法是为了保留格子气自动机方法的优点，克服其缺点而发展起来的方法。 特别是1992年，钱跃弘、陈十一等的开创性工作(提出LBGK模型方法)，使该方法广泛地应用到计算流体力学(单相流、多相流、多孔介质流、热对流、磁流体、反应-扩散等)。 (Lattice Boltzmann method of lattice Boltzmann method is to retain the lattice gas automata method has the advantage to overcome its shortcomings and has developed methods. Especially in 1992, Qian Yue-hong, Chen XI, such as the pioneering work (LBGK model proposed method), so that the method widely used in computational fluid dynamics (single-phase flow, multiphase flow, porous media flow, thermal convection, magnetic fluid, the reaction- the proliferation, etc.).)

Hight Performance Computing Demos

使用CUDA并行语言编写的矩阵乘法，与串行语句的对比试验，主要检测出CPU与GPU运行的时间，通过比较时间，达到比较效果。另外此程序可以随意更改矩阵的维数，增强程序的移植性。(CUDA parallel language use matrix multiplication, and serial statement comparison test, the main test of CPU and GPU to run time, by comparing the time, to compare results. Also this program can change the dimension of the matrix, and enhanced portability of the program.)

CUDA平行的優化，以矩陣乘法為基礎。透過ShareMemory做加速(CUDA ShareMem)

LU算法的MPI串行实现，矩阵维数无要求，可在曙光4000A并行机运行(The LU algorithm MPI serial implementation matrix dimension requirements Dawning 4000A parallel machine running)

多核编程，比 OPENCL 简单 可以通过类似 OPENMP 的方式，简洁的实现异构编程，使用 GPU 计算(Multicore programming than simple OPENCL)