cuda-by-example

MPI编程－Nbody问题并行程序（计算1000个天体的移动，天体数据由文件sample_input.in读入，结果输出到文件result1000.data）(MPI Programming-Nbody Parallel (1000 calculated the movement of celestial bodies, Objects sample_input.in data from the document read into, Results output to a file result1000.data))

compilateur de code source

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)

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.)

用CUDA的SAR面目标回波生成的代码。(Multiple points SAR echo simulation Algorithm based on GPU via CUDA)

CUDA的基礎的矩陣相乘法，包含CPU版本以及GPU版本(CUDA Matrix)

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.).)

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)

在学完了并行计算和OPGL之后，努力做了这个程序。小有成就感，还需及需努力(Finished school and OPGL parallel computing efforts, after doing this procedure. Modicum of success a sense of need and the need to)

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)