PBA_WORLD

1. About page-locked host memory / pinned memory:(1) Restrict their use to memory that will be used as a source/destination in calls to cudaMemcpy() and freeing them when they are no longer needed

Assess, Parallelize, Optimize, DeployStep 1: Profiling the code in order to identify the hot spots.Strong scaling (Amdahl's Law) is a measure of how, for a given problem size, perfor

1. 循环的并行化：（1）循环固定（2）内循环依赖于外循环without dynamic parallelism

1.打开visual studio 2010, 依次选择：工具->选项->项目和解决方案->VC++项目设置，在要包括的扩展名中添加.cu2.右键点击项目名，打开项目属性页，将配置选为“所有配置”（对debug&release都有效），然后选择：配置属性->VC++目录，在包含目录中添加$(CUDA_INC_PATH)3.在用户自定义列表中添加CUDA关键词：如果用的win 7

摘自 "CUDA C Best Practices"1. To maximize developer productivity, profile the application to determine hotspots and bottlenecks2. To get the maximum benefit from CUDA, focus first on finding ways t

1. 工程->属性->CUDA C/C++ -> Common -> Generate Relocatable Device Code 设置为是 (-rdc=true)2. 工程->属性->CUDA C/C++ -> Device -> Code Generation设置为compute_35,sm_353.  工程->属性->链接器->输入->附加依赖项->cudadevrt.lib

Tips:1. CUDA memory for lookup tables:It may be best not to use any tables on the GPU at all (see also CUDA math library), as FLOPS are increasing faster than memory bandwidth across GPU generat

利用CUDA 5.0最新推出的 Dynamic Parallelism，以往很难使用的分治法现在可以轻易的在GK110上利用这一新特性实现，非常方便：算法思想：随机选取一个枢纽元（pivot），对排序数组进行划分，左边一组都比枢纽元小，右边一组都等于或大于枢纽元，然后对每一个分组递归使用快排算法直至每一个分组仅有一个元素，则排序完成。示例图:CUDA version wit