http://www.cnblogs.com/dflower/archive/2009/10/16/1584827.html
首先介绍一下开发环境,Visual Studio 2008 + CUDA Wizard for Visual Studio. 确保显卡支持CUDA(GeForce 8系列之后,否则只能用模拟模式)并安装CUDA DDK及CUDA Toolkit。
安装完CUDA Wizard for Visual Studio之后,如果编译运行时出现"无法找到cutil32D.dll"的错误,则需要将安装的CUDA的sdk的路径加到系统环境变量中:
例如C:\Program Files\NVIDIA Corporation\NVIDIA CUDA SDK\bin\win32\
下的
├─Debug
├─EmuDebug
├─EmuRelease
└─Release
几个目录都加入到系统环境变量Path中,这样才能在运行程序的时候找到相应的dll库。
然后就可以在VS下直接编译CUDA程序了(*.cu)。以下是一个CUDA初始化的程序。
#include <stdio.h> #include <cuda_runtime.h> bool InitCUDA() { int count; cudaGetDeviceCount(&count); if(count==0) { fprintf(stderr,"There is no device.\n"); return false; } int i; for(i=0;i<count;i++) { cudaDeviceProp prop; if(cudaGetDeviceProperties(&prop,i) == cudaSuccess) { if(prop.major>=1) { //枚举详细信息 printf("Identify: %s\n",prop.name); printf("Host Memory: %d\n",prop.canMapHostMemory); printf("Clock Rate: %d khz\n",prop.clockRate); printf("Compute Mode: %d\n",prop.computeMode); printf("Device Overlap: %d\n",prop.deviceOverlap); printf("Integrated: %d\n",prop.integrated); printf("Kernel Exec Timeout Enabled: %d\n",prop.kernelExecTimeoutEnabled); printf("Max Grid Size: %d * %d * %d\n",prop.maxGridSize[0],prop.maxGridSize[1],prop.maxGridSize[2]); printf("Max Threads Dim: %d * %d * %d\n",prop.maxThreadsDim[0],prop.maxThreadsDim[1],prop.maxThreadsDim[2]); printf("Max Threads per Block: %d\n",prop.maxThreadsPerBlock); printf("Maximum Pitch: %d bytes\n",prop.memPitch); printf("Minor Compute Capability: %d\n",prop.minor); printf("Number of Multiprocessors: %d\n",prop.multiProcessorCount); printf("32bit Registers Availble per Block: %d\n",prop.regsPerBlock); printf("Shared Memory Available per Block: %d bytes\n",prop.sharedMemPerBlock); printf("Alignment Requirement for Textures: %d\n",prop.textureAlignment); printf("Constant Memory Available: %d bytes\n",prop.totalConstMem); printf("Global Memory Available: %d bytes\n",prop.totalGlobalMem); printf("Warp Size: %d threads\n",prop.warpSize); break; } } } if(i==count) { fprintf(stderr,"There is no device supporting CUDA.\n"); return false; } cudaSetDevice(i); return true; } void main() { if(!InitCUDA()) { getchar(); return; } printf("CUDA initialized.\n"); getchar(); }
运行结果如下:
在一般情况下,某事物个体发生具有其特有属性的负面现象,且无法以科学的角度得到合理有效的解释。我们通常称此类现象为“人品问题”(RPWT)。
——摘自《辞海》第314页