Command line build

Accelerated Computing CUDA CUDA Programming and Performance
1

Actually I’m not familar with command line compilation :"> but now I need compile and execute my project using it. In nvcc description I found “-run” option but I can’t understand it’s whole syntax. I make such thing:

>nvcc -run simpleTemplates.cu -ccbin=d:/MVS2k5/VC/bin

and I get numerous errors on linkage stage such as

tmpxft_00000d7c_00000000-12.obj : error LNK2019: unresolved external symbol __imp__cutCheckCmdLineFlag@12 referenced in function _main

What should I add to my command for successful compilation?

2

If the command line options are the same as on linux:
nvcc -o my_executable.exe file1.cu file2.cu …

3

Thanks, but your variant generates same errors on the linkage step (buy the way I add -ccbin option because without it your variant doesn’t work). I suppose that I didn’t add some directories to the rule file or something like that.

4

Oh, I guess I should have noticed the problem in your first post. The undefined reference is in the cutil library, it is in the NVIDIA SDK directory somewhere. I’m probably getting the file name wrong since I’m going on memory, but you should be able to figure it out. I don’t think you need any command line to specify a library to link, but I’ve never tried this myself.

nvcc -o executable /path/to/cutil/cutil.lib file.cu

5

Actually I found solution and this is my command:

D:\NVIDIA CUDA SDK\projects\simpleTemplates>nvcc -o test.exe simpleTemplates.cu -l cutil32  -I"d:\MVS2k5\VC\include" -L "d:\MVS2k5\vc\lib" -L  "d:\NVIDIA CUDA S

DK\common\lib"