I am using nsight to run example of NVIDIA SDK but getting some library referencing error.
I have installed cuda-5.0 and using nsight which comes with cuda-5.0
Some of the error:
Function ‘_ConvertSMVer2Cores’ could not be resolved
Function ‘checkCmdLineFlag’ could not be resolved
Please tell me about any configuration to fix these errors.
I have added some include paths and library paths in properties then most of the errors resolved but now the following error still there. please tell me what configuration need to be done to resolve this.
Error: macro “__cudaRegistryBinary” passed 1 arguments, but takes just 0.
Note: i have installed both 4.2 and 5.0 toolkit in my system. Is it the issue?
This might be an issue when you mix and match CUDA sdk samples 4.x with Nsight 5.0. What is the sample you are trying to import?
Yes, I am using CUDA sdk 4.2 samples. Can it possible to run these samples in Nsight 5.0? If yes, then how?
I have both Cuda 4.2 and 5.0 installed.
If you select File -> New CUDA / C++ Project and then give your project a name and select “Import CUDA Sample” as Project Type. You will then get the option of selecting a sample that you want to compile. I selected MonteCarloMultiGPU and it compiled and executed as expected. I am running NSight Eclipse Edition on OS X.
Nsight 5.0 has a built-in support for importing samples from CUDA SDK 5.0. Unfortunately, it cannot import samples from SDK4.2 as it needs some additional metadata to properly setup project from a CUDA sample (you may notice that every supported sample has a “NsightEclipse.xml” description file).
It is possible to configure a project from CUDA SDK 4.2 sample but it is a manual task. I would recommend creating a new empty project and try adding all the source files and headers the sample requires.