将变量从CUDA文件发送到CPP文件？

Question

I have a CPP project with 1 CUDA file in it.

I have a host and device array that I've declared in my CUDA main method

float* h_CovarianceMatrix[11][NumberOfFeatures];
float* d_CovarianceMatrix;

I did a cuda2DMemcpy host to device, did some calculations, and then did a cuda2DMemcpy device to host.

So now my h_CovarianceMatrix[][] is ready to be sent back to my cpp class.

My question is, how do i send this variable to a CPP file from CUDA?

Answer 1

I believe the global symbols declared in the the CUDA source file are visible to the CPP linker. You just need to move the declaration of the matrix symbol into a .h file and include the .h in both the CPP and the CUDA files.