Problems with cudaMemcpy. cudaErrorInvalidDevicePointer.

Accelerated Computing CUDA CUDA Programming and Performance
1

Hello, i’m getting problems with copying memory content. My program is combined c/c++ code and on start of a program i create my variables by cu file:

unsigned char* pbGlobalRGBData;

extern "C" void 

cuAllocateHostMemory(unsigned char **pbDataRGB, int w, int h, bool fWithpbData)

{

	cudaError_t result;

   CUDA_SAFE_CALL(result = cudaMallocHost((void**)pbDataRGB, w*h*3));

	

  CUDA_SAFE_CALL(result = cudaMalloc((void**)&pbGlobalRGBData, w*h*3*sizeof(unsigned char)));

}

After completing this variables initialization program returns to his tasks, and when i call the next function:

extern "C" void cuCopyRGBToGrayscale(unsigned char *pbData, unsigned char *pbRGBData, int w, int h)

{

	cudaError_t result;

   result = cudaMemcpy(pbGlobalRGBData, pbRGBData, w*h, cudaMemcpyHostToDevice);

/*	rest is not important

dim3 dimBlock(BLOCK_SIZE);

	dim3 dimGrid(h/BLOCK_SIZE);

	RGBtoGrayscale<<<dimGrid, dimBlock>>>(pbGlobalData, pbGlobalRGBData, w, h);

	

	cudaThreadSynchronize();

*/

	

}

after this cudaMemCopy i get the cudaErrorInvalidDevicePointer error. Maybe i don’t understand some basics of the CUDA programming, but why is it happening? I’ve also tried constant variables and copytosymbol, but nothing seem to work. Can anyone help?

2

Problem solved, I’ve created variable in one CPU thread and then i used it in other, what seems to be forbidden in CUDA.