CUDA grid launch failed: CUcontext: 2636322209376.....

Hello Guys,

I am launching a kernel which fails whenever I am trying to read a global device array inside it (i.e int a = devMem_Array[0]). Also, I got this error: CUDA grid launch failed: CUcontext: 2636322209376 CUmodule: 2636146208032 Function: _Z30MyKernelPdS_S_S_PiS_S0_S_S_S_ddi

I made sure that the device array has been well allocated and I don’t know why the reading on this memory will fail.

Please can you tell me what I should do?


Never mind ! I solved it now. So, I just did what Nvidia suggests:

__device__ float* devPointer;
float* ptr;
cudaMalloc(&ptr, 256 * sizeof(float));
cudaMemcpyToSymbol(devPointer, &ptr, sizeof(ptr));

However, instead of doing sizeof(ptr) on the cudaMemcpyToSymbol, I was putting the size explicitly (256 * sizeof(float)) and that’s why it did not work.