cudaArray invalid argument find the fault (please)

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi, when I try to run my test-program, I get a fault (invalid argument) in the line:

cutilSafeCall( cudaMemcpyToArray( cu_array, 0, 0, arr, size, cudaMemcpyHostToDevice));

Relevant part of the program:

int main(int argc, char** argv) 

{  

const int rows = 5;

const int columns = 5;

unsigned int size = rows * columns * sizeof(double);

//Initiate 2d array :

double **arr = (double**)malloc(rows * sizeof(*arr));

  for(int i = 0; i < rows; ++i)

	{

		arr[i] = (double*)malloc(columns * sizeof(*arr[0]));

	}

for(int i = 0; i < rows; ++i)

	{   for(int j = 0; j < columns; ++j)

	   {	 arr[i][j] = i*j+j;

			 printf("%lf\n", arr[i][j]);

	   }

	}

	

// allocate array and copy data

	cudaChannelFormatDesc channelDesc = cudaCreateChannelDesc(32, 0, 0, 0, cudaChannelFormatKindFloat);

	cudaArray* cu_array;

	cutilSafeCall( cudaMallocArray( &cu_array, &channelDesc, columns, rows )); 

	cutilSafeCall( cudaMemcpyToArray( cu_array, 0, 0, arr, size, cudaMemcpyHostToDevice));

}

Does anyone see per chance what I do wrong?

Thanks

2

anyone?

3

Did you ever get an answer? I’m having the same problem.
I get an invalid argument when I call cudaMemcpyToArray.
I have an idea of what might be causing your problem, although it’s just a guess.
You might try passing in arr instead of arr which is a double*.
I think the function calls for a const void * in that argument.

4

Because you used cudaChannelFormatKindFloat option, you should use float type instead of double type. Replacing the double data type with float, you can solve this problem.