cudamalloc with size variable on device memory

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi there,

I have four arrays:

x_max

x_min

y_max

y_min

Each ones first element is holding a integer value:

x_max[0]

x_min[0]

y_max[0]

y_min[0]

Now I want to call an cudamalloc like this:

cudaMalloc((void**) &picture, (x_max[0]-x_min[0])*(y_max[0]-y_min[0])*sizeof(float));

but that wont do, because I cannot access device memory from host like this.

How would you guys do that?

Thanks in advance for tips!

Julian

EDIT: Copy the four values to host like this:

cudaMemcpy( h_out_x_max, d_max_x, sizeof(float), cudaMemcpyDeviceToHost );

cudaMemcpy( h_out_x_min, d_min_x, sizeof(float), cudaMemcpyDeviceToHost );

cudaMemcpy( h_out_y_max, d_max_y, sizeof(float), cudaMemcpyDeviceToHost );

cudaMemcpy( h_out_y_min, d_min_y, sizeof(float), cudaMemcpyDeviceToHost );

I would like to circumvent because it takes 59ns for the four floats!!! Thats too much for my opinions.