4OH4

# run multiply function on GPU
multiply_func(    			
    result_gpu, a_gpu, b_gpu,
    block=(100, 1, 1), 
    grid=(1, 1, 1)
)

# Get data back from GPU
cuda.memcpy_dtoh(result, result_gpu)	

4OH4

# Check if PyCuda is installed
package_name = 'pycuda'
use_cuda = importlib.util.find_spec(package_name) is not None

if use_cuda:
    import pycuda.driver as cuda
    import pycuda.autoinit
    from pycuda.compiler import SourceModule

4OH4

#define IMW __IMW_keyword__   // image width (px)
#define IMH __IMH_keyword__   // image height (px)
#define BX __BX_keyword__     // block index x
#define BY __BY_keyword__     // block index y
#define nTPB (BX*BY)          // threads per block
#define IMN 21                // buffer length