a=np.random.rand(m*n).astype(np.float32)
# create a m*n array, initialized with random float32 precision data
import pyopencl as cl
ctx=cl.create_some_context() #create space for the context, which context, how big size and which devices??
queue=cl.CommandQueue(ctx) #request a chunk of memory in GPU, so a shown above can be transferred to the GPU