How computing in Numba works on Python

To understand the basics of GPU computing with Numba, it is important that we first learn how a @jit decorator works. It is a central feature of Numba. Using this decorator, Numba's JIT compiler can be used to optimize a function.

To import jit from Numba, we use the following syntax:

from numba import jit

In traditional Python, we use the following syntax for defining a function (recall our conventional multiply function-based program used earlier):

def multiply(p_cpu, q_cpu):
    for i in range(N):
        q_cpu[i] = p_cpu[i] * q_cpu[i]

To use JIT with Numba, we only have to add the following just before defining the function:

@jit
def multiply(p_cpu, q_cpu):
    for i in range(N):
        q_cpu[i] = p_cpu[i] * q_cpu[i]

The preceding change will allow the JIT compiler to take over the multiply function and translate it into fast machine...