Who knows why the running time of a module using opencl on the same platform decreased as the data array increased?

Question

Python

with a = np.arange(8000).astype(np.float32)

and with a = np.arange(80000).astype(np.float32) the

first one is 172409*7.33063e-07 s, and the second one is 154627*7.33063e -07 s, how can this be given that the size of the array of the first is 8000 and the second is 80000?
These calculations were performed on the following platform:
pyopencl.Device 'Intel(R) Core(TM) i3-2365M CPU @ 1.40GHz' on 'Intel(R) OpenCL'

Answer 1

Since this is a CPU, the overall CPU usage will affect the execution time. And for the CPU it is absolutely normal that it walks from time to time. In general, 0.07 seconds difference is not an indicator.