Cupy comes with an internal benchmarking function

from cupyx.profiler import benchmark

def my_func(a):
    return cp.sqrt(cp.sum(a**2, axis=-1))

a = cp.random.random((256, 1024))
print(benchmark(my_func, (a,), n_repeat=20))

with an output

my_func : CPU:   44.407 us   +/- 2.428 (min:   42.516 / max:   53.098) us     GPU-0:  181.565 us   +/- 1.853 (min:  180.288 / max:  188.608) us

Both benchmark appear to be very small in this example, but another RawKernel showed much higher values (~CPU: 50000 us) and (GPU-0: ~ 50000 us) for a very similiar operation.

What do long CPU and long GPU times indicate?