2024-heraklion-comp-arch/exercise.ipynb

In [ ]:

import numpy as np

In [ ]:

n_series = 32
len_one_series = 5*2**20
time_series = np.random.rand(n_series, len_one_series)
gap = 16*2**10

In [ ]:

print(f'Size of one time series: {int(time_series[0].nbytes/2**20)} M')
print(f'Size of collection: {int(time_series.nbytes/2**20)} M')
print(f'Gap size: {int(gap*8/2**10)} K')
print(f'Gapped series size: {int(time_series[0, ::gap].nbytes/2**10)} K')

In [ ]:

# compute a Taylor-like series
def taylor(time_series, mean, gap):
    for row, ts in enumerate(time_series):
        for pwr in range(1,20):
            mean[row] += (ts[::gap]**pwr).sum()
    return mean

Challenge¶

• Can you improve on the above implementation of the taylor function?
• Change the following taylor_improved function and see what you can do
• Remember: you can't change any other cell in this notebook!

In [ ]:

def taylor_improved(time_series, mean, gap):
    for row, ts in enumerate(time_series):
        for pwr in range(1,20):
            mean[row] += (ts[::gap]**pwr).sum()
    return mean

In [ ]:

# verify that they yield the same results
out1 = taylor(time_series, np.zeros(n_series), gap)
out2 = taylor_improved(time_series, np.zeros(n_series), gap)
np.testing.assert_allclose(out1, out2)

In [ ]:

mean = np.zeros(n_series, dtype='float64')
%timeit taylor(time_series, mean, gap)

In [ ]:

mean = np.zeros(n_series, dtype='float64')
%timeit taylor_improved(time_series, mean, gap)