Computational performance measurement tool

computational_performance.py

from typing import List
import matplotlib.pyplot as plt
import pandas as pd
import numpy as np
from timeit import repeat
from multiprocessing import Pool, cpu_count

def time_inline(function: str, *arg) -> float:
    stmt = f'{function}(*{arg})'
    return min(repeat(stmt, repeat=5, number=1, globals=globals()))
    
def time_it(functions: List[str], args):
    with Pool(cpu_count()) as pool:
        iterator = ((function, arg) for function in functions for arg in args)
        results = pool.starmap(time_inline, iterator)

    data = np.reshape(results, (len(functions), len(args)))
    result = pd.DataFrame(index=functions, columns=args, data=data)

    ax = result.T.plot()
    ax.set_xlabel('input size')
    ax.set_ylabel('time (seconds)');

    ax = result.div(result.min()).T.plot(loglog=True)
    ax.set_xlabel('input size')
    ax.set_ylabel('time (relative)');

functions = ['function1',
             'function2',
             'function3']
args = [10, 50, 100, 500, 1_000, 5_000, 10_000]
time_it(functions, args)