#!/usr/bin/env python3

from concurrent.futures import ThreadPoolExecutor, wait
import time


def one_thread(chunks, repeats):
    for repeat in range(repeats):
        b''.join(chunks)


def bench(threads, size, chunks, repeats):
    chunks = [b'x' * size for _ in range(chunks)]
    with ThreadPoolExecutor(threads) as pool:
        # warmup
        wait([pool.submit(one_thread, chunks, repeats//100+1) for _ in range(threads)])

        start = time.monotonic()
        wait([pool.submit(one_thread, chunks, repeats) for _ in range(threads)])
        elapsed = time.monotonic() - start

    return elapsed


def main():
    threads = 4
    print('size, chunks, rate[MB]')
    for size in (8, 16, 32, 64, 128, 256, 512, 1024):
        for chunks in (8, 16, 32, 64, 128, 256, 512, 1024):
            repeats = max(3, 10**8 // (size * chunks))
            elapsed = bench(threads, size, chunks, repeats)

            rate = size * chunks * repeats / elapsed
            rate_mb = rate / 1e6
            print(f'{size},{chunks},{rate_mb:.2f}')

main()