plot_parallel.py

import pandas as pd
import matplotlib.pyplot as plt
import sys

df = pd.read_fwf(sys.argv[1], comment=';')
df.drop(
    columns=[
        'seed', 'level', 'it', 'dijkstra', 'unordered', 'ordered', 'expanded',
        'm', 'many', 'm_expanded', 'bs'

        # 'seed', 'level', 'it', 'dijkstra', 'unordered', 'ordered', 'expanded',
        # 'm', 'many_avx', 'many', 'many_8', 'many_8_n', 'm_expanded',
        # 'dijkstra_pn', 'phast_pn', 'p_many_avx', 'p_many'
    ],
    inplace=True)

# df['concurrent_queue'] = df['work']

dfs = [(name, data)
       for _, (name, data) in zip(df.graph.unique(), df.groupby(df['graph']))]

for name, data in dfs:
    r1 = data.groupby('p').mean().at[1, 'rebuild']
    s1 = data.groupby('p').mean().at[1, 'spinning']
    c1 = data.groupby('p').mean().at[1, 'conditional']

    if int(sys.argv[2]) > 0:
        data['rebuild'] = r1 / data['rebuild']
        data['spinning'] = s1 / data['spinning']
        data['conditional'] = c1 / data['conditional']
    if int(sys.argv[2]) > 1:
        data['rebuild'] /= data['p']
        data['spinning'] /= data['p']
        data['conditional'] /= data['p']

    grouped = data.groupby('p')
    mean = grouped.mean()
    std = grouped.std()

    print(name)
    print(mean)

    mean.plot(fmt='o', yerr=std)
    # mean.plot()
    title = name
    plt.title(title)
    plt.xlabel("number of threads")
    plt.ylabel("time per tree [ms]")
    plt.show()