How to visualize threads in a multi-threading task/time graph?

Question

I'm trying to plot the threads of my multi-threading code in a meaningful way using matplotlib. I want that every thread is visualized by one color. In this way, the plot will clearly show which tasks are executed by which thread etc. So to be clear, I want to say (figure underneath) the yellow bars are processes executed by thread 1, the red bars processes executed by thread 2 and the blue bars processes executed by thread 3.

This seems quite hard and the best I could come up with is underneath (see figure and code). In the example we're having 3 threads and 12 tasks (the duration of every task is variable up to some point). Thread1 starts yellow, thread2 starts red and thread3 starts blue. I would love to keep them color-coded like that throughout the whole graph. What we are seeing however is that thread 1 executes tasks 0, 7 and 10 but it switches color from yellow to red to red. Same for thread 2: It executes tasks 2, 5, 8 and 11 but switches color from red to blue to blue to blue. Same for thread 3. So the colors appear actually in a cycle of 3 here and are independent of the threads. I want to have them dependent of the thread-number as I said earlier to make the multi-threading plot much more meaningful (because it isn't at the moment).

Anybody any ideas how to do this?

import threading
import multiprocessing
import math
import numpy as np
import time
import matplotlib.pyplot as plt
import glob
from PIL import Image
import random
from random import sample
import string
from concurrent.futures import ThreadPoolExecutor


cpu_workers = 3
nSim = 12

def generate_bar_colors(cpu_workers):
    colors = ['red', 'gold', 'royalblue']
    return colors

def visualize_runtimes(results, title):
    colors = generate_bar_colors(cpu_workers)
    plt.rcParams["font.family"] = "Times New Roman"
    plt.rcParams['axes.axisbelow'] = True
    start,stop = np.array(results).T
    plt.barh(range(len(start)),stop-start,left=start, color=colors)
    plt.grid(axis='x', color= 'lightgrey')
    plt.title("Tasks", rotation='horizontal', fontsize=12, horizontalalignment="left", x=0)
    plt.xlabel("Seconds", fontsize=12, horizontalalignment='right', x=1.0)

def multithreading(func, args, workers):
    begin_time=time.time()
    with ThreadPoolExecutor(max_workers = workers) as executor:
        res = executor.map(func, args, [begin_time for i in range (len(args))])
    return list(res)

def simulation(i, base):
    start = time.time() - base
    print(str(threading.current_thread().getName()) + ': '+ str(i))
    time.sleep(math.cos(i)+i*0.1+1)
    stop = time.time() - base
    return start, stop


if __name__ == '__main__':
    visualize_runtimes(multithreading(simulation, i, cpu_workers), "Multi-threading")
    plt.savefig('foo.png', bbox_inches='tight')

plt.show()

Answer 1

One way to achieve this (see code below). Now every thread has a color assigned and it is clear that calculations are multi-threaded as they should (in this case 3 threads were executing 12 tasks).

import threading

import math

from matplotlib.lines import Line2D


from concurrent.futures import ThreadPoolExecutor

from concurrent.futures import ProcessPoolExecutor

import matplotlib.pyplot as plt

from matplotlib.ticker import AutoMinorLocator, MultipleLocator, FuncFormatter
import time

import numpy as np




cpu_workers = 3

nSim = 12

i = range(nSim)



#Multi-threading function ----------------------------------------------------------------------------------------------

def multithreading(func, args, workers):

    with ThreadPoolExecutor(max_workers = workers) as executor:
        
        responses = executor.map(func, args)

    return list(responses)



#List of unique arguments in preserved order as in my_list -------------------------------------------------------------

def pres_uniq_list(my_list): #returns a unique list in preserved order

    seen = set()

    result = []

    for e in my_list:

        if e not in seen:

            result.append(e)

            seen.add(e)

    return result



#Get netto simulation start- and end-times as well as duration ---------------------------------------------------------

def sep_list_elements(list_of_lists, proc_start_t):

    start_values = [inner_list[0] for inner_list in list_of_lists]

    start_values = np.array(start_values) - proc_start_t


    end_values = [inner_list[1] for inner_list in list_of_lists]

    end_values = np.array(end_values) - proc_start_t

    return start_values, end_values



#Match colors with threads (one color per thread) ----------------------------------------------------------------------

def thread_colors(list_of_lists):

    thread_ids = [inner_list[2] for inner_list in list_of_lists]

    color_guide = ['red', 'royalblue', 'gold', 'darkgray', 'forestgreen', 'orangered', 'lightpink', 'teal']


    lookup = dict(zip(pres_uniq_list(thread_ids), color_guide))

    colors = [lookup[number] for number in thread_ids]

    return colors



#Graph legend to match with bars ---------------------------------------------------------------------------------------

def thread_legend(list_of_lists):

    thread_list = [inner_list[3] for inner_list in list_of_lists]

    color_guide = ['red', 'royalblue', 'gold', 'darkgray', 'forestgreen', 'orangered', 'lightpink', 'teal']


    lookup = dict(zip(pres_uniq_list(thread_list), color_guide))

    colors = [lookup[number] for number in thread_list]

    thread_legend = ([Line2D([0], [0], color=c, alpha=0.4, lw=4) for c in pres_uniq_list(colors)])

    names = [str('Thread-')+str(i) for i in range(cpu_workers)]

    return thread_legend, names



#Graph definition using MatPlotLib -------------------------------------------------------------------------------------

def graph_settings(start_t, end_t, title, colors, legend):


    plt.rcParams["font.family"] = "Times New Roman"

    plt.rcParams["font.size"] = 10

    plt.rcParams['axes.axisbelow'] = True

    plt.rcParams['axes.edgecolor'] = 'black'

    plt.rcParams['axes.linewidth'] = 0.8

    plt.rcParams['xtick.color'] = 'black'

    plt.rcParams['ytick.color'] = 'black'

    #plt.rcParams['font.weight']= 'heavy'



    fig = plt.figure(figsize=((12.0/2.54), (7.42/2.54)), facecolor='w', edgecolor='black')  #set (12,7.42) gulden snede but figsize is in inches

    fig, ax = plt.subplots()


    #ax.xaxis.set_major_locator(MultipleLocator(1.000))

    #ax.xaxis.set_minor_locator(AutoMinorLocator(4))

    #ax.yaxis.set_major_locator(MultipleLocator(1.000))

    #ax.yaxis.set_minor_locator(AutoMinorLocator(4))


    ax.set_xlim(-0.2, end_t[nSim-1])

    ax.set_ylim(-1, nSim)

    ax.spines['top'].set_visible(False) #to set color: ax.spines['top'].set_color('green')

    ax.spines['right'].set_visible(False)

    ax.spines['left'].set_smart_bounds(True)

    ax.spines['bottom'].set_smart_bounds(True)


    #ax.tick_params(which='major')

    #ax.tick_params(which='major', length=5)

    #ax.tick_params(which='minor', labelsize=10)

    #ax.tick_params(which='minor', length=2.5, labelsize=10, labelcolor='0.25')

    ax.margins(0.02) #margins in %, default is 5%

    ax.barh(range(len(start_t)), end_t-start_t, left=start_t, color=colors, alpha=0.4, edgecolor="none") #alpha adds vagueness to the bars

    ax.barh(range(len(start_t)), end_t-end_t+0.00001, left=end_t, color=colors, alpha=1, edgecolor="none") #alpha adds vagueness to the bars


    #w = pres_uniq_list(p.get_height() for p in ax.patches)
    #w_fl = float(i) for i in w]

    #ax.plot(end_t, range(nSim), linewidth=0, marker="o", markersize=w[0], color="grey")

    #print(end_t)

    #plt.grid(axis='x', color= 'lightgrey')

    #plt.ylabel("Tasks", fontsize=12, horizontalalignment='left', y=1.02, rotation='horizontal')

    ax.set_title(title, fontsize=14, fontweight="bold", horizontalalignment='center', y=1.04) #set title instead of y-label, bold does not work

    ax.annotate("Tasks", fontsize=12, xy=(0, 1), xytext=(0, 10), xycoords="axes fraction", textcoords="offset points", ha="right", ) #set ha="left" to have it above the axis

    ax.annotate(("Tasks: " + str(nSim) + " - Threads: " + str(cpu_workers)), fontsize=10, fontstyle='italic', xy=(0.5, 1), xytext=(0, 4), xycoords="axes fraction", textcoords="offset points", ha="center", color='gray' ) #set ha="left" to have it above the axis


    ax.set_xlabel("Time [s]", fontsize=12, horizontalalignment='right', x=1.0)
    #ax.set_suptitle(title, fontsize=16, fontweight='bold', y=1.005)

    thread_legend = reversed(legend[0])

    names = reversed(legend[1])



    leg = ax.legend(thread_legend, names, fancybox=True, loc='lower right', fontsize=10, edgecolor=None)

    #plt.legend(frameon=False)

    leg.get_frame().set_facecolor('none')

    leg.get_frame().set_edgecolor('none')


    plt.savefig('P1_5test.pdf', bbox_inches='tight')  #dpi=300 when png , bbox_inches='tight'

    plt.show()

    #plt.figure(figsize=(12, 10))

    #plt.savefig('foo1.pdf', , dpi = 300)

    return None



#Run code for if __name__ == '__main__' that links all functions -------------------------------------------------------

def plot_runtimes(workers, func, args, title):

    proc_start_t = time.time() # we track time from here

    runtimes = multithreading(func, args, workers) #collect task runtimes from function

    #net_runtimes = net_start_t(proc_start_t, runtimes) #extract net_runtimes from runtimes

    start_t, end_t = sep_list_elements(runtimes, proc_start_t) #seperate start_t and end_t

    colors = thread_colors(runtimes) #select thread colors

    legend = thread_legend(runtimes) #define thread legend

    graph_settings(start_t, end_t, title, colors, legend) #settings for the horizontal bar charts



#Tasks to perform ------------------------------------------------------------------------------------------------------

def simulation(i):

    #i = range(nSim)

    start_t = time.time()

    #print(str(i) + ': start')

    a = str(threading.current_thread().ident)

    b = str(threading.current_thread().getName())

    print(str(threading.current_thread().getName()) + ': '+ str(i))

    #print(str(i) + ': finish')

    time.sleep(math.cos(i)+i*0.1+1)

    end_t = time.time()

    return [start_t, end_t, a, b]

    #return a



#Definition ------------------------------------------------------------------------------------------------------------

if __name__ == '__main__':


    plot_runtimes(workers = cpu_workers,

                  func = simulation,

                  args = i,

                  title = "Multi-threading")