Why is my matplotlib script terribly slow on macbook desktop but works fine on external monitor? (Python, matplotlib, OSX)

Question

I am using a post processing plot script written in python, using python 3.7 and OSX 10.14

The script is probably not optimized, but it works good enough when I use my external monitors. When I try to run it on my macbook pro 15" monitor (2018), it freezes and becomes extremely slow and unresponsive unless I set the figure dimension small enough.

I have no idea of the reason, and I haven't found any explanation browsing the internet. Any clue on what can cause this behaviour and how I can fix it?

This is the 'cut' version of the script I'm running.

import numpy as np
import matplotlib
from matplotlib import pyplot as plt
from matplotlib.patches import Rectangle, Circle
import sys


def main(iter0,iterN,its,out_dir='output'):
    it0 = int(iter0)
    itN = int(iterN)
    itstep = int(its)
    a = list(range(it0,itN+1,itstep))
    loaded_steps=len(a)
    e = np.loadtxt(out_dir+'/energy_data_dynamics')
    #... load many other files

    plt.ion()
    fig = plt.figure(figsize=(20, 20),facecolor="white")  #THESE DIMENSIONS WORK FOR ANY EXTERNAL SCREEN 
#    fig = plt.figure(figsize=(10, 10),facecolor="white") #THESE DIMENSIONS WORK FOR MACBOOK SCREEN

    it = 0
    for index in range(0,loaded_steps):
        plt.cla()
        plt.axes().set_aspect('equal')
        #... plot many other subplots
        ax414=plt.subplot(414)
        ax414.plot(e[it0+it,0], e[it0+it,2], marker="o", color="crimson", ms=15)
        ax414.plot(e[it0:itN+1,0], e[it0:itN+1,2],'-r')
        ax414.set_title("Energy")

        plt.show()
        plt.pause(0.0000000001)
        
        it+=itstep
        if index == loaded_steps-1:
            input("Plot finished - Press [enter] to continue.")




if __name__ == '__main__':
    matplotlib.use("TkAgg")
    its = 1
    out_dir = 'output'
    if len(sys.argv)>=4:
        its = sys.argv[3]
    if len(sys.argv)>=5:
        out_dir = sys.argv[4]
    main(sys.argv[1],sys.argv[2],its,out_dir)

Answer 1

It is hard to give a direct answer without running your code but the symptoms you are describing (works on an external monitor but not the internal one) sounds like you may be running into a GPU bug. External monitors may use a different GPU (usually the second GPU vs the built-in GPU). But the fact this simple script shows that slowdown implies there is a bug (your code, the driver, somewhere) or the function calls are causing a problem.

The figures seem tiny compared to what your are plotting: a figure size of 10x10 with a marker size of 15? I would start commenting out lines one by one using the built-in monitor until it gets fast again. There may also be a more efficient way to clear the axes and replot -- I've found Python / plt is very slow when I try to clear a figure and redraw. You may also want to try creating the subplot outside of your loop since some implementations/OSs have a lot of overhead for creating subplots.