Scripts/Benchmarking/propagation_timing.py

0001 import ROOT
0002 import csv
0003 import matplotlib.pyplot as plt
0004 import numpy as np
0005
0006 # Data preparation
0007 dataDict = {}
0008
0009 # Open the output file
0010 with open("output.log", mode="r") as csv_file:
0011     csv_reader = csv.reader(csv_file, delimiter=",")
0012     # read lines and go for it
0013     for csv_row in csv_reader:
0014         if len(csv_row) > 1:
0015             # get the job id
0016             jobID = csv_row[0]
0017             # we need the exec time
0018             exectime = 0.0
0019             with open("timing_" + jobID + ".tsv") as tsv_file:
0020                 tsv_reader = csv.reader(tsv_file, delimiter="\t")
0021                 for tsv_row in tsv_reader:
0022                     if str(tsv_row[0]) == "Algorithm:PropagationAlgorithm":
0023                         exectime = float(tsv_row[2])
0024             # stepper and pt
0025             stepper = int(csv_row[1])
0026             ptvalue = float(csv_row[2])
0027             # now open the ROOT file and extract the numbers
0028             rfile = ROOT.TFile("propagation_steps_" + str(jobID) + ".root")
0029             stree = rfile.Get("propagation_steps")
0030             stree.Draw("@g_x->size()>>h_steps")
0031             h_steps = ROOT.gDirectory.Get("h_steps")
0032             steps = h_steps.GetMean()
0033             stepsSpread = h_steps.GetMeanError()
0034
0035             # Make sure you have all the keys ready
0036             try:
0037                 cdict = dataDict[ptvalue]
0038             except:
0039                 dataDict[ptvalue] = {}
0040                 cdict = dataDict[ptvalue]
0041
0042             # Now fill the sub dictionary
0043             try:
0044                 vdict = cdict[stepper]
0045             except:
0046                 cdict[stepper] = []
0047                 vdict = cdict[stepper]
0048
0049             vdict += [steps, stepsSpread, exectime, exectime / steps]
0050
0051 # plot the dataDict
0052 plt.figure(figsize=(16, 5))
0053
0054 ax = plt.subplot(131)
0055 plt.loglog(
0056     dataDict.keys(),
0057     [i[0][0] for i in np.array(list(dataDict.values()))],
0058     "+",
0059     label="line",
0060 )
0061 plt.loglog(
0062     dataDict.keys(),
0063     [i[1][0] for i in np.array(list(dataDict.values()))],
0064     "*",
0065     label="eigen",
0066 )
0067 plt.loglog(
0068     dataDict.keys(),
0069     [i[2][0] for i in np.array(list(dataDict.values()))],
0070     "o",
0071     label="atlas",
0072 )
0073 ax.set_xlabel("$p_T$ [GeV]")
0074 ax.set_ylabel("#steps")
0075 ax.set_xlim((-10, 150))
0076 plt.legend(numpoints=1)
0077
0078 ax = plt.subplot(132)
0079 plt.loglog(dataDict.keys(), [i[0][2] for i in np.array(list(dataDict.values()))], "+")
0080 plt.loglog(dataDict.keys(), [i[1][2] for i in np.array(list(dataDict.values()))], "*")
0081 plt.loglog(dataDict.keys(), [i[2][2] for i in np.array(list(dataDict.values()))], "o")
0082 ax.set_xlabel("$p_T$ [GeV]")
0083 ax.set_ylabel("time [a.u.]")
0084 ax.set_xlim((-10, 150))
0085
0086
0087 ax = plt.subplot(133)
0088 plt.loglog(dataDict.keys(), [i[0][3] for i in np.array(list(dataDict.values()))], "+")
0089 plt.loglog(dataDict.keys(), [i[1][3] for i in np.array(list(dataDict.values()))], "*")
0090 plt.loglog(dataDict.keys(), [i[2][3] for i in np.array(list(dataDict.values()))], "o")
0091 ax.set_xlabel("$p_T$ [GeV]")
0092 ax.set_ylabel("time/steps [a.u.]")
0093 ax.set_xlim((-10, 150))
0094
0095
0096 plt.suptitle("Stepper comparison: Constant Field")
0097
0098 plt.show()