Dear MatPlotLib users,
I am having trouble with the performance of matplotlib.
For data analysis, I want to be able to place multiple graphs on screen,
with multiple lines, each consisting of 16000 data points.
I have benchmarked my solution, but it did not perform too well.
For example: 6 graphs with 6 lines each, takes 12.5 seconds.
This graph indicates my benchmark:
http://img59.imageshack.us/img59/7613/graphpython.gif
In comparison, matlab takes only 2.48 seconds for drawing those.
I also noticed that memory usage during the benchmark rises to too high levels.
I have, during a different experiment, plotted 36 graphs with 1 line.
This is about 9MB of total (x,y) data alltogether, but execution of the benchmark
spikes 1GB of memory usage.
My question:
- Is this performance of matplotlib to be expected?
- Can my code (see below) be improved in any way?
Thank you very much in advance,
Mike
···
================================
The code I use for the benchmark
for nr_of_graphs in range (1,7):
for nr_of_lines in range(1,7):
root = Tk.Tk()
#nr_of_lines = int(argv[0])
#nr_of_graphs = int(argv[1])
m = myLinMultiPlot()
m.drawxy("test {0}L on {1}G".format(nr_of_lines, nr_of_graphs), nr_of_graphs, nr_of_lines)
root.mainloop()
The code that plots the actual lines
class myLinMultiPlot(Tk.Toplevel):
def drawxy(self, test_name, plots, lines):
pointsize = 16000
figure = Figure(figsize=(2,1), dpi=100)
storage = []
axes_arr = []
for p in range(0,plots):
for li in range(0,lines):
shift = li * 100
axes = figure.add_subplot(plots,1,1 + p)
axes_arr.append(axes)
xarr = xrange(0,16000)
yarr = []
for x in xarr:
yarr.append(math.sqrt(x + shift))
strg = [xarr,yarr]
storage.append(strg)
startdraw = timeit.default_timer()
for a in axes_arr:
for l in storage:
a.plot(l[0],l[1])
canvas = FigureCanvasTkAgg(figure, master = self)
canvas._tkcanvas.pack(side=Tk.TOP, fill=Tk.BOTH, expand=1)
canvas.show()
canvas.blit()
//This is the time depicted in my benchmark!
durationdraw = timeit.default_timer() - startdraw