PyplotImageViewer delays Experiment execution

[MarcusZuber]

When a PyPlotImageViewer is attached to an experiment the frame-rate drops. As far as I understood, the PyplotImageViewer should drop frames when the drawing of an old frame is not yet finished.

Example:

from concert.experiments.base import Acquisition, Experiment
from concert.experiments.addons import Consumer
from concert.devices.cameras.dummy import Camera
from concert.ext.viewers import PyplotImageViewer
from concert.quantities import q

class MyExp(Experiment):
   def __init__(self):
      from concert.experiments.base import Acquisition, Experiment
      acq = Acquisition("img", self.take_frames)
      Experiment.__init__(self, [acq])

   def take_frames(self):
      import numpy as np
      from time import time
      start = time()
      for i in range(100):
         yield np.zeros((2048, 2048))
      print time()-start

exp = MyExp()
viewer = PyplotImageViewer()
live_view = Consumer(exp.acquisitions,viewer)

Running exp.run() like this gives me 1.24s. If I detatch the live_view I get 0.0125s.
I tested it on my local machine (no ssh x-forwarding).

Can I somehow make the PyplotImageViewer less blocking? Also the asyncronous ImageWriter delays the acquisition (until now usually less than a frame with 15fps and 2k x 2k pixels). Is it possible to speed up the whole addon-consumer-pipeline?

@tfarago I thinkt this is also related to the delay that we had (have) with the pco.4000

[tfarago]

I have very inconsistent results with np.zeros. To me it seems that sometimes python optimizes that just for array reservation and sometimes it really allocates zeros. Try the following:

class MyExp(Experiment):
   def __init__(self, num_iterations=30, create_new=True):
      acq = Acquisition("img", self.take_frames)
      Experiment.__init__(self, [acq])
      self.num_iterations = num_iterations
      self.create_new = create_new

   def take_frames(self):
      start = time()
      create_duration = 0
      im = np.zeros((2048, 2048))

      for i in range(self.num_iterations):
        create_st = time()
        if self.create_new:
            im = np.random.normal(size=(2048, 2048))
        create_duration += time() - create_st
        yield im

      print 'Creation duration:', create_duration
      print ' Overall duration:', time() - start