Skip to content

scipy.signal.lfilter memory leak when size of signal is similar to size of filter (Trac #1190) #1717

scipy-gitbot opened this Issue Apr 25, 2013 · 3 comments

2 participants


Original ticket on 2010-06-02 by trac user hogu, assigned to unknown.

import numpy as np
import scipy.signal

temp = np.random.random(200)
for c in range(200000):
    print c
    temp2 = scipy.signal.lfilter(np.ones(180), [1], temp)

this causes my process memory to grow

I am on ubuntu lucid lynx with libatlas3gf-sse2 if that is relevant


@jpaalasm wrote on 2010-10-19

Try if upgrading to scipy 0.7.1 or 0.8 would solve the problem.

A memory leak in lfilter was fixed in 0.7.1 (#911) and that might apply to this problem, too.


@rgommers wrote on 2010-10-19

Memory doesn't grow for me with 0.8.0

@bfroehle bfroehle pushed a commit to bfroehle/scipy that referenced this issue Apr 29, 2013
@jseabold jseabold TST: fix linalg.qz test to avoid seg fault on Windows. Closes #1717.
(backported from baa8f73)
SciPy member

Somehow had a comment on this in my inbox that isn't here anymore. Anyway, tried again to reproduce - no luck on Ubuntu or OS X. Reference are wrong, this has nothing to do with linalg.qz. Closing.

@rgommers rgommers closed this Dec 16, 2013
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Something went wrong with that request. Please try again.