← 输入框|github.com|刷新
Skip to content

300% CPU on Raspberry Pi #7989

New issue
New issue
Closed
Closed
300% CPU on Raspberry Pi#7989

Description

@alexandruradovici
alexandruradovici
opened on Jan 30, 2017

Bug report

matplotlib (1.4.x and 2.0.0) make 300% CPU usage after import

  • I am using matplotlib on the Raspberry Pi (raspbian jessie - nu GUI). Once I import anything from the library, the CPU goes to 300%. It does not matter if I select any specific backend.

Code for reproduction

import matplotlib

Matplotlib version

  • Raspbian Jessie, 1.4.x or 2.0.0

  • Python 2.7 and Python 3

  • I installed it with

    sudo apt-get install python-matplotlib

or

sudo pip install matplotlib

Activity

  1. tacaswell commented on Jan 30, 2017

    @tacaswell
    tacaswell
    on Jan 30, 2017
    Member

    Does it go down if you wait for a while? The first import may be searching for fonts which can take a while.

  2. alexandruradovici commented on Jan 30, 2017

    @alexandruradovici
    alexandruradovici
    on Jan 30, 2017 via email
    Author

  3. itdaniher commented on Feb 1, 2017

    @itdaniher
    itdaniher
    on Feb 1, 2017

    Were you running this test on the original raspberry pi?

  4. alexandruradovici commented on Feb 2, 2017

    @alexandruradovici
    alexandruradovici
    on Feb 2, 2017 via email
    Author

  5. tacaswell commented on Feb 2, 2017

    @tacaswell
    tacaswell
    on Feb 2, 2017
    Member

    Can you also try just running fc-list from the command line?

  6. itdaniher commented on Feb 2, 2017

    @itdaniher
    itdaniher
    on Feb 2, 2017

    Now I'm surprised. Thanks for clarifying, I'm curious as to the cause and will check out matplotlib on a comparable computer I have on hand.

  7. itdaniher commented on Feb 2, 2017

    @itdaniher
    itdaniher
    on Feb 2, 2017

    also pip freeze would be useful information as I'll try my best to match version numbers.

  8. alexandruradovici commented on Feb 2, 2017

    @alexandruradovici
    alexandruradovici
    on Feb 2, 2017
    Author

    I ran fc-list, here is the output

    pi@raspberrypi:~ $ fc-list
    /usr/share/fonts/truetype/droid/DroidSerif-Italic.ttf: Droid Serif:style=Italic
    /usr/share/fonts/truetype/droid/DroidSerif-BoldItalic.ttf: Droid Serif:style=Bold Italic
    /usr/share/fonts/type1/gsfonts/a010015l.pfb: URW Gothic L:style=Demi
    /usr/share/fonts/truetype/dejavu/DejaVuSansCondensed-Bold.ttf: DejaVu Sans,DejaVu Sans Condensed:style=Condensed Bold,Bold
    /usr/share/fonts/truetype/dejavu/DejaVuSansMono-Bold.ttf: DejaVu Sans Mono:style=Bold
    /usr/share/fonts/type1/gsfonts/n019043l.pfb: Nimbus Sans L:style=Regular Condensed
    /usr/share/fonts/truetype/droid/DroidSerif-Regular.ttf: Droid Serif:style=Regular
    /usr/share/fonts/truetype/dejavu/DejaVuSans-Oblique.ttf: DejaVu Sans:style=Oblique
    /usr/share/fonts/type1/gsfonts/n019044l.pfb: Nimbus Sans L:style=Bold Condensed
    /usr/share/fonts/type1/gsfonts/a010033l.pfb: URW Gothic L:style=Book Oblique
    /usr/share/fonts/type1/gsfonts/n021003l.pfb: Nimbus Roman No9 L:style=Regular
    /usr/share/fonts/type1/gsfonts/s050000l.pfb: Standard Symbols L:style=Regular
    /usr/share/fonts/type1/gsfonts/b018035l.pfb: URW Bookman L:style=Demi Bold Italic
    /usr/share/fonts/type1/gsfonts/c059033l.pfb: Century Schoolbook L:style=Italic
    /usr/share/fonts/truetype/droid/DroidSansEthiopic-Regular.ttf: Droid Sans Ethiopic:style=Regular
    /usr/share/fonts/type1/gsfonts/a010035l.pfb: URW Gothic L:style=Demi Oblique
    /usr/share/fonts/truetype/lyx/stmary10.ttf: stmary10:style=Roman
    /usr/share/fonts/truetype/droid/DroidSansArabic.ttf: Droid Sans Arabic:style=Regular
    /usr/share/fonts/truetype/dejavu/DejaVuSerif-Italic.ttf: DejaVu Serif:style=Italic
    /usr/share/fonts/truetype/droid/DroidNaskh-Regular.ttf: Droid Arabic Naskh:style=Regular
    /usr/share/fonts/truetype/droid/DroidKufi-Regular.ttf: Droid Arabic Kufi:style=Regular
    /usr/share/fonts/truetype/droid/DroidSansHebrew-Bold.ttf: Droid Sans Hebrew:style=Bold
    /usr/share/fonts/type1/gsfonts/n019023l.pfb: Nimbus Sans L:style=Regular Italic
    /usr/share/fonts/type1/gsfonts/b018012l.pfb: URW Bookman L:style=Light
    /usr/share/fonts/truetype/lyx/esint10.ttf: esint10:style=regular
    /usr/share/fonts/type1/gsfonts/c059016l.pfb: Century Schoolbook L:style=Bold
    /usr/share/fonts/type1/gsfonts/n022004l.pfb: Nimbus Mono L:style=Bold
    /usr/share/fonts/type1/gsfonts/n019024l.pfb: Nimbus Sans L:style=Bold Italic
    /usr/share/fonts/type1/gsfonts/b018032l.pfb: URW Bookman L:style=Light Italic
    /usr/share/fonts/type1/gsfonts/p052003l.pfb: URW Palladio L:style=Roman
    /usr/share/fonts/truetype/droid/DroidSansEthiopic-Bold.ttf: Droid Sans Ethiopic:style=Bold
    /usr/share/fonts/truetype/lyx/cmr10.ttf: cmr10:style=Regular
    /usr/share/fonts/truetype/lyx/msam10.ttf: msam10:style=Regular
    /usr/share/fonts/truetype/droid/DroidSansGeorgian.ttf: Droid Sans Georgian:style=Regular
    /usr/share/fonts/type1/gsfonts/n019004l.pfb: Nimbus Sans L:style=Bold
    /usr/share/fonts/type1/gsfonts/b018015l.pfb: URW Bookman L:style=Demi Bold
    /usr/share/fonts/type1/gsfonts/n022024l.pfb: Nimbus Mono L:style=Bold Oblique
    /usr/share/fonts/type1/gsfonts/c059036l.pfb: Century Schoolbook L:style=Bold Italic
    /usr/share/fonts/type1/gsfonts/p052024l.pfb: URW Palladio L:style=Bold Italic
    /usr/share/fonts/truetype/dejavu/DejaVuSerif.ttf: DejaVu Serif:style=Book
    /usr/share/fonts/truetype/lyx/cmsy10.ttf: cmsy10:style=Regular
    /usr/share/fonts/truetype/lyx/cmmi10.ttf: cmmi10:style=Regular
    /usr/share/fonts/truetype/dejavu/DejaVuSans-BoldOblique.ttf: DejaVu Sans:style=Bold Oblique
    /usr/share/fonts/type1/gsfonts/n019003l.pfb: Nimbus Sans L:style=Regular

  9. alexandruradovici commented on Feb 2, 2017

    @alexandruradovici
    alexandruradovici
    on Feb 2, 2017
    Author

    The pip-freeze is here

    html5lib==0.999
    ino==0.3.6
    itsdangerous==0.24
    joblib==0.8.3
    matplotlib==1.4.2
    meld3==1.0.0
    mock==1.0.1
    msgpack-python==0.4.7
    ndg-httpsclient==0.3.2
    nose==1.3.4
    numexpr==2.4
    numpy==1.8.2
    oauth==1.0.1
    openpyxl==1.7.0
    ordereddict==1.1
    pandas==0.14.1
    picamera==1.12
    pyFirmata==1.0.3
    pyOpenSSL==0.13.1
    pyasn1==0.1.7
    pybass==0.55.0
    pyfacebook==0.1
    pyinotify==0.9.4
    pyparsing==2.0.3
    pyserial==2.7
    python-apt==0.9.3.12
    python-dateutil==2.2
    pytz==2012c
    redis==2.10.5
    requests==2.4.3
    scipy==0.14.0
    simplejson==3.6.5
    six==1.8.0
    smbus==1.1
    statsmodels==0.4.2
    supervisor==3.0
    tables==3.1.1
    tweepy==2.3
    twilio==3.6.5
    urllib3==1.9.1
    wheel==0.24.0
    wsgiref==0.1.2
    xlrd==0.9.2
    xlwt==0.7.5

  10. jkseppan commented on Feb 11, 2017

    @jkseppan
    jkseppan
    on Feb 11, 2017
    Member

    @alexandruradovici You could run "top" to see which process is taking all the CPU. Then something like "strace -p pid" would show what system calls the process is making.

  11. alexandruradovici commented on Feb 11, 2017

    @alexandruradovici
    alexandruradovici
    on Feb 11, 2017 via email
    Author

  12. alexandruradovici commented on Feb 14, 2017

    @alexandruradovici
    alexandruradovici
    on Feb 14, 2017
    Author

    The python process forks 3 times after import.

    The main REPL process stays in read () - as it should be
    The other three threads (processes) run gettimeofday

  13. alexandruradovici commented on Feb 17, 2017

    @alexandruradovici
    alexandruradovici
    on Feb 17, 2017
    Author

    Do you have any ideas about this?

  14. itdaniher commented on Feb 18, 2017

    @itdaniher
    itdaniher
    on Feb 18, 2017

    @alexandruradovici honestly, no. I'll try reproducing. It might be worth trying a pip install --upgrade matplotlib as 1.4.2 is a little behind what I've successfully used on the 'pi.

  15. itdaniher commented on Feb 18, 2017

    @itdaniher
    itdaniher
    on Feb 18, 2017

    @alexandruradovici depending upon what you're doing, https://pythonhosted.org/pycha/ may work. It's much smaller than mpl.

  16. itdaniher commented on Feb 18, 2017

    @itdaniher
    itdaniher
    on Feb 18, 2017

    Okay, finally got around to install matplotlib==1.42 - the first run took ~5.5 seconds, so something is clearly amiss with your test case.

    Try pip install --upgrade matplotlib (if you installed via pip)

  17. alexandruradovici commented on Feb 18, 2017

    @alexandruradovici
    alexandruradovici
    on Feb 18, 2017
    Author

    I installed matplotlib 2.0.0 using

    pip install matplotlib

    I get the same result using strace.

    I am testing using it in a REPL shell. The moment I import matplotlib, the CPU goes up and stays like that

  18. rickbase1 commented on Mar 17, 2017

    @rickbase1
    rickbase1
    on Mar 17, 2017

    I have the same problem. Did anyone get to the bottom of this one.
    Matplotlib 2.0.0, installed via pip3, python3, Rpi3
    CPU load goes to 100% for 5-6 seconds, then drops to 75-80%
    normal load is 5-7%

  19. rickbase1 commented on Mar 17, 2017

    @rickbase1
    rickbase1
    on Mar 17, 2017

    Additional information:

    python 3.4.3
    uname -a =
    Linux Antenna_tracker 4.4.38-v7+ #938 SMP Thu Dec 15 15:22:21 GMT 2016 armv7l GNU/Linux
    sample code used to test CPU load

    import matplotlib.pyplot as plt

x = [1, 2, 3, 4, 5, 6]
y = [3, 3, 2, 5, 3, 5]
x2 = x
y2 = [3, 4, 5, 6, 3, 2]

plt.plot
plt.plot(x, y,  label =" x y numbers",  color = 'blue')
plt.plot(x, y2,  label =" x2 y2 numbers",  color ='r')
plt.xlabel(" x variables")
plt.ylabel(" y variables")
plt.title('CPU load test plot')
plt.legend()

plt.show()

  20. peroveh commented on Apr 9, 2017

    @peroveh
    peroveh
    on Apr 9, 2017

    Similar issue just importing numpy
    import numpy as np
    makes cpu load go to 300
    Both Python2 and Python3 have same issue

  21. itdaniher commented on Apr 9, 2017

    @itdaniher
    itdaniher
    on Apr 9, 2017

    I can confirm there are no blatant issues with matplotlib or numpy on ARMv7. I have a full numpy / scipy / matplotlib stack working on armbian working on a 'pi competitor.

  22. peroveh commented on Apr 10, 2017

    @peroveh
    peroveh
    on Apr 10, 2017 via email

  23. efiring commented on Apr 11, 2017

    @efiring
    efiring
    on Apr 11, 2017
    Member

    Based on the conversation to date it seems most likely that this has nothing to do with matplotlib specifically, so I am closing this.

  24. efiring
    closed this as completedon Apr 11, 2017
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions