Currently Being Moderated

How-to: Reporting PyCharm performance problems and taking CPU/memory snapshots

VERSION 3

Created on: Oct 20, 2010 11:04 AM by Eugene Toporov - Last Modified:  Nov 25, 2013 2:19 PM by Serge Baranov

This document is out of date, the new version is available at

https://intellij-support.jetbrains.com/entries/29983118-Reporting-performance-problems

Common solutions for PyCharm performance problems

Before capturing and providing the CPU snapshot, please be aware of the following known issues (and possible solutions):


  • Antiviral tools monitoring project and caches directories (~/.PyCharmXX).  Exclude PyCharm project/library/caches directories from the monitor.
  • PyCharm caches on the slow or network drive,  check that ~/.PyCharmXX is on the local drive, adjust the  location in PYCHARM_HOME\bin\idea.properties if necessary (but be aware  of the first item in this list).
  • Project files/SDK installation on a network drive. Move them to your local drives.
  • Low free system memory and therefore swapping. Close other applications, free as much memory as possible. Install additional RAM, decrease the heap.
  • Very  large files (>10 000 lines). Refactor them if possible. Disable  inspections per file to save some resources using the Hector icon in the  status bar.
  • Heavily fragmented partition and full MFT, see the related blog post. Drive defragmentation can make PyCharm much faster, we have proven it on our development machines.
  • Heavy disk activity caused by another applications, such as torrent clients, file servers.  One thread can read data from the modern HDD at ~100 MB/sec. If there  are 2 or more applications reading/writing to the same HDD at the same  time, performance will drop to 2-3 MB/sec. Check that there are no  processes constantly reading/writing to the disk. On Windows you can use  SysInternals Process Monitor or Process Explorer.
  • External process modifying or generating lots of files/logs into the project directories. Synchronization will take lots of time in this case. Exclude such directories from the project.
  • When using version control, if there are many unversioned files (thousands), Changes view update may be slow. Exclude such directories  from the version control in Settings | Version Control | Directory  Version Control Settings by adding the directory with such files to the  list and setting VCS to NONE.

 

Automatic Thread Dumps Logging

If  the UI of PyCharm is unresponsive for more than 5 seconds, it starts  writing thread dumps to the logs directory. If you've encountered a  situation of PyCharm unresponsiveness or deadlock, please find the  threadDumps-xxx directory from the session in which you've experienced  the problem and send the contents to support@jetbrains.com

 

On Windows and Linux, the logs directory can be found at:

 

 

~/.PyCharmXX/system/log


 

On Mac OS X, the logs directory can be found at:

 

~/Library/Logs/PyCharmXX

 

CPU snapshot capturing instructions

If you have a performance problem not related to the issues mentioned above, please contact support@jetbrains.com and provide the following information:

  • You hardware configuration (CPU, Memory, HDD)
  • OS, free memory when you start PyCharm, PyCharm heap settings, file system, PyCharm caches location (if modified)

 

 

  1. If  you are using PyCharm on Windows under a 64-bit JVM, please follow instructions below. Otherwise please proceed to the following step.

     

    Please obtain the library(64-bit  for 64-bit JVM) version from the YourKit Java Profiler distribution (you need Linux .zip download of 8.x version containing library versions for different operating systems:  Linux, Mac, Solaris, Windows). You will find libraries in "bin" folder  of the distribution. Copy (or replace) yjpagent library in PYCHARM_HOME\bin with the appropriate version. Use 8.x YourKit version for PyCharm, otherwise toolbar buttons will not appear!

     

  2. Open PYCHARM_HOME/bin/pycharm.exe.vmoptions (PyCharm.vmoptions on Linux), add the following line at the bottom:


    -agentlib:yjpagent

     

    On Mac OS X open Info.plist located in /Applications/PyCharm-X.X.app/Contents, find the following:

     

    <key>VMOptions</key>

    <string>-Xms16m -Xmx192m -XX:MaxPermSize=120m -Xbootclasspath/p:../lib/boot.jar -ea</string>

     

    Modify the string tag by adding -agentlib:yjpagent option, like:

     

    <string>-Xms16m -Xmx192m -XX:MaxPermSize=120m -Xbootclasspath/p:../lib/boot.jar -ea -agentlib:yjpagent</string>

     

 

Start PyCharm and you should notice 2 new toolbar buttons: one for taking CPU snapshot,  another one for Memory snapshot. Press the CPU button and perform the  actions which take a lot of CPU resources on your machine. When  finished, press this button again to stop recording and save the snapshot. Then submit an issue to PyCharm issue tracker and attach the zipped snapshot file to it. In the e-mail please also describe what you were doing in detail. If you can reproduce the problem in a small project, please also provide one.

 

Memory snapshot

 

In  case of memory related issues (memory usage goes high, garbage is not  collected, etc) please use the Memory snapshot button in the menu near  the CPU snapshot button. If it's not possible to get the snapshot  because of the application crashing with OutOfMemory errors, please add  the

-XX:+HeapDumpOnOutOfMemoryError

option to the PyCharm JVM options(the same place as for "-agentlib:yjpagent" option in  above CPU snapshot section). On the next OOM error the .hrpof dump will  be produced and saved by the JVM (usually in the application working  directory which is PYCHARM_HOME\bin).

 

The memory snapshots should be uploaded to our FTP site: ftp://ftp.intellij.net/.uploads/

 

After uploading the snapshot, please notify us by e-mail or file a YouTrack issue and specify the name of the uploaded snapshot file.

Average User Rating
(0 ratings)




Einar Bergmundur Einar Bergmundur  says:

It might be useful for OS X user to know that the cache is in ~/Library/Caches/PyCharm[10|20]

More Like This

  • Retrieving data ...

More by Eugene Toporov