Faq improve performance with PLS Toolbx and Matlab on Mac

From Eigenvector Documentation Wiki
Jump to: navigation, search

Issue:

How can I improve performance with PLS_Toolbox and Matlab on the Mac platform?

Possible Solutions:

Some combinations of PLS_Toolbox and Matlab can experience performance issues. These issues can be the result of a combination of factors involving both PLS_Toolbox and or Matlab. We recommend one or more of the following:

  1. Use the latest version of PLS_Toolbox. In some cases we've been able fix problems associated with Mac performance.
  2. Try changing the Java Heap Size from the Matlab → Preferences → Java Heap Memory menu item. Select the largest amount available (e.g., 256MB).
  3. For Matlab versions 2011a and newer, disable Screen Menus via the java.opts file. Save all of your work then from the Matlab Command window enter:
    >> edit(fullfile(matlabroot,'bin','maci64','java.opts'))
    then add the following line:
    -Dapple.laf.useScreenMenuBar=false
    save the file and restart Matlab.
  4. Try using a different version of Matlab. Depending on the version of hardware and PLS_Toolbox, some [older] versions of Matlab may work better than others. Newer versions (2011a-2012a) seem to require disabling of screen menus and increased Heap size (steps 2 and 3). In general we recommend the newest version of Matlab available.
  5. Hide the model cache. Sometimes the java tree component seems to slow things down. Hiding the cache when not in use may help. From the Analysis Tools menu select View Cache → Hide Cache Viewer.
  6. Set javaopts to use Quartz rendering (via the java opts as above).
    -Dapple.awt.graphics.UseQuartz=true
  7. Restart Matlab often (daily). This is probably the most effective way to keep performance from dropping.


The Mathworks has stated to us in response to service requests (FEB 2012):

"Currently, MATLAB graphics performance can be better under Windows than Mac. The Win API supports "immediate mode" drawing, and the "deferred mode" on OS X limits the raw performance vs. Windows. Also Intel has not made available for OS X a version of their compiler which supports PGO."


Still having problems? Please contact our helpdesk at helpdesk@eigenvector.com