88888888888888888888888888888888888888888888888888888888888888888888888888888888888888888888
file:///J:/dl/pte/general-P-R/Doxygen/performance_issues/Doxygen%20Speed%20Issues%20_ts%20martin%20sweitzer.htm
SEARCH_INCLUDES = NO made a big difference. With it set to
NO, the documentation takes ~40 minutes to build. With it set to YES,
Doxygen ran for 8 hours before exhausting memory and stopping.
-----------------------------------------------------
I tried first setting SEARCH_INCLUDES to NO and then removing *.h from
the set of input files. Either change significantly reduced the time
taken but at the expense of documentation produced.
********************************************************************************************
SHOW_DIRECTORIES - setting this to NO reduces
generation time by two thirds.
--------------
Jake Colman reported 39.31 mins with SHOW_DIRECTORIES = NO
and 57.32 mins with it enabled. 18.01 min difference.
significant difference if your project takes hours to DG.
88888888888888888888888888888888888888888888888888888888888888888888888888888888888888888888
I also tried wrapping a shell script around dot, as I mentioned, and only
actually calling dot for graphs of 100 or fewer nodes. This works splendidly.
The 15.5 hour running time was reduced to 45min. If any of you have a tough
time with a long Doxygen job that is using dot, give it a try.
-Karl
******************************************************************
From: Karl Pfleger <kpfleger@cs...>.stanford.edu>
Re: how to avoid running dot for really big diagrams
2003-03-28 18:42
> I wouldn"t expect this to do anything other than scale down the image. I
> wouldn"t think this would save any time running dot. I tried this but 5.5
> hours into my trial run it doesn"t seem to be helping.
So it turns out it did actually help. Reducing dot"s width and height from
1024 to 512 for both reduced running time from 15.5 hours to 7 hours. Still,
not nearly as good as 45min from not running dot on the 100+ node graphs.
-Karl
