Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Performance issue: C++ project in Tg 2006 R2 Eclipse

2 views
Skip to first unread message

Michael Lerch

unread,
Nov 27, 2006, 10:38:26 AM11/27/06
to
Hi,

I'm using Together 2006 Release 2 for Eclipse (V 8.1). I'm working with
a C++ modelling project with approx. 250 classes. I don't think that
this is a very large project. But on my machine (2GHz P4, 1.24 GB RAM,
Linux Suse 10.0) it takes about 6 min (!) to open the project.
Apparently, most of the time is needed to parse the source files (~500).
I used Together Architect 1.1 before and the parser was able to read
source files of the same project in 30 sec.!!

As proposed in the readme, I tried to tweak the JVM settings - without
any success. My current settings are:
-vm
/home/mlh/Borland/Together/jre/bin/java
-vmargs
-Xms512M
-Xmx1024M
-XX:MaxPermSize=128M


Do you have any suggestions to improve the performance? Any comments are
welcome!


Best regards,

Mike.

ElenaA

unread,
Nov 28, 2006, 6:18:13 AM11/28/06
to
The way C++ files are parsed has changed since TCC: in TCC, each header was
parsed only once. Now, each file is parsed as many times as it is included
(i.e. as many times as it appears in the include directives that can be
reached from the entry points). This, of course, makes the performance much
slower, but parsing results are more correct.

I'd suggest to inspect 'Special Considerations for C++ Projects ' Help
reference - proper configuring of your project can significantly improve the
performance

"Michael Lerch" <michae...@methodpark.de> wrote in message
news:456b...@newsgroups.borland.com...

Michael Lerch

unread,
Nov 29, 2006, 7:51:45 AM11/29/06
to
Hi,

thank you for the quick answer.

Obviously, one first thing to try, is to remove all *.h files as entry
points, because usually they will be included by the corresponding *.cpp
files. This will save at most half the time needed for parsing the
project sources. Unfortunately, in my case, this is not enough. For
further performance improvement, source code changes are necessary. One
can think of introducing macros (defined for Together only) that prevent
the parser from including files. This means much work and is also
dangerous, since the model could become inconsistant.

Best regards,

Mike.


ElenaA schrieb:

0 new messages