[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kde-devel
Subject:    Re: reason behind fno-exceptions?
From:       Waldo Bastian <bastian () kde ! org>
Date:       2001-07-30 23:01:56
[Download RAW message or body]

On Monday 30 July 2001 03:44 pm, Daniel Schmitt wrote:
> On Tuesday 31 July 2001 00:28, Martijn Klingens wrote:
> > > What's wrong with C++ exceptions, besides the fact that they cause
> > > large memory overhead with gcc ?
> >
> > "besides" ? I think that single argument is convincing enough...
> > Actually, I also think it's the only argument, but I haven't heard
> > stronger arguments that often...
>
> This whole thread made me experiment a bit, and now I'm positively
> confused. On my system / compiler (x86, Linux 2.4.7, glibc 2.2.3, gcc
> 2.95.3), I can't get more than about a tenth of a percent reduction in
> resident set size through -fno-exceptions from a couple of toy projects,
> regardless of whether the code was using exceptions originally or not.

It might not show up in your resident set size. Try to have a look at the 
numbers reported by "free" (corrected for buffers/cache). You might need to 
start a few instances to make the difference visible. (If you have 128Mb 
free, you will hardly notice 200Kb)

> Under what circumstances does this overhead become noticeable? Maybe only
> in shared libraries?

Compiling Qt with or without exception should give a noticable difference. 
I'm not sure if it is limited to libs, but those are my biggest concern. 
(Since the libs in KDE are typically much larger than the code of 
applications themselves)

Cheers,
Waldo
-- 
Andrei Sakharov, Exiled 1980-1986, USSR, http://www.aip.org/history/sakharov/
Dmitry Sklyarov, Detained 2001-????, USA, http://www.elcomsoft.com/
 
>> Visit http://master.kde.org/mailman/listinfo/kde-devel#unsub to unsubscribe <<

[prev in list] [next in list] [prev in thread] [next in thread]