On Friday 29 September 2006 22:54, Roger Larsson wrote:
> On Friday 29 September 2006 14:31, David Faure wrote:
> 
> > Any mistakes in my test? ;)
> >
> 
> You did not tell what compiler (version) and options you used.
> But I guess it was unoptimized...

gcc-4.0.3, default kde-trunk buildsystem options with debugfull enabled.

Let's see what happens with -O2 instead of -g3 -fno-inline...

QDEBUG : KFileItemTest::testPerformance() CTOR: Simple struct: 16
QDEBUG : KFileItemTest::testPerformance() CTOR: With QSharedDataPointer: 18
QDEBUG : KFileItemTest::testPerformance() COPY CTOR: Simple struct: 5
QDEBUG : KFileItemTest::testPerformance() COPY CTOR: With QSharedDataPointer: 1

Hey that's pretty good. The performance hit for the default ctor is down to 12%
while the performance hit for the copy ctor when not using qshareddatapointer 
is still very high (around 500%, although with '1' the number is too small to mean anything).
This seems to confirm that I should go ahead with QSharedDataPointer.

-- 
David Faure, faure@kde.org, sponsored by Trolltech to work on KDE,
Konqueror (http://www.konqueror.org), and KOffice (http://www.koffice.org).

_______________________________________________
Kde-optimize mailing list
Kde-optimize@kde.org
https://mail.kde.org/mailman/listinfo/kde-optimize