From kde-core-devel  Thu Mar 15 12:46:30 2007
From: Chusslove Illich <caslav.ilic () gmx ! net>
Date: Thu, 15 Mar 2007 12:46:30 +0000
To: kde-core-devel
Subject: Re: Could kdecore depend on kjs?
Message-Id: <200703151346.33301.caslav.ilic () gmx ! net>
X-MARC-Message: https://marc.info/?l=kde-core-devel&m=117396285823456
MIME-Version: 1
Content-Type: multipart/mixed; boundary="--nextPart2225020.LndxCbx4OJ"

--nextPart2225020.LndxCbx4OJ
Content-Type: text/plain;
  charset="utf-8"
Content-Transfer-Encoding: quoted-printable
Content-Disposition: inline

> [: Thiago Macieira :]
> The overhead is not negligible. I can tell you that right now.
>
> Each extra library loaded incurs a performance penalty at load-time.

Hm, I'm getting some suspiciously favorable results, is there something=20
wrong with doing:

time for i in `seq 1000`; do kde4-config; done

and then measuring and averaging the walltime delta between linked and=20
non-linked kjs? ldd reports 37 libs for kde4-config, versus eg. 50 for=20
kdialog, so the relative overhead should be more pronounced.

Unless this benchmark is useless, the results are 23.9 ms for non-linked,=20
versus 24.5 ms for linked call, a 1.6 ms overhead. That's on a 2 GHz=20
AthlonXP, and it predictably went to about 4.1 ms when I downclocked it to=
=20
700 MHz.

=2D-=20
Chusslove Illich (=D0=A7=D0=B0=D1=81=D0=BB=D0=B0=D0=B2 =D0=98=D0=BB=D0=B8=
=D1=9B)

--nextPart2225020.LndxCbx4OJ
Content-Type: application/pgp-signature

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)

iD8DBQBF+UApMSGXgigGr3ERArXKAJ92ZnvTyKoYhL7uAxl9Jm3w1pkb3ACeIAb2
WaBLpbiCodpOYxZqPNoe6iA=
=Xqj9
-----END PGP SIGNATURE-----

--nextPart2225020.LndxCbx4OJ--