| 1 |
On Monday 29 May 2006 11:18, Eskej wrote: |
| 2 |
> On Mon, 29 May 2006 12:24:20 +0400, Dirk Heinrichs |
| 3 |
> |
| 4 |
> <ext-dirk.heinrichs@×××××.com> wrote: |
| 5 |
> > Read my first mail again, please. I didn't write about a single app, I |
| 6 |
> > wrote |
| 7 |
> > about KDE startup. |
| 8 |
> |
| 9 |
> Oh, I'm sorry but shouldn't a single app be affected? And KDE startup time |
| 10 |
> is hardly to measure. |
| 11 |
> So the question is open: has anybody noticed a _single application_ |
| 12 |
> performance increase with GCC 4.1 and hidden visibility. KDE I mean. |
| 13 |
> Or maybe Konsole is no good to make tests? But it is the only precise test |
| 14 |
> I know. When it is about milliseconds human cannot see the difference |
| 15 |
> really. |
| 16 |
|
| 17 |
kdeinit exists to 'preload' all libs. It was introduced, to keep the loading |
| 18 |
time of apps down to a bearable time. Without kdeinit or prelink, the startup |
| 19 |
time of every kde app would be insane. That is, why a single app should not |
| 20 |
be so much affected like the overall kde startup time. After the start, |
| 21 |
everything is 'prelinked' (not really, but kdeinit will be running), and the |
| 22 |
times of the apps are depend on how fast they 'link' against kdeinit.... |
| 23 |
|
| 24 |
-- |
| 25 |
gentoo-user@g.o mailing list |
Archives