Re: Results on Claasen's -nocairo patches



On Thu, 13 Jul 2006 13:31:19 -0400, "Morten Welinder" wrote:
> 
> I would call excluding the X server a feature.  Sometimes desireable,
> sometimes not.

If we're worried about cairo slowing down GTK+, it ends up being quite
important to be able to measure the X server time as well.

For example, one of the early performance bugs caused by cairo usage
in GTK+ resulted in nautilus taking a very, very long time to draw the
background window on the desktop.

With that bug, cairo and GTK+ themselves weren't taking any
significant amount of additional time---they were just sending a
different request to the X server. And as it so happened, the request
being sent triggered a nasty slow path in the X server.

> There are a couple of situation where you want to do just that:
> 
> 1. Remote X connection, even if we're just talking a local network.
> 2. Multi-CPU machines.
> 
> In either case, X is running truly is parallel with the program.

But you still care about the total system time required to get your
job done. Even if the X server is running on a different machine
across a remote network, it's still very interesting to know if cairo
is making the X server take longer to achieve results than it took in
the past to achieve the same results. (Granted, it can get harder to
measure the interesting things in setups like this, but it's still
what's interesting.)

-Carl

Attachment: pgpAQdQgzmW4f.pgp
Description: PGP signature



[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]