Re: radeon graphic speed regression 2.4 -> 2.6 test7 benh
- To: Soeren Sonnenburg <email@example.com>
- Cc: Michel Dänzer <firstname.lastname@example.org>, "email@example.com" <firstname.lastname@example.org>
- Subject: Re: radeon graphic speed regression 2.4 -> 2.6 test7 benh
- From: Benjamin Herrenschmidt <email@example.com>
- Date: Thu, 30 Oct 2003 19:09:41 +1100
- Message-id: <1067501381.7861.2.camel@gaston>
- In-reply-to: <1067494825.1533.9.camel@localhost>
- References: <1067411001.1578.4.camel@localhost> <1067417008.3494.14.camel@gaston> <1067417466.1999.2.camel@localhost> <firstname.lastname@example.org> <1067445595.2982.15.camel@localhost> <email@example.com> <1067466793.6144.4.camel@localhost> <1067467159.3494.33.camel@gaston> <1067494825.1533.9.camel@localhost>
On Thu, 2003-10-30 at 17:20, Soeren Sonnenburg wrote:
> On Wed, 2003-10-29 at 23:39, Benjamin Herrenschmidt wrote:
> > > It is not synthetic. It feels really sluggish in the shell compared to
> > > 2.4... Also I have CONFIG_PREEMPT enabled and now I do wonder that I did
> > > not yet see a random crash - hmmhhh well except for a try to get
> > > cryptoloop working. That was reliably crashing.
> > >
> > > Should I go for the old radeon driver ?!
> > No, same problem I'm sure. Just disable CONFIG_PREEMPT, and make sure
> > you are using the kernel's AGP and DRI drivers (they do help for 2D on
> > ATIs)
> with single head, agp+dri it is acceptable... but still 1m slower than
> 2.4 (also noticable on the 100% filled sysload applet).
> real 5m10.619s
> user 0m1.367s
> sys 0m13.029s
> (btw these are 254665 files)
> I really wonder what changed from 2.4 to 2.6 to make the same X config 8
> times slower... w.r.t. terminal speed.
Did you measure without CONFIG_PREEMPT ? It has bugs that could explain