Subject: Re: largepage success
To: Jason R Thorpe <thorpej@zembu.com>
From: Manuel Bouyer <bouyer@antioche.lip6.fr>
List: port-i386
Date: 09/12/2000 09:43:59
On Mon, Sep 11, 2000 at 11:32:33AM -0700, Jason R Thorpe wrote:
> Note, I didn't add *any* extra -j options ... just "make".  The extact
> process I followed was:
> 
> 	make depend && make clean
> 	reboot
> 	time make
> 	make clean
> 	reboot into 4MB kernel
> 	time make

Ok, with this test I get the same results as you. Looks like 16 procs is too
much for a PC :)
> 
>  > Sure, I guess for such patterns the gain can be big. Maybe also for buffer
>  > cache pages ?
> 
> Buffer cache pages are problematic -- pagemove().

Ok, I see.

--
Manuel Bouyer, LIP6, Universite Paris VI.           Manuel.Bouyer@lip6.fr
--