Re: classzone-31 vs -ac10 VM

From: t.n.vanderleeuw@chello.nl
Date: Wed Jun 07 2000 - 03:01:35 EST


Hello!

Andrea's advice (below) did indeed fix the OOPSen that I got with
classzone.

I have tried to compare the performance of -ac7+classzone-31 against
-ac10 (which includes all of Rik's latest VM enhancements).

Intuitevely I'd say that -ac10 is faster when the load is low. However,
under high loads, the system probably stays smoother with the classzone
patches (less skimpy MP3 playback).

I find it hard to say though. It seems, as if distortions in MP3
playback are more evenly spreaded with classzone, and they are probably
smaller. With -ac10 playback is generally smooth while there are
sometimes distortions, but the distortions that occur are more serious
than with the classzone patches.

With -ac10, when there are distortions in playback, other applications
don't get time either (mouse and scrolling stop, for instance). With
classzone, the system seems smoother.

I'd say that it appears as if Andrea's classzone patches manage to
spread out the nessecary swap-work better than the current VM.

When I get a chance, I will try John's patch that was posted the other
day, to see if that changes system behaviour for me.

The computer on which I conducted my tests is a PentiumMMX at 225MHz,
64MB SDRAM, IDE drives and a cheapo ALS-110 soundcard.

--Tim :-)

On 5 Jun, Andrea Arcangeli wrote:
> On Mon, 5 Jun 2000 t.n.vanderleeuw@chello.nl wrote:
>
>>Here's the BUG line:
>>kernel BUG at /usr/src/linux/include/linux/mm.h:325!
>
> It happened here:
>
> #ifndef CONFIG_DISCONTIGMEM
> extern inline struct page * alloc_pages(int gfp_mask, unsigned long order)
> {
> /* temporary check. */
> if (contig_page_data.node_gfpmask_zone[gfp_mask].gfp_mask != (gfp_mask))
> BUG();
> ^^^^^
> Such field is read only and it's initialized at boot time. Maybe I'm on
> the wrong track but just in case could you try to recompile the kernel
> from scratch after a `make distclean`?
>
> Andrea
>

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/



This archive was generated by hypermail 2b29 : Wed Jun 07 2000 - 21:00:27 EST