Will Riel's vmpatch make it into 2.4.0-test9?

From: Bruce A. Locke (blocke@shivan.org)
Date: Fri Sep 15 2000 - 18:03:37 EST


Hello... Just a question from a user... :)

I was just wondering if Rik van Riel's VM patches might possibly be
integrated into 2.4.0-testX anytime soon?

I have been having very good experiences with Riel's latest patches in
both a desktop and light server environment. On the desktop, Riel's
vmpatch against 2.4.0-test8 is significantly better then 2.4.0-test8's
performance. Operation seems much faster and smoother. In fact, I'd say
it could even be much better then whats floating around in 2.2.x now!

Just a couple of examples on where the differences in Riel's patches
become apparent (though one might argue that they are useless
indicators)....

I have recently been having difficulty with Loki Games's latest patches
for the game Unreal Tournament. The problem is it leaks memory very
quickly. After about two minutes of gameplay the process takes up 300+MB
of memory with an increase of around 10mb per minute. This machine only
has 256mb of ram so you can imagine that it starts to swap at that
point. Usually under 2.2.x it takes about 4 minutes of gameplay before
I hear my harddrive screetch and the game starts becoming
unresponsive. With Riel's patch and 2.4.0-test8, I can sometimes play for
up to 6 minutes before the swapping makes the game unusual. Is Riel's
patch more efficient at swapping or something?

Another example... Start Quake3, join a game on the internet and play for
about 5 minutes. Then quit and go right back into the game. With Riel's
patch the harddrive is only touched for a brief second. With 2.2.x the
harddrive is touched for a few seconds while quake3 loads the maps, etc
that were played before. Riel's patch seems to keep things cached in
memory much more efficiently.

And yet another example... mkisofs on my system is a dog with
2.4.0-test8. It kills performance of other applications and makes mp3
players skip etc. There is no skipping and "freezeups" with Riel's
patch. Although mkisofs seems slower under 2.4.0-test8+vmpatch then stock
2.2.x, Riel's VM seems to be much "smoother".

I tried Riel's patch on a lightly loaded webserver. Though because of the
lack of any substantial load on the system I don't think its worth
commenting on it... Perhaps other people could comment on their
experiences?

I am looking forward to vmpatch with the planned IO tweaks and out of
memory handler... I would hate to have to patch the VM system throughout
the 2.4.x stable tree with his patch so I'm lobbying for it to be
included. :)

Just my opinions and experiences... please feel free to flame away if I am
misunderstanding something... Thanks for your time.

--------------------------------------------------------------
Bruce A. Locke
blocke@shivan.org

"The Internet views censorship as damage and routes around it"
www.eff.org www.peacefire.org

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



This archive was generated by hypermail 2b29 : Fri Sep 15 2000 - 21:00:26 EST