new vm - OK but not THAT great
Hi,
I tested Rik's latest, tuned, VM code on my machine at home. Sorry
this isn't more detailed - I have to do some comparisons later.
However, I DO notice some problems: when I run netscape and a few other
apps, performance is OK, but when I then run 'tar -xf linux-2.4.0-pre4.tar' preformance drops.
Firstly, the page cache gets really big. A LOT of swap is suddenly
used, like 25 MB where before there was nothing, and running programs (netscape,gnomehack,the gnome panel) start getting unresponsive due to swapin.
Secondly, programs like xfs, which are NOT RUNNING AT ALL, do not get
swapped out as much as vanilla pre7-4.
So, the page aging code does not appear to be helping as much as it
should. Swapping seems, in some sense, to not as good as pre7-4...
Awaiting next tuned version!!
-BenRIP.S. could this block of code go into the mainstreat kernel w/o the rest of the code?
Quoted 12 lines
(quoted diff) +5
-2+ /* Reset to 0 when we reach the end of address space */ + mm->swap_address = 0; + mm->swap_cnt = 0; + +out_unlock: vmlist_access_unlock(mm); /* We didn't find anything for the process */ - mm->swap_cnt = 0; - mm->swap_address = 0; return 0; }
Signature · 7 lines
"For nature does not give virtue, It is an art to become good." - Seneca Benjamin Redelings I <>< http://www.bol.ucla.edu/~bredelin -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM, see: http://www.linux.eu.org/Linux-MM/
Quoted 6 lines · Benjamin Redelings
I tested Rik's latest, tuned, VM code on my machine at home. Sorry
this isn't more detailed - I have to do some comparisons later.However, I DO notice some problems: when I run netscape
and a few other apps, performance is OK, but when I then run 'tar -xf linux-2.4.0-pre4.tar' preformance drops.
Indeed, I've seen this here as well. I have some ideas on fixing this and once I figure out all the details I'll implement and tune them.
(but first, I have to fix the SMP bug where __alloc_pages hands out a page that has one of the PG_(in)active_* bits set...)
Quoted 5 lines
Secondly, programs like xfs, which are NOT RUNNING AT ALL,
do not get swapped out as much as vanilla pre7-4.
So, the page aging code does not appear to be helping as
much as it should. Swapping seems, in some sense, to not as good as pre7-4...
This is also something which needs to be tuned more. Remember that you're working with a VM patch that received only about 4 hours of tuning ;)
I hope any of the other VM hackers will use the quiet period around the vger problems to help tune this VM patch a bit ;)
regards,
Rik
--
"What you're running that piece of shit Gnome?!?!"
-- Miguel de Icaza, UKUUG 2000
http://www.conectiva.com/ http://www.surriel.com/
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux.eu.org/Linux-MM/