Ok, I know this is a little late, these are results for test1-ac22
Riel and ClassZone, but they may be of some use...(Alan asked to test VM
ideas in real world use and this is a very real for me)
First, simple and dirty test: time to load Netscape and
gnome-terminal with mem=16M. This is a Celeron 566, and the two kernels
have the same configuration. The test is: boot the machine in runlevel
5, logon, netscape and gnome-terminal.
There is also information for the same test on RedHat 6.2 stock
kernel (mem=16M), just for comparition. But the compile configuration is
a lot different and kernel 2.2.14 isn't using the UDMA features from VIA
chipset.
[times in min:sec]
2.4.0-test1-ac22-Riel
Time to load Netscape = 1:40, Time to load gnome-terminal = 0:24
2.4.0-test1-ac22-ClassZone
Time to load Netscape = 1:08, Time to load gnome-terminal = 0:20
2.2.14-5.0
Time to load Netscape = 0:40, Time to load gnome-terminal = 0:13
ClassZone is much better than Riel for low memory machines (I tried
the same with mem=24M with similar results), but neither one could beat
2.2.14. [WHY??]
In the second test I use all my memory (192M) uncompressing (rar) a
cd iso image with 650M. I also used binchunk to convert a 300M image
from bin to iso (binchunk output file was in a different drive than rar
output).
2.4.0-test1-ac22-Riel
UnRar 650M = 3:07, binchunk 300M = 0:40
UnRar+binchunk (at the same time) = 4:41
2.4.0-test1-ac22-ClassZone
UnRar 650M = 3:06, binchunk 300M = 0:40
UnRar+binchunk (at the same time) = 4:37
2.2.14-5.0
UnRar 650M = 5:52, binchunk 300M = 2:25
UnRar+binchunk (at the same time) = ages...
(the mouse became too sluggish with 2.2.14 - probably because lack of
UDMA)
IMHO we shouldn't choose right now the VM algorithm to use. Why not
leave it as a compiler option?
I'm using ClassZone for my everyday work since Andrea's patch for
2.3.99-pre7 saved me from the flaws of VM 2.3.99-pre's. It's stable and
it's very good. We are very lucky to have 2 quality and mature VM
algorithms in Linux, I think we should keep both for a while and let
Riel and Andrea improve them.
Regards,
Miguel Freitas
-
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 : Fri Jul 07 2000 - 21:00:13 EST