[Fedora-ia64-list] RFC: Thoughts on SPARSE memory
Luck, Tony
tony.luck at intel.com
Fri Jul 14 17:01:48 UTC 2006
> lmbench results for 4p/4G SGI 330, and 64p/128G SGI 3700bx2 ...
Thanks for running these and posting the results ... but this section
at least looks like a random number generator was used!
Context switching - times in microseconds - smaller is better
-------------------------------------------------------------
Host OS 2p/0K 2p/16K 2p/64K 8p/16K 8p/64K 16p/16K 16p/64K
ctxsw ctxsw ctxsw ctxsw ctxsw ctxsw ctxsw
--------- ------------- ----- ------ ------ ------ ------ ------- -------
4p-DISCON Linux 2.6.17- 1.700 20.3 22.1 20.1 25.3 17.8 25.9
4p-SPARSE Linux 2.6.17- 16.9 2.9700 6.2700 14.2 10.3 16.4 17.3
DISCONTIG is either 10 times better, or 6.8 times worse than SPARSE :-(
Did you run the new(ish) version of lmbench that can bind processes to
particular cpus to help cut down on such erratic results?
-Tony
More information about the Fedora-ia64-list
mailing list