On Wed, Jul 23, 2003 at 03:16:44PM -0400, Rick Mohr wrote: > > I believe you are right. I'm afraid that I just trusted the advice of > someone else when he said that the "ht" flag indicates > hyperthreading. And as I had no plain P4's to check this against, it > seemed to be the right way. > > After doing a google search, I got the impression that it is not easy to > tell whether there is hypertheading or not. Is that the case? That would > explain why this hasn't been added to ganglia. > > Maybe I need to go back and simply make this some kind of configuration > option to be specified manually.
My suggestion would be to make an option that is settable manually to say how many logical CPUs per physical CPU there are. This won't compleatly specify the problem space, but it would handle everything available today and most sane cases in the future. It wouldn't handle the case where Intel shipped a new CPU with more logical CPUs and the same bus interface, but that isn't a configuration any manufacture is going to support. (It is worth noting that you _can_ run at least PIIIs with different multipliers so long as they use the same bus speed. It's insane, but it does work.) -- Brooks
pgplOQVIoFUh1.pgp
Description: PGP signature