Messages in this thread | | | From | Brad.Larden@solution ... | Subject | SUMMARY-SMP on a single processor machine | Date | Fri, 26 Nov 1999 09:10:50 +1100 |
| |
Thanks to those that have responded !
I thought I would summarise this as it's quite simple and there was no disagreement in the replies. Please excuse any 'oddness' in the format, I am stuck using Lotus Notes .....
Those that responded were:- Alan Cox <alan@lxorguk.ukuu.org.uk> Mark Hahn <hahn@coffee.psychology.mcmaster.ca> Christian Stuke <christian@stuke.de> Peter Waltenberg <peterw@surf.dascom.com>
----- > Having just installed Redhat 6.1 on the machine I find that it has > installed 2.2.12-20smp by default (yes, I know that's a redhat-ism).
Many SMP boxes report both CPU's when only one is present. It isnt a big problem.
> My question is; Do you see a problem running the smp kernel with a single > processor installed ?
On SMP capable CPU's - no. The single CPU kernel is more efficient however (typically 5% faster)
----- > I have a Dell PowerEdge 1300 which has a dual processor motherboard but is > only fitted with one processor and some sort of 'dummy' board in the 2nd > processor slot.
it's a bus terminator. its improves the signal properties of the bus when there's no processor there to load it properly.
> My question is; Do you see a problem running the smp kernel with a single > processor installed ?
there is none. it'll be very slightly slower, possibly unmeasurable outside of a benchmark. ----- > I have a Dell PowerEdge 1300 which has a dual processor motherboard but is > only fitted with one processor and some sort of 'dummy' board in the 2nd > processor slot. That´s a bus-terminator, it belongs where it is! ;) It helps to keep the second slot from acting as an antenna ;)
> Having just installed Redhat 6.1 on the machine I find that it has > installed 2.2.12-20smp by default (yes, I know that's a redhat-ism). > > My question is; Do you see a problem running the smp kernel with a single > processor installed ? no, it´s just a little slower i guess. the smp-kernel has a little more overhead but it´ll work ok in my opinion.
> Excuse my ignorance, I have searched the archives and come up empty on this > subject. sure. normally nobody would want to do this ...
> I just have no idea what implications might be in running the smp kernel > and as this machine is going to be a rather well used production squid > server, I don't want it falling in a heap because I 'assumed' everything > would be fine. By the way, the machine won't have a 2nd processor installed > anytime soon. In this case i´d recommend installing a single cpu kernel!
> Any pointers or comments would be most appreciated. Welcome > Regards, > Brad. Chris
-----
There have been buggy kernels where this won't work, but if it works, fine.
There's a small (<1%) speed penalty in the kernel. Since most programs spend very little time in the kernel, it's really not noticeable.
Peter
- 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/
| |