From: Joe C. <jo...@sw...> - 2005-10-22 08:54:49
|
> Vern wrote: > >> In case anyone is interested it has to do multiprocessing which I >> don't even have so disabling it does nothing. It's been working for >> over a year and all of a sudden... sometimes I hate computers! Anyway, >> that is resolved. I'll add a few minor bits of info to this thread: P4 processors over 3.0 GHz, P4-based Xeon processors (and not the older PIII-based Xeons), and a few other Intel processors have a feature called hyper-threading. This is treated by Linux as an additional processor and so if you have one of those processors, Linux will treat it as a dual processor SMP machine if you boot into an SMP kernel. This isn't a bug, it's a feature, and increases performance by a small amount for some types of task (primarily tasks that require multiple simultaneous processes rather than a very demanding single process task--a server with a database-backed website would see an improvement while a Squid caching proxy server would lose performance). That's not to say there aren't some systems that might get misdetected as multi-CPU when they don't have hyperthreading or additional physical CPUs, but I'm not aware of any that do. It would be a bug for a non-hyperthread cabable single-core single-CPU system to be detected as SMP-capable. Intel has an overview of hyper-threading, including a specific list of processors that have this feature, here: http://www.intel.com/technology/hyperthread/ An SMP kernel should operate on a single CPU machine without disatrous consequences. Performance will likely suffer, as there is additional overhead in an SMP system due to the additional locking required to have more than one CPU share a single pool of memory. If you find instability when running an SMP kernel on a single CPU system (that is stable when run under a uniprocessor kernel), a bug report to that effect to your operating system vendor would be appropriate. But, there is also no reason to run an SMP kernel on a machine that does not have either multiple CPUs or hyper-threading. It does not improve performance (it does the opposite due to the extra locks), and increases memory usage by a small amount. Finally, kernels are way outside of the scope of the Webmin list (as clearly evidenced by the well-meaning but misleading and/or incomplete replies to this query to date). The newbie or general technical discussion list of your Linux distribution would be the place to start with a problem like this, possibly escalating to a bug report to your OS vendor. |