Subj : Re: RCU+SMR To : comp.programming.threads From : David Hopwood Date : Sat Jul 09 2005 02:04 am Randy Howard wrote: > Joe Seigh wrote: >>So while you could do some really neat things with hyperthreading, >>no one will because it will be gone by the time the OSes could >>provide an api to exploit it. > > I hope so. Hyperthreading is beneficial in a very few special > cases, but in general it sucks compared to SMP or multi-core > solutions. I suppose keeping it around inside of multi-core > processors isn't horrible, as long as you can turn it off. In > some cases, it makes things much worse enabled than disabled. I'm not a fan of hyperthreading either, but putting it in processors and then turning it off is the worst of all worlds: the processor designers have to make compromises to support HT and spend time debugging it, that could have been spent more profitably with other optimizations and with verification. (On the subject of excessive complexity in processors, watch this talk by Bob Colwell, formerly of Intel: .) > True. I've yet to hear of any commercial or popular OSS > application bragging about the massive performance benefits > achieved through the use of lock-free techniques. I very much doubt that there is enough benefit from using lock-free techniques in *applications* to justify the complexity cost, and the effort required to learn how to use them correctly. They belong in operating systems and language implementations, where these costs are amortized over many apps. -- David Hopwood .