David Xu wrote: > On Friday 27 October 2006 07:33, John Birrell wrote: >> [ replying to myself ] >> >> On Thu, Oct 26, 2006 at 09:33:43PM +0000, John Birrell wrote: >>> If you use a GENERIC kernel, then this change won't affect you >>> because the KSE option will be on by default in GENERIC on >>> all arches/machines except sun4v (which doesn't handle signals >>> properly with the KSE code in the kernel). >> scottl persuaded me to add the KSE option to DEFAULTS on all >> arches/machines (except sun4v) to avoid causing the same problem >> that the io/mem change from default to optional caused. >> >> This means that ou will get KSE by default in your kernel (as >> before it was an option) _unless_ you use add 'nooption KSE' >> to your kernel config. >> >> This isn't my preferred solution because it makes it too >> transparent, however scottl's point is that unnecessary grief >> will be caused if the change isn't completely transparent. >> >> Sorry for the confusion. >> >> And BTW, the commits are all in current now...and coming to >> a cvsup server near you. >> >> -- >> John Birrell > > By compiling kernel without KSE option, super-smack's select-key.smack with > 10 clients on Athlon64 dual-core 3800+ breaks 31000 q/s, otherwise it can only > get 29000 q/s. Mysql version is 5.0.24a, FreeBSD AMD64. > I assume that you were using libthr for both the before and after tests? ScottReceived on Thu Oct 26 2006 - 22:56:56 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:01 UTC