Date: Tue, 26 Aug 2003 13:22:32 +0800 From: David Xu <davidxu@viatech.com.cn> To: Julian Elischer <julian@elischer.org> Cc: "freebsd-java@freebsd.org" <freebsd-java@freebsd.org> Subject: Re: vmark hangs with libthr and libkse Message-ID: <3F4AEE98.4080607@viatech.com.cn> In-Reply-To: <Pine.BSF.4.21.0308252141330.27218-100000@InterJet.elischer.org> References: <Pine.BSF.4.21.0308252141330.27218-100000@InterJet.elischer.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Julian Elischer wrote: >On Tue, 26 Aug 2003, David Xu wrote: > > > >>Jeff Roberson wrote: >> >> >> >>>>>> >>>>>> >>>>>> >>>>>> >>>>>Why do you need to do adjustrunqueue() in sched_prio? I also don't >>>>>understand the case in sched_switchout(). Can you please explain that? >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>adjustrunqueue maintains kg_last_assigned and related things, when a >>>>thread's priority is changed, >>>>the thread might no longer can be in scheduler's run queue, instead it >>>>will be in ksegrp's runqueue, >>>>because there is higher priority thread, and a KSE it attached should be >>>>detached now, and the KSE >>>>will attach to another higher priority thread, ULE ignores this >>>>requirement, as I can understand, >>>>ULE is only aware of 1:1 between KSE and thread. >>>>It would be nice if scheduler interface is thread aware but not kse aware. >>>> >>>> >>>> >>>> >>>Yes, wouldn't it be nice.. I don't think it should be ksegrp aware >>>either. oh well, it wasn't my design. >>> >>> >>> >>> >>SA process doesn't rely on kse and ksegrp because I introduced a >>kse_upcall structure, >>so I don't care someone drops kse or ksegrp and makes them as scheduler >>specific data structure. >> >> > >Well, this is not quite true. >without KSEGRPS there is no possibility to make both >process scope and system scope threads. > >process scope threads require a rendevous structure of some sort >and it can not be the process. > >The fact that the 1:1 threads don't do this is why they can not >do process-scope threads and system scope threads but are system scope >only. > > OK, I just don't want to manage kse, what I want is to set a concurrent level for a ksegrp, for example, a process-scope thread's group has a concurrent level equals number of cpu, so please don't force me to create and destroy kse, make them transparent. >MACH didn't have the additional concept of the KSEGRP and the >contortions they had to go to to try do process scope threads (they >eventually gave up) (I was a MACH user at that time) was incredible. > > > >>>Will you commit this patch? >>> >>> >>> >>> >>Will do. >> >> >> >>_______________________________________________ >>freebsd-threads@freebsd.org mailing list >>http://lists.freebsd.org/mailman/listinfo/freebsd-threads >>To unsubscribe, send any mail to "freebsd-threads-unsubscribe@freebsd.org" >> >> >> > >_______________________________________________ >freebsd-threads@freebsd.org mailing list >http://lists.freebsd.org/mailman/listinfo/freebsd-threads >To unsubscribe, send any mail to "freebsd-threads-unsubscribe@freebsd.org" > > >
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3F4AEE98.4080607>