Date: Mon, 18 Dec 2000 09:40:15 -0800 (PST) From: John Polstra <TrimYourCc@polstra.com> To: smp@freebsd.org Cc: cp@bsdi.com Subject: Re: atomic increment? Message-ID: <200012181740.eBIHeFw65943@vashon.polstra.com> In-Reply-To: <200012180413.eBI4D9P17266@berserker.bsdi.com> References: <200012180413.eBI4D9P17266@berserker.bsdi.com>
next in thread | previous in thread | raw e-mail | index | archive | help
In article <200012180413.eBI4D9P17266@berserker.bsdi.com>, Chuck Paterson <cp@bsdi.com> wrote: > I am pretty sure that what you said about the effects of > volatile matches my understanding of what they do. I am less sure if > we are saying the same thing about where they are needed. > > > While the volatile can be left off of lots of operations, it > still must be left on operations which are used for locking because > there is no way the compile can no what the lock being acquired is > protecting and therefore all computations following the operation > must be done after the asm statement, discarding is partial results > that may be laying around in a register. Yes, I agree with you. The dependency between the lock and the data it protects is one of those side-effects that can't be expressed as an output or a clobber. Well, it could _maybe_ be expressed as a clobber of "memory" (which makes the compiler assume it can clobber any or all values in memory). But I wouldn't rely on that, personally. However, I still doubt that volatile is needed on the atomic ops (such as atomic increment) themselves. It instead belongs specifically on the code that uses them for locking / unlocking. In other words, atomic increment can be used for many things, and most of them don't require volatile. Note, I'm coming in late on this thread, so it's entirely possible I'm preaching to the choir. :-) John -- John Polstra jdp@polstra.com John D. Polstra & Co., Inc. Seattle, Washington USA "Disappointment is a good sign of basic intelligence." -- Chögyam Trungpa To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-smp" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200012181740.eBIHeFw65943>