Date: Mon, 30 Sep 2002 09:21:53 -0700 (PDT) From: Diane Bruce <db@db.net> To: gallatin@cs.duke.edu (Andrew Gallatin) Cc: freebsd-ppc@FreeBSD.ORG, jkh@FreeBSD.ORG Subject: Re: atomic.h vs atomic.s Message-ID: <200209301621.g8UGLrV28713@mtbaker.tfm.com> In-Reply-To: <15768.30116.381592.891890@grasshopper.cs.duke.edu> from "Andrew Gallatin" at Sep 30, 2002 12:02:44 PM
next in thread | previous in thread | raw e-mail | index | archive | help
Andrew Gallatin says: > > > When poking around, trying to figure out my mcl problem, I noticed > that the atomic.s code uses an eieio followed by a sync, whereas > the code in atomic.h uses just eieio. > > >From my (admittedly very limited) knowledge of ppc, I'd think we > should be using just sync (like netbsd does). Or it could be that > the atomic stwcx instruction implies a sync.. The biggest difference is, sync ensures all the loads and stores are synchronised. eieio ensures they happen in the right order. (enforce in-order execution of I/O) I think eieio is the favourite instruction of all powerpc coders. > should be using just sync (like netbsd does). Or it could be that > the atomic stwcx instruction implies a sync.. I suggest you get a copy of M961446248442collateral.pdf its easy to find on the Motorola site. pg. 1-11 for a start. > > Jordan: Are we allowed to use the Darwin sources as a reference for things > like this? > > > Drew -- Diane Bruce, http://www.db.net/~db db@db.net --- aphorisms are ephemeral, speling mistakes are not. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-ppc" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200209301621.g8UGLrV28713>