From owner-freebsd-alpha Sun Jun 27 15:23:39 1999 Delivered-To: freebsd-alpha@freebsd.org Received: from overcee.netplex.com.au (overcee.netplex.com.au [202.12.86.7]) by hub.freebsd.org (Postfix) with ESMTP id 4C68214CA1 for ; Sun, 27 Jun 1999 15:23:33 -0700 (PDT) (envelope-from peter@netplex.com.au) Received: from netplex.com.au (localhost [127.0.0.1]) by overcee.netplex.com.au (Postfix) with ESMTP id 8D80781; Mon, 28 Jun 1999 06:23:32 +0800 (WST) (envelope-from peter@netplex.com.au) X-Mailer: exmh version 2.0.2 2/24/98 To: Dmitrij Tejblum Cc: Doug Rabson , alpha@FreeBSD.ORG Subject: Re: Hmm!! In-reply-to: Your message of "Mon, 28 Jun 1999 01:41:29 +0400." <199906272141.BAA03092@tejblum.dnttm.rssi.ru> Date: Mon, 28 Jun 1999 06:23:32 +0800 From: Peter Wemm Message-Id: <19990627222332.8D80781@overcee.netplex.com.au> Sender: owner-freebsd-alpha@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.org Dmitrij Tejblum wrote: > Peter Wemm wrote: > > I went a bit further and it seems to have fixed my mystery problems: > > One line in your program completely change its behavior, and you claim > several lines in the kernel fix it? ;-) Cool! ;-) I was suspicious of something along those lines but the Alpha system is quite new to me. I also kept thinking about Jason Thorpe's "you aren't doing all the required imb's" comment before. I also note that DEC/Compaq have explicitly said that their upcoming CPU's are far more strict about certain things, I wonder if this is one? This one isn't a particularly new CPU though. > > [...] > > I'm pretty sure this is overkill, but it doesn't seem to hurt so far. I > > have not tried your single patch though, I got out the sledgehammer to see > > if I could get some sanity. I'm not sure if the alpha_pal_tbi() functions > > do an implied imb as well or not. If so, then the first chunks are > > irrelevant. > > I don't think so. To which part? :-) > While your patch is overkill, it still seem to miss imb in > pmap_enter_quick()... Yeah, seems so.. pmap_enter_quick() is, by definition, used on the current pmap. We only need alpha_pal_tbi() calls when a mapping is being changed or removed, yes? pmap_enter_quick() only seems to be used for preinitializing pt's and for page fault clustering.. > Dima Cheers, -Peter To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-alpha" in the body of the message