From owner-freebsd-arch Mon Oct 14 18:14:37 2002 Delivered-To: freebsd-arch@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id E17E637B401 for ; Mon, 14 Oct 2002 18:14:35 -0700 (PDT) Received: from canning.wemm.org (canning.wemm.org [192.203.228.65]) by mx1.FreeBSD.org (Postfix) with ESMTP id 9A98143EB3 for ; Mon, 14 Oct 2002 18:14:35 -0700 (PDT) (envelope-from peter@wemm.org) Received: from wemm.org (localhost [127.0.0.1]) by canning.wemm.org (Postfix) with ESMTP id 763AE2A88D; Mon, 14 Oct 2002 18:14:35 -0700 (PDT) (envelope-from peter@wemm.org) X-Mailer: exmh version 2.5 07/13/2001 with nmh-1.0.4 To: Terry Lambert Cc: Wes Peters , ticso@cicely.de, "M. Warner Losh" , hch@infradead.org, dillon@apollo.backplane.com, vova@sw.ru, nate@root.org, arch@FreeBSD.org Subject: Re: Database indexes and ram In-Reply-To: <3DAA695F.96E345E6@mindspring.com> Date: Mon, 14 Oct 2002 18:14:35 -0700 From: Peter Wemm Message-Id: <20021015011435.763AE2A88D@canning.wemm.org> Sender: owner-freebsd-arch@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG Terry Lambert wrote: > Peter Wemm wrote: > > And then there's the AGP remap table stuff to provide a window from > > anywhere in memory into the lower 4G of space that's within reach of 32 bit > > PCI devices... > > Peter, you are sneaky... in a good way. > > Do you really think this could be abused this way? How much, in the > way of windows, can you have lying around at one time? I'm not very > AGP-aware, other than "I need it for my X server". I cant claim credit for this.. The linux folks supposedly have it working - David O'Brien told me about it. Anyway, with this you then are required to have the kernel manage the aperture space. For a desktop, the real question is whether you want to slow down the agp-using X server in order to avoid bounces by using resources that could otherwise be used for doing graphics. But for a server with no agp card in use, it would presumably be a no-brainer. The way that I understand that it works is that the northbridge/memory controller/whatever provides an alternative "view" of system memory. > If you could have one per outstanding request to a card, that would > be enough, I think... I believe this is a function of the agp/memory subsystem (which often share the same silicon), not the graphics card itself. I mentioned this before and one of my misunderstandings was corrected. I'm still not sure that I have the details correct. If you read some of the pci/agp_* drivers (I read agp_amd.c), you will see that the chips have things like page tables, tlb's etc. And you can connect arbitary memory pages to the mapping tables to make them visible in the aperture (below 4G) so that busmaster devices can get to them. I'm not entirely sure what the graphics subsystem *does* with this stuff, but I assume it is useful for something. Cheers, -Peter -- Peter Wemm - peter@wemm.org; peter@FreeBSD.org; peter@yahoo-inc.com "All of this is for nothing if we don't go to the stars" - JMS/B5 To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-arch" in the body of the message