From owner-freebsd-hackers Mon Apr 30 21:18: 3 2001 Delivered-To: freebsd-hackers@freebsd.org Received: from silby.com (adam042-060.resnet.wisc.edu [146.151.42.60]) by hub.freebsd.org (Postfix) with ESMTP id A38C537B61A for ; Mon, 30 Apr 2001 21:17:59 -0700 (PDT) (envelope-from silby@silby.com) Received: (qmail 21325 invoked by uid 1000); 1 May 2001 04:17:58 -0000 Received: from localhost (sendmail-bs@127.0.0.1) by localhost with SMTP; 1 May 2001 04:17:58 -0000 Date: Mon, 30 Apr 2001 23:17:58 -0500 (CDT) From: Mike Silbersack To: Jan Mikkelsen Cc: Freebsd-Hackers Subject: Re: write() vs aio_write() In-Reply-To: <00bd01c0d1d1$106ccdf0$0901a8c0@haym.transactionsite.com> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG On Tue, 1 May 2001, Jan Mikkelsen wrote: > Mike Silbersack wrote: > [ On using aio on disks vs. sockets ] > >Sockets already support non-blocking IO, and have for a long while. > >Hence, the socket code is probably more optimized for non-blocking > >operation than AIO operation. As a plus, using non-blocking socket > >operations will allow your code to run on any platform; aio isn't as > >portable. > > I recall reading about possible zero copy I/O using the aio interface. Is > anyone thinking about this? And on a related note, how about something like > IRIX's O_DIRECT mode for files? > > I'm sure there are lots of issues, but I'm curious. > > Jan Mikkelsen I think the zero-copy patch for FreeBSD sets the page of the written COW and can thereby work with the standard write call. (Assuming that you don't re-write the page soon after doing the write.) Mike "Silby" Silbersack To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message