From owner-freebsd-hackers Wed Aug 4 14:42:42 1999 Delivered-To: freebsd-hackers@freebsd.org Received: from alcanet.com.au (border.alcanet.com.au [203.62.196.10]) by hub.freebsd.org (Postfix) with ESMTP id E7F3814CED for ; Wed, 4 Aug 1999 14:42:37 -0700 (PDT) (envelope-from jeremyp@gsmx07.alcatel.com.au) Received: by border.alcanet.com.au id <40331>; Thu, 5 Aug 1999 07:22:53 +1000 Date: Thu, 5 Aug 1999 07:42:17 +1000 From: Peter Jeremy Subject: Re: Results of investigating optimizing calloc()... To: hackers@FreeBSD.ORG Cc: des@flood.ping.uio.no Message-Id: <99Aug5.072253est.40331@border.alcanet.com.au> Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG Dag-Erling Smorgrav wrote: >Which reminds me - has anyone thought of using DMA for zeroing pages, This sounds reasonable. Some DMA engines support filling regions and memory-memory copies, but I'm not sure about what can be done with the DMA engine(s) in PCs. > The idea is to keep a chunk of zeroes on disk and DMA it into memory Have you looked at disk latencies recently? A modern CPU could zero- fill a decent fraction of its RAM in the time taken to fetch a page of zeroes from the platter. And if it was accessed frequently enough to keep the zeroed page in disk cache, you've just moved the bottleneck into that disk controller (and you've reduced the effective size of the disk's cache by a page). Peter To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message