Date: Wed, 17 Jul 1996 14:00:18 -0400 From: dennis@etinc.com (Dennis) To: "Michael L. VanLoon -- HeadCandy.com" <michaelv@HeadCandy.com> Cc: hackers@freebsd.org Subject: Re: mitsumi CD-ROM Message-ID: <199607171800.OAA24031@etinc.com>
next in thread | raw e-mail | index | archive | help
>>To no-doubt spurn another wide-eyed debate among the academics, I also >>dont want a bus-mastering controller stealing the bus from my more-critical >>communications in a router-system where disk functions are secondary. > >This makes no sense. The only time a bus-mastering controller >"steals" the bus is when you're doing I/O. An IDE drive not only >steals the bus by implication, but it steals the CPU as well! >Basically, IDE steals the entire machine until the I/O is finished. >At least with a bus-mastering controller, the CPU can run out of the >cache while a disk transfer is in progress. > >I can understand your opinion on this, as stated above. But to think >that you're not getting the bus, and the CPU, stolen, when >bus-mastering would steal only the bus, is a complete >misunderstanding. stealing the bus is stealing the cpu when a router is doing largely transfers from cards. Your cant process packets until you get them. And if the cpu is in a transfer cycle and cant get the bus then its idle. I dont want my packet transfers being interrupted by trivial disk activity. An IDE will only use the CPU when it has it already anyway.....a bus-master makes the system much less predictable and assumes the wrong priorities. db ---------------------------------------------------------------------------- Emerging Technologies, Inc. http://www.etinc.com Synchronous Communications Cards and Routers For Discriminating Tastes. 56k to T1 and beyond. Frame Relay, PPP, HDLC, and X.25 for BSD/OS, FreeBSD and LINUX
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199607171800.OAA24031>