Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/5/84; site terak.UUCP
Path: utzoo!linus!philabs!cmcl2!seismo!hao!noao!terak!doug
From: doug@terak.UUCP (Doug Pardee)
Newsgroups: net.arch
Subject: Re: Cached vs uncached MMUs
Message-ID: <666@terak.UUCP>
Date: Tue, 6-Aug-85 12:20:47 EDT
Article-I.D.: terak.666
Posted: Tue Aug  6 12:20:47 1985
Date-Received: Sat, 10-Aug-85 23:02:07 EDT
References: <5374@fortune.UUCP> <268@gcc-bill.ARPA> <563@sftig.UUCP>
Distribution: net
Organization: Calcomp Display Products Division, Scottsdale, AZ, USA
Lines: 11

> You pay to load the page table when you context switch with either system. 
> ...  The reference to main memory to
> load the DAT box cache line cost just as much as the reference to main
> memory to load the page table entry.
 
With the cached approach (IBM DAT or NS32082) you don't also have to pay
for the instructions the CPU would have to execute to load a page table.

Also, it takes one memory cycle to load the cache, not two (load/store).
-- 
Doug Pardee -- CalComp -- {seismo!noao,decvax!noao,ihnp4}!terak!doug