home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.electronics
- Path: sparky!uunet!utcsri!torn!nott!bnrgate!bcars6a8!bcarh1d7!clewis
- From: clewis@ferret.ocunix.on.ca (Chris Lewis)
- Subject: Re: 68000 and Cache?
- Message-ID: <1992Nov16.152737.23352@bcars6a8.bnr.ca>
- Originator: clewis@bcarh1d7
- Keywords: 68000
- Sender: usenet@bcars6a8.bnr.ca (Use Net)
- Nntp-Posting-Host: bcarh1d7
- Organization: fo
- References: <1992Nov6.225506.1973@nntp.hut.fi> <1dnt1fINNk0f@uranium.sto.pdb.sni.de> <1992Nov12.222122.12200@nntp.hut.fi>
- Date: Mon, 16 Nov 1992 15:27:37 GMT
- Lines: 28
-
- In article <1992Nov12.222122.12200@nntp.hut.fi> sakaria@vipunen.hut.fi (Sakari Aaltonen) writes:
- >In article <1dnt1fINNk0f@uranium.sto.pdb.sni.de> Josef Moellers <mollers.pad@sni.de> writes:
- >>You don't need any provisions for a cache, i.e. none in particular.
- >>All You need is a means to temporarily suspend operation of the CPU
- >>while You process a cache miss.
-
- >Yeah, well, I do know in a general way how caches operate. What I
- >don't know is how they are implemented with a 68000. What kind of
- >hardware will do the job, that sort of thing. Are any IC's available
- >for the purpose, for example? What do the accelerators use? I've
- >never seen one...
-
- Many of the dynamic RAM controller chips can do this. It's been a while,
- so I don't remember the numbers. This is the same as in the 80?86 community,
- so if you can find a 386 with caching (say), you can find out some of the
- part numbers.
-
- Should point out, however, that if you're talking about disk caching (or anything
- else that requires CPU intervention) that you can't do it without a lot of
- grief. The MC68000 has no way to save memory access state, so cannot be
- used to service the cache miss. The MC68010 and above *do* have this
- capability. (In essence, the 68000 doesn't have instruction restart or
- continue)
-
- Back in the old days, if you wanted to do demand paging with a 68000, you
- had to put the 68000 into HALT, and have an auxiliary processor wake up,
- service the page fault, and unHALT the 68000. Obviously, you don't
- get to timeslice processes during the fault processing...
-