home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.sys.amiga.programmer:17336 comp.sys.amiga.hardware:21461
- Newsgroups: comp.sys.amiga.programmer,comp.sys.amiga.hardware
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!cs.utexas.edu!sdd.hp.com!col.hp.com!fc.hp.com!koren
- From: koren@fc.hp.com (Steve Koren)
- Subject: Re: CISC and RISC
- Sender: news@fc.hp.com (news daemon)
- Message-ID: <Bz98Mw.7AK@fc.hp.com>
- Date: Mon, 14 Dec 1992 15:06:32 GMT
- References: <Bz8FD1.Dxt@ns1.nodak.edu>
- Organization: Hewlett-Packard Fort Collins Site
- X-Newsreader: Tin 1.1.3 PL5
- Lines: 58
-
- > Why should computer makers go from CISC to RISC?
- >
- > I was under the impression that CISC meant complicated instruction set
- > and RISC meant reduced instruction set. So wouldn't that mean larger
- > code size on a RISC machine compared to CISC.
- >
- > What are the advantages of CISC and RISC?
-
- This probably isn't quite the right place to ask about this - maybe one
- of the comp.* architecture groups would be better.
-
- Anyway, briefly, the answer is "it depends". :-)
-
- Today's commonly available hot RISC chips (ie, DEC Alpha, HP-PA, etc.),
- are faster than today's common CISC chips (68040, 486, etc.), by a good
- margin. That's why manufacturers use them. But people obviously still
- buy CISC machines, even when a faster RISC machine is available, so
- there are obviously good reasons to use CISC.
-
- Anyway, there are popular RISC chips today which are slower than 68040s,
- and for all I know, there is some CISC chip out there faster than a fast
- DEC Alpha or HP-PA chip. (The 586 and 68060 should come close to if not
- surpass today's lower end Alpha and PA-RISC machines. Already a 68040
- comes within about 4X of an HP 720 for floating point performance).
-
- The "larger code size" thing you mention shows up a little bit, but
- often its not too bad. I just did a quick comparison some of 68K vs
- HP-PA binaries from the same source code, and the RISC version tends to
- be about 25% larger on average. However, the worst case was almost 3X,
- for a small binary. And its certainly architecture and compiler
- dependent. You may get different results from an IBM RS/6000, say.
-
- Some reasons a manufacturer might want to stay with CISC? Binary
- compatibility with old applications. That's a big one. Or if you're
- doing assembly programming, CISC chips tend to be much easier to deal
- with than RISC, which are only designed to be programmed via compiler.
- (Not that it can't be done, it is just less straightforward, and you
- have to remember funky rules like what happens if you have an
- unconditional branch in the delay slot of a conditional branch, and
- other far-out things). Also, if you currently have a 68K based design,
- its a whole lot easier to come up with a new 68060 based system than it
- would be to go to a RISC chip of any sort.
-
- Some reasons a manufacturer might want to go to RISC? Cutting edge
- single-chip performance. Probably some others I haven't thought of.
-
- BTW, "RISC" is a bit of a misnomer given the chips we label as RISC
- today.
-
- Assuming Motorola comes through with the 68060, I think Commodore won't
- have any compelling reason to move to RISC. 68K won't give cutting edge
- performance, but heck, it doesn't do that _now_. It will be good enough
- for a long time, though, especially for the PC world which doesn't tend
- to be performance-critical. After all, how many people out there still
- get along just fine with 68000s and 286s?
-
- - steve "if you want leading edge workstation performance from a PC
- class machine, buy 4 or 5 of them"
-