home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.arch
- Path: sparky!uunet!utcsri!newsflash.concordia.ca!sifon!thunder.mcrcim.mcgill.edu!homer.cs.mcgill.ca!storm
- From: storm@cs.mcgill.ca (Marc WANDSCHNEIDER)
- Subject: Re: RISC goes CISC?
- Message-ID: <1992Nov8.193946.2210@cs.mcgill.ca>
- Sender: news@cs.mcgill.ca (Netnews Administrator)
- Organization: SOCS - McGill University, Montreal, Canada
- References: <1992Nov6.092012.19239@rhein-main.de>
- Date: Sun, 8 Nov 1992 19:39:46 GMT
- Lines: 66
-
- In article <1992Nov6.092012.19239@rhein-main.de> vhs@rhein-main.de writes:
- >Wasn't one of the principles of RISC the principle of putting as much work
- >as possible from execution to compile time?
-
- Only indirectly perhaps. The basic philosohpy is that if the
- instruction isn't used very frequently, then there is little reason to include
- it in the instruction set.
-
- However, it is entirely possible to get a little overzealous and
- reduce instructions to the point where they would be TOO small, and compilers
- would end up making the "Hello World" program 100k or so (exaggeration).
-
- >Under this assumption, how does e.g. SuperSPARC fit this principle? As far
- >as I know the processor does a lot of cross-checking between pipelines,
- >squashing instructions depending on non-trivial conditions, etc. Is that done
- >just to maintain compatibility with previous implementations?
-
- This is strictly a hardware IMPLEMENTATION detail, and doesn't really
- affect the programmer instruction set or model. The problem with the Super-
- Sparc is that it tries to issue three instructions per clock cycle. If there
- are dependancies between the instructions, then problems will arise, and
- special hardware must be added to the processor to take whatever actions
- are necessary so that the meaning of the program remains intact.
-
- >Could, theoretically, the most recent compilers take care of this
- >at compile-time, thus eliminating the need for run time
-
- Compilers are EXTREMELY difficult to write, and can only do a good
- job at best of reordering instructions so that pipeline dependancies are
- avoided. It is also enitrely possible that a situation will arise where
- it is just impossible to remove dependancies....
-
- >Aren't pipelines necessary so that instructions that take several cycles
- >(i.e. "complicated" instructions) can be issued one per cycle? Wouldn't
-
- Pipelines are one way of achieving the Ideal 1 instruction issued
- every clock cycle.
-
- >it conform more to the RISC principle to keep instructions so simple
- >that they only *need* one cycle to execute and concentrate on a *fast*
-
- Given current technologies, it could end up being EXTREMELY difficult
- to make instructions execute in one clock cycle, if the clock cycles starting
- dropping below a certain time. There is just so much that goes on, and gates
- can only go so fast. Thus, the compromize is to try and overlap execution.
-
- >Will PCs ever compare favorably to workstations?
-
- Intel is slowly finding that the CISC way of doing things is not
- always optimal. Pipelining a CISC chip is painstaking brutal, and they
- are thus focusing their efforts on the new Pentium chip, which will largely
- be a RISC chip which can execute the 486 instruction set if need be.
-
- >Will Jane marry Tarzan?
-
- No, she ran off to Vegas with a Lesbian Biker Lover named Despina.
- Last I heard, they were rumored to be Mud-Wrestling at the Hotel De Sleeze.
-
-
- ToodlepiP!
- Marc 'em.
- --
- storm@cs.mcgill.ca McGill University "-- Attack ships on
- Marc Wandschneider Montreal, CANADA fire "
- Any opinions expressed are not mine, but those of the Demon Lord
- Yeegeheeegenogohugu who possessed me whilst I munched on Raisin Bran.
-