home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!auspex-gw!guy
- From: guy@Auspex.COM (Guy Harris)
- Newsgroups: comp.sys.sun.hardware
- Subject: Re: SPARCClassic (and other Sun announcments)
- Message-ID: <15589@auspex-gw.auspex.com>
- Date: 20 Nov 92 19:31:56 GMT
- References: <1992Nov11.190331.12893@eskimo.celestial.com> <15455@auspex-gw.auspex.com> <BxzItA.Mq8@pix.com>
- Sender: news@auspex-gw.auspex.com
- Organization: Auspex Systems, Santa Clara
- Lines: 24
- Nntp-Posting-Host: auspex.auspex.com
-
- >My guess is that the cache on the clasic is either larger or smarter
- >then the SS2...
-
- Larger? I doubt it. The SS2 has a 64KB unified cache; microSPARC has a
- 4K I-cache and a 2K D-cache, so unless they've stuck an external cache
- on the microSPARC machines (which seems unlikely; I don't think they did
- so), the cache is smaller, not larger.
-
- Smarter? Maybe, although its smarts would have to compensate for the
- smaller size. Then again, the microSPARC cache may be big enough for
- Dhrystone, so that the smaller size doesn't hurt....
-
- > or the compiler got alot smarter and the SS2 numbers are old...
-
- "Smarter" on Dhrystone would presumably be enough; note that the SPEC
- numbers for the microSPARC machines aren't much different from the
- numbers for the SS2.
-
- How many multiplies/divides does Dhrystone do? MicroSPARC, being a
- SPARC Version 8 implementation, has multiply and divide instructions;
- if the compiler used to compile Dhrystone generated those instructions,
- and they run faster than does the procedure call to the multiply/divide
- routines plus the routines themselves, that may explain some of the
- speedup. (Not necessarily all of it, though.)
-