home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!inmos!wraxall.inmos.co.uk!frogland!des
- Newsgroups: comp.sys.sun.hardware
- From: des@inmos.co.uk (David Shepherd)
- Subject: Re: Is it Sunset yet?!?!
- Message-ID: <1993Jan7.104752.28485@wraxall.inmos.co.uk>
- X-Newsreader: TIN [version 1.1 PL8]
- References: <16269@auspex-gw.auspex.com>
- Date: Thu, 7 Jan 93 10:47:50 GMT
- Lines: 28
-
- Guy Harris (guy@Auspex.COM) wrote:
- : Maybe, maybe not. Those are, I think, Dhrystone MIPS; the SPEC92
- : numbers for the LX and Classic are closer to those for the SS2. I'm
- : curious whether Dhrystone is multiply-intensive or divide-intensive, in
- : which case the microSPARC may be faster because it has multiply and
- : divide instructions that run faster than the software version, or
- : whether it's due to compiler differences, or library differences, or
- : what.
-
- Dhrystone always used to be "copy null terminated string" intensive.
- (I think it was worse than that as the string being copying was a
- constant string). if you put in hardware support for strcpy then
- you got great results. when the t414 transputer came out some
- one here coded the dhrystone benchmark in occam to give some
- feel of the performance and got enourmously good figures (for that
- time that is) - the reason? occam represented (constant) strings
- using a representation that consisted of length and data so the
- string copy was handled purely by telling the transputers move instruction
- how many bytes to moves, which it would do at a rate of one word per
- cycle - hence very fast when compared with C strcpy which had to
- read each byte and compare against 0! great benchmark figures, but
- in this case fairly meaningless.
-
-
- --------------------------------------------------------------------------
- david shepherd: des@inmos.co.uk tel: 0454-616616 x 625
- inmos ltd, 1000 aztec west, almondsbury, bristol, bs12 4sq
- New Year Resolution for 1993: Start using capital letters.
-