home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.benchmarks:1865 comp.arch:11599
- Newsgroups: comp.benchmarks,comp.arch
- Path: sparky!uunet!spool.mu.edu!uwm.edu!zaphod.mps.ohio-state.edu!usc!enterpoop.mit.edu!eff!ssd.intel.com!ichips!ichips!glew
- From: glew@pdx007.intel.com (Andy Glew)
- Subject: Re: VMEbus PCs
- In-Reply-To: adams@pdv2.fmr.maschinenbau.th-darmstadt.de's message of Thu, 10 Dec 1992 19:52:26 GMT
- Message-ID: <GLEW.92Dec12151952@pdx007.intel.com>
- Sender: news@ichips.intel.com (News Account)
- Organization: Intel Corp., Hillsboro, Oregon
- References: <WAYNE.92Nov18185710@backbone.uucp> <1992Nov25.104313.12274@cadlab.sublink.org>
- <wilson.723435964@moonshine> <id.7ZLV.TF2@ferranti.com>
- <ADAMS.92Dec9012224@PDV2.pdv2.fmr.maschinenbau.th-darmstadt.de>
- <GLEW.92Dec9215057@pdx007.intel.com>
- <ADAMS.92Dec10195226@PDV2.pdv2.fmr.maschinenbau.th-darmstadt.de>
- Date: Sat, 12 Dec 1992 23:19:52 GMT
- Lines: 52
-
-
-
- Up to know, we did not see a trustworthy implementation
- of READ-MODIFY-WRITE cycles by VMEBus PCs, though
- VMEBus provides it, and "lock" is a valid prefix for operations
- on x86.
-
- Many VMEbus SBCs do not provide an entirely trustworthy
- implementation of READ-MODIFY-WRITEs.
-
- "Lock" is only a recent addition to VMEbus. It was not part of the
- standard for a long time (in fact, I'm not even sure that it is now
- part of the standard).
-
- Several machines use tricks like maintaining a continuous address
- strobe for RMWs that access a single location. Of course, this means
- that 68000 instructions that access more than one location, like
- compare and swap double, don't work.
-
- And, of course, the address strobe tricks reduce performance for
- secondary busses, like VSB.
-
-
- Moral: locking of data on VMEbus has very much been a catch as catch
- can world. See my synchronization survey for details.
-
-
-
-
-
-
- > Andy Glew, glew@ichips.intel.com
- > Intel Corp., M/S JF1-19, 5200 NE Elam Young Pkwy,
- ^^^^^
- BIASED???
-
- Of course you may think I'm biased (although I try not to be). That's
- why I put my company in my .signature -- so you can account for it.
-
- By the way, before Intel I worked for Motorola's VMEbus group.
-
-
- --
-
- Andy Glew, glew@ichips.intel.com
- Intel Corp., M/S JF1-19, 5200 NE Elam Young Pkwy,
- Hillsboro, Oregon 97124-6497
-
- This is a private posting; it does not indicate opinions or positions
- of Intel Corp.
-
- Intel Inside (tm)
-