KOSH [Kommunity Orientated Software Hardware]

Weekly Summary

Week:  22nd February 1999

Number: 007

Mailing List:  kosh-hardware-o

In the mailing list this week, the following items were discussed.
Please do not email the scribe regarding any of these topics, it is 
not his job to answer these questions, but merely to report the topics
of conversation. If you have any queries about this summary, please 
email summaries@kosh.convergence.org, stating the Summary Number, and
Mailing List Name and he will try to answer your queries.

a)

Subject:  Nearest Neighbors in Distributed Processing

Summary of Debate:  If distributed processing is being done over a
		    network, the topology of that network would
		    determine the set of "nearest neighbors."

b)

Subject:  Hardware Level CPU Independence

Summary of Debate:  The proposed model works in the oposite manner
		    from the one which it is believed that Transmeta
		    is designing.  It is believed that Transmeta is
		    designing a chip that can run native code from
		    various other processors.  In the proposed model
		    code would be in a specially designed instruction
		    set, and then at the hardware level translated
		    into the native code of whatever processor is
		    currently installed in the system.  (Note that
		    this translation doesn't have to be physically
		    done in hardware, as long as the rest of the
		    system, inlcluding the OS, view this translation
		    as a hardware level process.
		          Questions were raised about whether any
		    hardware level CPU independence implementation
		    would provide tolerable performance when compared
		    to native code.  One potential difficulty would
		    be the varying number of registers in various
		    CPU's.  One potential solution to the register
		    issue would be to create a "standard model" and
		    when a CPU is in use that has less registers
		    than the standard, emulate the missing registers
		    in "on board" memory.	  

c)

Subject: Hetereogeneous distributed processing

Summary of Debate:  One difficulty would be data being stored in
		    differenct formats on various differing 
		    prossessors.  To compensate for this some
		    standardized data format would be needed, to
		    enable passing data from one CPU to another.