Hi gang,
 
I ran across an article from Systems and Software Magazine, June 1985, that
I thought would be interesting to share.  Apologies to copyright fan(atic)s:
 
The article begins:
 
Risc concept is debatable.
 
The RISC versus-CISC tradeoffs have been considered objectively by a group
of Carnegie-Mellon
University (Pittsburgh) computer scientists led by Robert P. Colwell.
 
The main problem, according to the group, is that no one has been able to
prove which factor is more
important for RISC performance: the reduced-instruction set, or the
multiple-register set.  The group
also questions the validity of benchmark test indicating RISCs generally
outperform CISCs.
 
The group believes the benchmarks may have compared apples to oranges,
because the tests used
different compiler technologies, virtual-memory techniques, and
operating-system overheads.
 
Other computer scientists contend that the massive register sets increase
the semantic gap and
slow the system.
 
In their opinion, register-intensive architecture leads to more LOAD/STORE
instructions - which in turn,
reduce speed because they require additional memory references.  The critics
also believe register-intensive
architectures make it harder for compilers to generate optimal code, because
registers are addressed
differently than is memory.  Moreover, the present trend toward modular,
separately compiled programs
reduces the amount of register optimization that compilers can perform, and
leads to more register LOADS
and STORES across subroutine boundaries.
 
The upshot is that the future of RISCs is promising but the outcome is
uncertain.
 
It is still debatable whether any studies have been done that accurately
compare the performance of RISC
and CISC cpu's.
 
All that is known so far is that comparative performance varies depending on
whether, for instance,
trancendental routines are written in assembly language, or Fortran with a
VMS compiler, or C language
with a highly optimizing C compiler.
 
End of article
 
 
I don't really have any particular comments to make about this except for an
observation.  I think that it is
interesting in these days that we are doing all kinds of prognostications
about the future of this and the future
of that.  My observation is that we are doing a lot of prognostication with
much *less* actual technical information
than our predecessors of eleven years ago had, and many of them were just
simply wrong.
 
Maybe the market will pull a "Que sera, que sera" and be what it will be.
---
Bill Lancaster         Lancaster Consulting
(541)926-1542 (phone)  (541)917-0807 (fax)
[log in to unmask]       http://www.proaxis.com/~bill