[comp.sys.transputer] R: TDS Questions, and .....

zenith-steven@CS.Yale.EDU (Steven Ericsson Zenith) (02/09/90)

In article <268:vorbrueggen@mpih-frankfurt.mpg.dbp.de> 
                vorbrueggen@mpih-frankfurt.mpg.dbp.DE writes:
>Hello Transputer wizards,
[lots of fun technical questions, better left to those that know better
than I]

No time to provide technical support on Inmos' behalf these
days, so I'll leave all the above for someone at Inmos to answer,
... if they have the time. Yes, it's hard to believe I know, but
people at Inmos do read this group. Come on Roger! Do the man a 
favour.

However ..

> ... (Why dup wasn't there
>from the start, but was added as an afterthought to the T800, would
>probably make an amusing historical anecdote. Does anybody know the
>reason?)
>

Well, the rationale ran something like this: here's an instruction set
to enable the implementation of high level programming languages like
Occam. Now, there was a very clear expression evaluation model for use
by the Occam compiler which excluded dup , nor was there considered to
be a need for it to support other constructs in the language. It
always puzzled me.

NAAA, you wouldn't want to write a compiler for any "conventional"
(spit) language, would you? Or do things any other <<your>> way? 

It was very much the view that everything was going so much faster
anyway that we could afford to ignore any inefficiencies in the model.
However, this is not very pragmatic, and in the event was arrogant and 
foolish. In the end dup got smuggled in by silicon designers (Guy or Fred
in think) ignoring the architects and presenting them with a f-a-c. 

Generally some flawed statistical analysis here, same finger in the air
stuff that meant the device had a bloody slow shift, and why you'll
probably get a four cycle mul instead of the 1-2 cycle mul you'd
prefer (and the 1 cycle INMOS <should> work for).

Regards,
Steven

                                                                        .
Steven Ericsson Zenith              *            email: zenith@cs.yale.edu
Department of Computer Science      |            voice: (203) 432 1278
Yale University 51 Prospect Street New Haven CT 06520 USA.
     "All can know beauty as beauty only because there is ugliness"