Newsgroups: comp.sys.transputer
From: roger@pact.srf.ac.uk (Roger Shepherd)
Subject: Re: CJ comments
Organization: University of Bristol, England
Date: Tue, 12 Apr 1994 14:06:28 GMT
Message-ID: <Co5GIs.BDG@info.bris.ac.uk>

Kirk Bailey (bailey@fix.fenris.com) wrote:

: I'm not sure about the original design reason for this although I recall
: hearing that it was somehow related to the code generation strategy of an
: early OCCAM backend.  Perhaps someone who worked at INMOS at the time could
: comment?

Early versions of the transputer instruciton set had a conditional jump
instruction which was JUMP TRUE. This popped a TRUE (0) and kept the
non-zero value. At a late stage the sense of the jump was changed to match
what the compiler writers wanted. In making this change the event which caused
the Areg to be popped remained the same (jumping) - this preserved some
parts of the generation of contional expresison evaluation. However, the 
change in value created some problems which were just not picked up at the
time. A mistake.
--
Roger Shepherd

Email: roger@inmos.co.uk
Mail : 10 Priory Road, Clifton, Bristol, BS8 1TU. UK.
Phone: +44 272 707 189, Fax +44 272 707 171.


