Newsgroups: comp.sys.transputer
From: jan@neuroinformatik.ruhr-uni-bochum.de (Jan Vorbrueggen)
Subject: Re: jump destination anomalies?
Organization: Institut fuer Neuroinformatik, Ruhr-Universitaet Bochum, Germany
Date: 18 Jan 94 17:19:57 GMT
Message-ID: <JAN.94Jan18181957@thalia.neuroinformatik.ruhr-uni-bochum.de>

Most of these jumps are spurious. In those places, occam most often stores
constants (eg, of floating point numbers), or those are the jump table stubs
for library routines. Debug info (such as routine names) might also be in
there - this stuff is generally before the entry point of routines. Also, all
entry points are word-aligned.

The easiest way to debug/decode such a thing is to look for matching 
ajw <-n> / ajw <n> pairs. The ajw <-n> is the entry point, and the ajw <n> the
next-to-last instruction (followed by a ret :-)) of a routine. The code in the
middle should decode properly, and you can find those constants by evaluating
the pointer generated to them, usually by a ldpi <-n> near the routine's entry
point, which gets dumped somewhere in the workspace. Do you need templates for
the parallel process contructs? (PRI PAR has a nice little feature in its
trail-out code...)

HTH, Jan

