🚀 go-pugleaf

In article <6348@sersun1.essex.ac.uk> aghao@sersun2.essex.ac.uk () writes:
>Greetings.
>
>As part of my research, I need to be able to examine the workspace of one or 
>more Transputer processes.  Clearly, this is quite feasible, as I can get hold
>of the Wptr of a process.
>
I tried to mail you but it bounced (unknown address) so ...
I would be interested in your discussions on this topic, if (or when) feasable,
could you mail/email some of your results?

thanks in advance
Brian d'Auriol
University of New Brunswick  
  

In <6348@sersun1.essex.ac.uk> aghao@sersun2.essex.ac.uk writes:

>As part of my research, I need to be able to examine the workspace of one or 
>more Transputer processes.  Clearly, this is quite feasible, as I can get hold
>of the Wptr of a process.

>However, as I need to access the workspaces as separate, individual entities,
>I need to know what the boundaries are, ie where each particular process' 
>workspace starts and *ends*.  One possibility is to find out what the size
>of each workspace is, and that is what I am hoping to be able to do.

It is not entirely clear to me what you're looking for, i.e. how you
define the start and end of the workspace.  Are you interested in the
memory reserved for the entire stack of the current thread, or just in
the frame used by the current function invocation?

If the latter, you know the start by loading the Wptr (something
like `__asm ("ldlp 0")'), and you'll have to know a bit about your
compiler system to find out how to find the end.  In general, this
is were any arguments passed to this function are stored, so there
may be some tricks to get this info.  But again, this will most likely
differ (if only slightly) between compilers.  Please send me email
if you'd be interested in the solution for our PACT Parallel C compiler.

>Does anyone out there have any suggestions?  Most of my work is done using 
>GENESYS C, although suggestions with ref. to any other language would be 
>gratefully received.

If you're interested in finding out the start and end of the current
stack for this thread, the solution will depend very much on your
compiler or operating system, as the transputer process itself
stores no such information.  Most decent transputer compilers for
modern languages like C do stack checking in order to allow for
recursion, for which they need to pass some kind of thread
descriptor to each function.  This thread descriptor contains
information about the start and end of the current stack, as well
as info needed to get back to previous chunks of stack.  This info
is most likely given in your compiler documentation.

>	Obi Aghanya

Rob Kurver
PACT Parallel Architecture and Compiler Technology

--
     PACT                   Rob Kurver
    Foulkeslaan 87         rob@pact.nl
   2625 RB Delft     ph: +31 15 616864 
  The Netherlands   fax: +31 15 610032