🚀 go-pugleaf

In <142@autro1.UUCP> teig@autro1.UUCP (Oyvind Teig) writes:

>Recursion of occam-2 or -3

>I have seen enough of stack overflow errors to appreciate the lack
>of recursion in occam.

Recursion + stack-checking -> no stack overflow.

>Rob Kurver writes:
>> It is very well possible to introduce stack checking code during linking,
>> when a call-graph of the program can be used to determine if and when
>> it is possible.  This would result in ZERO overhead for a program written
>> without recursion, and minimum overhead for recursive programs.  We
>> are currently working on this feature in PACT Parallel C.

>Ok, but you still have not removed the stack overflow!

When I say "stack checking" I mean "stack checking plus expansion if
necessary".  By checking the stack and expanding it if necessary, you
remove the stack overflow.  The linker can build a call-graph for the
entire program, so it knows which functions call which other functions.
This allows it to detect recursion.  All non-recursive branches need
only have enough stack allocated once, at the start of the branch (i.e.
at process or program creation).  The recursive branches, if any, need
to have stack checking (+ expansion) code introduced in the loop
(unless you just want to restrict the depth at linktime, of course).

This way, there's no overhead if you don't use recursion, and there is
a small overhead if you do use recursion.  You don't have stack
overflows in any case.

>This comes close to the discussion of exception-handling: how do
>you solve that in Pact-C, and what would you recommend when stack overflow
>is encountered?

Upon stack overflow, the stack is extended by allocating a new block of
memory from the heap.  If this is not possible (out of memory), you get
a fatal error message and your program aborts.

>If occam had been written from the philosophy that "if you don't need it,
>don't use it", it definitively would have not been occam.

I guess it would not have been Occam-2, no.  It might have had data
structures :-).

>Is it possible to do multi-language design with the Inmos D7205 occam
>toolset and your Pact C-compiler, like it is with the Inmos C-toolset?

Yes.  You can link Occam (TCOFF) modules with PACT Parallel C modules.

>For me as an occam-user, your parallel C (:-) is of little interest because
>C is not interesting, if it were, I would not have used occam (recursive
>definition: interesting but little useful(%-|)

I C :-).  A lot of people do find C interesting, though.

>0yvind Teig
>Autronica, Trondheim, Norway

Rob

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