'Correctness of Dataflow

mstroebl@risc.uni-linz.ac.at (Markus Stroebl)
Wed, 29 Sep 1993 11:44:35 GMT

          From comp.compilers

Related articles
'Correctness of Dataflow mstroebl@risc.uni-linz.ac.at (1993-09-29)
| List of all articles for this month |

Newsgroups: comp.compilers
From: mstroebl@risc.uni-linz.ac.at (Markus Stroebl)
Keywords: dataflow, question
Organization: RISC, J.K. University of Linz, Austria
Date: Wed, 29 Sep 1993 11:44:35 GMT

Hello netusers, I've got the following problem: I am writing a compiler in
which it is posssible to 'describe' a program in a sense of dataflow:
There is a initial step, which has following step(s), each with a
condition. In the first, only the initial step is active. When now a
condition holds, the step with that condition is now active, the first one
inactive. There is the possibility of parallel branching and reuniting.
There must be a semantically check of this construction of data flow. As
it is quite similar to a CPM, I have to check if there is a unallowed
combination of steps and transitions (dead locks, steps that are never
reached, ..... ) (sounds hard, so here it is graphical).


Initial Step


/-----\
| |
\-----/
|
+
------------- parallel branching
-------------
|a1) |a2)
| |
| |
/-----\ ------------ alternative branching
| | +b) +c)
\-----/ | |
| | |
| | |
| | |
----------- | parallel uniting
----------- |
| |
                                        | |
                                        | |
                                        ------------------ alternative uniting
|
|
|


The flow goes via a1 and a2, and then either b or c. If it goes via c one
gets a deadlock. And this can only be checked semantically.


Thanks in advance for advice, or bibliographical help, Markus
--
Markus Stroebel RISC Linz, AUSTRIA
--


Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.