Systolic Nets Modelling with Data Flow Petri Nets
A systolic machine is a net of elementary processors in which data circulates with a constant flow. The word systol, usually used by physiologists to describe the rhythm of heart contractions, is employed to show the analogy between the operations of the heart and that of the systolic machine. Indeed, each processor pumps data from its entry to its exit, while carrying out certain operations, so that we obtain a regular data flow through the net.
The communication and calculation control in the processor is very simple. The memory place employed is little and constant, independently of the net size.
This kind of machine is interesting for two reasons: they offer an easy implementation and do not need complex commands, then they can be connected to general use machines, like Data Flow machines.
Among the possible approaches to systolic structure we have chosen their modelling with Data Flow Petri Nets. When a calculation is to be solved, first we work out calculation parallelism, then we model with Data Flow Petri Net and we synchronize the transitions.
Unable to display preview. Download preview PDF.
- 1.J. Almanah, Modelisation par reseaux de Petri a flux de donnees. Application a la synthese de l’operateur de Riccati rapide. These d’Etat — Marseille (1983).Google Scholar
- 2.G.W. Brahms, Reseaux de Petri, Theorie et pratique. Tomes 1 et 2. Masson (1983).Google Scholar
- 3.H.T. Kung, “Why systolic architecture”, IEEE Computer (1982).Google Scholar
- 4.A. Giulieri, H. Malki, Etude d’un operateur de Lyapunov a architecture systolique, LAIAT Toulon, IASTED (19–21 Juin, 1985).Google Scholar
- 5.A. Giulieri, G. Nolibe, Etude et realisation d’operateurs specialises multifonctions et application aux reseaux systoliques. Rapport interne LAIAT Toulon 83.268.Google Scholar
- 6.P. Quinton, The systematic design of systolic arrays. IRISA Report 193, (1983).Google Scholar