Banach R., MONSTR I Fundamental Issues and the Design of MONSTR. Submitted to New Generation Computing, (1993).  Home/Search   Document Details and Download   Summary   Related Articles   Check

....for executions involving redexes permitted to overlap sufficiently to allow the programming of deterministic synchronisations, and despite the failure of exact subcommutativity, a ChurchRosser theorem is proved to hold up to markings and garbage. 1 Introduction In the first MONSTR paper ( Banach (1996b)] hereafter referred to as M I) we introduced the MONSTR generalised term graph rewriting language, together with its operational semantics, and the architectural rationale behind its design. We also briefly described some other semantic models for MONSTR and the correctness problems that they ....

Banach. R., MONSTR I Fundamental Issues and the Design of MONSTR. J.UCS, 2, 164-216, (1996). http: //www. iicm. edu/jucs

....can signal their completion to their parents. MONSTR Syntactic Restrictions To make the above a computational model suited to distributed machines, a number of restrictions are imposed on the syntactic structure of systems so that some useful runtime properties hold. We paraphrase from [2], where there is a thorough study of why these are appropriate. Alphabets and Symbols. The alphabet of symbols S, is the disjoint union of three subalphabets S = F C V where: F is the alphabet of function symbols which may label the root of the LHS L of a rule, but not any subroot node of L, ....

....of each redex, depend solely on the symbol at the root (and so can be delegated to simple hardware) No pointer equivalence is required for matching any redex node, that is not evident from Map(s(root) ditto) All execution graphs are balanced and state saturated. The overwriting lemma ([2] Lemma 5.10) applies to most redirections, in practice enabling the convenient representation of rewriting by packet store manipulations, and particularly the representation of redirection by packet overwriting) 1. This explains the MONSTR acronym. It stands for: a Maximum of One Non root ....

[Article contains additional citation context not shown here]

R. Banach, "MONSTR I --- Fundamental Issues and the Design of MONSTR", submitted to the Journal of Universal Computer Science, 1995.

....the above ideas into a computational model suited to distributed machines, a number of restrictions are imposed on the syntactic structure of systems so that some useful runtime properties can be rigorously demonstrated. We point out the main ones now rather informally, referring the reader to [4,5] for a thorough study (in the context of the formal semantics of MONSTR) of why these are appropriate and what their consequences are. All nodes respect the arities of their symbols. Within rules; and by means of a simple induction, within all execution graphs. The alphabet of symbols is ....

....rules when either would match ( is the sequential rule selector in concrete syntax) Note that we have said nothing very specific about garbage collection. The general idea is that active and Root labelled nodes are live, and liveness is propagated down normal arcs and up notification arcs; see [4,5] for a more precise discussion. The implicit mark scan strategy that such a scheme embodies can be considerably simplified when we restrict to a linear subset (see below) 3 Interaction Nets Interaction Nets were invented for describing finegrained computations graphically. Their theory builds on ....

[Article contains additional citation context not shown here]

Banach R., "MONSTR I --- Fundamental Issues and the Design of MONSTR", Journal of Universal Computer Science, 2 (4), 1996, pp. 164-216, http://www.iicm. tu-graz.ac.at/jucs.

....are proved about the independence of suspending primitive actions, which are useful in the verification of MONSTR systems. Key Words: Intermediate Languages, Term Graph Rewriting, MONSTR, Semantic Models. Category: D.1.3, D.3.1, F.3.2, F.4. 2 1 INTRODUCTION In the first paper in this series, [Banach (1996)] we introduced the MONSTR generalised term graph rewriting language, a sublanguage of the DACTL language, and the architectural rationale behind its design. We also briefly described some other semantic models for MONSTR and the correctness problems that they engender when soundness with respect ....

....is when two execution steps of the model may be interchanged. This is needed when discussing serialisability properties of arbitrary executions of systems. The rest of this paper is structured as follows. Section 2] reviews the salient material from M I, the first paper in this series, [Banach (1996)] Although the present paper is self contained, a reasonable working knowledge of M I will be of benefit to the reader as we do not repeat every useful detail from M I in order to prevent the size of these Journal of Universal Computer Science, vol. 3, no. 7 (1997) 755 801 submitted: 21 09 95, ....

Banach. R., MONSTR I --- Fundamental Issues and the Design of MONSTR. J.UCS, 2, 164-216, (1996).

....[Milner (1989) to allow networks of processes to evolve dynamically. It is thus a process algebra language. MONSTR by contrast is a generalised term graph rewriting language that was used as the intermediate language for the Flagship machine. See [Banach et al. 1988) Banach and Watson (1989) Banach (1993a), Banach (1993b) Watson and Watson (1987) Watson et ah (1987) Watson et ah (1989) Since the machine needed a runtime system, whose implementation centred round MONSTR, the 339 connection with operating systems emerges. In fact MONSTR evolved as a restriction of a more general term graph ....

....by implementation issues. MONSTR therefore rejoices in the virtue of having been implemented in anger for a real machine. In particular, the directed arcs of a MONSTR graph are intended to be directly modeled by pointers in a conventional store in the overwhelming majority of instances, see [Banach (1993a)] for an exposition of exactly how) So translations of process algebra formalisms (or for that matter anything else) into MONSTR can give a reasonable idea of the practicality of the primitive notions inherent in these formalisms. In the present case we find that the atomicity and synchronisation ....

[Article contains additional citation context not shown here]

Banach R., MONSTR I Fundamental Issues and the Design of MONSTR. Submitted to New Generation Computing, (1993).

....in terms of traces, the latter relationship being only an opfibration over the former) 1 INTRODUCTION MONSTR is a generalised term graph rewriting language that was used as the intermediate language for the Flagship machine. See [Banach et al. 1988) Banach and Watson (1989) Banach (1993) Banach (1996), Banach (1995) Watson and Watson (1987) Watson et al. 1987) Watson et al. 1989) Since the expressiveness of the language was closely matched to the capabilities of a distributed store parallel machine, it is not surprising that MONSTR has proved itself well suited to describing features of ....

....by a rewrite is found to be non idle, the rewrite suspends until the subcomputation notifies in the normal way, although this is a feature of MONSTR semantics that we will not need below. The above constitutes a brief sketch of MONSTR and its abstract execution model. For a definitive account see [Banach (1996)] Banach (1995) Lack of space prevents us from saying more here. 3 THE PI CALCULUS The p calculus first appeared in [Milner et al. 1992) and since that time has been seen in a number of minor variants. We will use a version of the monadic calculus, Milner (1993a) since it proves to be ....

[Article contains additional citation context not shown here]

Banach R., MONSTR I --- Fundamental Issues and the Design of MONSTR. J.UCS, 2, (1996), 164-216. http://www.iicm.tu-graz.ac.at/jucs

....whose properties carry over to the corresponding MONSTR code. Thus, an Interaction Nets MONSTR sublanguage would be able to take advantage of those properties for the benefit of the underlying implementation. To further clarify some of these points, we show how the MONSTR compiler environment (Banach 1993, Glauert et al. 1990, Watson et al. 1988) could take advantage of the knowledge that the MONSTR code generated originates from Interaction Net code enjoying the usual properties, possibly coupled with knowledge derived from some static dataflow analysis. In particular, it is possible to enhance the MONSTR code with ....

Banach, R. (1996a) MONSTR I -- fundamental issues and the design of MONSTR. Journal of Universal Computer Science, 2(4), 164--216 (http://www.iicm.edu/jucs).

....object in the vicinity of the cut. In the world of the representing graphs, two agents joined by a connection which is principal for both of them is the analogue of the cut, and its elimination is a rewrite rule for such graphs. Thus arises the Interaction Net model of graph rewriting. MONSTR ([1,3]) originated from the desire to implement the generalised Term Graph Rewriting language DACTL ( 9] on a distributed parallel machine, the Flagship machine ( 13] The demands of (even an imperfectly adhered to notion of) serialisability for DACTL executions, necessitated curtailing the expressive ....

....the above ideas into a computational model suited to distributed machines, a number of restrictions are imposed on the syntactic structure of systems so that some useful runtime properties can be rigorously demonstrated. We point out the main ones now rather informally, referring the reader to [1,3] for a more detailed study of why these are appropriate and what their consequences are: All nodes respect the arities of their symbols. Within rules; and by means of a simple induction, within all execution graphs. The alphabet of symbols is divided into functions, constructors and ....

[Article contains additional citation context not shown here]

Banach R., "MONSTR I --- Fundamental Issues and the Design of MONSTR", Journal of Universal Computer Science, 2 (4), 1996, pp. 164-216, http://www.iicm.tu-graz.ac.at/jucs.

....its nature as a general purpose computational model, and languages based on TGRSs are suitable as compiler target (intermediate) languages. In particular, a number of specific TGR languages have been developed, all closely related, among which we may mention Clean ( 14] DACTL ( 9] and MONSTR ([2]) The last of these, MONSTR, can be viewed as a subset of DACTL, at least in the syntactic sense, and it is the machine language of the distributed architecture Flagship ( 11] MONSTR will be of considerable concern to us in this paper, being the target of our translation of a fragment of ....

....a rewrite is found to be non idle, the rewrite suspends until the subcomputation notifies in the normal way, although this is a feature of MONSTR semantics that we will not need below. The above constitutes a brief sketch of MONSTR and its abstract execution model; for a more detailed account see [2]. At the implementation level, term graphs may be represented by packets in memory, a packet being a data structure containing principally, a marking, a symbol, and a sequence of pointers to child packets. Thus each packet represents a node and its out arcs. Notification arcs are in fact ....

[Article contains additional citation context not shown here]

R. Banach, MONSTR I --- Fundamental Issues and the Design of MONSTR, submitted to New Generation Computing, 1993.

....depend solely on the symbol at the root (and so can be delegated to simple hardware) ffl No pointer equivalence is required for matching any redex node, that is not evident from Map(oe(root) ditto) ffl All execution graphs are balanced and state saturated. ffl The overwriting lemma (Banach 1996a, Lemma 5.10), applies to most redirections, in practice enabling the convenient representation of rewriting by packet store manipulations, and particularly the representation of redirection by packet overwriting) To ensure that rewriting conforms to the exigencies of MONSTR garbage collection, an issue we ....

Banach R. (1996a), MONSTR I --- Fundamental Issues and the Design of MONSTR, Journal of Universal Computer Science, to appear.

....[Milner (1989) to allow networks of processes to evolve dynamically. It is thus a process algebra language. MONSTR by contrast is a generalised term graph rewriting language that was used as the intermediate language for the Flagship machine. See [Banach et al. 1988) Banach and Watson (1989) Banach (1993a), Banach (1993b) Watson and Watson (1987) Watson et al. 1987) Watson et al. 1989) Since the machine needed a runtime system, whose implementation centred round MONSTR, the Journal of Universal Computer Science, vol. 1, no. 6 (1995) 335 395 submitted: 20 10 94, accepted: 26 6 95, appeared: ....

....by implementation issues. MONSTR therefore rejoices in the virtue of having been implemented in anger for a real machine. In particular, the directed arcs of a MONSTR graph are intended to be directly modeled by pointers in a conventional store in the overwhelming majority of instances, see [Banach (1993a)] for an exposition of exactly how) So translations of process algebra formalisms (or for that matter anything else) into MONSTR can give a reasonable idea of the practicality of the primitive notions inherent in these formalisms. In the present case we find that the atomicity and synchronisation ....

[Article contains additional citation context not shown here]

Banach R., MONSTR I --- Fundamental Issues and the Design of MONSTR. Submitted to New Generation Computing, (1993).

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC