Dynamic multirole session types

Deniélou, Pierre-Malo; Yoshida, Nobuko

doi:10.1145/1926385.1926435

Cited by 69 publications

(86 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In our setting they also control the reconfiguration phase, in which a session adapts itself to new environmental conditions. Exchanged values are marked by labels, as in [6].…”

Section: Global Typesmentioning

confidence: 99%

See 1 more Smart Citation

Data-driven adaptation for smart sessions

Bono

Coppo

Dezani‐Ciancaglini

et al. 2017

Journal of Logical and Algebraic Methods in Programming

View full text Add to dashboard Cite

This paper presents a formal framework of self-adaptation for multiparty sessions. The adaptation function contains the dynamic evolution policy, by prescribing how the session needs to reconfigure itself, based on critical changes in global data. A global type prescribes the overall communication choreography; its projections onto participants generate the monitors, which set-up the communication protocols. The key technical novelty of the calculus is the parallel operator for building global types and monitors, which allows the adaptation procedure to be rather flexible. The smart session is able to minimise its adaptation, by partially reconfiguring some of the communications and leaving all others unchanged, in case a part of the whole behaviour only needs to be modified. Furthermore, new participants can be added and/or some of the old participants can be removed. As a main result, we prove that this adaptation mechanism is safe, in order to guarantee that the communications will continue to evolve in a correct way after reconfiguration.

show abstract

“…In our setting they also control the reconfiguration phase, in which a session adapts itself to new environmental conditions. Exchanged values are marked by labels, as in [6].…”

Section: Global Typesmentioning

confidence: 99%

“…A projection onto a participant r not involved, as sender or receiver, in a choice is defined if either the projection onto r of all continuations are the same (condition G i q = G j q for all i, j ∈ I) or all these projections can be merged in an input monitor (last case), as in [6]. …”

Section: Monitorsmentioning

confidence: 99%

Data-driven adaptation for smart sessions

Bono

Coppo

Dezani‐Ciancaglini

et al. 2017

Journal of Logical and Algebraic Methods in Programming

View full text Add to dashboard Cite

show abstract

“…Dyadic session types have also been generalized to multiparty [9] interactions, where potentially more than two parties are involved in an interaction. This approach has further been generalized to a dynamically varying number of participants, based on assigning roles [5] to participants and describing generic protocols for each participant role. …”

Section: Related Workmentioning

confidence: 99%

Parameterized Dataflow (Extended Abstract)

Duggan

Yao

2016

Electron. Proc. Theor. Comput. Sci.

View full text Add to dashboard Cite

Dataflow networks have application in various forms of stream processing, for example for parallel processing of multimedia data. The description of dataflow graphs, including their firing behavior, is typically non-compositional and not amenable to separate compilation. This article considers a dataflow language with a type and effect system that captures the firing behavior of actors. This system allows definitions to abstract over actor firing rates, supporting the definition and safe composition of actor definitions where firing rates are not instantiated until a dataflow graph is launched. IntroductionDataflow or stream processing is becoming increasingly important, with the growing prevalence of signal, video and audio processing, particularly on mobile devices. Dataflow processing is a good match with multicore and GPGPU parallel architectures that are now prevalent on desktop computers, and will shortly be available on consumer mobile devices. The data parallelism of such architectures is at least potentially a good match with the demands of stream processing applications. The synthesis of these architectures with stream processing may provide a domain-specific solution to the challenge of programming the new generations of parallel computing architectures.Our starting point is a computational model similar to that originally proposed by Kahn [11]. This provides for a network of sequential actors, each implemented in a conventional sequential language such as C or Algol. Actors are connected by communication buffers on which they can send and receive data. A key point is that actors cannot nondeterministically select among inputs on several input channels, nor can they test input channels for available inputs (so polling cannot be implemented). This restricts each actor to a completely deterministic semantics. The combination of implicit parallelism and deterministic execution makes dataflow computation a good fit with some of the current thinking of how best to successfully exploit the parallelism available in modern multicore and GPGPU architectures, in those domains where the dataflow paradigm is applicable.In the embedded systems and digital signal processing community, a very useful class of restricted Kahn networks has been identified, the so-called synchronous dataflow (SDF) [14] networks. SDF networks enable static scheduling for multi-rate applications. More recently, domain-specific languages such as Streamit [15] have been defined, based on the principles of SDF, but also providing support for compiling programmer code to run on modern parallel architectures.Sessional dataflow provides a framework for providing compositional descriptions of dataflow networks [7]. A type and effect system captures the firing behavior of actor bodies, and this information is used to ensure that the composition of actors does not deadlock. For simplicity, that simple effect system did not consider variable firing rates for actors, so for example no communication was possible within a loop (finite loops were sti...

show abstract

“…The syntax of local types is standard [5], but for the types inv a and acc a , which are the counterparts of the calls in light global types: Definition 3 (Local types). We define local types, ranged over T , T , T i , .…”

Section: Local Types and Projectionsmentioning

confidence: 99%

“…Following [5][6][7], our definition of projection relies on a merge operator on local types to gather different branches from the same sender.…”

Section: Local Types and Projectionsmentioning

confidence: 99%

Lightening global types

Chen

2015

Journal of Logical and Algebraic Methods in Programming

View full text Add to dashboard Cite

Global session types prevent participants from waiting for never coming messages. Some interactions take place just for the purpose of informing receivers that some message will never arrive or the session is terminated. By decomposing a big global type into several light global types, one can avoid such kind of redundant interactions. This work proposes a framework which allows to easily decompose global types into light global types, preserving the interaction sequences of the original ones but for redundant interactions.

show abstract

Dynamic multirole session types

Cited by 69 publications

References 22 publications

Data-driven adaptation for smart sessions

Data-driven adaptation for smart sessions

Parameterized Dataflow (Extended Abstract)

Lightening global types

Contact Info

Product

Resources

About