Scheduling dynamic dataflow graphs with bounded memory using the token flow model

Buck, John A.; Lee, E.A.

doi:10.1109/icassp.1993.319147

Cited by 334 publications

(250 citation statements)

References 5 publications

Supporting

Mentioning

250

Contrasting

Order By: Relevance

“…It is known for instance [12] that the problem to decide for a given KPN (with unbounded buffer capacities) whether it is deadlock-free is undecidable. The proof of this fact [12] relies on a reduction from the Halting Problem of Turing Machines.…”

Section: Analizability Resultsmentioning

confidence: 99%

“…The proof of this fact [12] relies on a reduction from the Halting Problem of Turing Machines. It is shown that a Kahn Process Network can simulate a Universal Turing Machine; in other words KPN is Turing Complete.…”

Section: Analizability Resultsmentioning

confidence: 99%

“…Because the former is known to be an undecidable problem, so must the latter problem be undecidable. The details of the translation from Turing Machines to KPNs are given in [12].…”

Section: Analizability Resultsmentioning

confidence: 99%

“…Then it blocks and other processes take over, effectively regulating the relative speeds of different processes. As discussed previously, it is undecidable in general, how much buffer capacity is needed in every channel [12]. If buffers are chosen too large, memory is wasted.…”

Section: Run-time Scheduling and Buffer Managementmentioning

confidence: 99%

See 3 more Smart Citations

Kahn Process Networks and a Reactive Extension

Geilen

Basten

2013

Handbook of Signal Processing Systems

View full text Add to dashboard Cite

Kahn and MacQueen have introduced a generic class of determinate asynchronous data-flow applications, called Kahn Process Networks (KPNs) with an elegant mathematical model and semantics in terms of Scott-continuous functions on data streams together with an implementation model of independent asynchronous sequential programs communicating through FIFO buffers with blocking read and non-blocking write operations. The two are related by the Kahn Principle which states that a realization according to the implementation model behaves as predicted by the mathematical function. Additional steps are required to arrive at an actual implementation of a KPN to take care of scheduling of independent processes on a single processor and to manage communication buffers. Because of the expressiveness of the KPN model, buffer sizes and schedules cannot be determined at design time in general and require dynamic run-time system support. Constraints are discussed that need to be placed on such system support so as to maintain the Kahn Principle. We then discuss a possible extension of the KPN model to include the possibility for sporadic, reactive behavior which is not possible in the standard model. The extended model is called Reactive Process Networks. We introduce its semantics, look at analyzability and at more constrained data-flow models combined with reactive behavior.

show abstract

Section: Analizability Resultsmentioning

confidence: 99%

Section: Analizability Resultsmentioning

confidence: 99%

“…Because the former is known to be an undecidable problem, so must the latter problem be undecidable. The details of the translation from Turing Machines to KPNs are given in [12].…”

Section: Analizability Resultsmentioning

confidence: 99%

Section: Run-time Scheduling and Buffer Managementmentioning

confidence: 99%

See 2 more Smart Citations

Kahn Process Networks and a Reactive Extension

Geilen

Basten

2013

Handbook of Signal Processing Systems

View full text Add to dashboard Cite

show abstract

“…In parameterized dataflow [1], dynamic reconfiguration of actor and subsystem parameters is allowed through separation of functionality into subgraphs that perform reconfiguration and subgraphs that are periodically reconfigured. Boolean [4], bounded dynamic [6], and cyclo-dynamic [10] dataflow offer dynamically varying production and consumption rates by incorporating various other data-dependent modeling constructs.…”

Section: Related Workmentioning

confidence: 99%

Analysis of Dataflow Programs with Interval-limited Data-rates

Teich

Bhattacharyya

2006

J VLSI Sign Process Syst Sign Image Video Technol

View full text Add to dashboard Cite

In this paper, we consider the problem of analyzing data flow programs with the property that actor production and consumption rates are not constant and fixed, but limited by intervals. Such interval ranges may result from uncertainty in the specification of an actor or as a design freedom of the model. Major questions such as consistency and buffer memory requirements for singleprocessor schedules will be analyzed here for such specifications for the first time. MotivationThe role of data flow models of computation is becoming increasingly important for modeling and synthesis of digital processing applications. Our paper is concerned with analyzing dataflow graphs whose component data rates are not known precisely in advance. Such models are often given due to imprecise specifications or due to uncertainties in the implementation. Examples of imprecise specifications include unknown execution times of tasks, unknown data rates, unknown consumption and production behavior of modules and many more. For example, a speech compression system may have a fixed overall structure. However, the subsystem data rates are typically influenced by the size of the speech segment that is to be processed [1]. Here, we will propose a data flow model of computation that is able to model such uncertain behavior.To understand such models, it useful to first review principles of the synchronous data flow (SDF) model of computation, where production and consumption rates of data flow actors, representing computational blocks, consume fixed, known amounts of data (tokens) upon each invocation. For such graph models, often represented by a graph in which actors are connected by directed arcs that transport tokens, many interesting results have been shown such as 1) consistency, 2) memory bounds and memory analysis, and 3) scheduling algorithms. Consistency, e.g., is a static property of an SDF-graph specification which is necessary in order to guarantee the existence of a finite sequence of actor firings, also called a schedule. Typically, an SDF specification is compiled by constructing a valid schedule, i.e., a schedule that fires each actor at least once, does not deadlock, and produces no net change in the number of tokens queued on each edge. For each actor firing, a corresponding code block is instantiated from a library to produce machine code.In [5], efficient algorithms are presented to determine whether or not a given SDF graph is consistent or not and to determine the minimum number of firings of each actor §

show abstract

References

2011

Design for Embedded Image Processing on FPGAs

View full text Add to dashboard Cite

Scheduling dynamic dataflow graphs with bounded memory using the token flow model

Cited by 334 publications

References 5 publications

Kahn Process Networks and a Reactive Extension

Kahn Process Networks and a Reactive Extension

Analysis of Dataflow Programs with Interval-limited Data-rates

References

Contact Info

Product

Resources

About