J. Buck and E. A. Lee, "Scheduling dynamic dataflow graphs with bounded memory using the token flow model," International Conference on Acoustics, Speech, and Signal Processing, pp. 429-432, 1993.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....produced by the operat ors to one. The real restriction that we im pose is that we know the precise dependency graph. An extensive research has been made concerning the solution of balance equations [1] for the cases where multiple tokens are con sumed produced, using the methods presented in [18, 19] the firing order of the actors can be established, thus the precise dependency graph can be found and our system model can be ex tended to these cases as well. The precedence relations among the tasks can be represented as a directed acyclic graph. The target hardware system may contain ....

J. T. Buck and E. A. Lee, "Scheduling Dynamic Dataflow Graphs with Bounded Memory Using the Token-flow model," Proc. IEEE ICASSP95,

....implemented on an embedded processor, and (2) allow one to evaluate tradeoffs between memory size and execution speed of the final implementation. Several techniques for software synthesis from a concurrent functional specification, along the lines discussed above, have been proposed. Buck and Lee [5] have introduced the Boolean Data Permission to make digital hardcopy of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage, the copyright notice, the ....

J.Buck. Scheduling dynamic dataflow graphs with bounded memory using the token flow model. Ph.D dissertation. UC Berkeley, 1993.

....the workload of the Kahn processes. However, one can also co simulate the application and architecture models and schedule the IDF actors dynamically. But in this case, one cannot guarantee whether the execution is completed in finite time or whether it is performed within bounded memory [4]. In this paper, we are interested in refining grain sizes of both computation and communication events and subsequently model parallel execution (intra task parallelism) of these events at the architecture level. More specifically, as our Kahn application models often operate on blocks of data, ....

J. T. Buck. Scheduling Dynamic Dataflow Graphs with Bounded Memory using the Token Flow Model. PhD thesis, Dept. of Electrical Engineering and Computer Sciences, University of California, Berkeley, 1993.

....calls for combining multiple IDF sub models (one for each required refinement) in a hierarchy of switches that selects the appropriate refinement. Further study is needed to investigate how this affects our refinement method. Another important issue is the analyzability of our IDF models. In [4], it is shown that dynamic data flow models, such as boolean and integer controlled data flow models, are hard to analyze statically since many of their analysis problems are undecidable. In our IDF models, the incoming event trace at a root switch acts as a control token stream of which the ....

J. T. Buck. Scheduling Dynamic Dataflow Graphs with Bounded Memory using the Token Flow Model. PhD thesis, Dept. of Electrical Engineering and Computer Sciences, University of California, Berkeley, 1993.

....increased expressive power by allowing data dependent production and consumption rates. In exchange, some of the intuitive simplicity and appeal of SDF is lost, static scheduling cannot always be employed, and the problems of bounded memory verification and deadlock detection become undecidable [26], which means that, in general, they cannot be solved in finite time. However, heuristics have been developed for constructing efficient quasi static schedules, and attempting to verify bounded memory requirements. These heuristics have been shown to work well in practice [26] A natural extension ....

....become undecidable [26] which means that, in general, they cannot be solved in finite time. However, heuristics have been developed for constructing efficient quasi static schedules, and attempting to verify bounded memory requirements. These heuristics have been shown to work well in practice [26]. A natural extension of BDF, called integer controlled dataflow, that allows control tokens to take on arbitrary integer values has been explored in [27] Parameterized dataflow [29] 28] provides a general framework for incorporating powerful dynamic reconfiguration capabilities into arbitrary ....

J. T. Buck, "Scheduling dynamic dataflow graphs with bounded memory using the token flow model," Ph.D. dissertation, Dept. Elec. Eng. Comput. Sci., Univ. of California at Berkeley, Sept. 1993.

....platform architecture. Two real time embedded system examples are given in Section 4. Section 5 concludes the article and gives directions for future research work. 2 Previous Work Several techniques for sotvare synthesis from a concurrent functional specification have been proposed [6] 8] [9], 10] 11] 12] 13] 14] Buck and Lee [9] have introduced the Boolean Data Flow (BDF) networks model and proposed an algorithm to compute a quasi static schedule. However, the problem of scheduling BDF with bounded memory is undecidable, i.e. any algorithm may fail to find a schedule even ....

....system examples are given in Section 4. Section 5 concludes the article and gives directions for future research work. 2 Previous Work Several techniques for sotvare synthesis from a concurrent functional specification have been proposed [6] 8] 9] 10] 11] 12] 13] 14] Buck and Lee [9] have introduced the Boolean Data Flow (BDF) networks model and proposed an algorithm to compute a quasi static schedule. However, the problem of scheduling BDF with bounded memory is undecidable, i.e. any algorithm may fail to find a schedule even if the BDF is schedulable. Hence, the algorithm ....

J. Buck, Scheduling dynamic dataflow graphs with bounded memory using the token flow model, Ph. D, dissertation, UC Berkeley, 1993.

....our emulation platform architecture. Two embedded system examples are given in Section 4. Section 5 concludes the article and gives directions for future research work. 2 Previous Work Several techniques for software synthesis from a concurrent functional specification have been proposed [5] [8], 9] 10] 11] 12] 15] 16] Buck and Lee [8] have introduced the Boolean Data Flow (BDF) networks model and proposed an algorithm to compute a quasi static schedule. However, the problem of scheduling BDF with bounded memory is undecidable, i.e. any algorithm may fail to find a schedule ....

....system examples are given in Section 4. Section 5 concludes the article and gives directions for future research work. 2 Previous Work Several techniques for software synthesis from a concurrent functional specification have been proposed [5] 8] 9] 10] 11] 12] 15] 16] Buck and Lee [8] have introduced the Boolean Data Flow (BDF) networks model and proposed an algorithm to compute a quasi static schedule. However, the problem of scheduling BDF with bounded memory is undecidable, i.e. any algorithm may fail to find a schedule even if the BDF is schedulable. Hence, the algorithm ....

J. Buck, Scheduling dynamic dataflow graphs with bounded memory using the token flow model, Ph. D, dissertation, UC Berkeley, 1993.

....The importance of such techniques is more and more evident as complex control operations are required in recent dataflow applications. However, a known problem of such techniques is that the scheduling problem becomes undecidable in general when data dependent controls are taken into account [1]. We have proposed earlier an algorithm for this problem [2] in which precise data dependency of the controls are abstracted as non deterministic choices. With this abstraction, exactly how each control is resolved is not considered during the scheduling. Instead, the algorithm searches for a ....

....operations cannot be determined in general until run time, when values of data used by the constructs become known. The concurrent processes are sequentialized in order to reduce run time overhead maximally, while data dependent control constructs are resolved at run time (Quasi Static Scheduling [1, 9, 10]) The schedule is then transformed into tasks by traversing the graph to generate software code using the original code annotated with the transition nodes. 2.1. The false path problem The false path problem arises when data dependent controls are abstracted as non deterministic choices. Figure ....

[Article contains additional citation context not shown here]

J. Buck. Scheduling dynamic dataflow graphs with bounded memory using the token flow model. PhD thesis, U.C. Berkeley, 1993.

....size, code size, and optimizing the schedule. 1 Introduction Synchronous dataflow graphs have become a powerful and widespread programming abstraction for DSP environments. Originally proposed by Lee and Messerchmitt in 1987 [18] synchronous dataflow has since evolved into a number of varieties [5, 1, 26, 6] which provide a natural framework for developing modular processing blocks and composing them into a concurrent network. A dataflow graph consists of a set of nodes with channels running between them; data values are passed as streams over the channels. The synchronous assumption is that a node ....

J. Buck. Scheduling dynamic dataflow graphs with bounded memory using the token flow model. PhD thesis, University of California, Berkeley, 1993. 15

....of data driven and demanddriven scheduling [12, 2] in which a process can produce output (in a data driven way) until the channel it is writing to is full. Then it blocks and other processes take over. Unfortunately, it is undecidable in general, how much bu#er capacity is needed in every channel [4]. If bu#ers are chosen too large, memory is wasted. If bu#ers are chosen too small, artificial deadlocks (i.e. processes being permanently blocked on a full channel) may occur that are not in the original KPN. Therefore, a scheduler is needed that determines the order of execution of processes ....

....not fair maximal in the KPN. 4.2 Deadlocks A PNB behaves the same as the corresponding KPN, except for the fact that certain write actions may be disabled by full channels. This may lead to artificial deadlock situations. Since the amount of memory required for channels cannot be decided upfront [4], a scheduler must provide the means to dynamically (at run time) increase channel capacity where needed. If a process is blocked trying to read from an empty channel or trying to write to a full channel, this situation can only be resolved by a unique other process in the network or, in the ....

J.T. Buck. Scheduling Dynamic Dataflow Graphs with Bounded Memory Using the Token Flow Model. PhD thesis, University of California, EECS Dept., Berkeley, CA, 1993.

No context found.

J. Buck and E. A. Lee, "Scheduling dynamic dataflow graphs with bounded memory using the token flow model," International Conference on Acoustics, Speech, and Signal Processing, pp. 429-432, 1993.

No context found.

J.T. Buck and E.A. Lee, "Scheduling Dynamic Dataflow Graphs with Bounded Memory Using the Token Flow Model," in Proc. Int. Conf. Acoustics Speech & Signal Processing, pp. 429-432, Apr. 1993.

No context found.

J. T. Buck. Scheduling Dynamic Dataflow Graphs with Bounded Memory Using the Token Flow Model. PhD thesis, Department of Electrical Engineering and Computer Science, University of California at Berkeley, 1993.

No context found.

J. Buck. Scheduling dynamic dataflow graphs with bounded memory using the token flow model. PhD thesis, University of Berkeley, 1993.

No context found.

J. T. Buck. Scheduling Dynamic Dataflow Graphs with Bounded Memory using the Token Flow Model. PhD thesis, Dept. of EECS, UC Berkeley, 1993.

No context found.

J. T. Buck, "Scheduling dynamic dataflow graphs with bounded memory using the token flow mode", Ph.D. thesis, Electrical Engineering and Computer Sciences, University of California Berkeley, 1993.

No context found.

J. T. Buck, "Scheduling dynamic dataflow graphs with bounded memory using the token flow model," Ph.D. dissertation, Univ. California, Elec. Eng. Comput. Sci., Berkeley, 1993.

No context found.

J. T. Buck and E. A. Lee. Scheduling dynamic dataflow graphs with bounded memory using the token flow model. In Proceedings of IEEE International Conference on Acoustics, Speech, and Signal Processing, pages 429--432, Minneapolis, MN, 1993.

No context found.

J. T. Buck. Scheduling Dynamic Dataflow Graphs with Bounded Memory Using the Token Flow Model. PhD thesis, EECS Department, University of California at Berkeley, CA, 1993.

No context found.

J. Buck. Scheduling dynamic dataflow graphs with bounded memory using the token flow model.PhD thesis, U.C. Berkeley, 1993.

No context found.

J. T. Buck and E. A. Lee. Scheduling dynamic dataflow graphs with bounded memory using the token flow model. In Proceedings of IEEE International Conference on Acoustics, Speech, and Signal Processing, pages 429--432, Minneapolis, MN, 1993.

No context found.

J. T. Buck. Scheduling Dynamic Dataflow Graphs with Bounded Memory Using the Token Flow Model. PhD thesis, EECS Department, University of California at Berkeley, CA, 1993.

No context found.

J. T. Buck. Scheduling Dynamic Dataflow Graphs with Bounded Memory using the Token Flow Model. PhD thesis, Department of Electrical Engineering and Computer Sciences, University of California at Berkeley, September 1993.

No context found.

J. T. Buck, "Scheduling dynamic dataflow graphs with bounded memory using the token flow model," Ph.D. Dissertation, Dept. EECS, Univ. California, Berkeley, CA, 1993; Tech. Rep. UCB/ERL 93/69, .

No context found.

J. T. Buck, Scheduling Dynamic Dataflow Graphs with Bounded Memory using the Token Flow Model, Ph.D. thesis, Dept. of EECS, UC Berkeley, 1993.

First 50 documents  Next 50

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