G. Bilsen, M.Engels, R. Lauwereins, and J. A. Peperstraete. Cyclo-static data flow. In Proceedings of the IEEE International Conference on Acoustics, Speech, and Signal Processing, pages 3255--3258, 1995.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....context of a multiprocessor implementation [11] Using a linear pro gramming framework, they minimize the buffer size across schedules that have optimal throughput. There are some streaming computation models which are less constrained than SDF. The most relevant is Cyclo Static Data Flow (CSDF) [7, 21]. CSDF actors have multiple work functions, each of which can produce consume a different number of data items. Phased scheduling could be viewed as a generalization of CSDF to the case where hierarchical stream containers not just leaf nodes have cyclic phases. In addition, incorporating child ....

Greet Bilsen, Marc Engels, Rudy Lauwereins, and Jean Peperstraete. Cyclo-Static Dataflow. IEEE Trans. on Signal Processing, pages 397 408, February 1996.

.... on coarse grain dataflow graphs is widespread in the DSP design community, and a large and growing set of DSP design tools support such dataflow semantics [2] Since a variety of dataflow model ing styles and accompanying semantic constructs have been developed for DSP design tools (e.g. see [1, 4, 5, 8, 12, 13]) a critical problem in the process of tech nology transfer to, from, and across such tools is a cormnon, vendor independent language, and associated suite of intermediate representations and algorithms for DSP oriented dataflow modeling. This paper describes our first version of a data 1. This ....

.... support for synchronous dataflow (SDF) semantics [12] which have emerged as an important cormnon denominator across many DSP design tools and support powerful algorithms for analysis and software synthesis [3] DIF also includes support for the closely related cyclo static dataflow (CSDF) model [4], and has specialized support for various restricted versions of SDF, in particular, homogeneous and single rate dataflow, which are often used in multiprocessor scheduling and hardware synthesis. Additionally, support for dynamic, variable parameter dataflow quantities (production rates, ....

[Article contains additional citation context not shown here]

G. Bilsen, M. Engels, R. Lauwereins, and $. A. Peperstraete. Cyclo-static data flow. In Proc. ICASSP, pages 3255-3258, May 1995.

....Dataflow (CSDF) CSDF and scalable synchronout data flow (SSDF) described in Section II A 6) are presently the most widely used extensions of SDF. In CSDF, the number of tokens produced and consumed by an actor is allowed to vary as long the variation takes the form of a fixed periodic pattern [16], 17] More precisely, each actor in a CSDF graph has associated with it a fundamental period , which specifies the number of phases in one minimal period of the cyclic production consumption pattern of . For each input edge to , the scalar SDF attribute cns is replaced by a tuple , where each ....

G. Bilsen, M. Engels, R. Lauwereins, and J. A. Peperstraete, "Cyclostatic data flow," in Proc. Int. Conf. Acoustics, Speech, and Signal Processing, May 1995, pp. 3255--3258.

....integrate these components into a processing subsystem that can be easily integrated into a software application environment such as Python or MATLAB. Some of the following tools address data synchronization and static rate, dataflow modeling (e.g. using synchronous or cyclo static dataflow [23] [24]) in order to meet predictability requirements for automated assembly or hardware software cosynthesis algorithms. The Logic Foundry does not seek to integrate FPGA components with microprocessor programs automatically, but rather to facilitate the development of FPGA subsystems that can be easily ....

G. Bilsen, M. Engels, R. Lauwereins, and J. A. Peperstraete, "Cyclo-static dataflow", IEEE Transactions on Signal Processing, 44(2), pp. 397-408, Feb. 1996.

.... GG q e ( src ( q prd e ( e ( snk ( q cns e ( eG SJS ( SAJS ( A ( q ( JS ( G eG TNSE G e ( TNSE G e ( e ( src ( q prd e ( e ( snk ( q cns e ( ABCDE , q 102112, TNSE G AD , 10= TNSE G BC , 2= DE , receiver systems, just to name a few [2, 11, 12, 34, 46, 51]. Commercial tools that employ SDF semantics include Simulink by The Math Works, SPW by Cadence, and ADS by Hewlett Packard. SDF based research tools include Gabriel [32] and several key domains in Ptolemy [16] from U.C. Berkeley; and ASSIGN from Carnegie Mellon [40] Except where otherwise ....

....and expressive power that the bounded memory and deadlock problems become undecidable; and the dataflow meta models, which are model independent mechanisms for adding expressive power to broad classes of dataflow modeling approaches. Decidable dataflow models include SDF; cyclo static dataflow [12] and scalable synchronous dataflow [44] which we discuss in Sections 2.4 and 6, respectively; and multidimensional synchronous dataflow [35] for expressing multidimensional DSP applications, such as those arising in image and video processing. Dynamic dataflow models include boolean dataflow and ....

[Article contains additional citation context not shown here]

G.Bilsen, M.Engels, R.Lauwereins, and J. A. Peperstraete, "Cyclo-Static Dataflow," IEEE Transactions on Signal Processing, vol. 44, pp. 397--408, 1996.

....(The reverse is not necessarily true, as we will see shortly. If this condition holds there is a positive integer vector q in the nullspace of the topology matrix called a repetition vector for G. The repetition vector for G with the smallest norm is called the basic repetition vector (BRV) for G [1]. For example, the BRV for the SDFG in Figure 4(a) is q = The elements of a BRV q indicate that q j copies of node v j must be executed during every iteration of the static schedule. In our example we must schedule two copies of A and one copy each of B and C each time; see Figure 5(a) ....

....so that each edge carries at most one token. Since each node is expecting to either produce or consume more data than this, an EHG compensates by inserting multiple edges between nodes. An algorithm for creating a graph s EHG appears as Algorithm 1 below. It is adapted from the method of [1] for constructing the EHG of cyclostatic DFGs, which not only permit multiple tokens to pass along edges but also specifies the pattern of their production or consumption. The algorithm first creates enough copies of each node to satisfy the specifications of the BRV. It then inserts edges. If ....

[Article contains additional citation context not shown here]

G. Bilsen, M. Engels, R. Lauwereins, and J. Peperstraete. Cyclo-static dataflow. IEEE Transactions on

....number of tokens consumed and produced by an actor is constant for each firing. This property makes it possible to statically construct a finite schedule that can be periodically repeated to implement a process network that operates on infinite streams of data tokens. Cyclo static dataflow (CSDF) [3, 1] generalizes SDF by allowing the number of tokens consumed and produced by an actor to vary from one firing to the next in 1 1 2 1 C (a) 1,0] 0,1] 1] C (b) C (c) Figure 1: The commutator in synchronous, cyclostatic and state dependent forms. a cyclic pattern. Because these ....

....that can be periodically repeated to implement a process network that operates on infinite streams of data tokens. Cyclo static dataflow (CSDF) 3, 1] generalizes SDF by allowing the number of tokens consumed and produced by an actor to vary from one firing to the next in 1 1 2 1 C (a) [1,0] [0,1] 1] C (b) C (c) Figure 1: The commutator in synchronous, cyclostatic and state dependent forms. a cyclic pattern. Because these patterns are periodic and predictable, it is still possible to statically construct periodic schedules using techniques based on those developed for SDF ....

[Article contains additional citation context not shown here]

G. Bilsen, M. Engels, R. Lauwereins, and J. A. Peperstraete. Cyclo-static data flow. In IEEE Int. Conf. ASSP, pages 3255-- 3258, Detroit, Michigan, May 1995.

....scheduling since in such cases, the cost must be incurred at least once for each clusterization step, and there may be up to total clusterization steps, where is the number of edges in the original SDF graph. A third exact condition is described by Bilsen, Engels, Lauwereins, and Peperstraete [31] for a computational model called cyclo static dataflow, which is an extension of SDF. As in the approach of Karp and Miller, evaluating this condition for a general SDF graph requires examining each simple cycle separately. For a cycle that is not deadlocked, the tightest known upper bound, time ....

....upper bound, time complexity expression for the test applied to each cycle is , where is the repetitions vector of and is the number of nodes traversed by . It is easily shown that there is no upper bound on the minimum repetitions vector component that is polynomial in , and thus, the approach of [31] exhibits the potential for a similar form of two level Figure 10. A family of SDF graphs for which the running time required by the loop scheduling algorithm of [18] is . W number of SDF nodes ( 2 ( A 0 A 1 A 2 A 3 2D 2D 2D 2 1 2 1 2 1 2 2 1 2 1 A n 1 A n 2D 2 2 1 1 1 W x ( q x ....

G. Bilsen, M. Engels, R. Lauwereins, and J. A. Peperstraete, "Cyclo-static data flow," Proceeding of the International Conference on Acoustics, Speech, and Signal Processing, Detroit, May, 1995.

....concepts can be incorporated into any dataflow model in which there is a notion of a graph iteration. For example, the parameterized framework is compatible with many of the existing dataflow models such as synchronous dataflow [19] scalable synchronous dataflow [25] cyclo static dataflow [9], and Boolean dataflow [11] For clarity and uniformity, and because SDF is currently the most popular and widely studied dataflow model for DSP, we develop parameterized dataflow formally in the context of SDF [called parameterized synchronous dataflow (PSDF) In addition to the formal model, ....

....set, and associated with each , there are two properties and such that , C. Synchronous Dataflow Many successful commercial tools for DSP employ synchronous dataflow (SDF) semantics [19] or closely related alternative models, such as scalable synchronous dataflow [25] and cyclo static dataflow [9]. SDF is a restricted form of dataflow in which the numbers of tokens produced and consumed by each actor execution are restricted to be constant and statically known (known at compile time) This restriction provides SDF important benefits such as predictability, static scheduling, and powerful ....

[Article contains additional citation context not shown here]

G. Bilsen, M. Engels, R. Lauwereins, and J. A. Peperstraete, "Cyclostatic dataflow," IEEE Trans. Signal Processing, vol. 44, pp. 397--408, Feb. 1996.

....of trajectories. In a more restricted case, when the transition function observes only # to determine the next function state, it always describes a single path in # . Furthermore, if this path is finite, the transition function determines a cycle, corresponding to a cyclo static schedule [5]. At each state, a specific function needs to be evaluated as determined by the binding function . This binding function associates a function from the set with a particular function state . Only one function can be associated with a particular function state. ....

....or more firing rules. These rules are checked by a scheduler. Therefore, the scheduler determines the ordering of the actors at either compile time or at run time. The most well known dataflow models are homogeneous dataflow (HDF) 21] synchronous dataflow (SDF) 18] cyclo static dataflow (CSDF) [5] and dynamic dataflow (DDF) 11] The SBF model can act as dataflow models HDF, SDF, and CSDF. In the case of HDF and SDF, the set of functions contains only one function. The transition function is equal to g , to which the binding function binds the only function present. In case of SDF, ....

G. Bilsen, M. Engels, R. Lauwereins, and J.A. Peperstraete. Cyclo-static data flow. In IEEE International Conference ASSP, pages 3255 -- 3258, May 1995.

....The router table R is specified by an S N matrix. The rows of R are denoted 2 In future work we will consider the production and consumption of more than one token per arc per firing to model multi rate execution. In order to avoid dummy communication we might switch to cyclostatic dataflow [9], 10] as well. s 3 (1) s 4 (1) 2 6 6 4 0 1 0 1 3 7 7 5 2 6 6 4 1 1 0 0 1 1 3 7 7 5 s 1 ; s 2 s 1 ; s 3 s 2 ; s 4 R 2 R 1 s 2 (2) s 1 (2) Fig. 3. by vectors r s , s = 0; 1; S 1. The slot a router is in is called the current slot of that router. ....

G. Bilsen, M. Engels, R. Lauwereins, and J. PeperStaete, "Cyclostatic data flow," in ICASSP-95., 1995 International Conference on Acoustics, Speech, and Signal Processing, 1995, vol. 5, pp. 3255--3258.

....scheduling since in such cases, the cost must be incurred at least once for each clusterization step, and there may be up to total clusterization steps, where is the number of edges in the original SDF graph. A third exact condition is described by Bilsen, Engels, Lauwereins, and Peperstraete [31] for a computational model called cyclo static dataflow, which is an extension of SDF. As in the approach of Karp and Miller, evaluating this condition for a general SDF graph requires examining each simple cycle separately. For a cycle that is not deadlocked, the tightest known upper bound, time ....

....required by the loop scheduling algorithm of [18] is . W number of SDF nodes ( 2 ( A 0 A 1 A 2 A 3 2D 2D 2D 2 1 2 1 2 1 2 2 1 2 1 A n 1 A n 2D 2 2 1 1 1 W x ( q xT T T W n 2 ( W n 2 ( e e Onx( q x is traversed by C ( min ( q C n C 21 mial in , and thus, the approach of [31] exhibits the potential for a similar form of two level running time explosion as is present in the approach of Karp and Miller. Two sufficient but not necessary clustering conditions that were developed previously the merge pass conditions of Buck [26] and How s clustering of ....

G. Bilsen, M. Engels, R. Lauwereins, and J. A. Peperstraete, "Cyclo-static data flow," Proceeding of the International Conference on Acoustics, Speech, and Signal Processing, Detroit, May, 1995.

....can be represented vis. a vis. SDF, while maintaining its compile time predictability properties as much as possible, and at the same time, allowing representations that expose more optimization opportunities to a compiler. Multidimensional dataflow (MD SDF) 24] and cyclo static dataflow (CSDF) [15] lie in the latter category. MD SDF allows efficient modeling of multidimensional applications, and exposes parallelism more effectively than pure SDF. It also addresses the issue of re initialization of delays on graph edges, which allows certain applications that cannot be represented in SDF to ....

....Parameterized Modeling and Scheduling of Dataflow Graphs. Page 16 of 224. ated with that arc. Thus for 2D SDF as shown in the figure, it specifies the number of initial rows and columns. 1.3. 4 Cyclo Dynamic Dataflow Cyclo dynamic dataflow [35] is an extension of Cyclo static dataflow (CSDF) [15]. In cyclo static dataflow, the number of tokens produced and consumed by an actor can vary between firings, as long as the variations form a certain type of periodic pattern. Each time an actor is fired, a different piece of code called a phase is executed. CSDF specification of a distributor ....

M. Engels, G. Bilsen, R. Lauwreins, J. Peperstraete, "Cyclo-Static dataflow," IEEE Transactions on Signal Processing, Vol. 44, No. 2, February 1996.

....represent a cyclical firing of the frames inside the case structure. This corresponds to a cyclostatic dataflow (CSDF) program. Such a diagram can be statically scheduled by transforming it into a cyclostatic graph and applying the appropriate scheduling algorithms for cyclostatic dataflow [2, 10]. 0 3 4 5 6 10 2 0 3 Figure 2a Cyclo static G Diagram Figure 2b CSDF Representation 6.4 Scheduling algorithm The homogeneity and structured programming constructs in G make the creation of a static schedule relatively simple, as iterations within a periodic schedule are not an issue ....

G. Bilsen, M. Engels, R. Lauwereins, "Cyclo-Static Dataflow," IEEE Transactions on Signal Processing, vol. 44, no. 2, pp. 397-408, Feb. 1996.

....significantly in multirate FIR filters, which are common building blocks in multirate DSP applications. As we show in the following section, a multirate FIR filter that performs a sample rate conversion of factor (in reduced fraction form) can be implemented with an efficient polyphase realization [2, 3] for which the CBP parameter can be set to . EQ 5) CBP A a i a o , a o ( prod a o ( prod 1 a o ( prod 2 . 0 , a 0 ( prod A a i a o , 1 CBP A a i a o , a o ( prod 1 CBP A a i a o , a o ( prod ....

....of Technology [13] since the absence of nested loops in the vectorized schedules allows for more flexible merging of input output buffers. Buffer merging is also compatible with buffer access enhancements such as polyphase filter implementation [4, 6] and cyclo static dataflow specification [3]. Vanhoof, Bolsens and H. De Man have observed that in general, the full address space of an array does not always contain live data [15] Thus, they define an address reference window as the maximum distance between any two live data elements throughout the lifetime of an array, and fold ....

G. Bilsen, M. Engels, R. Lauwereins, and J. A. Peperstraete, "Cyclo-Static Dataflow," IEEE Transactions on Signal Processing, vol. 44, pp. 397-408, 1996.

....Cyclo static dataflow (CSDF) and scalable synchronous dataflow (described in section 2.1.6) are presently the most widelyused extensions of SDF. In CSDF, the number of tokens produced and consumed by an actor is allowed to vary as long the variation takes the form of a fixed, periodic pattern [16, 17]. More precisely, each actor A in a CSDF graph has associated with it a fundamental period (A) 2 f1; 2; g, which specifies the number of phases in one minimal period of the cyclic production consumption pattern of A. For each input edge e to A, the scalar SDF attribute cns(e) is replaced by ....

G. Bilsen, M. Engels, R. Lauwereins, and J. A. Peperstraete. Cyclo-static data flow. In Proceedings of the International Conference on Acoustics, Speech, and Signal Processing, pages 3255--3258, May 1995.

....summary, realizing synchronous state update in a dataflow model incurs the following problems. P1) Block reusability (figure 3(a) b) c) d) P2) Redundant copy (figure 3(a) P3) Synchronization (figure 3(d) P4) Buffer size (figure 3(b) In the literature, cyclo static dataflow model [5] gives a slightly improved solution with the same topology of figure 3 (c) Since it allows a cyclic pattern of sample rates, at the first firing of node PD, one SU sample is generated, while the next 9 firings do not generate any sample on the SU arc. This cyclo static solution solves (P2) and ....

Greet Bilsen, Marc Engels, Rudy Lauwereins, J.A. Peperstraete, "Cyclo-Static Dataflow", IEEE Transactions on Signal Processing, Vol. 44, No. 2, Feb. 1996, pp. 397-408.

....main distinction is that GRAPE is targeted at real time execution whereas the other environments mainly target simulation. GRAPE s design flow consists of four phases. In the specification phase, the application is described using an extended data flow model, called cyclo static data flow (CSDF) [4], which is an extension of Lee s Synchronous Data Flow [5] In short, the application is represented as a directed graph G= N,E) where the nodes N represent computation tasks, and the edges E the communication of the results (called tokens) from a producing to a consuming task. The functionality ....

Greet Bilsen, Marc Engels, Rudy Lauwereins, J.A. Peperstraete, "Cyclo-Static Dataflow", IEEE Transactions on Signal Processing, Vol. 44, No. 2, Feb. 1996, pp. 397-408.

....Tel. 32 16 32 11 42, Fax. 32 16 32 19 86 email: Piet.Wauters esat.kuleuven.ac.be 1. Abstract In this paper we present cyclo dynamic data flow (CDDF) a new data flow model for real time digital signal processing (DSP) applications. CDDF is an extension of cyclo static dataflow (CSDF) [1,2] which keeps the interesting properties like analyzability and efficient compile time scheduling, while introducing data dependent control flow to improve the expressivity. The semantics are constructed such that extra knowledge about the internals of the actors, which is known to the programmer, ....

....in the number and or the value of tokens present in the FIFOs. Depending on how the consumption and production together with the firing rules are specified, we can divide graphs or subgraphs into different classes, ranging from synchronous data flow (SDF) 4]and cyclostatic data flow (CSDF) [1,2] over Boolean and integer controlled dataflow (BDF [3] IDF[5] to dynamic data flow (DDF) In a single graph any combination of these types can occur. In the following sections we briefly describe these different dataflow models based on the token transfers and the firing rule. Synchronous data ....

[Article contains additional citation context not shown here]

Greet Bilsen, Marc Engels, Rudy Lauwereins, J.A. Peperstraete, "Cyclo-Static Dataflow", submitted for IEEE Transactions on Signal Processing.

No context found.

G. Bilsen, M.Engels, R. Lauwereins, and J. A. Peperstraete. Cyclo-static data flow. In Proceedings of the IEEE International Conference on Acoustics, Speech, and Signal Processing, pages 3255--3258, 1995.

No context found.

Bilson G.; Engels, M.; Lauwereins, R.; Cyclo-static dataflow. IEEE Trans. on Signal Processing, Vol. 44, No. 2, pp. 397408, February. 1996.

No context found.

G. Bilsen, M. Engels, R. Lauwereins, and J. Peperstraete, "Cyclo-static dataflow," IEEE Trans. Signal Processing, vol. 44, pp. 397--408, Feb. 1996.

No context found.

G. Bilsen, M. Engels, R. Lauwereins, and J. A. Peperstraete. Cyclo-static data flow. In Proceedings of the International Conference on Acoustics, Speech, and Signal Processing, pages 3255--3258, May 1995.

No context found.

G. Bilsen, M. Engels, R. Lauwereins, and J. Peperstraete. Cyclo-static dataflow. IEEE Trans. on Signal Processing, 1996.

No context found.

G. Bilsen, M. Engels, R. Lauwereins, and J. A. Peperstraete, "Cyclostatic dataflow," IEEE Trans. Signal Processing, vol. 44, pp. 397--408, Feb. 1996.

First 50 documents

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