A technique for synthesizing distributed burst-mode circuits

Kudva, P.; Gopalakrishnan, Ganesh; Jacobson, H.

doi:10.1109/dac.1996.545548

Cited by 11 publications

(8 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [21,18], Vogler and Wollowski presented a formal method for decomposition based on the net contraction approach of [7], which is then validated through bisimulation. Kapoor and Josephs have proposed a simple source language-level decomposition technique, equivalent to Petri net decomposition, which significantly improves runtime, but (as with [11]) can only decompose for fixed series-parallel constructs such as forks and joins [10]. A direct mapping approach by Bystrov and Yakovlev [5], targeted for low-latency, is also a form of Petri-net decomposition.…”

Section: Related Workmentioning

confidence: 97%

“…However, none of these methods focuses on individual controllers, and each is either template-based (Tangram, Balsa) or fairly coarse-grained. Kudva et al [11] have proposed a high-level synthesis flow which maps a control-dataflow graph into separate datapath and control. The control is then partitioned into burst-mode sub-controllers, where several may be simultaneously active.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

A Cycle-Based Decomposition Method for Burst-Mode Asynchronous Controllers

Agyekum

Nowick

2007

13th IEEE International Symposium on Asynchronous Circuits and Systems (ASYNC'07)

View full text Add to dashboard Cite

In this paper, a systematic and automated methodology is proposed for decomposing an asynchronous burst-mode (BM) controller into smaller sub-controllers, where each resulting subcontroller is activated on a communication channel. The proposed approach consists of a new decomposition algorithm, control micro-architecture and inter-controller communication protocol. This method has also been broadened to handle extended burstmode (XBM) controllers. For both controller types, only a moderate amount of auxiliary hardware is required, and optimizations are proposed to eliminate or simplify this hardware.Initial runtime results for both burst-mode and extended burstmode controllers are promising. Two of the largest BM benchmarks (dean-cache, scsi) were run using the Minimalist CAD tool and an optimized script. While the original controllers each timed out after 10 hours, the decomposition runs each completed in under 84 seconds. Further attempts to synthesize the original controllers using a suboptimal script succeeded, but with 16-200x greater runtime. Several XBM benchmarks were synthesized using the 3D CAD tool; one large complex controller (cdp-p1) was unable to complete while the decomposed run succeeded in under 197 seconds.

show abstract

Section: Related Workmentioning

confidence: 97%

Section: Related Workmentioning

confidence: 99%

A Cycle-Based Decomposition Method for Burst-Mode Asynchronous Controllers

Agyekum

Nowick

2007

13th IEEE International Symposium on Asynchronous Circuits and Systems (ASYNC'07)

View full text Add to dashboard Cite

show abstract

“…The other examples are classic burst-mode benchmarks from various publications. The burst-mode controllers ackcd-player, ack-fibonacci, ack-diffeq, ack-barcode, ack-gcd, and ack-factorial are all generated automatically from a procedural language description by the high-level synthesis framework ACK [30,6,31,5]. All benchmarks are run on a 333 MHz Ultrasparc-2 processor with 1 GB of physical memory and 840 MB of virtual memory running the Solaris operating system.…”

Section: Benchmark Resultsmentioning

confidence: 99%

“…For example, different protocols and state assignments could lead to substantial improvements in circuit area and delay. Since existing exact synthesis methods are slow, and may not even complete for large circuits, designers are often forced to use heuristic methods, or partitioning [6], to interactively explore the design space in a reasonable amount of time often yielding sub-optimal circuits. The bottleneck of extended burst-mode finite state machine synthesis is typically in the hazard-free logic minimization step.…”

Section: Introductionmentioning

confidence: 99%

Achieving fast and exact hazard-free logic minimization of extended burst-mode gC finite state machines

Jacobson

Myers

Gopalakrishman³

IEEE/ACM International Conference on Computer Aided Design. ICCAD - 2000. IEEE/ACM Digest of Technical Papers (Cat. No.00CH3714

Self Cite

View full text Add to dashboard Cite

show abstract

“…In section 3 we show that the restriction of output concurrency by the FS rule does not hinder a proper GHFRPSRVL WLRQ RI WKH SULPLWLYH )60 that would even reinstate output concurrency in a V\VWHP of XBMMs. (If a system of circuits is the design goal, decomposition at the STG level has already been proposed by [3], [5], and [11].) Instead of applying the FM-and FS-restricted firing rule to the STG, the SpG_FM can also be obtained from the SpG by removing all arcs that violate the FM and FS restrictions.…”

Section: B the Forced-output-simultaneity Restriction (Fs Rule)mentioning

confidence: 99%

From STG to extended-burst-mode machines

Beister

Eckstein²,

Wollowski³

Proceedings. Fifth International Symposium on Advanced Research in Asynchronous Circuits and Systems

View full text Add to dashboard Cite

URP 67* WR SULPLWLYH 0HDO\)60 &RQFHSW DQG EDVLF GHILQLWLRQV On the way from the STG to an XBM solution we first derive an FSM that represents the causally specified behaviour in the EHVW SRVVLEOH VHTXHQWLDO way, i.e. completely, without additions, and such that the original causal relations can still be identified. The most general FSM of this kind is one which has what will be called the process-preserving (PP) property: It can generate -with the addition of state transitions -DOO WKRVH DQG RQO\ WKRVH level-oriented WHPSR UDO LQSXWRXWSXW VHTXHQFHV that are consistent with the causal specification, conform to fundamental-mode operation, and specify only a single suitable (quasi-)ordering for each set of concurrent output edges. -The present section will explain how to construct a SULPLWLYH 0HDO\-type FSM with the PP property from the STG.A Mealy-FSM is a sextuple M = (,,2,6,δ,λ,S 1 ), where ,, 2, and 6 are finite sets of inputs, outputs, and internal states, respectively; δ: , x 6 → 6 is the next-state function, λ: , x 6 → 2 is the output function, and S 1 ∈ 6 is the initial state. -The Mealy-FSM is primitive iff every S i ∈ 6 is stable under at most one I k ∈ , (i.e. iff there is at most one stable total state (S i , I k ) per row of the flow table). -The flow table of a primitive Mealy-FSM will be called primitive flow table (PFT) [1], and its graph a primitive state transition graph (PSTG).The Mealy type is chosen with regard for the subsequent XBM synthesis; but a Moore-type FSM with the PP property could also be constructed. Primitivity is, however, essential for capturing the PP property, regardless of the FSM type. The construction algorithm consists of four steps which are explained in subsections 2.2 to 2.5. 6WHS )URP 67* WR 6S*We assume the currently established STG model (e.g.[9]) where the underlying Petri net is n-safe (n ≥ 1) and all arcs have weight 1. This STG allows the specification of causal dependence, causal independence (concurrency) and conflict (choice, exclusion). Of course, the STG should show that rising and falling edges of every signal must always alternate (consistency). We will use Petri net notation for STGs, with full rectangles for input transitions and hollow ones for output transitions. - Fig. 2.1.a shows the STG for what originally was an extended-burst-mode example from [7], but has been modified by additional signals a and z. The circuit is to answer each φ+ edge either by x+ (if conditional input c is 0) or by y+ (if c is 1 when φ+ occurs). Besides, the output edges x-and z+ have been made concurrent. -Executing such a Petri net generates causal processes represented by SRVHWV of input and output edges.

show abstract

A technique for synthesizing distributed burst-mode circuits

Cited by 11 publications

References 12 publications

A Cycle-Based Decomposition Method for Burst-Mode Asynchronous Controllers

A Cycle-Based Decomposition Method for Burst-Mode Asynchronous Controllers

Achieving fast and exact hazard-free logic minimization of extended burst-mode gC finite state machines

From STG to extended-burst-mode machines

Contact Info

Product

Resources

About