Constraint-driven Synthesis of System-on-a-Chip Communication Architectures

Luca Carloni and Alessandro Pinto
(Professor Alberto L. Sangiovanni-Vincentelli)
Semiconductor Research Corporation

The design of the communication architecture is an essential step in the realization of a system-on-a-chip (SOC). The semiconductor industry is experiencing a paradigm shift from "computation-bound design" to "communication-bound design." The number of transistors that can be reached in a clock cycle, and not those that can be integrated on a chip, will drive the design process. In the context of a communication-based design methodology for SOC, we are developing the idea of constraint-driven communication synthesis. As logic synthesis was the key to the success of ASIC design in the VLSI era, we predict that the automatic synthesis of the on-chip communication architecture will be the key to the success of SOC design in the gigascale integration era. The goal of this research is to start with a system-level view of the SOC communication problem and automatically synthesize the best communication architecture among the SOC modules while considering both high level issues, such as communication topologies, protocol interfacing, resource sharing, as well as low level ones, such as metal layer allocation, wire buffering, and wire pipelining.

The first step in constraint-driven communication synthesis is the specification of the key communication parameters that govern the interactions among the SOC modules as a set of constraints for the synthesis problem. Then, the synthesis process proceeds in two stages. First, the optimal high-level architecture structure is derived while considering the different communication topologies, the possibility of mixing and matching them, the role of resource sharing (e.g., time multiplexing), the choice of the various communication protocols, and the needs of interfacing incompatible protocols. Notice that since the notion of optimality may vary from one design to another, depending on the designers' priorities (area, power, performance, complexity), the cost figure of a communication topology must be a configurable parameter. Once the best communication architecture is established, the focus shifts to finding the best low-level implementation based on the features that are available with the given technology process. In other words, this step corresponds to designing the global interconnect among the SOC modules and includes exploring the characteristics of the various metal layers, the insertion of buffers (stateless repeaters) on long wires, and the pipelining of (even) longer wires by means of latches (stateful repeaters).

The present research project relies on several results from our previous work. In particular, we have recently presented a generic theoretical framework to support constrained-driven communication synthesis as well as its application to the case of bus-based communication topologies [1]. Also, we plan to incorporate key results from our research on latency-insensitive design [2,3] and interface-based design [4].

[1]
A. Pinto, L. P. Carloni, and A. L. Sangiovanni-Vincentelli, "Constraint-driven Communication Synthesis," Proc. Design Automation Conf., Montreal, Canada, September-October 2002.
[2]
L. P. Carloni, K. L. McMillan, and A. L. Sangiovanni-Vincentelli, "Theory of Latency-Insensitive Design," IEEE Trans. Computer-Aided Design of Integrated Circuits and Systems, Vol. 20, No. 9, September 2001.
[3]
L. P. Carloni, K. L. McMillan, A. Saldanha, and A. L. Sangiovanni-Vincentelli, "A Methodology for Correct-by-Construction Latency-Insensitive Design," ed. A. Kuehlmann et al., The Best of ICCAD: 20 Years of Excellence in Computer-Aided Design, Kluwer Academic Publishers, 2003.
[4]
R. Passerone, J. A. Rowson, and A. L. Sangiovanni-Vincentelli, "Automatic Synthesis of Interfaces between Incompatible Protocols," Proc. Design Automation Conf., San Francisco, CA, June 1998.

More information (http://www-cad.eecs.berkeley.edu/~lcarloni) or

Send mail to the author : (lcarloni@eecs.berkeley.edu)


Edit this abstract