Summary and Future Work

Một phần của tài liệu Reconfigurable computing the theory and practice of FPGA based computation~tqw~ darksiderg (Trang 375 - 378)

Part III: Mapping Designs to Reconfigurable Platforms 275

15.8 Summary and Future Work

This chapter presented an overview of some of the many issues to consider when realizing datapaths on reconfigurable logic devices. The aspect ofregularityis a crucial one and must be considered both at the level of the target device archi- tecture and during the operation of the EDA tools. Module generators are an efficient means to actually create the circuits making up the datapath. However, in addition they must offer sufficient metadata to the rest of the tool flow as a base for effective transformation and optimization steps.

With increasing requirements on datapath performance, tool flows and algorithms must keep up with improvements in device architectures. All of the techniques described here have the potential for further refinement. Refinement opportunities include module generators that better support specialization, floorplanning with constrained two-dimensional placement, and a compaction technique in which the best of these refinements is combined.

References

[1] C. Ababei, K. Bazargan. Non-contiguous linear placement for reconfigurable fabrics.Proceedings of the of the Reconfigurable Architectures Workshop, 2004.

[2] V. Betz, J. Rose, A. Marquardt. Architecture and CAD for Deep-Submicron FPGAs, Kluwer, 1999.

[3] T. J. Callahan, P. Chong, A. DeHon, J. Wawrzynek. Fast module mapping and place- ment for datapaths in FPGAs. Proceedings of the of the International Symposium on Field-Programmable Gate Arrays, 1998.

[4] T. Callahan, R. Hauser, J. Wawrzynek. The GARP architecture and C compiler.

IEEE Computer33(4), 2000.

[5] J. M. Emmert, D. Bhati. A methodology for fast FPGA floorplanning. Proceedings of the International Symposium on Field-Programmable Gate Arrays, 1999.

[6] B. Hutchings, P. Bellows. J. Hawkins, S. Hemmert. A CAD suite for high- performance FPGA design. Proceedings of the of the IEEE Symposium on Field- Programmable Custom Computing Machines, 1999.

[7] N. Kasprzyk, A. Koch. High-level-language compilation for reconfigurable comput- ers.Proceedings of the International Conference on Reconfigurable Communication- centric SoCs, 2005.

[8] A. Koch. Module compaction in FPGA-based regular datapaths. Proceedings of the Design Automation Conference, 1996.

15.8 Summary and Future Work 345 [9] A. Koch. Structured design implementation—A strategy for implementing regular datapaths on FPGAs. Proceedings of the International Symposium on Field- Programmable Gate Arrays, 1996.

[10] A. Koch.Regular Datapaths on Field-Programmable Gate Arrays.CS doctoral thesis, technical, University of Braunschweig, 1997.

[11] A. Koch. On tool integration in high-performance FPGA design flows.Proceedings of the International Conference on Field-Programmable Logic and Applications, 1999.

[12] T. Kutzschebauch, L. Stok. Regularity-driven logic synthesis. Proceedings of the International Conference on Computer-Aided Design, 2000

[13] J. Li, J. Lillis, L. T. Liu, C. K. Cheng. New spectral linear placement and clustering approach.Proceedings of the Design Automation Conference, 1996.

[14] T. Neumann, A. Koch. A generic library for adaptive computing environments.

Proceedings of the International Conference on Field-Programmable Logic and Appli- cations, 2001.

[15] R. Nijssen, J. Jess. Two-dimensional datapath regularity extraction.Proceedings of the ACM SIGDA Physical Design Workshop, 1996.

[16] E. M. Sentovich, K. J. Singh, L. Lavagno, C. Moon, et al. SIS: A system for sequential circuit synthesis.EECS Memorandum No. UCB/ERL M92/41, University of California, Berkeley, 1992.

[17] R. Tessier. Frontier: A fast placement system for FPGAs.Proceedings of the Inter- national Conference on VLSI, 1999.

[18] F. Thorns. CLAP—Clustering and placement. Diploma thesis, Technical University of Braunschweig, 2003.

[19] C. C. Vi, D. Lewis. Area-speed trade-offs for hierarchical field-programmable gate arrays. Proceedings of the International Symposium on Field-Programmable Gate Arrays, 1996.

[20] C. Wewetzer. A Universal Generator for Logic Circuits on FPGAs. Diploma thesis, Technical University of Braunschweig, 2005.

[21] A. G. Ye.Field-Programmable Gate Array Architectures and Algorithms Optimized for Implementing Datapath Circuits. Doctoral thesis, University of Toronto, 2004.

[22] T. T. Ye, G. De Micheli. Data path placement with regularity. Proceedings of the International Conference on Computer-Aided Design, 2000.

[23] A. G. Ye, J. Rose, D. Lewis. Synthesizing datapath circuits for FPGAs with empha- sis on area minimization. Proceedings of the International Conference on Field- Programmable Technology, 2002.

[24] A. G. Ye, J. Rose. Measuring and utilizing the correlation between signal connectiv- ity and signal positioning for FPGAs containing multibit building blocks.Proceed- ings of the International Conference on Field-Programmable Logic and Applications, 2005.

C H A P T E R 16

S PECIFYING C IRCUIT L AYOUT ON FPGA S

Satnam Singh

Programming Principles and Tools Group Microsoft Research Cambridge

Typically, the layout of a circuit implemented on a field-programmable gate array (FPGA) is computed automatically by vendor design tools. This computation often results in an acceptable mapping of logical wires in the design onto actual physical routing resources on the FPGA that meets the designer’s performance requirements. Instead of relying on automated tools, however, a designer could try to use an FPGA by explicitly stating the configuration of individual logic blocks and explicitly specifying the routing between them. One almost never needs to program an FPGA at this basic and raw level, and often the proprietary nature of programming information makes it difficult or impossible to take this approach. Still, the FPGA design flow provides a powerful set of abstractions that allow a designer to think in terms of structural circuit netlists, which can be automatically converted into programming information for FPGAs. Structural netlists are abstracted further by the synthesis flow, which allows designers to think of circuit functions in an algorithmic or sequential manner.

Một phần của tài liệu Reconfigurable computing the theory and practice of FPGA based computation~tqw~ darksiderg (Trang 375 - 378)

Tải bản đầy đủ (PDF)

(945 trang)