CUBA: an architecture for efficient CPU/co-processor data communication

ABSTRACT

Data-parallel co-processors have the potential to improve performance in highly parallel regions of code when coupled to a general-purpose CPU. However, applications often have to be modified in non-intuitive and complicated ways to mitigate the cost of data marshalling between the CPU and the co-processor. In some applications the overheads cannot be amortized and co-processors are unable to provide benefit. The additional effort and complexity of incorporating co-processors makes it difficult, if not impossible, to effectively utilize co-processors in large applications.

This paper presents CUBA, an architecture model where co-processors encapsulated as function calls can efficiently access their input and output data structures through pointer parameters. The key idea is to map the data structures required by the co-processor to the co-processor local memory as opposed to the CPU's main memory. The mapping in CUBA preserves the original layout of the shared data structures hosted in the co-processor local memory. The mapping renders the data marshalling process unnecessary and reduces the need for code changes in order to use the co-processors. CUBA allows the CPU to cache hosted data structures with a selective write-through cache policy, allowing the CPU to access hosted data structures while supporting efficient communication with the co-processors. Benchmark simulation results show that a CUBA-based system can approach optimal transfer rates while requiring few changes to the code that executes on the CPU.

REFERENCES

• 1: AMD Staff. AMD64 Architecture Programmer's Manual. AMD Corporation, Sept. 2006.
• 2: D. Anderson. Hyper-Transport System Architecture. Addison-Wesley Professional, 2003.
• 3: R. Enzler, M. Platzer, C. Plessl, L. Thiele, and G. Troester. Reconfigurable processors for handhelds and wearables: Application analysis. In Reconfigurable Technology, pages 135146, Denver, CO, USA, Aug. 2001.
• 4: M. Fahey, S. Alam, T. Dunigan Jr, J. Vetter, and P. Worley. Early Evaluation of the Cray XD1. Cray User Group Conference, 2005.
• 5: Michael Gschwind , H. Peter Hofstee , Brian Flachs , Martin Hopkins , Yukio Watanabe , Takeshi Yamazaki, Synergistic Processing in Cell's Multicore Architecture, IEEE Micro, v.26 n.2, p.10-24, March 2006 [doi>10.1109/MM.2006.41]
• 6: Zhi Guo , Betul Buyukkurt , Walid Najjar, Input data reuse in compiling window operations onto reconfigurable hardware, ACM SIGPLAN Notices, v.39 n.7, July 2004
• 7: Zhi Guo , Walid Najjar , Frank Vahid , Kees Vissers, A quantitative analysis of the speedup factors of FPGAs over processors, Proceedings of the 2004 ACM/SIGDA 12th international symposium on Field programmable gate arrays, February 22-24, 2004, Monterey, California, USA [doi>10.1145/968280.968304]
• 8: S. Hauck , T. W. Fry , M. M. Hosler , J. P. Kao, The Chimaera reconfigurable functional unit, Proceedings of the 5th IEEE Symposium on FPGA-Based Custom Computing Machines, p.87, April 16-18, 1997
• 9: J. R. Hauser , J. Wawrzynek, Garp: a MIPS processor with a reconfigurable coprocessor, Proceedings of the 5th IEEE Symposium on FPGA-Based Custom Computing Machines, p.12, April 16-18, 1997
• 10: M. Hummel, M. Krause, and D. O'Flaherty. AMD and HP: Protocol enhacements for tightly coupled accelerators. Press Release, 2007.
• 11: Intel Staff. Intel 64 and IA-32 Architectures Software Developer's Manuals. Intel, May 2007.
• 12: Jeffrey A. Jacob , Paul Chow, Memory interfacing and instruction specification for reconfigurable processors, Proceedings of the 1999 ACM/SIGDA seventh international symposium on Field programmable gate arrays, p.145-154, February 21-23, 1999, Monterey, California, United States [doi>10.1145/296399.296446]
• 13: John H. Kelm , Isaac Gelado , Mark J. Murphy , Nacho Navarro , Steve Lumetta , Wen-mei Hwu, CIGAR: Application Partitioning for a CPU/Coprocessor Architecture, Proceedings of the 16th International Conference on Parallel Architecture and Compilation Techniques, p.317-326, September 15-19, 2007 [doi>10.1109/PACT.2007.21]
• 14: Donglok Kim , Ravi Managuli , Yongmin Kim, Data Cache and Direct Memory Access in Programming Mediaprocessors, IEEE Micro, v.21 n.4, p.33-42, July 2001 [doi>10.1109/40.946678]
• 15: D. A. Koufaty , X. Chen , D. K. Poulsen , J. Torrellas, Data forwarding in scalable shared-memory multiprocessors, Proceedings of the 9th international conference on Supercomputing, p.255-264, July 03-07, 1995, Barcelona, Spain [doi>10.1145/224538.224569]
• 16: MIPS Staff. MIPS32 Architecture for Programmers. MIPS Technologies, Mar. 2001.
• 17: J. Renau, B. Fragela, J. Tuck, W. Liu, L. Ceze, S. Sarangi, P. Sack, K. Strauss, and P. Montesinos. SESC simulator. http://sesc.sourceforge.net, Jan. 2005.
• 18: Shane Ryoo , Christopher I. Rodrigues , Sara S. Baghsorkhi , Sam S. Stone , David B. Kirk , Wen-mei W. Hwu, Optimization principles and application performance evaluation of a multithreaded GPU using CUDA, Proceedings of the 13th ACM SIGPLAN Symposium on Principles and practice of parallel programming, February 20-23, 2008, Salt Lake City, UT, USA [doi>10.1145/1345206.1345220]
• 19: David Seal, ARM Architecture Reference Manual, Addison-Wesley Longman Publishing Co., Inc., Boston, MA, 2000
• 20: Hartej Singh , Ming-Hau Lee , Guangming Lu , Nader Bagherzadeh , Fadi J. Kurdahi , Eliseu M. Chaves Filho, MorphoSys: An Integrated Reconfigurable System for Data-Parallel and Computation-Intensive Applications, IEEE Transactions on Computers, v.49 n.5, p.465-481, May 2000 [doi>10.1109/12.859540]
• 21: Xilinx. Virtex-II Pro and Virtex-II Pro X Plaform FPGAs: Complete Data Sheet, Oct. 2005.

INDEX TERMS

Primary Classification:
  C. Computer Systems Organization
  C.0 GENERAL

General Terms:
Design

Keywords:
co-processors