Understanding software approaches for GPGPU reliability

ABSTRACT

Even though graphics processors (GPUs) are becoming increasingly popular for general purpose computing, current (and likely near future) generations of GPUs do not provide hardware support for detecting soft/hard errors in computation logic or memory storage cells since graphics applications are inherently fault tolerant. As a result, if an error occurs in GPUs during program execution, the results could be silently corrupted, which is not acceptable for general purpose computations. To improve the fidelity of general purpose computation on GPUs (GPGPU), we investigate software approaches to perform redundant execution. In particular, we propose and study three different, application-level techniques. The first technique simply executes the GPU kernel program twice, and thus achieves roughly half of the throughput of a non-redundant execution. The next two techniques interleave redundant execution with the original code in different ways to take advantage of the parallelism between the original code and its redundant copy. Furthermore, we evaluate the benefits of providing hardware support, including ECC/parity protection to on-chip and off-chip memories, for each of the software techniques. Interestingly, our findings, based on six commonly used applications, indicate that the benefits of complex software approaches are both application and architecture dependent. The simple approach, which executes the kernel twice, is often sufficient and may even outperform the complex ones. Moreover, we argue that the cost is not justified to protect memories with ECC/parity bits.

REFERENCES

• 1: AMD Stream Computing, http://ati.amd.com/technology/streamcomputing/index.html
• 2: H. Gao, M. Dimitrov, J. Kong, and H. Zhou, "Experiencing Various Massively Parallel Architectures and Programming Models for Data-Intensive Applications", Workshop on Computer Architecture Education (WCAE-08), in conjunction with ISCA-35, 2008.
• 3: Nvidia Developing with CUDA, http://www.nvidia.com/object/cuda_develop.html
• 4: Qimonda GDDR5 -- White Paper, http://www.qimonda-news.com/download/Qimonda_GDDR5_whitepaper.pdf
• 5: Samsung Electronics: 256Mbit GDDR3 SDRAM: Revision 1.8, April 2005
• 6: Jie S. Hu , Feihui Li , Vijay Degalahal , Mahmut Kandemir , N. Vijaykrishnan , Mary J. Irwin, Compiler-Directed Instruction Duplication for Soft Error Detection, Proceedings of the conference on Design, Automation and Test in Europe, p.1056-1057, March 07-11, 2005 [doi>10.1109/DATE.2005.98]
• 7: A. Munshi, "OpenCL: Parallel computing o the GPU and CPU", tutorial, SIGGRAPH, 2008.
• 8: N. Oh, P. Shirvani, and E. McCluskey, "Error detection by duplicating instruction in super-scalar processors", IEEE Trans. on Reliability, 2002.
• 9: M. Qureshi, O. Mutlu, and Y. Patt, "Microarchitecture-Based Introspection: A Technique for Transient-Fault Tolerance in Microprocessors", DSN, 2005
• 10: M. Rebaudengo, M. Sonza Reorda, M. Torchiano, M. Violante, "A source-to-source compiler for generating dependable software", IEEE International Workshop on Source Code Analysis and Manipulation, 2001.
• 11: Steven K. Reinhardt , Shubhendu S. Mukherjee, Transient fault detection via simultaneous multithreading, Proceedings of the 27th annual international symposium on Computer architecture, p.25-36, June 2000, Vancouver, British Columbia, Canada
• 12: George A. Reis , Jonathan Chang , Neil Vachharajani , Ram Rangan , David I. August, SWIFT: Software Implemented Fault Tolerance, Proceedings of the international symposium on Code generation and optimization, p.243-254, March 20-23, 2005 [doi>10.1109/CGO.2005.34]
• 13: Jeremy W. Sheaffer , David P. Luebke , Kevin Skadron, The visual vulnerability spectrum: characterizing architectural vulnerability for graphics hardware, Proceedings of the 21st ACM SIGGRAPH/EUROGRAPHICS symposium on Graphics hardware, September 03-04, 2006, Vienna, Austria [doi>10.1145/1283900.1283902]
• 14: Jeremy W. Sheaffer , David P. Luebke , Kevin Skadron, A hardware redundancy and recovery mechanism for reliable scientific computation on graphics processors, Proceedings of the 22nd ACM SIGGRAPH/EUROGRAPHICS symposium on Graphics hardware, August 04-05, 2007, San Diego, California
• 15: Jayanth Srinivasan , Sarita V. Adve , Pradip Bose , Jude A. Rivers, The Impact of Technology Scaling on Lifetime Reliability, Proceedings of the 2004 International Conference on Dependable Systems and Networks, p.177, June 28-July 01, 2004
• 16: N. Wang and S. Patel, "ReStore: Symptom Based Soft Error Detection in Microprocessors", in DSN, 2005.
• 17: Christopher Weaver , Joel Emer , Shubhendu S. Mukherjee , Steven K. Reinhardt, Techniques to Reduce the Soft Error Rate of a High-Performance Microprocessor, Proceedings of the 31st annual international symposium on Computer architecture, p.264, June 19-23, 2004, München, Germany

INDEX TERMS

Primary Classification:
  B. Hardware
  B.8 Performance and Reliability
      B.8.1 Reliability, Testing, and Fault-Tolerance

Additional Classification:
  I. Computing Methodologies
  I.3 COMPUTER GRAPHICS
      I.3.1 Hardware architecture
          Subjects: Graphics processors

General Terms:
Performance, Reliability

Keywords:
GPGPU, reliability