A Timing Error Mitigation Technique for High Performance Designs

Dynamic flip-flop conversion (DFFC) is a time borrowing method which converts the critical flip-flops into transparent latches to allow timing slacks pass between pipeline stages of given circuits. However, our previous DFFC methods [13] [14] suffer from false error prediction. It means even when there is no setup time violation, our previous method incorrectly issues timing error. In this paper we present an improved DFFC method which consumes less power and unlike previous DFFC methods does not suffer from false errors. Also we investigate the effectiveness of our proposed DFFC method on different benchmarks by considering all existing critical paths, instead of applying our method only to one of critical paths as done in [13][14]. The results show that our proposed method improves maximum allowable frequency 7.54% on average while the previous method in [13] [14] increases maximum allowable frequency 4.24% on average. Furthermore, it consumes on average 19.59 microwatt less power per critical path compared to previous DFFC methods [13][14].

[1]  Jason Helge Anderson,et al.  Latch-Based Performance Optimization for Field-Programmable Gate Arrays , 2013, IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems.

[2]  David Blaauw,et al.  Razor: A Low-Power Pipeline Based on Circuit-Level Timing Speculation , 2003, MICRO.

[3]  David Blaauw,et al.  Razor: a variability-tolerant design methodology for low-power and robust computing , 2009 .

[4]  Stephen Dean Brown,et al.  Integrated retiming and placement for field programmable gate arrays , 2002, FPGA '02.

[5]  Youngsoo Shin,et al.  Retiming and time borrowing: Optimizing high-performance pulsed-latch-based circuits , 2009, 2009 IEEE/ACM International Conference on Computer-Aided Design - Digest of Technical Papers.

[6]  Kwanyeob Chae,et al.  Timing error prevention using elastic clocking , 2011, 2011 IEEE International Conference on IC Design & Technology.

[7]  David Blaauw,et al.  Timing yield enhancement through soft edge flip-flop based design , 2008, 2008 IEEE Custom Integrated Circuits Conference.

[8]  David Blaauw,et al.  Bubble Razor: Eliminating Timing Margins in an ARM Cortex-M3 Processor in 45 nm CMOS Using Architecturally Independent Error Detection and Correction , 2013, IEEE Journal of Solid-State Circuits.

[9]  David Blaauw,et al.  Soft-edge flip-flops for improved timing yield: design and optimization , 2007, ICCAD 2007.

[10]  K. Roy,et al.  A time borrowing selectively clocked skewed logic for high-performance circuits in scaled technologies , 2003, ESSCIRC 2004 - 29th European Solid-State Circuits Conference (IEEE Cat. No.03EX705).

[11]  Yuan Xie,et al.  Tolerating process variations in high-level synthesis using transparent latches , 2009, 2009 Asia and South Pacific Design Automation Conference.

[12]  Josep Torrellas,et al.  ReCycle:: pipeline adaptation to tolerate process variation , 2007, ISCA '07.

[13]  Robert C. Aitken,et al.  Time-Borrowing Circuit Designs and Hardware Prototyping for Timing Error Resilience , 2014, IEEE Transactions on Computers.

[14]  Ali Afzali-Kusha,et al.  Dynamic Flip-Flop Conversion: A Time-Borrowing Method for Performance Improvement of Low-Power Digital Circuits Prone to Variations , 2015, IEEE Transactions on Very Large Scale Integration (VLSI) Systems.

[15]  Ali Afzali-Kusha,et al.  Dynamic Flip-Flop conversion to tolerate process variation in low power circuits , 2014, 2014 Design, Automation & Test in Europe Conference & Exhibition (DATE).

[16]  N. Ranganathan,et al.  Dynamic clock stretching for variation compensation in VLSI circuit design , 2012, JETC.