Finally, we will conclude our method and give some future work in this chapter.
5.1. Conclusions
In this thesis, we propose a method named amount-driven encoding method (ADEM) for low-power on-chip data bus design. It reduces the number of self- and coupling transitions by using encoding technique and decreases the value of coupling capacitance by applying Spacing. In summary, ADEM has the following features and contributions.
(1) In order to avoid the influence of neighbor coupling transitions, ADEM considers the coupling transitions in two phases. In phase one while ADEM encodes the pairs, it first ignores the power dissipation caused by the coupling transitions between each pair of lines. Then in phase two, the unresolved coupling transitions between each pair will be dealt with Spacing. This two phases design avoids the influence upon two adjacent coupling capacitances, thus our encoding method design can only concentrate on the improvement of pair transitions. Therefore, for the power dissipation caused by pair transition, ADEM save more power than those of OEBI and CBBI.
(2) Through our bus encoding method design, ADEM separates all the pairs into four types. Each type of pair will be encoded with four encoding methods. During encoding, ADEM first recognizes the order of appearance number for each type to process encoding without calculating total power dissipation. The experiment result shows that ADEM which applies four encoding methods outperforms only one in OEBI and CBBI. Moreover, because the order of the appearance number is independent of the
bus width, ADEM is well suited for high-performance devices which contain wider bus system. Even as the fabrication trends advancing, ADEM saves more power and is still the most effective method compared to OEBI and CBBI.
(3) The overhead of delay time and circuit complexity is less than that of OEBI.
The reason is that ADEM can encode by recognizing the appearance number of each type of pair instead of calculating total power dissipation. Meanwhile, the coupling transition between each pair of lines can be ignored during encoding.
5.2. Future work
In addition to our previous features, there are still some attractive issues worthy of further investigations. First, ADEM is mainly designed for data buses. It may not perform well on instruction or address buses, because it cannot exploit the high correlation of instruction or address data streams. In the future, we will try to find the un-correlated part of instruction or address stream, and then modify ADEM applying to the part. For instance, the un-significant part of address stream almost has weak correlation, so that we can apply ADEM to the un-significant part of address stream.
Second, we assume that the signals on the all lines are synchronized, i.e. there is no delay skew between them. However, for on-chip bus, a relative delay between data on neighbor lines usually occurs due to process, voltage and temperature variation.
Therefore, delay skew between neighbor lines may change the charge, discharge, or toggle events of coupling capacitance [9, 19]. For example, the toggling event 10Æ01 will become two separate events 10Æ00 followed by 00Æ01, so that we can’t exactly estimate the number of coupling transitions by our proposed method. In the future, we may consider that the signals with delay skew of normal distribution, and then re-estimate the number of self- and coupling transitions.
Finally, it should be noted that our proposed method doesn’t consider the coupling transitions between nonadjacent lines, because its effects on power dissipation seems relatively small. However, in an ultra deep-submicron technology design in the future, the events of coupling transition is affected not only by two adjacent lines but also by two nonadjacent lines. In the case, even the coupling transition is not affected by adjacent lines instead of nonadjacent lines, we also need to estimate it
Bibliography
[1] P. P. Sotiriadis and A. Chandrakasan, “Bus energy minimization by transition pattern coding (TPC) in deep sub-micron technologies,” Proc. of 2000 IEEE/ACM Int. Conf. Computer-Aided Design, pp. 322–327. 2000.
[2] M. R. Stan and W. P. Burleson, “Bus-invert coding for low-power I/O”, IEEE Trans. on VLSI Systems, Volume 3, pp. 49–58, Mar. 1995.
[3] Y. Shin, S. Chae, and K. Choi, “Partial bus-invert coding for power optimization of application-specific systems”, IEEE Trans. on VLSI Systems, Volume 9, pp.
377–383, Apr. 2001.
[4] L. Benini, G. De Micheli, A. Macii, E. Macii, and M. Poncino, “Reducing power consumption of dedicated processors through instruction set encoding”, Proc. of IEEE 8th Great Lakes Symposium on VLSI, pp.8–12, Feb. 1998.
[5] L. Benini, G. D. Micheli, E. Macii, D. Sciuto, and C. Silvano, “Asymptotic zero transition activity encoding for address busses in low-power microprocessor-based systems”, Proc. of IEEE 7th Great Lakes Symposium on VLSI, pp. 77–82, Mar.
1997.
[6] Y. Aghaghiri, F. Fallah, and M. Pedram, “Irredundant address bus encoding for low-power”, Proc. of IEEE International Symposium on Low-Power Electronics and Design, pp. 182–187, Aug. 2001.
[7] W. F. Yang, “Instruction level scheduling for low-power on VLIW DSP”, Master Thesis, National Chiao-Tung University, June 2005.
[8] T. Lindkvist, J. Löfvenberg, H. Ohlsson, K. Johansson, and L. Wanhammar, “A power-efficient, low-complexity, memoryless coding scheme for buses with dominating inter-wire capacitances”, Proc. of the 4th IEEE international Workshop on System-on-Chip for Real-Time Application, Jul. 2004.
[9] Y. Zhang, J. Lach, K. Skadron, and M. R. Stan, “Odd/even bus invert with two-phase transfer for busses with coupling”, Proc. of IEEE International Symposium on Low-Power Electronics and Design, pp. 754–757, Aug. 2002.
[10] P. Petrov and A. Orailoglu, “Low-power instruction bus encoding for embedded processors”, IEEE Trans. on VLSI Systems, Volume 12, Issue 8, pp. 812-826, Jul.
2004.
[11] S. K. Wong and C. Y. Tsui, “Re-configurable bus encoding scheme for reducing power consumption of the cross coupling capacitance for deep sub-micron instruction bus”, Proc. of IEEE Annual Symposium on VLSI, pp. 167-172, Feb.
2003.
[12] T. Lv, J. Henkel, H. Lekatsas, and W. Wolf, “A dictionary-based en/decoding scheme for low-power data buses”, IEEE Trans. on VLSI systems, Volume 11, No.
5, Oct. 2003.
[13] M. Ghoneima and Y. Ismail, “Low Power Coupling-Based Encoding for On-Chip Buses”, Proc. of the 2004 Int. Symposium on ISCAS’04, Volume 2, pp. 23-26, May.
2004.
[14] R. Arunachalam, E. Acar, and S. Nassif, “Optimal shielding/spacing metrics for low power design”, Proc. of IEEE Annual Symposium on VLSI, pp. 167-172, Feb.
2003.
[15] E. Musoll, T. Lang, and J. Cortadella, “Working-Zone encoding for reducing the energy in microprocessor address buses,” IEEE Trans. on VLSI System, Volume 6, pp. 568–572, Dec. 1998.
[16] R. B. Lin, “Coupling reduction analysis of bus-invert coding”, IEEE Symposium on Circuit and System 2005, ISCAS’05, Volume 6, pp. 23-26, May. 2005.
[17] Enrico M., Massimo P., Sabino S., “Combining wire swapping and spacing for low-power deep-submicron buses”, Proc. of .the 13th ACM Great Lakes symposium on VLSI, pp. 198-202, 2003.
[18] S. J. Ruan, Edwin N., Uwe S., “Simultaneous wire permutation, inversion, and spacing with genetic algorithm for energy-efficient bus design”, Proc. of the 19th IEEE Symposium on IPDPS'05, Volume 12, pp. 233.1, 2005.
[19] M. Ghoneima and Y. I. Ismail, “Delayed line bus scheme: a low-power bus scheme for coupled on-chip buses”, Proc. of the 2004 International Symposium on Low Power Electronics and Design, ISLPED ’04, pp. 66-69, 2004.