一种低功耗抗串扰的自适应时空总线编码方法  

A Low Power Adaptive Spatio-temporal Bus Coding for Crosstalk Avoidance

在线阅读下载全文

作  者:刘毅[1,2] 钟广德[1] 杨银堂[1,2] 

机构地区:[1]西安电子科技大学微电子学院,西安710071 [2]宽禁带半导体材料与器件教育部重点实验室,西安710071

出  处:《电子与信息学报》2011年第4期945-950,共6页Journal of Electronics & Information Technology

基  金:国家杰出青年基金(60725415);国家自然科学基金(60476046;60676009);中央高校基本科研业务费专项资金(K50510250004)资助课题

摘  要:深亚微米片上总线的功耗、布线面积约束和线间串扰是限制总线数据吞吐率的关键因素,为此该文提出一种自适应时空编码方法以降低总线的串扰延迟和功耗。该方法首先采用空间编码将总线分割为两个子总线,从而减小了恶性串扰发生几率;然后通过恶性串扰判决器分别判断子总线的原码数据及反码数据是否存在恶性串扰:对于任意子总线的原码数据与反码数据均存在恶性串扰的情况,传送屏蔽字;否则,选取无恶性串扰且动态功耗小的总线数据形式并传送。采用SPEC标准数据源对算法进行了评估,该方法在消除恶性串扰的同时使总线数据吞吐率提高了62.59%~81.62%,功耗比同类方法降低14.63%~54.67%,对于32位数据总线,仅需7根冗余线,在动态功耗、布线资源和性能方面获得了有效的优化。The power consumption,wiring overhead and crosstalk in deep sub-micron on-chip buses are the main facts restricting the bus throughput.An adaptive spatio-temporal bus coding scheme is proposed to reduce crosstalk induced delay and power consumption in the buses.Firstly,on-chip bus is partitioned into two sub-buses by spatio coding to reduce the Worst-Case-Crosstalk(WCC).Then,decisions are made respectively in the two sub-buses whether the original code and inverted code should incur WCC through an arbiter CCA(Crosstalk Class Arbiter).In the case of both original and inverted codes in any sub-bus incurring WCC,a shielding pattern is transmitted;otherwise the WCC-free and energy saving code is transmitted.The proposed scheme is evaluated using the SPEC benchmarks.The results show that the proposed scheme improves the throughput by 62.59% to 81.62% over the un-coded approach and reduces the power consumption by 14.63% to 54.67% compared to the other similar schemes while eliminating the WCC with only 7 wires overhead for a 32 bit bus.The scheme achieves a good enhancement in dynamic power,wiring overhead,and performance.

关 键 词:总线编码 时空编码 串扰延迟 吞吐率 低功耗 

分 类 号:TN402[电子电信—微电子学与固体电子学]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象