Proceedings of the 39th annual Design Automation Conference
Repeater insertion and wire sizing optimization for throughput-centric VLSI global interconnects
Proceedings of the 2002 IEEE/ACM international conference on Computer-aided design
Networks on chip
Interconnect-power dissipation in a microprocessor
Proceedings of the 2004 international workshop on System level interconnect prediction
Replacing global wires with an on-chip network: a power analysis
ISLPED '05 Proceedings of the 2005 international symposium on Low power electronics and design
ISQED '07 Proceedings of the 8th International Symposium on Quality Electronic Design
INTACTE: an interconnect area, delay, and energy estimation tool for microarchitectural explorations
CASES '07 Proceedings of the 2007 international conference on Compilers, architecture, and synthesis for embedded systems
Prediction of high-performance on-chip global interconnection
Proceedings of the 11th international workshop on System level interconnect prediction
Low Power Design Essentials
A global wiring paradigm for deep submicron design
IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
Hi-index | 0.00 |
Due to the ever increasing demand for computing capacity, throughput-centric design for on-chip global interconnects has played an important role in the emerging parallel computing architectures. In this paper, we explore the performance of flip-flop-based pipelined global interconnects with more design freedoms under the voltage and technology scaling for different applications. Based on the derived accurate voltage-scaled models of pipelined interconnects, we propose a general evaluation flow using numerical experiments to study the impact of pipelining depth, voltage scaling, and different processes on the performance of pipelined interconnects under four different design objectives. Our experimental results show that, with the dedicated throughput-centric optimization, at 45 nm node, up to 25x overall throughput-per-energy-area (TPEA) improvement can be obtained with only 4x increase on the interconnect latency compared with the conventional minimum-latency design, making this new design methodology more promising in the future nodes.