Parallel Multipath Transmission for Burst Traffic Optimization in Point-to-Point NoCs

Abstract

Network-on-chip (NoC) is a promising solution to connect more than hundreds of processing elements (PEs). As the number of PEs increases, the high communication latency caused by the burst traffic hampers the speedup gained by computation acceleration. Although parallel multipath transmission is an effective method to reduce transmission latency, its advantages have not been fully exploited in previous works, especially for emerging point-to-point NoCs since: (i) Previous static message splitting strategy increases contentions when traffic loads are heavy, degrading NoC performance. (ii) Only limited shortest paths are chosen, ignoring other possible paths without contentions. (iii) The optimization of hardware that supports parallel multipath transmission is missing, resulting in additional overhead. Thus, we propose a software and hardware collaborated design to reduce latency in point-to-point NoCs through parallel multipath transmission. Specifically, we revise hardware design to support parallel multipath transmission efficiently. Moreover, we propose a reinforcement learning-based algorithm to decide when and how to split messages, and which path should be used according to traffic loads. Experiments show that our algorithm achieves a remarkable performance improvement (+12.1% ∼ +21.0%) when compared with the state-of-the-art dual-path algorithm. Also, our hardware decreases power and area consumption by 23.2% and 10.3% over the dual-path hardware.

MUGNoC

“…If the XY and YX routes are used, the performance is greatly inuenced by the scheduling order, as shown in Fig.8(a) and (b). Moreover, multipath parallel transmission which has been investigated in[4] inspires us to transmit data to one destination through dierent paths, e.g., memory takes the path 08-07-04 and 08-05-04 to transmit data to PE 04, as shown in 8(c). Although it cannot further decrease runtime compared with the case in 8(b), it performs well in any scheduled orders since content-free is guaranteed by selecting two discontented routes for one message.…”

“…Although it cannot further decrease runtime compared with the case in 8(b), it performs well in any scheduled orders since content-free is guaranteed by selecting two discontented routes for one message. Also, the hardware to support multipath parallel transmission in ArSMART has been presented in[4] already.However, unicast with a xed source needs a dedicated routing strategy. Thus, we propose the routing algorithm as shown in Algo.…”