Carpool: a bufferless on-chip network supporting adaptive multicast and hotspot alleviation

Modern chip multiprocessors (CMPs) employ on-chip networks to enable communication between the individual cores. Operations such as coherence and synchronization generate a significant amount of the on-chip network traffic, and often create network requests that have one-to-many (i.e., a core multicasting a message to several cores) or many-to-one (i.e., several cores sending the same message to a common hotspot destination core) flows. As the number of cores in a CMP increases, one-to-many and many-to-one flows result in greater congestion on the network. To alleviate this congestion, prior work provides hardware support for efficient one-to-many and many-to-one flows in buffered on-chip networks. Unfortunately, this hardware support cannot be used in bufferless on-chip networks, which are shown to have lower hardware complexity and higher energy efficiency than buffered networks, and thus are likely a good fit for large-scale CMPs. We propose Carpool, the first bufferless on-chip network optimized for one-to-many (i.e., multicast) and many-to-one (i.e., hotspot) traffic. Carpool is based on three key ideas: it (1) adaptively forks multicast flit replicas; (2) merges hotspot flits ; and (3) employs a novel parallel port allocation mechanism within its routers, which reduces the router critical path latency by 5.7% over a bufferless network router without multicast support. We evaluate Carpool using synthetic traffic workloads that emulate the range of rates at which multithreaded applications inject multicast and hotspot requests due to coherence and synchronization. Our evaluation shows that for an 8×8 mesh network, Carpool reduces the average packet latency by 43.1% and power consumption by 8.3% over a bufferless network without multicast or hotspot support. We also find that Carpool reduces the average packet latency by 26.4% and power consumption by 50.5% over a buffered network with multicast support, while consuming 63.5% less area for each router.