The Adaptive

The design of a new adaptive virtual cut-through router for torus networks is presented in this paper. With much lower VLSI costs than adaptive wormhole routers, the adaptive Bubble router is even faster than deterministic wormhole routers based on virtual channels. This has been achieved by combining a low-cost deadlock avoidance mechanism for virtual cut-through networks, called Bubble flow control, with an adequate design of the router's arbiter. A thorough methodology has been employed to quantify the impact that this router design has at all levels, from its hardware cost to the system performance when running parallel applications. At the VLSI level, our proposal is the adaptive router with the shortest clock cycle and node delay when compared with other state-of-the-art alternatives. This translates into the lowest latency and highest throughput under standard synthetic loads. At system level, these gains reduce the execution time of the benchmarks considered. Compared with current adaptive wormhole routers, the execution time is reduced by up to 27%. Furthermore, this is the only router that improves system performance when compared with simpler static designs. 1. INTRODUCTION Parallel and distributed computing performance is generally a fraction of the sum of the computational power provided by the processors that make up the parallel system. This performance degradation can be attributed to two sources: i) the impossibility of adequately balancing the computing load among processors, and ii) the overheads involved in communication between processors and memories. An efficient mapping can be found for a number of applications, resulting in a uniform computational load among processors. However, the cost of communication is unavoidable and we can only do our best to minimize it. Scalable distributed shared-memory multiprocessors with hardware data cache coherence (cc-NUMA) are nowadays the trend in building parallel computers because they provide the most desirable programmability [20]. Other existing distributed shared-memory parallel