No abstract

No abstract

Recommended by Maurizio PalesiNetwork-on-chip-(NoC-) based application-specific systems on chip, where information traffic is heterogeneous and delay requirements may largely vary, require individual capacity assignment for each link in the NoC. This is in contrast to the standard approach of on-and off-chip interconnection networks which employ uniform-capacity links. Therefore, the allocation of link capacities is an essential step in the automated design process of NoC-based systems. The algorithm should minimize the communication resource costs under Quality-of-Service timing constraints. This paper presents a novel analytical delay model for virtual channeled wormhole networks with nonuniform links and applies the analysis in devising an efficient capacity allocation algorithm which assigns link capacities such that packet delay requirements for each flow are satisfied.

Abstract-Modern scientific discovery is driven by an insatiable demand for computing performance. The HPC community is targeting development of supercomputers able to sustain 1 ExaFlops by the year 2020 and power consumption is the primary obstacle to achieving this goal. A combination of architectural improvements, circuit design, and manufacturing technologies must provide over a 20× improvement in energy efficiency. In this paper, we present some of the progress NVIDIA Research is making toward the design of Exascale systems by tailoring features to address the scaling challenges of performance and energy efficiency. We evaluate several architectural concepts for a set of HPC applications demonstrating expected energy efficiency improvements resulting from circuit and packaging innovations such as low-voltage SRAM, low-energy signaling, and on-package memory. Finally, we discuss the scaling of these features with respect to future process technologies and provide power and performance projections for our Exascale research architecture.

No abstract