Design Space Optimization of Shared Memory Architecture in Accelerator-rich Systems

Shared memory architectures, as opposed to private-only memories, provide a viable alternative to meet the ever-increasing memory requirements of multi-accelerator systems to achieve high performance under stringent area and energy constraints. However, an impulsive memory sharing degrades performance due to network contention and latency to access shared memory. We propose the Accelerator Shared Memory (ASM) framework to provide an optimal private/shared memory configuration and shared data allocation under a system’s resource and network constraints. Evaluations show ASM provides up to 34.35% and 31.34% improvement in performance and energy, respectively, over baseline systems.

Contention-Aware Mapping and Scheduling Optimization for NoC-Based MPSoCs (Student Abstract)

We consider spacial and temporal aspects of communication to avoid contention in Network-on-Chip (NoC) architectures. A constraint model is constructed such that the design concerns can be evaluated, and an efficient evolutionary algorithm with various heuristics is proposed to search for better solutions. Experimentations from random benchmarks demonstrate the efficiency of our method in multi-objective optimization and the effectiveness of our techniques in avoiding network contention.

PIF and ReCiF: Efficient Interest-Packet Forwarding Mechanisms for Named-Data Wireless Mesh Networks

In this paper, we propose three mechanisms to reduce the broadcast storm problem in wireless mesh networks based on the Named-Data Network (NDN) architecture. The goal of our mechanisms is to reduce the number of content requests forwarded by nodes and consequently, increase the network efficiency. The first proposed mechanism, called Probabilistic Interest Forwarding (PIF), randomly forwards content requests. The second mechanism, called Retransmission-Counter-based Forwarding (ReCIF), decides to forward content requests based on the number of retransmissions by adding a counter to the header of requests. The third mechanism, called ReCIF+PIF, combines the features of PIF and ReCIF to suppress content requests. We compare the performance of our mechanisms with both the NDN default forwarding mechanism and the Listen First Broadcast Later (LFBL) mechanism. Our proposals outperform the default NDN forwarding mechanism by up to 21% regarding the data delivery rate in dense networks and provide a 25% lower delivery delay than the default NDN. Our mechanisms accomplish this performance by only reducing the number of content requests forwarded by nodes. One of our mechanisms, PIF, outperforms LFBL regarding the data delivery rate and delivery delay by up to 263% and 55%, respectively, for high network contention levels.

A new adaptive data transfer library for model coupling

Data transfer means transferring data fields from a sender to a receiver. It is a fundamental and frequently used operation of a coupler. Most versions of state-of-the-art couplers currently use an implementation based on the point-to-point (P2P) communication of the message passing interface (MPI) (referred to as “P2P implementation” hereafter). In this paper, we reveal the drawbacks of the P2P implementation when the parallel decompositions of the sender and the receiver are different, including low communication bandwidth due to small message size, variable and high number of MPI messages, as well as network contention. To overcome these drawbacks, we propose a butterfly implementation for data transfer. Although the butterfly implementation outperforms the P2P implementation in many cases, it degrades the performance when the sender and the receiver have similar parallel decompositions or when the number of processes used for running models is small. To ensure data transfer with optimal performance, we design and implement an adaptive data transfer library that combines the advantages of both butterfly implementation and P2P implementation. As the adaptive data transfer library automatically uses the best implementation for data transfer, it outperforms the P2P implementation in many cases while it does not decrease the performance in any cases. Now, the adaptive data transfer library is open to the public and has been imported into the C-Coupler1 coupler for performance improvement of data transfer. We believe that other couplers can also benefit from this.