English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Neuromorphic Systems-on-Chip (NSoCs) are becoming heterogeneous byintegrating general-purpose processors (GPPs) and neural processing units(NPUs) on the same SoC. For embedded systems, an NSoC may need to execute userapplications built using a variety of machine learning models. We propose areal-time scheduler, called PRISM, which can schedule machine learning modelson a heterogeneous NSoC either individually or concurrently to improve theirsystem performance. PRISM consists of the following four key steps. First, itconstructs an interprocessor communication (IPC) graph of a machine learningmodel from a mapping and a self-timed schedule. Second, it creates atransaction order for the communication actors and embeds this order into theIPC graph. Third, it schedules the graph on an NSoC by overlappingcommunication with the computation. Finally, it uses a Hill Climbing heuristicto explore the design space of mapping operations on GPPs and NPUs to improvethe performance. Unlike existing schedulers which use only the NPUs of an NSoC,PRISM improves performance by enabling batch, pipeline, and operationparallelism via exploiting a platform's heterogeneity. For use-cases withconcurrent applications, PRISM uses a heuristic resource sharing strategy and anon-preemptive scheduling to reduce the expected wait time before concurrentoperations can be scheduled on contending resources. Our extensive evaluationswith 20 machine learning workloads show that PRISM significantly improves theperformance per watt for both individual applications and use-cases whencompared to state-of-the-art schedulers.

Real-Time Scheduling of Machine Learning Operations on Heterogeneous Neuromorphic SoC