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 Plus annual

Presents the access of premium content as premium feature

Premium Content

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 Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Single computation engines have become a popular design choice for FPGA-basedconvolutional neural networks (CNNs) enabling the deployment of diverse modelswithout fabric reconfiguration. This flexibility, however, often comes withsignificantly reduced performance on memory-bound layers and resourceunderutilisation due to suboptimal mapping of certain layers on the engine'sfixed configuration. In this work, we investigate the implications in terms ofCNN engine design for a class of models that introduce a pre-convolution stageto decompress the weights at run time. We refer to these approaches ason-the-fly. To minimise the negative impact of limited bandwidth onmemory-bound layers, we present a novel hardware component that enables theon-chip on-the-fly generation of weights. We further introduce an inputselective processing element (PE) design that balances the load between PEs onsuboptimally mapped layers. Finally, we present unzipFPGA, a framework to trainon-the-fly models and traverse the design space to select the highestperforming CNN engine configuration. Quantitative evaluation shows thatunzipFPGA yields an average speedup of 2.14x and 71% over optimised status-quoand pruned CNN engines under constrained bandwidth and up to 3.69x higherperformance density over the state-of-the-art FPGA-based CNN accelerators.

unzipFPGA: Enhancing FPGA-based CNN Engines with On-the-Fly Weights Generation