SDK Helps Simplify FPGA-Based AI Implementation
Empowering sparse neural networks to improve performance and enable more efficient Edge AI on PolarFire FPGAs and SoCs, Microchip Technology's VectorBlox 3.0 Accelerator Software Development Kit (SDK) helps simplify implementation and speed time-to-market.
Available free of charge, the VectorBlox 3.0 SDK and associated CoreVectorBlox IP is an integrated toolchain that streamlines optimization, compilation, and deployment of convolutional-neural-network (CNN) models on PolarFire FPGA- and SoC-based platforms.
The solution scales efficiently across model sizes and supports multiple AI workloads on a single device. It can consolidate various vision or sensor-based AI functions on a single low-power FPGA. VectorBlox 3.0 builds on sparsity-based model compression from the company's Neuronix acquisition to reduce compute demands while preserving accuracy.
VectorBlox 3.0 promotes efficient execution of vision-based CNN models by skipping zero-valued operations, helping developers accelerate inference performance while reducing power consumption. Sparsity-based model compression reduces compute and memory demands, while preserving accuracy.
VectorBlox SDK v3.0, supported by Microchip’s Libero SoC Design Suite, integrates with CoreVectorBlox IP and the company’s full portfolio of FPGAs and design resources.
Related links:
About the Author
Alix Paultre
Editor-at-Large, Microwaves & RF
Alix is Editor-at-Large for Microwaves & RF.
An Army veteran, Alix Paultre was a signals intelligence soldier on the East/West German border in the early ‘80s, and eventually wound up helping launch and run a publication on consumer electronics for the U.S. military stationed in Europe. Alix first began in this industry in 1998 at Electronic Products magazine, and since then has worked for a variety of publications, most recently as Editor-in-Chief of Power Systems Design.
Alix currently lives in Wiesbaden, Germany.