Artykuł techniczny

BLUEDOT: Accelerating NN-based DeepField-PQO design using Catapult HLS

Picture of BLUEDOT’s AI-based DeepField-PQO IP

Recently, the proportion of video in mobile traffic has been increasing exponentially. This video traffic growth trend is due to the development of Internet/video technology and changes in video consumption patterns.

Video service platform companies must bear a lot of costs to encode with better picture quality and smaller capacity. To solve this problem, BLUEDOT developed DeepField-PQO, an AI-based CODEC preprocessing filter. With the existing development method and as AI-based algorithms become more complex, design and verification take a lot of time, limiting the development period. To solve this, we introduced HLS to FPGA and developed it, and we introduced Catapult HLS to target ASIC.

With the introduction of Catapult HLS, we were able to flexibly respond to spec changes to improve performance, and shorten the overall development period through easier collaboration, verification, and code reuse.

Udostępnij

Powiązane treści

NVIDIA: High-Level Synthesis in Agile System-on-Chip Flows: Overview and Techniques
Webinar

NVIDIA: High-Level Synthesis in Agile System-on-Chip Flows: Overview and Techniques

This talk provides a brief overview of NVIDIA Research’s use of Catapult HLS and highlights some useful features and flows of the Connections library, such as the ability to back-annotate SystemC simulations.

LG Electronics: Video Encoder IP Design Optimization and Verification Using Catapult
Webinar

LG Electronics: Video Encoder IP Design Optimization and Verification Using Catapult

In this session, LGE describes a new design approach which concluded that adopting Catapult in IP development increases efficiency in time and cost, and they plan to increase usage in future IP projects.