iWave has incorporated the DRP-AI accelerator into its RZ/V2L System on Module (iW-RainboW-G56M), providing an optimized solution for edge AI applications. The DRP-AI accelerator dynamically reconfigures itself to optimize AI workload execution, ensuring high-speed inference performance while consuming significantly less power than conventional processing solutions and having the ability to respond to evolution. This innovative approach makes iWave’s RZ/V2L SoM an ideal choice for AI-powered embedded systems requiring real-time, efficient, and cost-effective AI processing.

What is DRP-AI?

DRP-AI is an advanced AI accelerator integrated to enhance AI processing efficiency. It dynamically reconfigures itself to execute AI tasks with high speed and minimal power consumption, ensuring optimal performance for embedded applications. With its intelligent workload distribution, DRP-AI enables real-time AI inference for applications such as machine vision, predictive analytics, and smart automation, where both efficiency and performance are critical.

Implementing DRP-AI on RZ/V2L SMARC SoM

Integrating the DRP-AI (Dynamic Reconfigurable Processor for AI) on the RZ/V2L SMARC System on Module (SoM) requires careful configuration of memory regions and device settings to ensure seamless AI inference performance. This process involves reserving dedicated memory spaces to support secure operations and efficient DRP-AI execution, optimizing data flow and computation.

A key aspect of this implementation is the configuration of a Direct Memory Access (DMA) buffer, which facilitates high-speed data transfers between memory and processing units, minimizing latency, and improving overall performance. Additionally, the DRP-AI device must be correctly linked to these allocated memory regions to enable proper functionality.

To activate DRP-AI on the RZ/V2L platform, essential software packages must be integrated into the YOCTO build system and compiled accordingly. Once successfully implemented, this setup allows developers to leverage hardware-accelerated AI inference on embedded systems, enabling advanced applications such as image recognition, object detection, and industrial automation.

Software Support: DRP-AI Support Package

The DRP-AI Support Package is a comprehensive development environment provided by Renesas to accelerate AI inference on DRP-AI hardware. It includes:

• DRP-AI Translator: Converts AI models from ONNX format into a DRP-AI executable format for efficient deployment.
• Linux Package: Provides necessary drivers and runtime libraries for Linux-based AI application development.
• Application Sample Code: Offers ready-to-use sample code and application development guidance for Linux.
• AI Evaluation Software: Assists in evaluating and optimizing AI models for DRP-AI execution.
• Implementation Guide: Covers the entire AI development workflow, from model conversion to deployment.
• DRP-AI Driver: Ensures seamless integration of AI processing on Linux-based platforms.

With these tools, developers can experience seamless AI development, from working with open software (ONNX) to deploying AI applications on embedded hardware.