2020 BRN Discussion, page-28212

Hi Slade,
We looked at the R-Car-V3U a few days ago.
It uses several off-the-shelf processors and , tantalisingly, "dedicated CNN IP".

https://www.renesas.com/sg/en/products/automotive-products/automotive-system-chips-socs/r-car-v3u-best-class-r-car-v3u-asil-d-system-chip-automated-driving

R-Car-V3U Active Best-in-Class R-Car V3U ASIL D System-on-Chip for Automated Driving

The R-Car V3U SoC, based on the R-Car Gen 4 architecture, is the latest member of the open and innovative Renesas autonomy platform for ADAS and AD. The platform offers scalability from entry-level NCAP applications up to highly automated driving systems with the R-Car V3U providing up to 96k DMIPS and 60TOPS. The SoC integrates multiple sophisticated safety mechanisms that provide high coverage with fast detection and response for random hardware faults, achieving ASIL D metrics for the majority of the SoC processing chain, as well as reducing design complexity, time to market and system cost.

It delivers highly flexible DNN (Deep Neural Processing) and AI machine learning functions. Its flexible architecture can handle any state-of-the-art neural networks providing up to 60 TOPS with low power consumption allowing air cooling systems.

The R-Car V3U comes with an open and integrated development environment that enables customers to take advantage of the R-Car platform’s built-in hardware benefits, as well as low power consumption and deterministic real-time software to enable fast time-to-market for computer vision and deep learning-based solutions.

Features
8 Arm® Cortex®-A76 for application programming, configurable in LockStep Mode
Dual Cortex-R52 LockStep cores to run AUTOSAR
Dedicated CNN IP delivering 60 TOPS, capable of handling any state-of-the-art neural networks for automated driving
High-speed processing of algorithms like optical flow, object detection and classification
Dual DSP for Radar processingIntegrated ISP enables high level of integration for reduced cost