RISC-V embedded cores for AI/ML edge customisation
Two RISC-V cores have been added to the low power embedded RISC-V processor cores by Codasip. The L31 and L11 are optimised for customisation. Processor designs can be customised using Codasip Studio tools to support tasks such as neural networks (artificial intelligence / machine learing or AI / ML) in small, power-constrained applications, such as IoT edge.
AI/ML running in edge IoT/IIoT devices improve security and power consumption, and reduce latency for real-time processing, said Codasip. Algorithms for AI/ML are computationally intensive and custom processors are needed to deliver sufficient performance with the limited resources available in embedded systems. The L31 and L11 embedded cores run Google’s TensorFlowLite for Microcontrollers with Codasip Studio tools to customise a new breed of Codasip’s Embedded AI cores for IoT applications where both space and power are at a premium.
AI and ML applications are not well suited to off-the-shelf processors, explained Codasip. The data types, the quantisation and performance needs of the devices differ significantly from application to application. Codasip’s Studio tools can be used to customise the processor for its specific system, software and application requirements. Similarly, embedded devices in low power IoT applications are extremely resource-constrained in terms of limited memory and a limited instruction set. Yet these devices need to be low power, inherently secure, and able to respond and communicate in real-time.
Custom instructions enabled via Codasip Studio RISC-V design tools are suitable to develop processors for AI/ML. The company said using TensorFlow Lite for Microcontrollers (TFLite Micro), RISC-V custom instructions and Codasip processor design tools can reduce latency, improve security, accelerate communication and reduce power consumption in edge neural networking processing. These benefits are essential for emerging IoT and Industrial IoT (IIoT) edge applications where the ability to run real-time AI/ML tasks is rapidly becoming a standard SoC feature, Codasip commented.
The L31 and L11 processor cores are the first to feature TFLite Micro support, although this support is available across Codasip’s portfolio of RISC-V cores.
