Support Tengine, Rock960 based on RK3399 performance optimization, speed increased by 50%!

Based on micro-network news, Tengine is a lightweight modular high-performance neural network inference engine developed by OPEN AI LAB. It is optimized for Arm embedded devices and does not need to rely on third-party libraries. It can be used across platforms to support Android, Liunx. Rockchip RK3399 based Rock960 is the first to support Tengine, performance improvement, running speed increased by more than 50%.

Rock960 platform, based on Rockchip RK3399, has powerful performance and rich expansion interface. It adopts exquisite and integrated design and can be applied in many innovative fields, such as IOT equipment, intelligent interactive equipment, personal computer, robot, etc. It has the following characteristics:

CPU: RK3399, dual-core Cortex-A72+ quad-core Cortex-A53, optional 2G/16G, 4G/32G

GPU: Mali T860MP4, OpenGL 3.2, OpenCL 1.2

HDMI 2.0 4K@60

USB type C (supports DP 1.2 protocol, can connect DP display in typeC, and can also use TypeC to HDMI to realize dual HDMI screen dual screen display)

Support three systems:

AndroidASOP+debian+Ubuntu

Small size (85mm x 55mm)

Tengine supports a variety of common convolutional neural networks, including SqueezeNet, MobileNet, AlexNet, ResNet, etc., support layer fusion, 8-bit quantization and other optimization strategies. And by calling HCL libraries optimized for different CPU micro-architectures, the performance of the Arm CPU is sufficient. Excavation. As in the RK3399 platform Cortex-A72 single-threaded mobile terminal commonly used MobileNet only needs 111ms.

As an open computing platform, Tengine has a very strong expansion capability. In addition to supporting the built-in HCL library, it also supports the common Openblas and Arm Compute Library libraries. Customers can easily expand their own operator libraries and give users freedom. In addition, TEngine can also adapt to various hardware resources such as GPU on the chip or even DLA hardware accelerator. Simply register the device on TEngine and mount the driver, you can make full use of the hardware on the chip. Resource improvement computing power.

Tengine's blessing is a milestone for the AI ​​open source ecosystem, which will greatly promote the development of technology and product development in the AI ​​field.

Part of the data is taken from the OPEN AI LAB Open Intelligence Lab

2016 GoodChinaBrand | ICP: 12011751 | China Exports