// 5. Setup CAN interface, I use 125k baudrate. $ sudo ip link set can0 up type can bitrate 125000 $ sudo ifconfig can0 Link encap:UNSPEC HWaddr 00-00-00-00-00-00-00-00-00-00-00-00-00-00-00-00 UP RUNNING NOARP MTU:16 Metric:1 RX packets:0 errors:0 dropped:0 overruns:0 frame:0 TX packets:0 errors:0 dropped:0 overruns:0 carrier:0 collisions:0 txqueuelen:10 RX bytes:0 (0.0 B) TX bytes:0 (0.0 B)
1.1 Overview Knowledge Processing Unit (KPU, aka Neural network Processing Unit). KPU is a general-purpose neural network processor that implements convolutional neural network calculations with low power consumption. It can acquire the size, coordinates and types of detected objects in real time, and detect and classify faces or objects. When using kpu, the neural network model must be generated with the model compiler. 1.2 Features KPU has the following characteristics: • Supports the fixed training model that the common training framework trains according to specific restriction rules • There is no direct limit on the number of network layers, which supports separate configuration of each layer of convolutional neural network parameters, including the number of input and output channels, input and output line width and column height. • Support for two convolution kernels 1x1 and 3x3 • Support for any form of activation function • The maximum supported neural network parameter size in real-time work is 5.5MiB to 5.9MiB • Maximum support network parameter size when working in non-real time is (Flash capacity) - (software size)
2.1 Overview Advanced Encryption Standard (AES) acceleration engine. The AES module uses hardware to implement AES operation acceleration. 2.2 Features K210 have a built-in AES(Advanced Encryption Standard acceleration engine). Compared with software, it can greatly improve the speed of AES operation. The AES accelerator supports multiple encryption/decryption modes (ECB, CBC, GCM) and multiple length of KEY (128, 192, 256).
7.1 Overview Digital Video Port (DVP) unit is a camera interface unit that supports forwarding camera input image data to KPU or memory. 7.2 Features The DVP unit has the following features: • Support RGB565, RGB422 and single channel Y gray scale input mode • Support for setting frame interrupt • Support setting transfer address • Supports writing data to two addresses at the same time (output format is RGB888 and RGB565 respectively) • Support for discarding frames that do not need to be processed
8.1 Overview Fast Fourier Transform (FFT) Accelerator. The FFT accelerator implements the radix-2 decimation-in-time (DIT) Cooley–Tukey FFT algorithm* 1 acceleration in hardware. 8.2 Features The FFT accelerator currently supports 64-point, 128-point, 256-point, and 512- point FFTs and IFFTs. Inside the FFT accelerator, there are two SRAMs with a size of 512 * 32 bits. After the configuration is completed, the FFT sends a TX request to the DMA, and the data sent by the DMA is placed in one of the SRAMs until the data volume satisfies the current FFT operation needs, and the FFT operation begins at this point. The butterfly operation unit reads data from the SRAM which containing the valid data, and writes the data to another SRAM after the end of the operation. The next butterfly operation reads the data from the SRAM just written, when the operation ends write to another SRAM. This process alternates this way until the entire FFT operation is completed.
9.1 Overview Secure Hash Algorithm (SHA) accelerator supports hardware acceleration of the sha256 algorithm. 9.2 Features • Supports hardware acceleration of the sha256 algorithm