Hardware codec acceleration CSI interface camera on TX2

1.TX2 Upper CSI Hardware

1. 1. The official development board puts the six routes CSI Received a 60*2 The so-called interface of camera development Given an onboard camera , Trouble . If you want to use these six interfaces by yourself , A transfer interface 145 bye . So you have to spend some money on an expanded version .
   
   

2. About camera driver .
   Nvidia There will be an official TX2 deriver package, There is a part of the camera driver, which can be used in NVL4T docs View in the document .
   
   You can buy cameras according to these models . Although they are cheap .
   NVidia It provides a convenient library so that we can easily install the corresponding sensor The driver . As an invitation to students to buy only 48 One of them ov. I studied it how to add a module.
   

In applications that support a direct V4L2 interface, use this interface to communicate to the NVIDIA V4L2 driver without having to use the camera core user mode library. Use this path for capturing RAW data from sensor or for validating sensor drivers.
The application uses the kernel mode V4L2 drivers as follows:

• To add one or more camera modules to a device tree
• 1Locate or create a tegra-camera-platform device node in the kernel source tree at:

/hardware/nvidia/platform/t19x/common/kernel-dts/t19x-common-modules/tegra194-camera-imx185-a00.dtsi

• 2In a Tegra-camera-platform device node, create a module table (modules) with one or more modules.
  Each module must contain its basic information and the definition of the devices that are inside that module.

2. About hardware codec .

• 1. Video codec process
• 
• 2.Deepstream Hardware codec routines on .

stay deepstream userGuide There is an introduction to this routine in .Nvidia It has independent hardware decoding module DEC
samples Inside DECPERF It is about the routine of hardware codec .

Workflow ：

1. Feed packets to the pipeline
2. Add decoding task
3. Analyze decoding performance
   
   1. First step : Create a DeviceWorker.

DeviceWorker Responsible for the whole DeepStream Workflow . It includes multi-channel decoding and analysis pipeline maintenance .DeviceWorker from createDeviceWorker Function creation , With multiple channels （g_nChannels） and GPU ID（g_devID） As a parameter .`

IDeviceWorker *pDeviceWorker = createDeviceWorker(g_nChannels, g_devID, g_maxBatchSize); 
 

2. The second step ： data input

more data can be load. 
bool getData(uint8_t **ppBuf, int *pnBuf);
 // reset the dataProvider to reload from the video file from the beginning. 
void reload(); 

3. The third step ： Add decoding task

// Add decode task, the parameter is the format of codec. 
pDeviceWorker->addDecodeTask(cudaVideoCodec_H264); 
 

4 function DeviceWorker

// Start and stop the DeviceWorker. 
pDeviceWorker->start();
 pDeviceWorker->stop();