Jetson TK1 Four: Reinstall the system (brush machine)

Reprint: http://blog.sina.com.cn/s/blog_bab3fa030102vk21.html

Jetson TK1 is a low-cost development board based on Tegra K1 developed by NVIDIA, on board a Tegra K1 32-bit (Logan) chip, an HDMI output on the development Board, an Ethernet port, a USB 3.0, a micro USB port, Sata,mini PCIE,SD card slots, the debug port has a serial port DB9 and JTAG, as well as a large number of IO interface leads. Can be said to be very perfectly formed.

The Tegra K1 has a Kepler display core with a desktop GPU architecture and a 192-core architecture, with peak floating-point arithmetic being nearly as many as dozens of times times the best single-core DSP available. After opening the Cuda interface of the mobile platform, Tegra K1 as a performance monster, will have more and better applications in the broader market.

When you get the JTK1 Development Board, it should be the default l4t (Linux for Tegra), the GStreamer pack and the Cuda 6.0 SDK are all installed. If you want to play it again, you can refer to the following steps:

1 Download and write L4T system
The latest release is 19.2, and all of this is done on the Linux machine.

Driver Package tegra124_linux_r19.2.0_armhf.tbz2

Root File System Package tegra_linux_sample-root-filesystem_r19.2.0_armhf.tbz2

Download the two tar packages from a PC and put them in a folder to extract the driver package first

sudo tar--numeric-owner-jxpf tegra124_linux_r19.2.0_armhf.tbz2
Note the red part, which needs to be added, is not given in the current release document. If you do not join, it will lead to the completion of the burn write, the system Ubuntu user rights problems. When finished, a Linux_for_tegra folder is generated in the current directory.

CD Linux_for_tegra/rootfs
sudo tar--numeric-owner-jxpf. /.. /tegra_linux_sample-root-filesystem_r19.2.0_armhf.tbz2
The root file System package is decompressed into the Rootfs folder generated by the driver package, sudo is best to add, otherwise there will be warning.

Cd.. /
sudo./apply_binaries.sh
Now use a micro USB cable to connect to the Development Board, press the recovery key does not release, and then press the Reset key, enter the burn-write mode, in the Linux_for_tegra directory

sudo./flash.sh-s 8GiB jetson-tk1 Mmcblk0p1
Patiently waiting for the burn to finish, restart the Development Board, you can enter the Ubuntu desktop environment, the default user name is Ubuntu, password Ubuntu. The corresponding l4t documentation and source code packages can be found here.

2 Download and install Cuda SDK for Tegra
If you re-burn the system, you need to find a USB card to connect to the JTK1 Development Board for the current step. Because after re-burning the system, through the onboard network connection, DNS resolution is problematic, cannot resolve the domain name, only through the IP address to access the network. You can access it directly on the board or download the Cuda SDK for Tegra on your PC before you copy it to the Development Board. But you need to register the NV developer account first, and apply for the development of Tegra and Cuda, are free of charge.
The file I downloaded is cuda-l4t-r19.2_6.0-42_armhf.deb, which is executed on the Development Board:

$ sudo dpkg-i cuda-l4t-r19.2_6.0-42_armhf.deb
$ sudo apt-get update
$ sudo apt-get install cuda-toolkit-6-0
$ sudo usermod-a-G video Ubuntu
When you are finished, add the/usr/local/cuda-6.0/bin path to path, add the/usr/local/cuda-6.0/lib path to the Ld_library_path, or configure Pkgconfig.

The following can compile the Cuda sample program,

$ cuda-install-samples-6.0.sh

If the previous environment has been set successfully, then make it into the generated folder.

3 Test Video Codec
L4T uses the standard GStreamer as the codec frame, while providing the NV hardware acceleration plug-in, here are a few examples

Coding

gst-launch-0.10 filesrc location=! Videoparse width=640 height=480 format=1 FRAMERATE=30/1! Nv_omx_h264enc! Qtmux! Filesink location=-V
Play

Nvgstplayer-i

Jetson TK1 Four: Reinstall the system (brush machine)