Prerequisites for Generating Code for NVIDIA Boards
To use NVIDIA® Jetson™ or NVIDIA DRIVE® boards with MATLAB®, you must install required software and hardware. You must use a supported hardware board and install an operating system on the board. You must also install required third-party libraries and set up environment variables. To target GPU devices, you must also install GPU Coder™.
On Jetson boards, use the Hardware Setup tool to install required third-party libraries and set environment variables on the board. You can use the tool to install optional software that provides additional functionality. (since R2026a)
Supported Hardware Boards
The MATLAB Coder™ Support Package for NVIDIA Jetson and NVIDIA DRIVE Platforms supports these NVIDIA development kits:
NVIDIA Jetson AGX Thor™
NVIDIA Jetson AGX Orin™
NVIDIA Jetson Orin™ Nano
NVIDIA Jetson Xavier™ NX
NVIDIA Jetson Nano™
NVIDIA Jetson AGX Xavier
NVIDIA Jetson TX2
NVIDIA IGX Orin™
NVIDIA Clara™ AGX™
NVIDIA DRIVE PX 2
To execute commands, build generated code, and run generated code on the Jetson or DRIVE platform, the support package uses an SSH connection over TCP/IP. Connect the target hardware to the same network as the host computer, or use an Ethernet crossover cable to connect the board directly to the host computer.
Note
To connect to the MATLAB server running on the board from a Windows® platform, you must open port 18735 in the
Windows Firewall settings.
Custom Hardware Boards
You can also connect to and deploy code to a custom hardware board that contains one of these Jetson modules:
Jetson AGX Thor
Jetson Orin NX
Jetson Orin Nano
Jetson AGX Xavier
Jetson Xavier NX
Jetson TX2
Jetson Nano
When you connect to a hardware board, the support package uses the GPIO pin mapping from the NVIDIA developer kit for the Jetson module. To use GPIO features, use an NVIDIA developer kit board instead of a custom board.
NVIDIA Operating System and Libraries
Use the NVIDIA SDK Manager software to install the OS image, developer tools, and the libraries required for developing applications on Jetson or DRIVE platforms. To download the NVIDIA SDK Manager, see SDK Manager on the NVIDIA website.
NVIDIA Jetson
For Jetson boards, use the SDK Manager to install NVIDIA JetPack™. You must install:
Jetson Linux®
CUDA® Toolkit
GStreamer library version 1.0 or higher
Optionally, to generate code that uses NVIDIA deep learning libraries, install the NVIDIA CUDA Deep Neural Network (cuDNN) and NVIDIA TensorRT™ libraries.
The MATLAB Coder Support Package for NVIDIA Jetson and NVIDIA DRIVE Platforms supports these NVIDIA JetPack versions.
| Hardware Platform | Software Version |
|---|---|
Jetson AGX Thor | JetPack 7.0 Note Generating code that uses the NVIDIA TensorRT library is not supported. |
| IGX Orin | IGX-SW 1.1 To install IGX software, see IGX Download Center on the NVIDIA website. |
| Jetson AGX Orin | JetPack 6.0 By default, JetPack 6 sets the GPIO pins of the board to only receive input. To use a GPIO pin for output, learn how to change the pinmux of the board in the NVIDIA documentation. |
Jetson Orin™ Nano | |
Jetson Orin NX | |
Jetson AGX Xavier | JetPack 5.1.1 |
Jetson Xavier NX | |
Clara AGX | Holoscan SDK Note
|
Jetson Nano | JetPack 4.6.1 |
Jetson TX2 |
To deploy CUDA code generated by GPU Coder to a Jetson GPU with compute capability 11.0, such as Jetson AGX Thor, specify the compute capability in the code generation
configuration. To set the compute capability in the GPU Coder app, set the Custom
compute capability (GPU Coder) parameter to
-arch=sm_110. Alternatively, in Simulink®, set the Custom
compute capability (GPU Coder) model configuration parameter to
-arch=sm_110.
NVIDIA DRIVE
For DRIVE boards, install:
NVIDIA DriveOS™ SDK version 5.0.10.3-12606092
CUDA Toolkit
Required Software and Environment Variables
After installing the operating system, you must install a minimum set of software libraries on the hardware board. You must also set the required environment variables so the support package can locate the required libraries. The process to install the software differs for Jetson and DRIVE boards:
For Jetson boards, use the Hardware Setup tool to install the required libraries on the board.
For DRIVE boards, open the terminal on the board and install the libraries by using terminal commands. See Install Required Libraries on NVIDIA Boards Using the Terminal.
Input Devices
To record video or audio on the hardware board, connect:
A camera to the USB or CSI port
An audio device to the USB port
Development Host Requirements
To use a hardware board with MATLAB, install this software on the development host computer:
MATLAB
MATLAB Coder
MATLAB Coder Support Package for NVIDIA Jetson and NVIDIA DRIVE Platforms
For processor-in-the-loop (PIL) execution, you must also install a supported host compiler. For supported compilers, see Supported and Compatible Compilers.
Optional MathWorks Software
To target the GPU on the hardware board, generate code for deep learning applications, or use the board with Simulink, you must install additional MathWorks® software. To generate CUDA code targeting the hardware board, install:
Parallel Computing Toolbox™
GPU Coder
To verify the GPU code generation environment on the hardware board, use the
GPU
Environment Check (GPU Coder) app or the coder.checkGpuInstall (GPU Coder) function.
For deep learning applications, install:
Deep Learning Toolbox™
MATLAB Coder Interface for Deep Learning
GPU Coder Interface for Deep Learning
To deploy Simulink models on the board, install:
Simulink
Simulink Coder
Embedded Coder®, Computer Vision Toolbox™, and Image Processing Toolbox™ are recommended.
