Compile with TensorRT Lib on Jetson Orin Nano Dev Kit in C++

Issue Overview

Users are encountering issues when attempting to use the TensorRT library in C++ programs on the Nvidia Jetson Orin Nano Developer Kit. Specifically, the error arises during the configuration phase with CMake, where the following error message is displayed:

CMake Error at CMakeLists.txt:18 (find_package):
  By not providing "FindTensorRT.cmake" in CMAKE_MODULE_PATH this project has
  asked CMake to find a package configuration file provided by "TensorRT",
  but CMake did not find one.
  
  Could not find a package configuration file provided by "TensorRT" with any
  of the following names:
  
    TensorRTConfig.cmake
    tensorrt-config.cmake

  Add the installation prefix of "TensorRT" to CMAKE_PREFIX_PATH or set
  "TensorRT_DIR" to a directory containing one of the above files.

Context

• Hardware: Jetson Orin Nano Developer Kit
• Software: Jetpack 5.1.1, CUDA 11.4.315, cuDNN 8.6.0.166, TensorRT 8.5.2.2
• Installation Method: SD Card Image Method 1
• Frequency: The issue appears consistently when users attempt to compile their projects using TensorRT.
• Impact: This issue prevents users from successfully compiling their applications that rely on TensorRT, hindering development and functionality.

Possible Causes

• Missing Configuration Files: The absence of TensorRTConfig.cmake or tensorrt-config.cmake files could indicate that TensorRT is not properly installed or configured.

• Incorrect Installation Path: The installation prefix for TensorRT may not be correctly set in CMAKE_PREFIX_PATH, leading CMake to fail in locating the necessary files.

• Development Package Not Installed: Users may not have installed the development package for TensorRT, which includes essential configuration files.

• User Configuration Errors: Users might not have correctly configured their CMakeLists.txt or may have overlooked necessary steps in setting up their environment.

Troubleshooting Steps, Solutions & Fixes

Step-by-Step Instructions

1. Verify TensorRT Installation:

   • Check if TensorRT is installed by running:

     dpkg-query -W tensorrt
     

   • Ensure it returns a version number (e.g., 8.5.2.2).

2. Locate Required Files:

   • Check if the required files exist:

     ls /usr/lib/aarch64-linux-gnu/libnvinfer*
     ls /usr/include/aarch64-linux-gnu/NvInfer*
     

   • If these files are missing, it indicates an incomplete installation.

3. Set CMAKE_PREFIX_PATH:

   • Add the installation prefix of TensorRT to your CMAKE_PREFIX_PATH:

     export CMAKE_PREFIX_PATH=/usr/lib/aarch64-linux-gnu/
     

4. Create FindTensorRT.cmake File:

   • If the above steps do not resolve the issue, create a custom FindTensorRT.cmake file similar to the one found here:
     FindTensorRT.cmake.
   • Place this file in the same directory as your CMakeLists.txt.

5. Modify CMakeLists.txt:

   • Ensure your CMakeLists.txt includes:

     find_package(TensorRT REQUIRED)
     

6. Run CMake Again:

   • After making these changes, run the following commands again in your project directory:

     mkdir build && cd build
     cmake ..
     make
     

Additional Recommendations

• Ensure all dependencies are installed and up-to-date.
• Regularly check for updates to Jetpack and associated libraries.
• Consult Nvidia’s documentation for any specific configurations required for your version of Jetpack and TensorRT.

Unresolved Aspects

While many users have successfully resolved their issues by following these steps, some may still encounter challenges related to specific configurations or environmental factors that require further investigation. Testing with different setups or consulting additional resources may be necessary for persistent issues.