GPU Coder Main File

Hello, I am currently trying to deploy some codes from MATLAB to Jetson Nano Developer Kit, I am using GPU Coder Application, however; I am encountering a very serious issue right here. I tried working with some basic codes firstly, addition or anything, It works perfectly because the matlab example has a main.cu file attached with it, which should be easy to use. Basically the main.cu is a code that allows me to enter the point function, now.

What I am actually going to deploy to JETSON in reality is

• Camera Fusion with Lidar Function

How am I going to write that main.cu file now? Can someone help me that if any application takes the matlab function and generates the main.cu file?

Reference to what I mean by main.cu --> https://www.mathworks.com/help/gpucoder/ug/build-run-on-nvidia-hardware-using-app.html

Answer 1

[Ram Kokku]

Hi Ahmed,

The NVIDIA hardware support package provides functionality for accessing various I/O devices, including cameras and Velodyne LiDAR. You can refer to the examples listed below to write MATLAB code for your I/O and generate code using GPU Coder.

Camera example: https://www.mathworks.com/help/gpucoder/ug/sobel-edge-detection-using-webcam-on-nvidia-jetson.html

Lidar doc and example: https://www.mathworks.com/help/coder/nvidia/ref/velodynelidar.html?s_tid=doc_ta

[Ahmed Tamer]

I will give them a try tommorow.

• Can I run the generated excutable file on Jetson itself?
• Is there an ability to generate an excutable code for a code I wrote myself? The issue here is that I would need to create a main.cu file for my own code, which I dont know how to, you get my issue?

[Ram Kokku]

Hi Ahmed,

Yes, you can run the generated executable on the Jetson. There are several ways to do this. You can manually copy the executable from the 'codegen' directory to the Jetson board and run it, or you can utilize the functionality of the NVIDIA Hardware Support Package to do so.

Regarding generating an executable, yes, you can also accomplish this by creating an executable code generation configuration using cfg = coder.gpuConfig('exe'). For more information, refer to the documentation - https://www.mathworks.com/help/gpucoder/ref/coder.gpucodeconfig.html

[Ahmed Tamer]

Yes I do get it, however my main issue still persists. I do need a main.cu code for each function that I generate an excutable for, look here at this step, I have got to have the main.cu file. For this example, myAdd function, the main.cu was attached to the example.

Now, for the own function of mine, how am I going to create a main.cu file so I can attach it?

[Ahmed Tamer]

I fixed it. Thanks