Error with codegen for Jetson Nano; issue with main.cu creation.

I am able to connect to the Jetson Nano (Developers Kit) and retrieve an image to display on host computer:

https://www.mathworks.com/help/supportpkg/nvidia/ref/camera.html

Problem comes when trying to run this example:

https://www.mathworks.com/help//supportpkg/nvidia/ref/imagedisplay.imagedisplay.html

Wrote the liveCapture.m and the "codegen" step fails with:

    ??? codegen\exe\liveCapture\examples\main.cu specified in custom source
    files string does not exist in any of the following search directories:
    C:\Users\charl\Documents\MATLAB\Jetson\codegen\exe\liveCapture\examples
    C:\Users\charl\Documents\MATLAB\Jetson\codegen\exe\liveCapture\examples\codegen\exe\liveCapture
    C:\Users\charl\Documents\MATLAB\Jetson\codegen\exe\liveCapture\examples\codegen\exe\liveCapture\examples
    C:\Users\charl\Documents\MATLAB\Jetson\codegen\exe\liveCapture\examples
    
    Code generation failed: View Error Report
    Error using codegen
    
    
    Error in Code2Jetson (line 16)
    codegen -config cfg liveCapture

I have tried:

1) Running example as described without creating any main.cu

2) Putting in a blank "main.cu" into "codegen\exe\liveCapture\examples"

3) Putting in a populated "main.cu" into "codegen\exe\liveCapture\examples"

- Used this as a guide: https://www.mathworks.com/help//coder/ug/generate-and-modify-an-example-cc-main-function.html

- When the codegen step was run, and failed, the main.cu I wrote had been erased

All attempts produced same output error.

Addons Installed

GPU Coder Interface for Deep Learning Libraries

GPU Coder Support Package for NVIDIA GPUs

Compiler being used:

mex -setup
MEX configured to use 'Microsoft Visual C++ 2017 (C)' for C language compilation.

Environment variables and packages have been installed:

https://www.mathworks.com/help/gpucoder/gs/setting-up-the-toolchain.html

Compatible GPU           : PASSED 
CUDA Environment         : PASSED 
	Runtime   : PASSED 
	cuFFT     : PASSED 
	cuSOLVER  : PASSED 
	cuBLAS    : PASSED 
cuDNN Environment        : PASSED 
TensorRT Environment     : PASSED 
Basic Code Generation    : PASSED 
Basic Code Execution     : PASSED 
Deep Learning (TensorRT) Code Generation: PASSED 
00007FF9EE15F6E8cuBLAS 10.2.0 but is linked against cuBLAS 10.1.0
00007FF9EE15F6E8cuBLAS 10.2.0 but is linked against cuBLAS 10.1.0
00007FF9EE15F6E8cuBLAS 10.2.0 but is linked against cuBLAS 10.1.0
Deep Learning (TensorRT) Code Execution: PASSED 

Answer 1

[Charles Taylor]

Update, the sobelEdgeDetection example ran fine. It uses a slightly different "cfg" object description and declared build directory. It doesn't use the CustomInclude or CustomSource fields.

Made changes to the liveCapture.m and codegen to emulate the sobelEdgeDetection example and it worked just fine:

function liveCapture() %#codegen

% Copyright 2018-2020 The MathWorks, Inc.

hwobj = jetson;
camObj = camera(hwobj,"vi-output, imx219 8-0010",[1280 720]);
dispObj = imageDisplay(hwobj);


% Main loop
for k = 1:10000
    % Capture the image from the camera on hardware.
    img = snapshot(camObj);
    
    % Display image.
    image(dispObj,img);
end

end

With the codegen:

%% Setup

disp('Building cfg object....')

cfg = coder.gpuConfig('exe');
cfg.Hardware = coder.hardware('NVIDIA Jetson');
cfg.Hardware.BuildDir = '~/remoteBuildDir';
cfg.GenerateExampleMain = 'GenerateCodeAndCompile';

disp('...done with cfg object')

%% compile
disp('Starting codegen....')

% codegen -config cfg liveCapture  %from example on site

codegen('-config ',cfg,'liveCapture','-report');

disp('...completed codegen')


%% Run on Target

hwobj.setDisplayEnvironment('1.0');
pid = runApplication(hwobj,'liveCapture');