Why does my GPU set-up error with 'emlc:compilationError'?

When I run the setup command "coder.checkGpuInstall", it fails with "gpuEnvConfig" set to "host" but not "Jetson" when running a code generation test.

Answer 1

[MathWorks Support Team]

The host and Jetson are two different environments. In this workflow, code generation is performed on the host, and the resulting directory is copied over to the Jetson and built.

To view more information, run the check from the

GPU Coder setup GUI

to view the HTML report, or run the following checks on the command line to test basic and deep learning code generation:

% Basic Code gen
gpuEnvObj = coder.gpuEnvConfig;
gpuEnvObj.GpuId = 0;
gpuEnvObj.BasicCodegen = 1;
gpuEnvObj.BasicCodeexec = 1;
coder.checkGpuInstall(gpuEnvObj)

% Deep code gen
gpuEnvObj = coder.gpuEnvConfig;
gpuEnvObj.GpuId = 0;
gpuEnvObj.DeepLibTarget = 'cudnn';
gpuEnvObj.DeepCodegen = 1;
gpuEnvObj.DeepCodeexec = 1;
coder.checkGpuInstall(gpuEnvObj)

If  a check fails, there may be an issue with the host environment. Check that all of the following are present and correctly configured in the host environment: 

        1. MATLAB Coder and Parallel Computing Toolbox

        2. Required 3rd party products 

        3. Properly set up environment variables 

        4. A CUDA-enabled GPU (see the NVIDIA documentation). You can find the GPU information by calling the "gpuDeviceTable" function.

        5. The correct CUDA Toolkit Version for the MATLAB release 

        6. Up-to-date NVIDIA GPU card drivers