Error executing command "touch -c /home/jets​on/remoteB​uildDir/MA​TLAB_ws/R2​019b/C/Use​rs/JSH/Doc​uments/MAT​LAB/codege​n/exe/MyNe​tJetson/*.​*;make -f MyNetJetson_rtw.mk all MATLAB_

hi guys I have big problem, The codegen task to transfer from Matlab to jetson nano does not work with the error below. Can you solve this problem?

Error executing command "touch -c /home/jetson/remoteBuildDir/MATLAB_ws/R2019b/C/Users/JSH/Documents/MATLAB/codegen/exe/MyNetJetson/*.*;make -f MyNetJetson_rtw.mk all MATLAB_WORKSPACE="/home/jetson/remoteBuildDir/MATLAB_ws/R2019b" -C /home/jetson/remoteBuildDir/MATLAB_ws/R2019b/C/Users/JSH/Documents/MATLAB/codegen/exe/MyNetJetson". Details:

STDERR: /home/jetson/remoteBuildDir/MATLAB_ws/R2019b/C/Users/JSH/Documents/MATLAB/codegen/exe/MyNetJetson/MyNetJetson.cu: In function 쁵oid MyNetJetson():

/home/jetson/remoteBuildDir/MATLAB_ws/R2019b/C/Users/JSH/Documents/MATLAB/codegen/exe/MyNetJetson/MyNetJetson.cu:672:10: warning: format not a string literal and no format arguments [-Wformat-security]

printf(cv);

^

/usr/bin/ld: skipping incompatible /usr/local/cuda/bin/../targets/aarch64-linux/lib/libcudnn.so when searching for -lcudnn

MWFusedConvReLULayerImpl.o: In function `MWFusedConvReLULayerImpl::getConvAlgoNoWorkSpace()':

tmpxft_00003e0d_00000000-5_MWFusedConvReLULayerImpl.cudafe1.cpp:(.text+0xbfc): undefined reference to `cudnnGetConvolutionForwardAlgorithm'

collect2: error: ld returned 1 exit status

make: *** [/home/jetson/remoteBuildDir/MATLAB_ws/R2019b/C/Users/JSH/Documents/MATLAB/MyNetJetson.elf] Error 1

nvcc -lm -lrt -ldl -Xlinker -rpath,/usr/lib32 -Xnvlink -w -lcudart -lcuda -Wno-deprecated-gpu-targets -lcudnn -lcublas -arch sm_35 -lSDL -lcufft -lcublas -o /home/jetson/remoteBuildDir/MATLAB_ws/R2019b/C/Users/JSH/Documents/MATLAB/MyNetJetson.elf MWDepthConcatenationLayer.o MWElementwiseAffineLayer.o MWFusedConvReLULayer.o MWSplittingLayer.o cnn_api.o MWCNNLayerImpl.o MWDepthConcatenationLayerImpl.o MWElementwiseAffineLayerImpl.o MWElementwiseAffineLayerImplKernel.o MWFusedConvReLULayerImpl.o MWSplittingLayerImpl.o MWTargetNetworkImpl.o rt_nonfinite.o rtGetNaN.o rtGetInf.o v4l2_cam.o availableWebcam.o MW_SDL_video_display.o MyNetJetson_rtwutil.o MyNetJetson_data.o MyNetJetson_initialize.o MyNetJetson_terminate.o MyNetJetson.o DeepLearningNetwork.o predict.o fileManager.o main.o MWCudaDimUtility.o -lm -lm

MyNetJetson_rtw.mk:320: recipe for target '/home/jetson/remoteBuildDir/MATLAB_ws/R2019b/C/Users/JSH/Documents/MATLAB/MyNetJetson.elf' failed

make: Leaving directory '/home/jetson/remoteBuildDir/MATLAB_ws/R2019b/C/Users/JSH/Documents/MATLAB/codegen/exe/MyNetJetson'

------------------------------------------------------------------------

??? Build error: C++ compiler produced errors. See the Build Log for further details.

More information

Code generation failed: View Error Report

[Ramakrishna Mandalapu]

Hi Se Hwan,

This can be an issue with the DL library mismatch on the target which means that the generated code expects a cuDNN version but the target has a different version. To understand the problem cleary, we need the software details on this board.

Can you provide the software details of this board, majorly Jetpack, CUDA, cuDNN and TensorRT versions.

Thanks,

Ramakrishna

[Se Hwan Lee]

Thank you for your answer!

This is Jetson nano board software. Do you need windows Desktop software detail too?

Answer 1

[R]

Hello,

From the ouput log, I understand that you are using MATLAB R2019b and cuDNN version 8.0. The error "undefined reference to `cudnnGetConvolutionForwardAlgorithm'" occurs because the "cudnnGetConvolutionForwardAlgorithm" API is deprecated in cuDNN version 8.0. Please refer to the table in the link below, which details the API functions and data types removed in cuDNN 8.0.0:

https://docs.nvidia.com/deeplearning/cudnn/api/index.html#release-800:~:text=API%20functions%20and%20data%20types%20that%20were%20removed%20in%20cuDNN%208.0.0%20Preview

Further, MATLAB R2019b has been tested on cudNN version 7.0. I recommend you either upgrade MATLAB or downgrade the cudNN version to resolve this issue. Support for cudNN version 8.0 is provided from MATLAB R2021a onwards.

Here are the install prerequisites of MATLAB R2021a for your reference:

https://in.mathworks.com/help/releases/R2021a/gpucoder/gs/install-prerequisites.html#:~:text=NVIDIA%20CUDA%20deep%20neural%20network%20library%20(cuDNN)%20for%20NVIDIA%20GPUs