Matlab GPU coder for Jetson TX2

Question

Vasyl Varvolik 2017-11-22

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/368633-matlab-gpu-coder-for-jetson-tx2

评论： Vasyl Varvolik 2018-4-6

采纳的回答： nam dinh

I cannot run the generated project in Matlab on Jetson TX2. Someone worked with GPU coder?

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

nam dinh 2018-4-3

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/368633-matlab-gpu-coder-for-jetson-tx2#answer_313243

Hi Vasyl I just install matlab 2018a on ubuntu 16.04 I check my GPU coder and got the result in this photo below, could you share with me about: - what matlab version? - cuda toolkit version?, and Cuda DNN version? - after inter Matlab, Cuda, cudnn...do you need configuration more( please, share with me?) Thank you so much and hope to receive your answer soon!

5 个评论
显示 3更早的评论隐藏 3更早的评论

Vasyl Varvolik 2018-4-3

You need to install Jetpack https://developer.nvidia.com/embedded/jetpack

Then the first task is configuring a toolchain for the hardware target. A cross-compiled toolchain executes on the host system (in this case, 64-bit Linux), but generates binary code for the specified hardware target. For the NVIDIA Jetson TX2 and TX1 boards, use the Linaro 4.9 Linux-to-AARCH64 toolchain. Download the gcc-linaro-4.9-2016.02-x86_64_aarch64-linux-gnu release tarball from here: https://releases.linaro.org/components/toolchain/binaries/4.9-2016.02/aarch64-linux-gnu/ Once downloaded, extract the files. The extracted folder contains the subfolders bin, lib, and so on. Set the environment variable LINARO_TOOLCHAIN_4_9 to point to the root folder of the toolchain. The next task is configuring the target-specific CUDA environment for cross-compilation. This environment is the set of include files and libraries provided by Nvidia for the hardware target. Copy these files directly from the target device. For either of the Nvidia Tegra boards, copy all the files (including subfolders) from /usr/local/cuda on the target hardware to a location on your development computer. Be sure not to overwrite any of the CUDA files on the development computer . From a Linux terminal, you can perform the copy operation by using the scp command. $>scp -r ubuntu@tegra-ubuntu:/usr/local/cuda ~/cuda-tegra-arm Then, set the appropriate environment variable to point to the target CUDA implementation. For Nvidia Tegra X2, set NVIDIA_CUDA_TX2. For Nvidia Tegra X1, set NVIDIA_CUDA_TX1 and you need set NVIDIA_CNN.

nam dinh 2018-4-5

hi Vasyl! could you let me know your matlab version, ubuntu version, jetpack version,...etc. some detail of your host pc? hope to hear from you soon, thank you so much

Vasyl Varvolik 2018-4-6