Cross compiling using 'gcc-4.8.5-aarch64' to target Jetson tx2

Hi,

I have a project using tensorflow c++ API’s. I have downloaded the ‘gcc-4.8.5-aarch64’ jetpack 4.2 for cross compilation on amd64 to target Jetson TX2. However, I am not able to use the toolchain with bazel build. Are there any other dependencies which needs to be specified during the build?

Thank you,
Abhijeet

Hi,

JetPack4.2 is using gcc-7.3 and all the package are built with gcc-7.3 also.
Any special reason to use gcc-4.8.5?

Thanks.

Hi,

I downloaded the package from ‘GCC Tool Chain for 64-bit BSP v32.1 - https://developer.nvidia.com/embedded/downloads#?search=cross&tx=$product,jetson_tx2

Thank you.

Hi,

There is a default gcc 7.3.0 integrated into the L4T.
You should be able to use it with bazel without issues.

Thanks.

Hi,

Thank you for your response. I am trying to build a custom project which uses Tensorflow 1.13 C++ API and deployed on the Jetson TX2 platform which is flashed with Jetpack 4.2. I am facing build issues when I use Bazel 0.19, 0.21 and 0.24 as well to build my project (inside the Tensorflow 1.13 workspace)

I have referred the following links and used nonccl config parameter:
https://devtalk.nvidia.com/default/topic/1049100/general/tensorflow-installation-on-drive-px2-/

Since the build failed, I was trying to cross compile the same project on the host PC (to speed up the build process) which has all the Cuda Installations required for the host but however I could not succeed in the same.

Can you please provide the solutions to the above two approaches?

Thank you.

Is there syntactically a different command to use other than ‘gcc’? If so, can you please tell me. Thanks.

Duplicate question. See:
https://devtalk.nvidia.com/default/topic/1030446/jetson-tx2/gcc-crosscompiler-update-for-jetson-tx2-/post/5350711/#5350711