Build TensorRT samples on DRIVE PX 2

Hello,

We are working on DRIVE PX 2 AutoChauffeur platform. We are currently in the process of deploying a neural network on DPX2. We are referring to TensorRT 3.0.2 documentation (https://docs.nvidia.com/deeplearning/sdk/tensorrt-archived/tensorrt_302/tensorrt-developer-guide/index.html#samples). The documentation does not clearly specify how to build the TRT samples. We have observed a makefile in the samples directory, so we assume a ‘make’ command should do the trick.

I apologize if this sounds like a dumb request, but I would appreciate it if somebody here can confirm the procedure for me.

Thank you.

Dear sagar,
Yes. You can build using make. Please check makefile of any TRT sample

Dear Siva,

Thank you for your quick reply.

For apps targeting DPX2, NVIDIA asks to cross-compile apps using Nsight Eclipse on host PC, transfer them to DPX2 and then execute them. Just out of curiosity, has anyone tried this procedure for TRT samples?

I ask this because the TRT libs for aarch64 were installed on my host PC when I used DriveInstall. So even though the PLAN are device specific, using the aarch64 libs should create the proper PLAN, isn’t it?

Cross compilation allows you get aarch64 executable. The plan gets created when you run the executable on drivePX2 and serialize the plan to reuse for future purpose.
Any TRT sample that gets compiled on Host PC can be compiled on DrivePX2 out of box with the Makefile structure.

Got it. Thanks for clarifying.