Cross compilation support

[gfvvz]

https://github.com/NVIDIA/TensorRT-Edge-LLM/blob/main/docs/source/developer_guide/01.2_Quick_Start_Guide.md\

From above page, it introduce the host and device compilation, how about one command compilation to support for both host and device code by using cross compilation, the push the binary to device and execution directly?

[nvluxiaoz]

Thanks for your suggestion. For platforms where native compilation is not supported, we should and will support cross build. However, for platforms like Jetson, native compilation is supported well with good compilation time. TensorRT Edge-LLM perform ONNX export on host (with no C++ and only Python), and the pure C++ runtime is compiled and run on device.

[nvluxiaoz]

DriveOS is supported with Cross build in 0.6.1. Closing this discussion. If you need further assistance please let us know. Thanks!