Crosscompiling Nerves on Jetson

[zacky1972]

It will be required to adjust into this issue:

nerves-project/toolchains#56

[zacky1972]

Perhaps, there may be the case that CUDA doesn't exist on compile time but exist on runtime.
In this case, NVCC environment variable will be set to the cross compiler of nvcc on compile time.
This requires to modify CpuInfo.cuda/1.

[celaxodon]

@zacky1972, I can try to help with this issue. It will likely come up soon once I'm a little farther along in the Buildroot port for the Nano and start figuring out how to include CUDA and related libraries as Buildroot packages. Ideally it would be will be baked into the firmware we generate and exist at compile time.

[zacky1972]

@celaxodon Thank you, a lot!