Getting Started with TX1

I am having trouble getting started. I’ve run the JetPack-L4T-2.0-linux-x64.run file. It almost finished but hung towards the end. It appeared to be compiling examples at the time. I control’d c out of it and the GUI dialog seemed to finish from there. I then went to the TX1 and ran some of the examples fine. Now I’m at a point where I want to write a small Cuda program on the host and run it on the TX1. This is where I am currently and desperately trying to get some traction with a simple example. I got a feeling its one of those situations where you have to read 15 documents on 15 different subjects and piece it together by myself. Or I’m just a bozo and missed the one piece of information to get me moving. I also noted I can’t find any information on more technical descriptions of the GPU itself. For example, I know the registers are one clock cycle away from alu, but how does it multiplex the data,address lines between the 32 SPs accessing at the same time. The type of information I want is the kind that would allow me to understand exactly how the chip works. In a previous capacity over 20 years ago, I was tasked with writing the graphical interface for running microcode on a 512 node SIMD computer, and I had to understand it at the register transfer level. I guess I’m looking for the same level documentation.

Sounds like you’re used to working closer to the metal than is typically expected (or supported) on the Jetson. The Jetson devices work more like general-purpose x86 systems than your typical embedded system, with the various levels of abstractions that have accumulated in Linux for the past 20 years or so. Most tutorials for getting code up and running that work for an x86 machine should be mostly portable to the TX1, changing library architectures when needed of course.

If you’re trying to learn CUDA programming, there’s lots of documentation from various sources, though NVIDIA has one here: http://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#axzz3x9ESBFuZ (this is a newer version of the guide, but it should mostly apply). If nothing there helps, you might want to post to the CUDA forum: https://devtalk.nvidia.com/default/board/57/cuda-programming-and-performance/

Oh, and you might want to rerun that JetPack session to make sure the install completes.