gpgpu

I think we should drop the many references to ancient John or Jumbo versions, and simply describe the current Jumbo version.

Current implementation of join can be improved by performing the operation in a single call to the backend kernel instead of multiple calls.

This is a fairly easy kernel and may be a good issue for someone getting to know CUDA/ArrayFire internals. Ping me if you want additional info.

In python, you can write python -c 'print(42)' to start an interpreter on the input. Similarly, it would be nice to be able to write futhark run -c 'iota 42' or echo "iota 42" | futhark run - to run the interpreter on some simple input.

Add Javadoc to document the examples in TornadoVM.

This affects the packages under the examples module:

https://github.com/beehive-lab/TornadoVM/tree/master/examples/src/main/java/uk/ac/manchester/tornado/examples

The documentation is at the class-level and it will contain a description of how the TornadoVM API is used for each example. Additionally, it contains how to run the example

Open issue to openly discuss potential ideas or improvements, whether on documentation, interfaces, examples, bug fixes, etc.

Just an FYI whilst I was trawling through the ROCm GitHub page:

https://rocmdocs.amd.com/en/latest/Programming_Guides/Programming-Guides.html#