gpgpu

Somehow some of these names start with fmt_ocl_ while most start with fmt_opencl_. Is this intentional? It causes the fmt_ocl_ to be listed/tested/benchmarked first. Then there's fmt_opencl_1otus5 with a 1 (one) in there. So the first OpenCL formats become:

john_register_one(&fmt_ocl_cryptosafe);
john_register_one(&fmt_ocl_cryptsha1);
john_register_one(&fmt_ocl_KeePass);
john

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 order to test manually altered IR, it would be nice to have a --skip-compilation flag for futhark test, just like we do for futhark bench.

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#