A simple and incomplete linear algebra library which uses OpenCL on a GPU for all calculations.
Reducing the amount of threads helps to make the output more readable.
$ RUST_LOG=debug cargo test -- --nocapture --test-threads 1
Note: Kernels for the f16 and f64 types might not work because the GPU driver doesn't support them yet. (Looking at you Mesa...)