# Julia introduction

## cases where julia would have been nice irl
- DBCSR generating kernels for specific matrix sizes (in julia generated functions or just function specialization)
- MultiFloats.jl (Float64x{N}) that works on GPUs too?
- LoopVectorization: @avx? / VectorizationBase: Vec to show off @code_native
- Autodiff?
- Interfacing with C, (templated) C++ https://github.com/JuliaInterop/Cxx.jl (Eigen Vector Wrapper in X lines?)

TODO:
- Check julia version, CUDA.jl and MPI.jl support on Daint / JupyterHub
- Check out Dagger.jl? Flux.jl?
- See if COSMA.jl works on Daint with DistributedArrays.jl
- Debugger