Performance wasn't the aim of this presentation so it wasn't what I was focussed on. But it's instructive to play around with functions to see what's generated

Performance wasn't the aim of this presentation so it wasn't what I was focussed on. But it's instructive to play around with functions to see what's generated under the covers. For adding two numbers together, for example, it can boil down to just a few x86 instructions. Since it compiles down to IL/IR there is a lot of common optimisations from the LLVM/Clang tool chains that are inherited for free. I would expect to see the performance improve over time too.