Not the OP, but was mentioned they were open to optimizations of implementations that were used as comparisons.
1 Like
I don't think there's much interesting here. This is just a demonstration that if you write fortran-like code and give it to LLVM, it doesn't matter much where you start so long as you give it types. (Note that the non-ascribed, non-generic version of the Julia code is 7x slower.)
3 Likes