> But will Julia be able to output the necessary instructions for the future hardware accelerators, which could be totally different architectures? I'm thinking

> But will Julia be able to output the necessary instructions for the future hardware accelerators, which could be totally different architectures? I'm thinking of all the new neural networks cores, the DSP, fpgas, and heterogeneous computing from different rival vendors. It seems Julia is deeply dependant of LLVM. Yes. Watch HN in the next week or two for an announcement that may interest you ;).