Yeah that's what I would have thought, but I would like to see it quantified. Does all the target-specific info give you 10% or 100% improvement? Maybe they'r

Yeah that's what I would have thought, but I would like to see it quantified. Does all the target-specific info give you 10% or 100% improvement? Maybe they're onto something? The simplification does sound drastic. He does actually mention the AES instruction in the talk. I think they just have some one-offs to deal with those specific cases. I guess Go is different than C because it comes with a big standard library. They can just make sure that their standard library crypto uses the AESNES instruction, whereas C compilers have no idea what crypto library you're using.