From the article: "Even Apple’s LLVM fork lacks scheduling annotations for Apple Silicon. How am I supposed to write efficient code when Apple doesn’t bother t

From the article: "Even Apple’s LLVM fork lacks scheduling annotations for Apple Silicon. How am I supposed to write efficient code when Apple doesn’t bother to tune their own compiler?" In addition to its public fork of LLVM, Apple also keeps a very private fork where (one assumes) they keep all the really juicy stuff. I agree that it is frustrating not to have the data, but don't confuse "I don't have the data" with "no one has the data".