### Problem cranelift is pretty bad at recognizing manual implementation of such intrinsics, leading to suboptimal assembly ### Proposed Solution add more builtin cranelift functions like `popcnt` , `clz`, etc. for a list of builtin cranelift functions see [the cranelift docs](https://docs.rs/cranelift-codegen/0.109.0/cranelift_codegen/ir/trait.InstBuilder.html) ### Notes _No response_