// offset is not a multiple of `stride`, the input pointer was misaligned and no pointer
// offset will be able to produce a `p` aligned to the specified `a`.
//
- // The naive `-p (mod a)` equation inhibits LLVM's ability to select instructions
+ // The naive `-p (mod a)` equation inhibits LLVM's ability to select instructions
// like `lea`. We compute `(round_up_to_next_alignment(p, a) - p)` instead. This
// redistributes operations around the load-bearing, but pessimizing `and` instruction
// sufficiently for LLVM to be able to utilize the various optimizations it knows about.