Tag
This blog post explores an optimization for LLVM's SmallVector::push_back by tail-calling the grow-and-push path, which eliminates callee-saved register spills and improves the fast path performance.