Optimizing LLVM's Bump Allocator · HackerLangs