We should see how axom::Array performs relative to std::vector and C++ arrays, and optimize where we can.