Currently, Iris uses Java NIO for vertex writing when building chunks with Sodium, which is fast enough, but it could be faster if we used Unsafe or LWJGL MemoryUtil. This is mainly a reminder and won't be implemented until we can address code duplication issues.