Automatically Exploiting the Memory Hierarchy of GPUs through Just-in-Time Compilation