As Kunle noted in lecture, by adding further granularity to the log, write, and read calls required to run a chunk of code atomically, we can trust the compiler to optimize this using a technique called common subexpression elimination.