What is one way to avoid low-level parallelization details?
monads
partially applied functions
literal functions
parallel collections