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