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