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