Splits lines and counts the words in them using specialized object-to-long hashmap (to avoid boxing-unboxing overhead of Long in java/scala HashMap)
Gets the top k words in terms of word counts.
Gets the top k words in terms of word counts. Assumes that each word exists only once
in the data
iterator (that is, the counts have been reduced).
Warms up the SparkContext in master and slave by running tasks to force JIT kick in before real workload starts.