java - Why is map.values().stream() much slower than Array.stream(array) -

creating 2 constructs university second semester computer science count words in text. 1 implementation uses array word-objects save word string , frequency int. other uses hashmap word key , frequency value. function "totalwords" should return sum of frequency.

in hashmap variant:

return _map.values().stream().reduce(0, (a, b) -> + b); 

in array variant:

return arrays.stream(_words)             .map((word) -> word != null ? word.count() : 0)             .reduce(0, (a, b) -> + b); 

my problem is: in junit test short test text array variant need approximately 0.001s , map variant needs 0.040s , not understand why map need more time. has explanation , maybe better solution?

one of reason iterating hashmap can slower array, reason locality. computation bottleneck of modern processor dominated memory access, , that's why cache used. array stores data in contiguous chunk of memory, means when swap chunk of memory cache, more using in cache, or cache hits, cache likes contiguous memory of data. on other hand, each element of hashmap stored in different place in memory, when traverse hashmap, lot of cache misses, end swapping data in , out of cache time, dramatically slows down program.

although actual implementation of hashmap in optimized way such data in memory clustered together, in case, (@radiodef)since hashmap uses sort of linked list, each element of hashmap contains pointers, hashmap consumes more memory array, more memory means more cache misses , more page faults, hashmap in general slower array.