Odd Even Merge Sort

Hi I get a code sample from this blog: http://m.blog.csdn.net/blog/luckyboy101/9079271 in CUDA Sorting Networks Category. There is something that I can’t understand.

At execution for N= 1048576 the array length of 16 is sorted with 10ms.
When N is increased to N*2 = 2097152 the array length of 16 elements is sorted 54 ms. Why is this can anyone explain to me!

Thnx in advance.

Sorting serially is n * log n which means the sort time grows exponentially as the number of elements increase.

This probably still holds true for parallel version…