The Example Code for C# is making a block division instead a Cyclic Division. I debugged the code, made desktop testing and found that implementation in the example code is making a Block Division.

I modified the code with the following:

I deleted the lines:

int chunkSize = arr.Length / NUM_THREADS; int start = threadId * chunkSize; int end = (threadId == NUM_THREADS - 1) ? arr.Length : start + chunkSize;

I modified the statement:

for (int i = start; i < end; ++i)

And I changed it for:

for (int i = threadId; i < arr.Length; i = i + NUM_THREADS)

And with that modification the access to the indexes is cyclic.

I didn't test the code in Java, but I think the same thing is happening.