Question

Why is not the same answer for following?

Normal coding:

long sum = 0;

for (long i = 1; i <= 10; i++)
{
    long result = 1;

    for (long j = 1; j <= i; j++)
    {
        result = result*j;
    }
    sum = sum + result;
}

Parallel Coding:

long sum = 0;

Parallel.For(1, 10, delegate(int i)
    {
        long result = 1;

        Parallel.For(1, i, delegate(int j)
            {
                result = result*j;
            });

        sum = sum + result;
    });

Please show me the right way

for (long i = 1; i <= 5; i++)
        {
            sum = sum * i;
        }

and

Parallel.For(1, 5, delegate(int i)
    {
        sum = sum * i;
    });

Result of parallel = 24

Result of normal = 120

Answer 1

The answer in parallel version is arbitrary because sum and result are accessed and modified by different threads. So what you should do is separating between computing each step and summing up results. To be able to sum up results correctly, you need to obtain a lock so that threads modify sum exclusively. One way to fix could be:

long sum = 0;
object monitor = new object(); 
Parallel.For(1, 11, () => 0L, (i, state, local) => 
{ 
    long result = 1;

    for (long j = 1; j <= i; j++)
    {
        result = result*j;
    } 
    return local + result;
}, local => { lock (monitor) sum += local; }); 

Note that you rarely need two nested Parallel.For loops, because they often lead to bad performance. So one Parallel.For loop at the outermost level and keeping inner for loops as they are is recommended. Also to obtain some speedup, you have to test with numbers much bigger than 10.