Changing foreach loops to Parallel.ForEach and weird bugs

Question

I have several loops that I wrote using the traditional syntax ``` foreach(x in xs) {....} ``` Some of these loops are pretty intensive in terms of computation and I just changed them using the parallel syntax like this: ``` Parallel.ForEach(x, xs => {...}); ``` I see a major increase in performance!! Now my question is this: Am I introducing bugs using parallel multithreading? I read that thread safety is complex and can create weird bugs; what should I be concerned about?

Accepted Answer

Accessing shared state will most likely not produce the desired result. Simple example:

```
int sum = 0;
for (int i = 0; i < 1000000; i++)
{
    sum++;
}
```

change this to

```
Parallel.For(0, 1000000, i => { sum++; });
```

and you'll see `sum` will have some random value because multiple threads are reading/writing `sum`.

If you lock around the update, you'll solve the problem, but you'll essentially turn the operation into a sequential operation again.

You need to make sure that whatever happens in the loop is safe.

Microsoft Patterns and Practices did a book that explains all this and more. You should check that out before you simply change the code to use parallel loops.

Changing foreach loops to Parallel.ForEach and weird bugs

Solution

Problem