Parallel Performance Analysis

. Net4.5 introduces Task-based parallel programming. Parallel Programming is complicated for everyone. This time,. Net introduced a new programming model, and I believe it has aroused everyone's attention. During my learning, I made a small experiment and shared it with you.

In the past, when we were doing performance analysis, we used a timer that was amazing. This time, I found that the System. Diagnostics. Stopwatch class is very useful. It takes only a few steps to solve the time statistics problem:

Stopwatch stop = new Stopwatch ();

Stop. Start (); // Start statistics

Stop. Stop (); // end statistics

Stop. ElapsedMilliseconds // return the elapsed time in milliseconds

Stop. Reset (); // clears the statistical time

Stop. Restart (); // clear and start statistics

Second, I found that writing parallel tasks is much simpler (although extremely basic) than I imagined ).

You only need:

Foreach (var e in data source) {execute Operation}

Convert to the following form to automatically execute the operation tasks of each element in parallel:

Parallet. ForEach (data source, e =>{ execute Operation })

Parallet is a class in the space of System. Threading. Tasks.

If the execution is not a foreach loop, but a for loop, it is also possible:

For (int I = 0; I <1000; I ++) {execution operation }=> Parallet. For (0, 1000, I =>{ execution operation })

It's so easy! I believe that those who don't really want to understand parallel programming can also slightly improve their program concurrency through the above syntax.

Then I made a performance analysis on my E7200 dual-core cpu. When each task runs for about 1 ms, the parallel ratio synchronization is: 1.3: 1; when 10 ms, 1.7: 1; 100: 1 in 1.9 milliseconds. if it is less than 1 ms, the performance will be inferior to synchronization.

In view of rp, most of the code I write is run at extremely high speed, so this parallel can be used rarely. But I hope it will be useful to everyone.

Finished!