.net - Benchmarking small code samples in C#, can this implementation be improved? -

quite on find myself benchmarking small chunks of code see implemnetation fastest.

quite see comments benchmarking code not take account jitting or garbage collector.

i have following simple benchmarking function have evolved:

  static void profile(string description, int iterations, action func) {         // warm          func();         // clean         gc.collect();          var watch = new stopwatch();         watch.start();         (int = 0; < iterations; i++) {             func();         }         watch.stop();         console.write(description);         console.writeline(" time elapsed {0} ms", watch.elapsedmilliseconds);     } 

usage:

profile("a descriptions", how_many_iterations_to_run, () => {    // ... code being profiled }); 

does implementation have flaws? enough show implementaion x faster implementation y on z iterations? can think of ways improve this?

edit pretty clear time based approach (as opposed iterations), preferred, have implementations time checks not impact performance?

here modified function: recommended community, feel free amend community wiki.

static double profile(string description, int iterations, action func) {     //run @ highest priority minimize fluctuations caused other processes/threads     process.getcurrentprocess().priorityclass = processpriorityclass.high;     thread.currentthread.priority = threadpriority.highest;      // warm      func();      var watch = new stopwatch();       // clean     gc.collect();     gc.waitforpendingfinalizers();     gc.collect();      watch.start();     (int = 0; < iterations; i++) {         func();     }     watch.stop();     console.write(description);     console.writeline(" time elapsed {0} ms", watch.elapsed.totalmilliseconds);     return watch.elapsed.totalmilliseconds; } 

make sure compile in release optimizations enabled, , run tests outside of visual studio. last part important because jit stints optimizations debugger attached, in release mode.