0

I have a Timer class that measures a method execution time: 

 public class Timer {
            private long startTime;

            public long start()  {
                    startTime=System.currentTimeMillis();
                    return startTime;
            }

            public long getElapsedTime()  {
                    return System.currentTimeMillis()-startTime;
            }
}

And I have a method that does some memory-heavy operations (like, copies into array millions of references to new different objects/millions of references to an existing object).

  public static void Test {
        private static Timer timer=new Timer();
        private static String s0="STRING";
        private static long size=5000_000L;

        public static void main(String[] args)  {
              System.gc();
              foo();
              //bar();  //executed if foo() is commented
        }

        public static void foo()  {
              System.out.println("Foo starts: ");
              timer.start(); 
              for (int i=0; i<size; i++)  {
                   arrayList.add(s0);   
              }
              System.out.println("Foo ends: "+timer.getElapsedTime()+"ms");  
        }     

        public static void bar()  {
            //some code here
        }
}

When foo() is executed, a GC happens. It takes different time from run to run, for instance [Times: user=0.80 sys=0.08, real=0.39 secs] and foo() takes 1545 ms. Then I have a bar() that does a similar thing and I want to find out if it's just as efficient. It takes [Times: user=0.84 sys=0.06, real=0.31 secs] and 1825 ms.

What should I do to find out which method is effective, if I don't have enough RAM to avoid garbage collection during the run? Can I just subtract it like 1825-310 and 1545-390 and then compare the results?

EDIT: I run the methods separately: one is always commented out. System.gc() is used at the very beginning.

parsecer
  • 4,758
  • 13
  • 71
  • 140
  • Jon Skeet [recommends](http://stackoverflow.com/a/504119/1553851) calling `System.gc()` between runs. You can also run with the `-verbose:gc` option so you tell what's actually happening. – shmosel Mar 23 '17 at 00:48
  • @shmosel, I forgot to mention - I use `System.gc()` as a first command in `main` and there is always only one method being executed. – parsecer Mar 23 '17 at 00:53
  • @shmosel, As to `-verbose` it's already enabled. From what I can make of it, it collects garbage when the heap space is up. – parsecer Mar 23 '17 at 01:33
  • Is the purpose benchmarking or monitoring? – the8472 Mar 23 '17 at 19:15
  • @the8472, The puspose is benchmarking, finding out the best method and its success margin, relative to another method. – parsecer Mar 23 '17 at 20:53
  • Possible duplicate of [How do I write a correct micro-benchmark in Java?](http://stackoverflow.com/questions/504103/how-do-i-write-a-correct-micro-benchmark-in-java) – the8472 Mar 23 '17 at 21:53
  • write a proper microbenchmark, look at its statistical output. *min* should represent performance without GC and *avg* performance with amortized GC which actually is more realistic. – the8472 Mar 23 '17 at 21:55

0 Answers0