Benchmark keeps running

[Yann Le Tallec]

Which falls under the category: "I'm probably missing something obvious".
Sorry about that and thanks.


On 8 May 2013 08:17, Aleksey Shipilev <aleksey.shipilev at oracle.com> wrote:

> On 05/08/2013 07:21 AM, Yann Le Tallec wrote:
> > @OutputTimeUnit(TimeUnit.NANOSECONDS)
> > public class CAS {
> >
> >     @State(Scope.Benchmark)
> >     public static class AIState {
> >         private final AtomicInteger ai = new AtomicInteger();
> >     }
> >
> >     @GenerateMicroBenchmark(BenchmarkType.AverageTimePerOp)
> >     public int cas(CAS.AIState state) {
> >         int i = state.ai.get();
> >         while (!state.ai.compareAndSet(i, i + 1)) { }
> >         return state.ai.get();
> >     }
> > }
>
> Your benchmark is probably incorrect, you have to re-read the AI value
> in the CAS'ed loop: if you had failed to CAS, that means the current
> value is not "i" anymore:
>
>     @GenerateMicroBenchmark(BenchmarkType.AverageTimePerOp)
>     public int cas(AIState state) {
>         int i;
>         while (!state.ai.compareAndSet((i = state.ai.get()), i + 1)) { }
>         return state.ai.get();
>     }
>
> Otherwise some thread can obviously be stuck. If that is the behavior
> you are after, you have to let JMH terminate the loop, see
>
> http://hg.openjdk.java.net/code-tools/jmh/file/tip/jmh-samples/src/main/java/org/openjdk/jmh/samples/JMHSample_18_Control.java
>
> -Aleksey.
>