Let me clarify.
The discussion is around removing the volatile read in the inc method.
A counter could be incremented >100k times a second and we don't want a
counter to cause too much of a performance degradation. A counter will
always be incremented by the same thread.
The get method will only be called every X seconds. So a volatile read is
fine.
On Saturday, October 29, 2016 at 11:13:38 AM UTC+3, Peter Veentjer wrote:
>
> How expensive is a volatile read for implementing a single writer
> performance counter?
>
> One can easily create some kind of performance counter using an AtomicLong
> or using a LongAdder if there is contention.
>
> some example code:
>
> public class Counter {
>
> private final AtomicLong c = new AtomicLong();
>
> public void inc(){
> c.addAndGet(1);
> }
>
> public long get(){
> return c.get();
> }
> }
>
> But in some cases one can create a counter where there is only a single
> writer and multiple readers. In these case one can make use of
> AtomicLong.lazySet. This prevent the cpu to wait for the store buffer to be
> drained. Martin Thompson explained this in his excellent training.
>
> So you get something like this:
>
> public class Counter {
>
> private final AtomicLong c = new AtomicLong();
>
> public void inc(){
> long newValue = c.get()+1;
> c.lazySet(newValue);
> }
>
> public long get(){
> return c.get();
> }
> }
>
> But in this case you still need to do a volatile read to read the actual
> value. In theory one could optimize this to:
>
> public class Counter {
>
> private final AtomicLong c = new AtomicLong();
> private long local;
>
> public void inc(){
> long newValue = local+1;
> this.local = newValue;
> c.lazySet(newValue);
> }
>
> public long get(){
> return c.get();
> }
> }
>
> In this case there is no volatile read involved, however more data needs
> to be written to memory.
>
> As soon as I have time, I'll create a JMH benchmark. But I'm interested in
> the thoughts behind volatile reads in this particular case.
>
--
You received this message because you are subscribed to the Google Groups
"mechanical-sympathy" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to mechanical-sympathy+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
