[concurrency-interest] Lock-free implementation of min including assignment

[Markus Jevring]

So, by paraphrasing what you have there, if I wanted something like 
getAndMin(AtomicLong var, long l),
I would define transform() as: "return Math.min(current, l)" to get the 
desired effect, right?
In essence, I'd be able to skip the cas operation in my else-case?

Similarly, minAndGet(AtomicLong var, long l) would simply return "next" 
instead of "current", right?

As other people have pointed out, both 'current' and 'min', as it were, 
may have become obsolete by the time the function returns, but as long 
as we keep the smallest value in 'var', we should be fine, right?

Markus

On 18/01/2012 21:23, Martin Buchholz wrote:
>
>
> On Wed, Jan 18, 2012 at 06:16, Markus Jevring 
> <m.jevring at iontrading.com <mailto:m.jevring at iontrading.com>> wrote:
>
>     If it does, I'm surprised it's not plastered all over the internet.
>
>
> We've written something here:
> http://g.oswego.edu/cgi-bin/viewcvs.cgi/jsr166/src/main/java/util/concurrent/atomic/package-info.java?view=co
>
>  * <p>It is straightforward to define new utility functions that, like
>  * {@code getAndIncrement}, apply a function to a value atomically.
>  * For example, given some transformation
>  * <pre> {@code long transform(long input)}</pre>
>  *
>  * write your utility method as follows:
>  * <pre> {@code
>  * long getAndTransform(AtomicLong var) {
>  *   while (true) {
>  *     long current = var.get();
>  *     long next = transform(current);
>  *     if (var.compareAndSet(current, next))
>  *         return current;
>  *   }
>  * }}</pre>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://cs.oswego.edu/pipermail/concurrency-interest/attachments/20120119/fb3d3c10/attachment.html>