Removing intrinsic of Thread.isInterrupted()

Karen Kinnear karen.kinnear at oracle.com
Tue Feb 25 11:05:07 PST 2014


Yumin,

Sounds like you are getting a lot of pushback removing the intrinsic.

So I think that instead of doing that, you might modify it as we discussed - 
to leave fast-path 1, but not fast-path 2 since that has a potential timing hole.

thanks for all your efforts here,
Karen

On Feb 25, 2014, at 1:45 PM, Aleksey Shipilev wrote:

> On 02/25/2014 10:23 PM, Karen Kinnear wrote:
>> Performance team folks -
>> 
>> I think it would be very helpful to identify any "real" benchmarks, rather than
>> a small hand-rolled micro-benchmark.
> 
> I think microbenchmarks are fine to understand the performance
> implications. My tests show order of magnitude difference gained by the
> intrinsic. Larger workloads can delude you into believing the change is
> safe, while the users who often write:
> 
> while(Thread.currentThread().isInterrupted()) {
>   // do stuff
> }
> 
> ...will have some questions for us.
> 
> -Aleksey.



More information about the hotspot-compiler-dev mailing list