https://issues.apache.org/bugzilla/show_bug.cgi?id=55325

--- Comment #10 from Rainer Jung <[email protected]> ---
(In reply to Sebb from comment #9)

> It might just possibly be worth trying disabling the nano-second time
> resolution by setting the JMeter property:
> 
> sampleresult.useNanoTime=false
> 
> I suppose if the nano-second clock is very badly behaved it might cause some
> issues, but I'm not sure how the negative times can come about.

There are some indications that the nano second timers can be problematic:

http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6458294
http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6440250

and an article about how it's supposed to work:

https://blogs.oracle.com/dholmes/entry/inside_the_hotspot_vm_clocks

Some of this indicates that on Windows XP before SP2 a CPU local counter is
used that is not synchronized between cores and can vary in frequency between
cores due to power management. So when threads are switched between cores bad
things can happen. This specific problem should have been fixed in Windows XP
SP 2.

Measuring time is really something complicated :(

-- 
You are receiving this mail because:
You are the assignee for the bug.

Reply via email to