ForkJoin wakeup update
gregg at wonderly.org
Thu Oct 18 12:57:23 PDT 2012
On 10/18/2012 2:32 PM, Aleksey Shipilev wrote:
> On 10/18/2012 10:38 PM, Gregg Wonderly wrote:
>>> * The main systems-level goal should be to find ways to make
>>> wakeups faster. Possibly entailing support of some sort of bulk
>>> signalling facilities?
>> Might there be some opportunity to use a PID mechanism here so that some
>> amount of dynamic tuning might happen automatically?
> I've played with these a bit back in the days.
> While sounding good in theory, they fail to account warmup effects
> properly, and especially are bad with non-linear behaviors (which is the
> usual behavior for general application under resource constraints).
> Maybe it would help in pathological cases though.
I haven't done anything with PID in this environment, so I have no real
experiences to draw on. What goal(s) did you tune against?
More information about the lambda-dev