luke warm methods?

[Tom Rodriguez]

On Apr 29, 2010, at 2:30 AM, Vikram A wrote:

> Please wait for Thomas's comments on this, but I would like to know...
> Shouldn't TieredCompilation help perform better in such cases...
> 
> http://blogs.sun.com/fatcatair/entry/tiered_compilation_almost_live

Tiered definitely would help since you'd get to C1 quality code very quickly but you still might not transition to C2 quality code, leaving some performance on the floor.  We'll also probably revisit the decay logic since we've changed how invocation counters operate a bit and I think simply decaying them doesn't really make sense.  I'm not sure it makes a lot of sense in the first place I guess.

tom

> 
> thanks,
> Vikram.
> 
> On Thu, Apr 29, 2010 at 2:32 PM, Ben Evans <ben.evans at db.com> wrote:
>> 
>> hotspot-dev-bounces at openjdk.java.net wrote on 22/04/2010 00:42:28:
>> 
>>> We decay counters periodically since over time counters should only
>>> go up so maybe that's the problem.  If you run with -XX:-
>>> UseCounterDecay it will turn it off but all the other flags that
>>> control it aren't available in product mode. This is certainly one
>>> of the pathologies of an invocation counter based system with
>>> relatively high thresholds.
>> 
>> Just so I'm clear, this doesn't affect the compiled status of methods,
>> right?
>> 
>> What's being discussed here is that if an initial burst of activity results
>> in a method being called say 4096 times (vs a threshold of 10000), then
>> after 1 minute, then that method has had its invocation counter reduced to
>> 1024.
>> 
>> So the usual scenario whereby HotSpot reaches a steady state in terms of
>> compilation activity can be disrupted by a sudden flurry of activity on
>> rarely-used methods (which will be effectively cold due to the decay
>> process), which pushes the invocation counts of these methods above
>> compilation threshold. If these methods then go quiet again, then HS will
>> *not* deoptimize and drop these methods back to interpreted mode, even if
>> their invocation counters subsequently decay below threshold. Correct?
>> 
>> Thanks,
>> 
>> Ben
>> --
>> Ben Evans
>> GFFX Auto Trading
>> Deutsche Bank, London
>> Office: +44 (0)20 7541 3953
>> 
>> ---
>> 
>> This e-mail may contain confidential and/or privileged information. If you
>> are not the intended recipient (or have received this e-mail in error)
>> please notify the sender immediately and delete this e-mail. Any
>> unauthorized copying, disclosure or distribution of the material in this
>> e-mail is strictly forbidden.
>> 
>> Please refer to http://www.db.com/en/content/eu_disclosures.htm for
>> additional EU corporate and regulatory disclosures.
>>