Stream perf test

[Remi Forax]

On 12/13/2012 10:59 AM, Aleksey Shipilev wrote:
> Hi Remi,
>
> We do have lots of perf tests internally (due to legal at this point to open source them), and yes, these are tricky to get right. There seem to be nothing special about the streams per se, just usual micro benchmarking troubles. If you can send your tests along, we can figure whether you are dealing with something we already know about.
>
> -Aleksey.

My main issue is that depending on the context the OSR is triggered 
before or after the method compilation resulting is code that are very 
different.
Anyway, when the other part of my code will be stabilized I will publish 
it so you can take a look.

Rémi

>
> On 13.12.2012, at 13:26, Remi Forax <forax at univ-mlv.fr> wrote:
>
>> I've tried to write several tests that use the stream pipeline and I've
>> troubled to get stable values, sometimes the fastest time is one order
>> magnitude faster than the slowest time.
>> Do we have any perf tests written somewhere to see what I did wrong ?
>>
>> Rémi
>>