You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@openwebbeans.apache.org by Mark Struberg <st...@yahoo.de> on 2013/01/15 13:15:21 UTC
new proxy benches
10 million invocations on 6 different methods each time:
with the new proxies + new interceptors:
Starting the real bench
took: 6530479 (6.5 seconds)
--------
with the new proxies + OLD interceptors:
Starting the real bench
took: .. I gave up...
for 1 million invocations it takes: 91909823 (92 seconds for 1/10 of the invocations)
-> factor 110...
There is something bloody broken in this situation...
--------
I was almost worried about the next benchmark:
OLD proxies + OLD interceptors
But it seems that the caching of intercepted methods helped a lot.
For 10 million invocations I got
took: 37578725 (37.5 seconds)
The new proxies don't look bad after all. And we didn't even start performance tuning...
See InterceptorProxyChainTest.
For enabling the new proxies you need to set
WebBeansContext#TODO_USING_NEW_INTERCEPTORS to true
LieGrue,
strub
Re: new proxy benches
Posted by Gerhard Petracek <ge...@gmail.com>.
that's awesome!
regards,
gerhard
2013/1/15 Mark Struberg <st...@yahoo.de>
> Have something else. Removed all the Interceptors and benched just the
> NormalScoping:
>
> new/new
> Starting the real bench
> took: 33909 (34 ms)
>
>
> vs current 1.1.x time:
> Starting the real bench
> took: 5275520 (5.3 seconds)
>
>
> No shit, I'll commit the test ;)
>
> LieGrue,
> strub
>
>
> >________________________________
> > From: Jean-Louis MONTEIRO <je...@gmail.com>
> >To: dev@openwebbeans.apache.org
> >Sent: Tuesday, January 15, 2013 1:29 PM
> >Subject: Re: new proxy benches
> >
> >OLD proxy and NEW interceptor?
> >
> >BTW, looks good!
> >Congrats
> >
> >
> >2013/1/15 Thomas Andraschko <an...@gmail.com>
> >
> >> Great improvement, cool! :)
> >>
> >> 2013/1/15 Mark Struberg <st...@yahoo.de>
> >>
> >> > 10 million invocations on 6 different methods each time:
> >> >
> >> > with the new proxies + new interceptors:
> >> > Starting the real bench
> >> > took: 6530479 (6.5 seconds)
> >> >
> >> > --------
> >> >
> >> > with the new proxies + OLD interceptors:
> >> > Starting the real bench
> >> > took: .. I gave up...
> >> > for 1 million invocations it takes: 91909823 (92 seconds for 1/10 of
> the
> >> > invocations)
> >> > -> factor 110...
> >> > There is something bloody broken in this situation...
> >> >
> >> > --------
> >> >
> >> > I was almost worried about the next benchmark:
> >> > OLD proxies + OLD interceptors
> >> > But it seems that the caching of intercepted methods helped a lot.
> >> > For 10 million invocations I got
> >> > took: 37578725 (37.5 seconds)
> >> >
> >> >
> >> >
> >> > The new proxies don't look bad after all. And we didn't even start
> >> > performance tuning...
> >> >
> >> >
> >> > See InterceptorProxyChainTest.
> >> >
> >> > For enabling the new proxies you need to set
> >> >
> >> > WebBeansContext#TODO_USING_NEW_INTERCEPTORS to true
> >> >
> >> > LieGrue,
> >> > strub
> >> >
> >>
> >
> >
> >
> >--
> >Jean-Louis
> >
> >
> >
>
Re: new proxy benches
Posted by Jean-Louis MONTEIRO <je...@gmail.com>.
Cool thx.
JLouis
2013/1/15 Mark Struberg <st...@yahoo.de>
> took: 33909 (34 ms)
>
>
> vs current 1.1.x time:
> Starting the real bench
> took: 5275520 (5.3 seconds)
>
>
> No shit, I'll commit the test ;)
>
> LieGrue,
> strub
>
--
Jean-Louis
Re: new proxy benches
Posted by Mark Struberg <st...@yahoo.de>.
Have something else. Removed all the Interceptors and benched just the NormalScoping:
new/new
Starting the real bench
took: 33909 (34 ms)
vs current 1.1.x time:
Starting the real bench
took: 5275520 (5.3 seconds)
No shit, I'll commit the test ;)
LieGrue,
strub
>________________________________
> From: Jean-Louis MONTEIRO <je...@gmail.com>
>To: dev@openwebbeans.apache.org
>Sent: Tuesday, January 15, 2013 1:29 PM
>Subject: Re: new proxy benches
>
>OLD proxy and NEW interceptor?
>
>BTW, looks good!
>Congrats
>
>
>2013/1/15 Thomas Andraschko <an...@gmail.com>
>
>> Great improvement, cool! :)
>>
>> 2013/1/15 Mark Struberg <st...@yahoo.de>
>>
>> > 10 million invocations on 6 different methods each time:
>> >
>> > with the new proxies + new interceptors:
>> > Starting the real bench
>> > took: 6530479 (6.5 seconds)
>> >
>> > --------
>> >
>> > with the new proxies + OLD interceptors:
>> > Starting the real bench
>> > took: .. I gave up...
>> > for 1 million invocations it takes: 91909823 (92 seconds for 1/10 of the
>> > invocations)
>> > -> factor 110...
>> > There is something bloody broken in this situation...
>> >
>> > --------
>> >
>> > I was almost worried about the next benchmark:
>> > OLD proxies + OLD interceptors
>> > But it seems that the caching of intercepted methods helped a lot.
>> > For 10 million invocations I got
>> > took: 37578725 (37.5 seconds)
>> >
>> >
>> >
>> > The new proxies don't look bad after all. And we didn't even start
>> > performance tuning...
>> >
>> >
>> > See InterceptorProxyChainTest.
>> >
>> > For enabling the new proxies you need to set
>> >
>> > WebBeansContext#TODO_USING_NEW_INTERCEPTORS to true
>> >
>> > LieGrue,
>> > strub
>> >
>>
>
>
>
>--
>Jean-Louis
>
>
>
Re: new proxy benches
Posted by Jean-Louis MONTEIRO <je...@gmail.com>.
OLD proxy and NEW interceptor?
BTW, looks good!
Congrats
2013/1/15 Thomas Andraschko <an...@gmail.com>
> Great improvement, cool! :)
>
> 2013/1/15 Mark Struberg <st...@yahoo.de>
>
> > 10 million invocations on 6 different methods each time:
> >
> > with the new proxies + new interceptors:
> > Starting the real bench
> > took: 6530479 (6.5 seconds)
> >
> > --------
> >
> > with the new proxies + OLD interceptors:
> > Starting the real bench
> > took: .. I gave up...
> > for 1 million invocations it takes: 91909823 (92 seconds for 1/10 of the
> > invocations)
> > -> factor 110...
> > There is something bloody broken in this situation...
> >
> > --------
> >
> > I was almost worried about the next benchmark:
> > OLD proxies + OLD interceptors
> > But it seems that the caching of intercepted methods helped a lot.
> > For 10 million invocations I got
> > took: 37578725 (37.5 seconds)
> >
> >
> >
> > The new proxies don't look bad after all. And we didn't even start
> > performance tuning...
> >
> >
> > See InterceptorProxyChainTest.
> >
> > For enabling the new proxies you need to set
> >
> > WebBeansContext#TODO_USING_NEW_INTERCEPTORS to true
> >
> > LieGrue,
> > strub
> >
>
--
Jean-Louis
Re: new proxy benches
Posted by Thomas Andraschko <an...@gmail.com>.
Great improvement, cool! :)
2013/1/15 Mark Struberg <st...@yahoo.de>
> 10 million invocations on 6 different methods each time:
>
> with the new proxies + new interceptors:
> Starting the real bench
> took: 6530479 (6.5 seconds)
>
> --------
>
> with the new proxies + OLD interceptors:
> Starting the real bench
> took: .. I gave up...
> for 1 million invocations it takes: 91909823 (92 seconds for 1/10 of the
> invocations)
> -> factor 110...
> There is something bloody broken in this situation...
>
> --------
>
> I was almost worried about the next benchmark:
> OLD proxies + OLD interceptors
> But it seems that the caching of intercepted methods helped a lot.
> For 10 million invocations I got
> took: 37578725 (37.5 seconds)
>
>
>
> The new proxies don't look bad after all. And we didn't even start
> performance tuning...
>
>
> See InterceptorProxyChainTest.
>
> For enabling the new proxies you need to set
>
> WebBeansContext#TODO_USING_NEW_INTERCEPTORS to true
>
> LieGrue,
> strub
>