Skip to main content


Eclipse Community Forums
Forum Search:

Search      Help    Register    Login    Home
Home » Modeling » M2T (model-to-text transformation) » [XPand]: 0.8.0 Performance surprise..
[XPand]: 0.8.0 Performance surprise.. [message #495961] Fri, 06 November 2009 19:11 Go to next message
Miles Parker is currently offline Miles ParkerFriend
Messages: 1341
Registered: July 2009
Senior Member
Hi folks,

I'm just experimenting with moving to 0.8.0 from 0.7.2 and I've noticed an, um.. significant change in performance characteristics. By significant, I mean that a model that was taking 1 second to generate is now taking 45 seconds . In general I'm seeing a 25-40x slow-down. This is with 0.8.0 v 2009092913 of Xpand, MWE, etc.. I've upgraded xtext as well but this happens on purely EMF based projects as well. Nothing else has changed (except for some API fixes for xtext in unrelated plugins) -- I've verified by reverting to each configuration.

Any guesses about what might be causing this? Areas where performance has been optimized for different cases? My templates are quite complex and use aspects and a lot of xtend calls.

I don't this this the class loader issues as the whole process seems to be much slower and again I haven't changed anything else.

cheers,

Miles

Re: [XPand]: 0.8.0 Performance surprise.. [message #495972 is a reply to message #495961] Fri, 06 November 2009 20:15 Go to previous messageGo to next message
Miles Parker is currently offline Miles ParkerFriend
Messages: 1341
Registered: July 2009
Senior Member
One additional data point here. I've just tried it on my desktop machine (MacPro 8-core) and the performance delta is less compared to my laptop, but still in the 5x-12x + range. So perhaps there is some kind of caching or other heap issue here. It's a bit difficult to switch back and forth but I can prob. set up two configurations if more testing would be helpful. But hopefully there is something simple going on here.
Re: [XPand]: 0.8.0 Performance surprise.. [message #496123 is a reply to message #495961] Sun, 08 November 2009 18:39 Go to previous messageGo to next message
Sebastian Zarnekow is currently offline Sebastian ZarnekowFriend
Messages: 3118
Registered: July 2009
Senior Member
Hi Miles,

that not a nice suprise.
Did you change the execution environment from plain Java VM to OSGi?
If you are in an OSGi environment, please have a look at
https://bugs.eclipse.org/bugs/show_bug.cgi?id=278753
Please file a bug with a short sample project if you do not use OSGi.

Regards,
Sebastian
--
Need professional support for Eclipse Modeling?
Go visit: http://xtext.itemis.com

Miles Parker schrieb:
> Hi folks,
>
> I'm just experimenting with moving to 0.8.0 from 0.7.2 and I've noticed
> an, um.. significant change in performance characteristics. By
> significant, I mean that a model that was taking 1 second to generate is
> now taking 45 seconds . In general I'm seeing a 25-40x slow-down. This
> is with 0.8.0 v 2009092913 of Xpand, MWE, etc.. I've upgraded xtext as
> well but this happens on purely EMF based projects as well. Nothing else
> has changed (except for some API fixes for xtext in unrelated plugins)
> -- I've verified by reverting to each configuration.
>
> Any guesses about what might be causing this? Areas where performance
> has been optimized for different cases? My templates are quite complex
> and use aspects and a lot of xtend calls.
>
> I don't this this the class loader issues as the whole process seems to
> be much slower and again I haven't changed anything else.
>
> cheers,
>
> Miles
>
Re: [XPand]: 0.8.0 Performance surprise.. [message #496698 is a reply to message #496123] Wed, 11 November 2009 01:40 Go to previous messageGo to next message
Miles Parker is currently offline Miles ParkerFriend
Messages: 1341
Registered: July 2009
Senior Member
Sebastian, see below please..

Sebastian Zarnekow wrote on Sun, 08 November 2009 13:39
Hi Miles,

that not a nice suprise.
Did you change the execution environment from plain Java VM to OSGi?
If you are in an OSGi environment, please have a look at
https://bugs.eclipse.org/bugs/show_bug.cgi?id=278753
Please file a bug with a short sample project if you do not use OSGi.



Well, I'm executing it from an Eclipse plugin using WorkflowRunner, so I guess that's an OSGi environment. Wink However, this is what I've been doing all along, so unless the performance changes for the bug above have somehow worked at cross purposes to my own lousy attempts at managing buddies and class loading, I don't think that's it. Also, when I run it, it seems slow throughout entire process. But OTOH, I do use helper ext Java statics. But actually, my class-loading must have been working pretty well because up to 0.8.0 I think I've been getting identical performance with my plugin launched and workflow launched templates. Or at least I've been able to generate 30k lines plus within user interactive time (~10sec).

It doesn't make sense to do a small project because this is happening with my very large system. But..You could install the AMF plugins and *then* execute a very small example project I could bundle up for you. (This is just part of my plot to get people to play with AMF. Wink) If you don't feel like doing that I'll try to take another look at this when I return from modeling days next week.

cheers,

Miles
Re: [XPand]: 0.8.0 Performance surprise.. [message #499998 is a reply to message #496698] Tue, 24 November 2009 16:37 Go to previous messageGo to next message
ekkehard gentz is currently offline ekkehard gentzFriend
Messages: 118
Registered: July 2009
Location: rosenheim, Germany, bavar...
Senior Member

This is a multi-part message in MIME format.
--------------070008080205070900020608
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit

Miles Parker schrieb:
> Sebastian, see below please..
>
> Sebastian Zarnekow wrote on Sun, 08 November 2009 13:39
>> Hi Miles,
>>
>> that not a nice suprise.
>> Did you change the execution environment from plain Java VM to OSGi?
>> If you are in an OSGi environment, please have a look at
>> https://bugs.eclipse.org/bugs/show_bug.cgi?id=278753
>> Please file a bug with a short sample project if you do not use OSGi.
>
>
> Well, I'm executing it from an Eclipse plugin using WorkflowRunner, so
> I guess that's an OSGi environment. ;) However, this is what I've been
> doing all along, so unless the performance changes for the bug above
> have somehow worked at cross purposes to my own lousy attempts at
> managing buddies and class loading, I don't think that's it. Also,
> when I run it, it seems slow throughout entire process. But OTOH, I do
> use helper ext Java statics. But actually, my class-loading must have
> been working pretty well because up to 0.8.0 I think I've been getting
> identical performance with my plugin launched and workflow launched
> templates. Or at least I've been able to generate 30k lines plus
> within user interactive time (~10sec).
>
> It doesn't make sense to do a small project because this is happening
> with my very large system. But..You could install the AMF plugins and
> *then* execute a very small example project I could bundle up for you.
> (This is just part of my plot to get people to play with AMF. ;)) If
> you don't feel like doing that I'll try to take another look at this
> when I return from modeling days next week.
>
> cheers,
>
> Miles
Hi Miles,

just curious - did you find the reason of your performance surprise ?

ekke

--

ekke (ekkehard gentz)
independent software-architect
senior erp-consultant
eclipse | osgi | equinox | mdsd | oaw | emf | uml
max-josefs-platz 30, D-83022 rosenheim, germany
mailto:ekke@ekkes-corner.org
homepage (de): http://gentz-software.de
blog (en): http://ekkes-corner.org
twitter: @ekkescorner
skype: ekkes-corner
Steuer-Nr: 156/220/30931 FA Rosenheim, UST-ID: DE189929490


--------------070008080205070900020608
Content-Type: text/html; charset=UTF-8
Content-Transfer-Encoding: 8bit

<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=UTF-8" http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
Miles Parker schrieb:
<blockquote cite="mid:hdd4m4$f1s$1@build.eclipse.org" type="cite">Sebastian,
see below please..
<br>
<br>
Sebastian Zarnekow wrote on Sun, 08 November 2009 13:39
<br>
<blockquote type="cite">Hi Miles,
<br>
<br>
that not a nice suprise.
<br>
Did you change the execution environment from plain Java VM to OSGi?
<br>
If you are in an OSGi environment, please have a look at
<a class="moz-txt-link-freetext" href="https://bugs.eclipse.org/bugs/show_bug.cgi?id=278753">https://bugs.eclipse.org/bugs/show_bug.cgi?id=278753</a>
<br>
Please file a bug with a short sample project if you do not use OSGi.
<br>
</blockquote>
<br>
<br>
Well, I'm executing it from an Eclipse plugin using WorkflowRunner, so
I guess that's an OSGi environment. ;) However, this is what I've been
doing all along, so unless the performance changes  for the bug above
have somehow worked at cross purposes to my own lousy attempts at
managing buddies and class loading, I don't think that's it. Also, when
I run it, it seems slow throughout entire process. But OTOH, I do use
helper ext Java statics. But actually, my class-loading must have been
working pretty well because up to 0.8.0 I think I've been getting
identical performance with my plugin launched and workflow launched
templates. Or at least I've been able to generate 30k lines plus within
user interactive time (~10sec).
<br>
<br>
It doesn't make sense to do a small project because this is happening
with my very large system. But..You could install the AMF plugins and
*then* execute a very small example project I could bundle up for you.
(This is just part of my plot to get people to play with AMF.  ;)) If
you don't feel like doing that I'll try to take another look at this
when I return from modeling days next week.
<br>
<br>
cheers,
<br>
<br>
Miles
<br>
</blockquote>
Hi Miles,<br>
<br>
just curious - did you find the reason of your performance surprise ?<br>
<br>
ekke<br>
<br>
<div class="moz-signature">-- <br>
<p style="margin-bottom: 0cm;"> <font ,="" color="#808080"
face="Verdana, sans-serif"> <font size="4">ekke (ekkehard gentz)<br>
</font> <font size="3"> independent software-architect<br>
senior erp-consultant<br>
</font> <font size="2">eclipse | osgi | equinox | mdsd | oaw | emf |
uml<br>
max-josefs-platz 30, D-83022 rosenheim, germany<br>
<a class="moz-txt-link-freetext" href="mailto:ekke@ekkes-corner.org">mailto:ekke@ekkes-corner.org</a><br>
homepage (de): <a class="moz-txt-link-freetext" href="http://gentz-software.de">http://gentz-software.de</a><br>
blog (en): <a class="moz-txt-link-freetext" href="http://ekkes-corner.org">http://ekkes-corner.org</a><br>
twitter: @ekkescorner<br>
skype: ekkes-corner<br>
</font> <font size="1"> Steuer-Nr: 156/220/30931 FA Rosenheim, UST-ID:
DE189929490<br>
</font> </font> </p>
</div>
</body>
</html>

--------------070008080205070900020608--


ekke
independent software architect

blog: http://ekkes-corner.org
twitter: @ekkescorner
Re: [XPand]: 0.8.0 Performance surprise.. [message #500043 is a reply to message #499998] Tue, 24 November 2009 19:08 Go to previous messageGo to next message
Miles Parker is currently offline Miles ParkerFriend
Messages: 1341
Registered: July 2009
Senior Member
Hi Ekke,

This is very high on my list of very high things on my list to address. Smile On the theory that it might be related to use of Java external calls from Xtend, I'm planning to get rid of those and see how it affects things. I'll report back as soon as I can.

cheers,

Miles
Re: [XPand]: 0.8.0 Performance surprise.. [message #501442 is a reply to message #495972] Wed, 02 December 2009 11:07 Go to previous messageGo to next message
ekkehard gentz is currently offline ekkehard gentzFriend
Messages: 118
Registered: July 2009
Location: rosenheim, Germany, bavar...
Senior Member

This is a multi-part message in MIME format.
--------------050003090907080001080506
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit

Miles Parker schrieb:
> One additional data point here. I've just tried it on my desktop
> machine (MacPro 8-core) and the performance delta is less compared to
> my laptop, but still in the 5x-12x + range. So perhaps there is some
> kind of caching or other heap issue here. It's a bit difficult to
> switch back and forth but I can prob. set up two configurations if
> more testing would be helpful. But hopefully there is something simple
> going on here.
Miles, you're not alone,

just tested 0.8.0 - both versions: 0.8.0M3 and N-Builds 2009-12-01
I got nearly same values as you: 18x - 40x slower
I reverted to 0.7.2

ekke
--

ekke (ekkehard gentz)
independent software-architect
senior erp-consultant
eclipse | osgi | equinox | mdsd | oaw | emf | uml
max-josefs-platz 30, D-83022 rosenheim, germany
mailto:ekke@ekkes-corner.org
homepage (de): http://gentz-software.de
blog (en): http://ekkes-corner.org
twitter: @ekkescorner
skype: ekkes-corner
Steuer-Nr: 156/220/30931 FA Rosenheim, UST-ID: DE189929490


--------------050003090907080001080506
Content-Type: text/html; charset=UTF-8
Content-Transfer-Encoding: 7bit

<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=UTF-8" http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
Miles Parker schrieb:
<blockquote cite="mid:hd2048$50b$1@build.eclipse.org" type="cite">One
additional data point here. I've just tried it on my desktop machine
(MacPro 8-core) and the performance delta is less compared to my
laptop, but still in the 5x-12x + range. So perhaps there is some kind
of caching or other heap issue here. It's a bit difficult to switch
back and forth but I can prob. set up two configurations if more
testing would be helpful. But hopefully there is something simple going
on here.
<br>
</blockquote>
Miles, you're not alone,<br>
<br>
just tested 0.8.0 - both versions: 0.8.0M3 and N-Builds 2009-12-01<br>
I got nearly same values as you: 18x - 40x slower<br>
I reverted to 0.7.2<br>
<br>
ekke<br>
<div class="moz-signature">-- <br>
<p style="margin-bottom: 0cm;"> <font ,="" color="#808080"
face="Verdana, sans-serif"> <font size="4">ekke (ekkehard gentz)<br>
</font> <font size="3"> independent software-architect<br>
senior erp-consultant<br>
</font> <font size="2">eclipse | osgi | equinox | mdsd | oaw | emf |
uml<br>
max-josefs-platz 30, D-83022 rosenheim, germany<br>
<a class="moz-txt-link-freetext" href="mailto:ekke@ekkes-corner.org">mailto:ekke@ekkes-corner.org</a><br>
homepage (de): <a class="moz-txt-link-freetext" href="http://gentz-software.de">http://gentz-software.de</a><br>
blog (en): <a class="moz-txt-link-freetext" href="http://ekkes-corner.org">http://ekkes-corner.org</a><br>
twitter: @ekkescorner<br>
skype: ekkes-corner<br>
</font> <font size="1"> Steuer-Nr: 156/220/30931 FA Rosenheim, UST-ID:
DE189929490<br>
</font> </font> </p>
</div>
</body>
</html>

--------------050003090907080001080506--


ekke
independent software architect

blog: http://ekkes-corner.org
twitter: @ekkescorner
Re: [XPand]: 0.8.0 Performance surprise.. [message #501460 is a reply to message #501442] Wed, 02 December 2009 12:53 Go to previous messageGo to next message
Sven Efftinge is currently offline Sven EfftingeFriend
Messages: 1823
Registered: July 2009
Senior Member
Hi Ekke,

could you please submit a bug report?

Thanks,
Sven

ekke schrieb:
> Miles Parker schrieb:
>> One additional data point here. I've just tried it on my desktop
>> machine (MacPro 8-core) and the performance delta is less compared to
>> my laptop, but still in the 5x-12x + range. So perhaps there is some
>> kind of caching or other heap issue here. It's a bit difficult to
>> switch back and forth but I can prob. set up two configurations if
>> more testing would be helpful. But hopefully there is something simple
>> going on here.
> Miles, you're not alone,
>
> just tested 0.8.0 - both versions: 0.8.0M3 and N-Builds 2009-12-01
> I got nearly same values as you: 18x - 40x slower
> I reverted to 0.7.2
>
> ekke
> --
>
> ekke (ekkehard gentz)
> independent software-architect
> senior erp-consultant
> eclipse | osgi | equinox | mdsd | oaw | emf | uml
> max-josefs-platz 30, D-83022 rosenheim, germany
> mailto:ekke@ekkes-corner.org
> homepage (de): http://gentz-software.de
> blog (en): http://ekkes-corner.org
> twitter: @ekkescorner
> skype: ekkes-corner
> Steuer-Nr: 156/220/30931 FA Rosenheim, UST-ID: DE189929490
>


--
Need professional support for Xtext and EMF?
Go to: http://xtext.itemis.com
Re: [XPand]: 0.8.0 Performance surprise.. [message #501462 is a reply to message #501460] Wed, 02 December 2009 12:55 Go to previous messageGo to next message
ekkehard gentz is currently offline ekkehard gentzFriend
Messages: 118
Registered: July 2009
Location: rosenheim, Germany, bavar...
Senior Member

This is a multi-part message in MIME format.
--------------050502010200050809080007
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit

Sven Efftinge schrieb:
> Hi Ekke,
>
> could you please submit a bug report?
of course, but will take some days to prepare a project to test with

ekke
>
> Thanks,
> Sven
>
> ekke schrieb:
>> Miles Parker schrieb:
>>> One additional data point here. I've just tried it on my desktop
>>> machine (MacPro 8-core) and the performance delta is less compared
>>> to my laptop, but still in the 5x-12x + range. So perhaps there is
>>> some kind of caching or other heap issue here. It's a bit difficult
>>> to switch back and forth but I can prob. set up two configurations
>>> if more testing would be helpful. But hopefully there is something
>>> simple going on here.
>> Miles, you're not alone,
>>
>> just tested 0.8.0 - both versions: 0.8.0M3 and N-Builds 2009-12-01
>> I got nearly same values as you: 18x - 40x slower
>> I reverted to 0.7.2
>>
>> ekke
>> --
>>
>> ekke (ekkehard gentz)
>> independent software-architect
>> senior erp-consultant
>> eclipse | osgi | equinox | mdsd | oaw | emf | uml
>> max-josefs-platz 30, D-83022 rosenheim, germany
>> mailto:ekke@ekkes-corner.org
>> homepage (de): http://gentz-software.de
>> blog (en): http://ekkes-corner.org
>> twitter: @ekkescorner
>> skype: ekkes-corner
>> Steuer-Nr: 156/220/30931 FA Rosenheim, UST-ID: DE189929490
>>
>
>


--

ekke (ekkehard gentz)
independent software-architect
senior erp-consultant
eclipse | osgi | equinox | mdsd | oaw | emf | uml
max-josefs-platz 30, D-83022 rosenheim, germany
mailto:ekke@ekkes-corner.org
homepage (de): http://gentz-software.de
blog (en): http://ekkes-corner.org
twitter: @ekkescorner
skype: ekkes-corner
Steuer-Nr: 156/220/30931 FA Rosenheim, UST-ID: DE189929490


--------------050502010200050809080007
Content-Type: text/html; charset=UTF-8
Content-Transfer-Encoding: 8bit

<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=UTF-8" http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
Sven Efftinge schrieb:
<blockquote cite="mid:hf5nvj$pp4$1@build.eclipse.org" type="cite">Hi
Ekke,
<br>
<br>
could you please submit a bug report?
<br>
</blockquote>
of course, but will take some days to prepare a project to test with<br>
<br>
ekke<br>
<blockquote cite="mid:hf5nvj$pp4$1@build.eclipse.org" type="cite"><br>
Thanks,
<br>
Sven
<br>
<br>
ekke schrieb:
<br>
<blockquote type="cite">Miles Parker schrieb:
<br>
<blockquote type="cite">One additional data point here. I've just
tried it on my desktop machine (MacPro 8-core) and the performance
delta is less compared to my laptop, but still in the 5x-12x + range.
So perhaps there is some kind of caching or other heap issue here. It's
a bit difficult to switch back and forth but I can prob. set up two
configurations if more testing would be helpful. But hopefully there is
something simple going on here.
<br>
</blockquote>
Miles, you're not alone,
<br>
<br>
just tested 0.8.0 - both versions: 0.8.0M3 and N-Builds 2009-12-01
<br>
I got nearly same values as you: 18x - 40x slower
<br>
I reverted to 0.7.2
<br>
<br>
ekke
<br>
-- <br>
<br>
ekke (ekkehard gentz)
<br>
independent software-architect
<br>
senior erp-consultant
<br>
eclipse | osgi | equinox | mdsd | oaw | emf | uml
<br>
max-josefs-platz 30, D-83022 rosenheim, germany
<br>
<a class="moz-txt-link-freetext" href="mailto:ekke@ekkes-corner.org">mailto:ekke@ekkes-corner.org</a>
<br>
homepage (de): <a class="moz-txt-link-freetext" href="http://gentz-software.de">http://gentz-software.de</a>
<br>
blog (en): <a class="moz-txt-link-freetext" href="http://ekkes-corner.org">http://ekkes-corner.org</a>
<br>
twitter: @ekkescorner
<br>
skype: ekkes-corner
<br>
Steuer-Nr: 156/220/30931 FA Rosenheim, UST-ID: DE189929490
<br>
<br>
</blockquote>
<br>
<br>
</blockquote>
<br>
<br>
<div class="moz-signature">-- <br>
<p style="margin-bottom: 0cm;"> <font ,="" color="#808080"
face="Verdana, sans-serif"> <font size="4">ekke (ekkehard gentz)<br>
</font> <font size="3"> independent software-architect<br>
senior erp-consultant<br>
</font> <font size="2">eclipse | osgi | equinox | mdsd | oaw | emf |
uml<br>
max-josefs-platz 30, D-83022 rosenheim, germany<br>
<a class="moz-txt-link-freetext" href="mailto:ekke@ekkes-corner.org">mailto:ekke@ekkes-corner.org</a><br>
homepage (de): <a class="moz-txt-link-freetext" href="http://gentz-software.de">http://gentz-software.de</a><br>
blog (en): <a class="moz-txt-link-freetext" href="http://ekkes-corner.org">http://ekkes-corner.org</a><br>
twitter: @ekkescorner<br>
skype: ekkes-corner<br>
</font> <font size="1"> Steuer-Nr: 156/220/30931 FA Rosenheim, UST-ID:
DE189929490<br>
</font> </font> </p>
</div>
</body>
</html>

--------------050502010200050809080007--


ekke
independent software architect

blog: http://ekkes-corner.org
twitter: @ekkescorner
Re: [XPand]: 0.8.0 Performance surprise.. [message #501591 is a reply to message #501460] Wed, 02 December 2009 22:52 Go to previous messageGo to next message
ekkehard gentz is currently offline ekkehard gentzFriend
Messages: 118
Registered: July 2009
Location: rosenheim, Germany, bavar...
Senior Member

This is a multi-part message in MIME format.
--------------090405000102080700050901
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit

after discussion with Karsten Thoms: there's some work-in-progress and
no bugzilla needed at the moment

ekke

Sven Efftinge schrieb:
> Hi Ekke,
>
> could you please submit a bug report?
>
> Thanks,
> Sven
>
> ekke schrieb:
>> Miles Parker schrieb:
>>> One additional data point here. I've just tried it on my desktop
>>> machine (MacPro 8-core) and the performance delta is less compared
>>> to my laptop, but still in the 5x-12x + range. So perhaps there is
>>> some kind of caching or other heap issue here. It's a bit difficult
>>> to switch back and forth but I can prob. set up two configurations
>>> if more testing would be helpful. But hopefully there is something
>>> simple going on here.
>> Miles, you're not alone,
>>
>> just tested 0.8.0 - both versions: 0.8.0M3 and N-Builds 2009-12-01
>> I got nearly same values as you: 18x - 40x slower
>> I reverted to 0.7.2
>>
>> ekke
>> --
>>
>> ekke (ekkehard gentz)
>> independent software-architect
>> senior erp-consultant
>> eclipse | osgi | equinox | mdsd | oaw | emf | uml
>> max-josefs-platz 30, D-83022 rosenheim, germany
>> mailto:ekke@ekkes-corner.org
>> homepage (de): http://gentz-software.de
>> blog (en): http://ekkes-corner.org
>> twitter: @ekkescorner
>> skype: ekkes-corner
>> Steuer-Nr: 156/220/30931 FA Rosenheim, UST-ID: DE189929490
>>
>
>


--

ekke (ekkehard gentz)
independent software-architect
senior erp-consultant
eclipse | osgi | equinox | mdsd | oaw | emf | uml
max-josefs-platz 30, D-83022 rosenheim, germany
mailto:ekke@ekkes-corner.org
homepage (de): http://gentz-software.de
blog (en): http://ekkes-corner.org
twitter: @ekkescorner
skype: ekkes-corner
Steuer-Nr: 156/220/30931 FA Rosenheim, UST-ID: DE189929490


--------------090405000102080700050901
Content-Type: text/html; charset=UTF-8
Content-Transfer-Encoding: 8bit

<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=UTF-8" http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
after discussion with Karsten Thoms: there's some work-in-progress and
no bugzilla needed at the moment<br>
<br>
ekke<br>
<br>
Sven Efftinge schrieb:
<blockquote cite="mid:hf5nvj$pp4$1@build.eclipse.org" type="cite">Hi
Ekke,
<br>
<br>
could you please submit a bug report?
<br>
<br>
Thanks,
<br>
Sven
<br>
<br>
ekke schrieb:
<br>
<blockquote type="cite">Miles Parker schrieb:
<br>
<blockquote type="cite">One additional data point here. I've just
tried it on my desktop machine (MacPro 8-core) and the performance
delta is less compared to my laptop, but still in the 5x-12x + range.
So perhaps there is some kind of caching or other heap issue here. It's
a bit difficult to switch back and forth but I can prob. set up two
configurations if more testing would be helpful. But hopefully there is
something simple going on here.
<br>
</blockquote>
Miles, you're not alone,
<br>
<br>
just tested 0.8.0 - both versions: 0.8.0M3 and N-Builds 2009-12-01
<br>
I got nearly same values as you: 18x - 40x slower
<br>
I reverted to 0.7.2
<br>
<br>
ekke
<br>
-- <br>
<br>
ekke (ekkehard gentz)
<br>
independent software-architect
<br>
senior erp-consultant
<br>
eclipse | osgi | equinox | mdsd | oaw | emf | uml
<br>
max-josefs-platz 30, D-83022 rosenheim, germany
<br>
<a class="moz-txt-link-freetext" href="mailto:ekke@ekkes-corner.org">mailto:ekke@ekkes-corner.org</a>
<br>
homepage (de): <a class="moz-txt-link-freetext" href="http://gentz-software.de">http://gentz-software.de</a>
<br>
blog (en): <a class="moz-txt-link-freetext" href="http://ekkes-corner.org">http://ekkes-corner.org</a>
<br>
twitter: @ekkescorner
<br>
skype: ekkes-corner
<br>
Steuer-Nr: 156/220/30931 FA Rosenheim, UST-ID: DE189929490
<br>
<br>
</blockquote>
<br>
<br>
</blockquote>
<br>
<br>
<div class="moz-signature">-- <br>
<p style="margin-bottom: 0cm;"> <font ,="" color="#808080"
face="Verdana, sans-serif"> <font size="4">ekke (ekkehard gentz)<br>
</font> <font size="3"> independent software-architect<br>
senior erp-consultant<br>
</font> <font size="2">eclipse | osgi | equinox | mdsd | oaw | emf |
uml<br>
max-josefs-platz 30, D-83022 rosenheim, germany<br>
<a class="moz-txt-link-freetext" href="mailto:ekke@ekkes-corner.org">mailto:ekke@ekkes-corner.org</a><br>
homepage (de): <a class="moz-txt-link-freetext" href="http://gentz-software.de">http://gentz-software.de</a><br>
blog (en): <a class="moz-txt-link-freetext" href="http://ekkes-corner.org">http://ekkes-corner.org</a><br>
twitter: @ekkescorner<br>
skype: ekkes-corner<br>
</font> <font size="1"> Steuer-Nr: 156/220/30931 FA Rosenheim, UST-ID:
DE189929490<br>
</font> </font> </p>
</div>
</body>
</html>

--------------090405000102080700050901--


ekke
independent software architect

blog: http://ekkes-corner.org
twitter: @ekkescorner
Re: [XPand]: 0.8.0 Performance surprise.. [message #501816 is a reply to message #501591] Thu, 03 December 2009 22:31 Go to previous messageGo to next message
Miles Parker is currently offline Miles ParkerFriend
Messages: 1341
Registered: July 2009
Senior Member
ekke (ekkehard gentz) wrote on Wed, 02 December 2009 17:52
This is a multi-part message in MIME format.
--------------090405000102080700050901
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit

after discussion with Karsten Thoms: there's some work-in-progress and
no bugzilla needed at the moment

ekke



Oh good. Smile I was working for a few hours yesterday trying to put something together but my stuff is so API based that I was having trouble putting together a quick workflow to test things -- not to mention that you need half a dozen plugin dependencies. Any hints?

cheers!

Miles

[Updated on: Thu, 03 December 2009 22:32]

Report message to a moderator

Re: [XPand]: 0.8.0 Performance surprise.. [message #502196 is a reply to message #495961] Mon, 07 December 2009 08:03 Go to previous messageGo to next message
Daniel Weber is currently offline Daniel WeberFriend
Messages: 51
Registered: July 2009
Member
Hi all,

> [...]
> significant, I mean that a model that was taking 1 second to generate is
> now taking 45 seconds . In general I'm seeing a 25-40x slow-down. This
> [...]

just tried to reproduce this using a workflow from the emfmodelvisualizer project [1]. On my machine, this workflow takes about 1.5 seconds with 0.7.2 and ~3 seconds using a 0.8 version. I was able to track it down to the following change in the EmfRegistryMetaModel:

@@ -262,20 +263,33 @@ public class EmfRegistryMetaModel implements MetaModel {
return getTypeForEClassifier(el.getEEnum());
}
final EClass clazz = ((EObject) obj).eClass();
return getTypeForEClassifier(clazz);
}
+
+ final Set<Type> types = getKnownTypes();
if (obj instanceof Enumerator) {
- final Set<Type> types = getKnownTypes();
for (Type t : types) {
+ // check for EEnumType because the EDataTypeType for
+ // ecore::EEnumerator matches all enumerator instances
if (t instanceof EEnumType) {
if (t.isInstance(obj)) {
return t;
}
}
}
+ } else {
+ // some other non-enumerated EDataType
+ for (Type t : types) {
+ if (t instanceof EDataTypeType) {
+ if (t.isInstance(obj)) {
+ return t;
+ }
+ }
+ }
}
+
return null;
}

public Set<Type> getKnownTypes() {
final Set<Type> result = new HashSet<Type>();


without the else branch (which is executed for tons of objects, e.g. all Strings, Integers,...), it only took 1.5 seconds again. But simply removing it won't be a solution, as it has been added to fix a bug [2].

Although this is "only" factor 2 (in contrast to 25-40x) for this workflow, it probably is at least one of the causes of what you observed.

> --- [ekke wrote] ---
> after discussion with Karsten Thoms: there's some work-in-progress and no bugzilla needed at the moment

Is this still true? Or should I put this information into a new bug report?

Regards,
Daniel

[1] http://code.google.com/p/emfmodelvisualizer/source/browse/ga lileo/trunk/org.openarchitectureware.vis.graphviz.demo/src/v isualizeEcore.mwe
[2] https://bugs.eclipse.org/bugs/show_bug.cgi?id=281269
Re: [XPand]: 0.8.0 Performance surprise.. [message #502226 is a reply to message #499998] Mon, 07 December 2009 10:07 Go to previous messageGo to next message
Sven Efftinge is currently offline Sven EfftingeFriend
Messages: 1823
Registered: July 2009
Senior Member
Well, others might want to be up to date:

https://bugs.eclipse.org/bugs/show_bug.cgi?id=297053


ekke schrieb:
> Miles Parker schrieb:
>> Sebastian, see below please..
>>
>> Sebastian Zarnekow wrote on Sun, 08 November 2009 13:39
>>> Hi Miles,
>>>
>>> that not a nice suprise.
>>> Did you change the execution environment from plain Java VM to OSGi?
>>> If you are in an OSGi environment, please have a look at
>>> https://bugs.eclipse.org/bugs/show_bug.cgi?id=278753
>>> Please file a bug with a short sample project if you do not use OSGi.
>>
>>
>> Well, I'm executing it from an Eclipse plugin using WorkflowRunner, so
>> I guess that's an OSGi environment. ;) However, this is what I've been
>> doing all along, so unless the performance changes for the bug above
>> have somehow worked at cross purposes to my own lousy attempts at
>> managing buddies and class loading, I don't think that's it. Also,
>> when I run it, it seems slow throughout entire process. But OTOH, I do
>> use helper ext Java statics. But actually, my class-loading must have
>> been working pretty well because up to 0.8.0 I think I've been getting
>> identical performance with my plugin launched and workflow launched
>> templates. Or at least I've been able to generate 30k lines plus
>> within user interactive time (~10sec).
>>
>> It doesn't make sense to do a small project because this is happening
>> with my very large system. But..You could install the AMF plugins and
>> *then* execute a very small example project I could bundle up for you.
>> (This is just part of my plot to get people to play with AMF. ;)) If
>> you don't feel like doing that I'll try to take another look at this
>> when I return from modeling days next week.
>>
>> cheers,
>>
>> Miles
> Hi Miles,
>
> just curious - did you find the reason of your performance surprise ?
>
> ekke
>
> --
>
> ekke (ekkehard gentz)
> independent software-architect
> senior erp-consultant
> eclipse | osgi | equinox | mdsd | oaw | emf | uml
> max-josefs-platz 30, D-83022 rosenheim, germany
> mailto:ekke@ekkes-corner.org
> homepage (de): http://gentz-software.de
> blog (en): http://ekkes-corner.org
> twitter: @ekkescorner
> skype: ekkes-corner
> Steuer-Nr: 156/220/30931 FA Rosenheim, UST-ID: DE189929490
>


--
Need professional support for Xtext and EMF?
Go to: http://xtext.itemis.com
Re: [XPand]: 0.8.0 Performance surprise.. [message #502227 is a reply to message #502196] Mon, 07 December 2009 10:10 Go to previous messageGo to next message
Sven Efftinge is currently offline Sven EfftingeFriend
Messages: 1823
Registered: July 2009
Senior Member
Thanks, that is of great help.

Cheers,
sven


Daniel Weber schrieb:
> Hi all,
>
>> [...]
>> significant, I mean that a model that was taking 1 second to generate
>> is now taking 45 seconds . In general I'm seeing a 25-40x slow-down.
>> This [...]
>
> just tried to reproduce this using a workflow from the
> emfmodelvisualizer project [1]. On my machine, this workflow takes about
> 1.5 seconds with 0.7.2 and ~3 seconds using a 0.8 version. I was able to
> track it down to the following change in the EmfRegistryMetaModel:
>
> @@ -262,20 +263,33 @@ public class EmfRegistryMetaModel implements
> MetaModel {
> return getTypeForEClassifier(el.getEEnum());
> }
> final EClass clazz = ((EObject) obj).eClass();
> return getTypeForEClassifier(clazz);
> }
> +
> + final Set<Type> types = getKnownTypes();
> if (obj instanceof Enumerator) {
> - final Set<Type> types = getKnownTypes();
> for (Type t : types) {
> + // check for EEnumType because the EDataTypeType for
> + // ecore::EEnumerator matches all enumerator instances
> if (t instanceof EEnumType) {
> if (t.isInstance(obj)) {
> return t;
> }
> }
> }
> + } else {
> + // some other non-enumerated EDataType
> + for (Type t : types) {
> + if (t instanceof EDataTypeType) {
> + if (t.isInstance(obj)) {
> + return t;
> + }
> + }
> + }
> }
> +
> return null;
> }
>
> public Set<Type> getKnownTypes() {
> final Set<Type> result = new HashSet<Type>();
>
>
> without the else branch (which is executed for tons of objects, e.g. all
> Strings, Integers,...), it only took 1.5 seconds again. But simply
> removing it won't be a solution, as it has been added to fix a bug [2].
> Although this is "only" factor 2 (in contrast to 25-40x) for this
> workflow, it probably is at least one of the causes of what you observed.
>
>> --- [ekke wrote] ---
>> after discussion with Karsten Thoms: there's some work-in-progress and
>> no bugzilla needed at the moment
>
> Is this still true? Or should I put this information into a new bug report?
>
> Regards,
> Daniel
>
> [1]
> http://code.google.com/p/emfmodelvisualizer/source/browse/ga lileo/trunk/org.openarchitectureware.vis.graphviz.demo/src/v isualizeEcore.mwe
>
> [2] https://bugs.eclipse.org/bugs/show_bug.cgi?id=281269


--
Need professional support for Xtext and EMF?
Go to: http://xtext.itemis.com
Re: [XPand]: 0.8.0 Performance surprise.. [message #503375 is a reply to message #502227] Sat, 12 December 2009 15:00 Go to previous messageGo to next message
ekkehard gentz is currently offline ekkehard gentzFriend
Messages: 118
Registered: July 2009
Location: rosenheim, Germany, bavar...
Senior Member

This is a multi-part message in MIME format.
--------------000604090806060001050400
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit

just tested using N200912111914 MWE, Xpand, Xtext

the good news:
performance problem is fixed -
runs same speed as before with 0.7.2 :-)

the bad news:
now I'm getting hundreds of those warnings:

15:25:52.299 WARN [o.e.xtend.typesystem.emf.EClassType] Couldn't
resolve type
for EDiagnosticChain
15:25:52.299 WARN [o.e.xtend.typesystem.emf.EClassType] Couldn't
resolve type
for EMap

but my workflow seems to run well

--

ekke (ekkehard gentz)
independent software-architect
senior erp-consultant
eclipse | osgi | equinox | mdsd | oaw | emf | uml
max-josefs-platz 30, D-83022 rosenheim, germany
mailto:ekke@ekkes-corner.org
homepage (de): http://gentz-software.de
blog (en): http://ekkes-corner.org
twitter: @ekkescorner
skype: ekkes-corner
Steuer-Nr: 156/220/30931 FA Rosenheim, UST-ID: DE189929490


--------------000604090806060001050400
Content-Type: text/html; charset=UTF-8
Content-Transfer-Encoding: 8bit

<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=UTF-8" http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
just tested using N200912111914 MWE, Xpand, Xtext<br>
<br>
the good news:<br>
performance problem is fixed -<br>
runs same speed as before with 0.7.2 :-)<br>
<br>
the bad news:<br>
now I'm getting hundreds of those warnings:<br>
<br>
15:25:52.299 WARN  [o.e.xtend.typesystem.emf.EClassType] Couldn't
resolve type<br>
for EDiagnosticChain<br>
15:25:52.299 WARN  [o.e.xtend.typesystem.emf.EClassType] Couldn't
resolve type<br>
for EMap<br>
<br>
but my workflow seems to run well<br>
<br>
<div class="moz-signature">-- <br>
<p style="margin-bottom: 0cm;"> <font ,="" color="#808080"
face="Verdana, sans-serif"> <font size="4">ekke (ekkehard gentz)<br>
</font> <font size="3"> independent software-architect<br>
senior erp-consultant<br>
</font> <font size="2">eclipse | osgi | equinox | mdsd | oaw | emf |
uml<br>
max-josefs-platz 30, D-83022 rosenheim, germany<br>
<a class="moz-txt-link-freetext" href="mailto:ekke@ekkes-corner.org">mailto:ekke@ekkes-corner.org</a><br>
homepage (de): <a class="moz-txt-link-freetext" href="http://gentz-software.de">http://gentz-software.de</a><br>
blog (en): <a class="moz-txt-link-freetext" href="http://ekkes-corner.org">http://ekkes-corner.org</a><br>
twitter: @ekkescorner<br>
skype: ekkes-corner<br>
</font> <font size="1"> Steuer-Nr: 156/220/30931 FA Rosenheim, UST-ID:
DE189929490<br>
</font> </font> </p>
</div>
</body>
</html>

--------------000604090806060001050400--


ekke
independent software architect

blog: http://ekkes-corner.org
twitter: @ekkescorner
Re: [XPand]: 0.8.0 Performance surprise.. [message #504455 is a reply to message #503375] Sat, 19 December 2009 14:19 Go to previous message
ekkehard gentz is currently offline ekkehard gentzFriend
Messages: 118
Registered: July 2009
Location: rosenheim, Germany, bavar...
Senior Member

This is a multi-part message in MIME format.
--------------050600030101040607070805
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit

using 0.8.0M4 this is fixed.
THX
ekke

ekke schrieb:
> just tested using N200912111914 MWE, Xpand, Xtext
>
> the good news:
> performance problem is fixed -
> runs same speed as before with 0.7.2 :-)
>
> the bad news:
> now I'm getting hundreds of those warnings:
>
> 15:25:52.299 WARN [o.e.xtend.typesystem.emf.EClassType] Couldn't
> resolve type
> for EDiagnosticChain
> 15:25:52.299 WARN [o.e.xtend.typesystem.emf.EClassType] Couldn't
> resolve type
> for EMap
>


--------------050600030101040607070805
Content-Type: text/html; charset=UTF-8
Content-Transfer-Encoding: 8bit

<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=UTF-8" http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
using 0.8.0M4 this is fixed.<br>
THX<br>
ekke<br>
<br>
ekke schrieb:
<blockquote cite="mid:hg0b6t$7hq$1@build.eclipse.org" type="cite">
<meta content="text/html;charset=UTF-8" http-equiv="Content-Type">
just tested using N200912111914 MWE, Xpand, Xtext<br>
<br>
the good news:<br>
performance problem is fixed -<br>
runs same speed as before with 0.7.2 :-)<br>
<br>
the bad news:<br>
now I'm getting hundreds of those warnings:<br>
<br>
15:25:52.299 WARN  [o.e.xtend.typesystem.emf.EClassType] Couldn't
resolve type<br>
for EDiagnosticChain<br>
15:25:52.299 WARN  [o.e.xtend.typesystem.emf.EClassType] Couldn't
resolve type<br>
for EMap<br>
<br>
</blockquote>
<br>
</body>
</html>

--------------050600030101040607070805--


ekke
independent software architect

blog: http://ekkes-corner.org
twitter: @ekkescorner
Previous Topic:[XPAND] [MWE] 0.8.0 regression
Next Topic:QVT code for xsd Annotations
Goto Forum:
  


Current Time: Thu Oct 10 20:43:04 GMT 2024

Powered by FUDForum. Page generated in 0.05964 seconds
.:: Contact :: Home ::.

Powered by: FUDforum 3.0.2.
Copyright ©2001-2010 FUDforum Bulletin Board Software

Back to the top