Eclipse Community Forums
Forum Search:

Search      Help    Register    Login    Home
Home » Modeling » Modeling (top-level project) » performance difference between dynamic and "static" emf
performance difference between dynamic and "static" emf [message #379978] Mon, 12 March 2007 20:31 Go to next message
Pradeep is currently offline Pradeep
Messages: 61
Registered: July 2009
Member
Hi,

Could someone tell me, what are the performance implications of dynamic emf.

How does it compare to that of using generated classes?


Thanks
Pradeep
Re: performance difference between dynamic and "static" emf [message #379980 is a reply to message #379978] Mon, 12 March 2007 23:00 Go to previous message
Ed Merks is currently offline Ed Merks
Messages: 25748
Registered: July 2009
Senior Member
This is a multi-part message in MIME format.
--------------000609060907010700040001
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

Pradeep,

Please use the EMF newsgroup for EMF-specific questions. I've added it
to the "to" list, so you can continue to use this thread.

Dynamic models are slower and will use more space. Here's some old
data that was focused on SDO performance improvements but does have
measurements that let you compare dynamic and static access:

http://www.eclipse.org/modeling/emf/docs/performance/EMFPerf ormanceTestsResults.html

Clearly a getX() will be faster than eGet(featureX) since the later
calls the former. Since a dynamic model doesn't have a generated getX,
that approach requires using eGet(featureX). Dynamic models don't use
primitive types, but rather use box types, i.e., instead of int they use
Integer, so that leads to more memory use. Also, a dynamic model uses
an array of Object to store the values, but static models use variables
of the correct type, including variables of primitive type, so that
makes static models both faster and smaller. It's probably best to make
measurements yourself on the actual JVM you will be using, since the JVM
often has a big impact on the performance of different things...


Pradeep wrote:
> Hi,
>
> Could someone tell me, what are the performance implications of
> dynamic emf.
>
> How does it compare to that of using generated classes?
>
>
> Thanks
> Pradeep


--------------000609060907010700040001
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit

<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
Pradeep,<br>
<br>
Please use the EMF newsgroup for EMF-specific questions.&nbsp; I've added it
to the "to" list, so you can continue to use this thread.<br>
<br>
Dynamic models are slower and will use more space.&nbsp;&nbsp; Here's some old
data that was focused on SDO performance improvements but does have
measurements that let you compare dynamic and static access:<br>
<blockquote><a
href=" http://www.eclipse.org/modeling/emf/docs/performance/EMFPerf ormanceTestsResults.html"> http://www.eclipse.org/modeling/emf/docs/performance/EMFPerf ormanceTestsResults.html</a><br>
</blockquote>
Clearly a getX() will be faster than eGet(featureX) since the later
calls the former.&nbsp; Since a dynamic model doesn't have a generated getX,
that approach requires using eGet(featureX).&nbsp; Dynamic models don't use
primitive types, but rather use box types, i.e., instead of int they
use Integer, so that leads to more memory use.&nbsp; Also, a dynamic model
uses an array of Object to store the values, but static models use
variables of the correct type, including variables of primitive type,
so that makes static models both faster and smaller.&nbsp; It's probably
best to make measurements yourself on the actual JVM you will be using,
since the JVM often has a big impact on the performance of different
things...<br>
<br>
<br>
Pradeep wrote:
<blockquote cite="midet4rda$aip$1@utils.eclipse.org" type="cite">Hi,
<br>
<br>
Could someone tell me, what are the performance implications of dynamic
emf.
<br>
<br>
How does it compare to that of using generated classes?
<br>
<br>
<br>
Thanks
<br>
Pradeep
<br>
</blockquote>
<br>
</body>
</html>

--------------000609060907010700040001--
Re: performance difference between dynamic and "static" emf [message #590016 is a reply to message #379978] Mon, 12 March 2007 23:00 Go to previous message
Ed Merks is currently offline Ed Merks
Messages: 25748
Registered: July 2009
Senior Member
This is a multi-part message in MIME format.
--------------000609060907010700040001
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

Pradeep,

Please use the EMF newsgroup for EMF-specific questions. I've added it
to the "to" list, so you can continue to use this thread.

Dynamic models are slower and will use more space. Here's some old
data that was focused on SDO performance improvements but does have
measurements that let you compare dynamic and static access:

http://www.eclipse.org/modeling/emf/docs/performance/EMFPerf ormanceTestsResults.html

Clearly a getX() will be faster than eGet(featureX) since the later
calls the former. Since a dynamic model doesn't have a generated getX,
that approach requires using eGet(featureX). Dynamic models don't use
primitive types, but rather use box types, i.e., instead of int they use
Integer, so that leads to more memory use. Also, a dynamic model uses
an array of Object to store the values, but static models use variables
of the correct type, including variables of primitive type, so that
makes static models both faster and smaller. It's probably best to make
measurements yourself on the actual JVM you will be using, since the JVM
often has a big impact on the performance of different things...


Pradeep wrote:
> Hi,
>
> Could someone tell me, what are the performance implications of
> dynamic emf.
>
> How does it compare to that of using generated classes?
>
>
> Thanks
> Pradeep


--------------000609060907010700040001
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit

<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
Pradeep,<br>
<br>
Please use the EMF newsgroup for EMF-specific questions.&nbsp; I've added it
to the "to" list, so you can continue to use this thread.<br>
<br>
Dynamic models are slower and will use more space.&nbsp;&nbsp; Here's some old
data that was focused on SDO performance improvements but does have
measurements that let you compare dynamic and static access:<br>
<blockquote><a
href=" http://www.eclipse.org/modeling/emf/docs/performance/EMFPerf ormanceTestsResults.html"> http://www.eclipse.org/modeling/emf/docs/performance/EMFPerf ormanceTestsResults.html</a><br>
</blockquote>
Clearly a getX() will be faster than eGet(featureX) since the later
calls the former.&nbsp; Since a dynamic model doesn't have a generated getX,
that approach requires using eGet(featureX).&nbsp; Dynamic models don't use
primitive types, but rather use box types, i.e., instead of int they
use Integer, so that leads to more memory use.&nbsp; Also, a dynamic model
uses an array of Object to store the values, but static models use
variables of the correct type, including variables of primitive type,
so that makes static models both faster and smaller.&nbsp; It's probably
best to make measurements yourself on the actual JVM you will be using,
since the JVM often has a big impact on the performance of different
things...<br>
<br>
<br>
Pradeep wrote:
<blockquote cite="midet4rda$aip$1@utils.eclipse.org" type="cite">Hi,
<br>
<br>
Could someone tell me, what are the performance implications of dynamic
emf.
<br>
<br>
How does it compare to that of using generated classes?
<br>
<br>
<br>
Thanks
<br>
Pradeep
<br>
</blockquote>
<br>
</body>
</html>

--------------000609060907010700040001--
Previous Topic:performance difference between dynamic and "static" emf
Next Topic:EMF genmodel properties documentation??
Goto Forum:
  


Current Time: Mon Apr 21 08:52:26 EDT 2014

Powered by FUDForum. Page generated in 0.01645 seconds