Eclipse Community Forums: EMF "Technology" (Ecore Tools, EMFatic, etc) » [EMFStore] Performance of large change packages

Help

Home

Home » Modeling » EMF "Technology" (Ecore Tools, EMFatic, etc) » [EMFStore] Performance of large change packages

Show: Today's Messages :: Show Polls :: Message Navigator

[EMFStore] Performance of large change packages [message #1449901]

Tue, 21 October 2014 23:44

Scott Dybiec

Messages: 148
Registered: July 2009

Senior Member

I noticed that the CPU was pegged when applying large EMFStore change
packages operations (updating), and since my application is somewhat
performance-sensitive, I decide to profile the execution of the EMFStore
code that does this work.

Applying 100 or so changes takes 25 seconds my quad core machine. The
attached output breaks down where the CPU time was spent by method call.

Hand-written EMF code that performs that adds the same elements to the
same lists takes less than a second. I expected that EMFStore would be
2x or 3x slower than the equivalent EMF hand-written code, but not two
orders of magnitude slower.

Is this expected?

Attachment: emfstore-profiled-performance-table.gif
(Size: 112.38KB, Downloaded 145 times)

Report message to a moderator

Re: [EMFStore] Performance of large change packages [message #1451329 is a reply to message #1449901]

Thu, 23 October 2014 17:17

Maximilian Koegel

Messages: 253
Registered: July 2009

Senior Member

Hi Scott,

no this is definitely not expected. Can you provide an isolated test
case maybe?

Cheers,
Maximilian

Am 22.10.2014 01:44, schrieb scott@xxxxxxxx:
> I noticed that the CPU was pegged when applying large EMFStore change
> packages operations (updating), and since my application is somewhat
> performance-sensitive, I decide to profile the execution of the EMFStore
> code that does this work.
>
> Applying 100 or so changes takes 25 seconds my quad core machine. The
> attached output breaks down where the CPU time was spent by method call.
>
> Hand-written EMF code that performs that adds the same elements to the
> same lists takes less than a second. I expected that EMFStore would be
> 2x or 3x slower than the equivalent EMF hand-written code, but not two
> orders of magnitude slower.
>
> Is this expected?

--
Maximilian Kögel

Get Professional Eclipse Support: http://eclipsesource.com/munich

Report message to a moderator

Re: [EMFStore] Performance of large change packages [message #1454063 is a reply to message #1451329]

Mon, 27 October 2014 23:02

Scott Dybiec

Messages: 148
Registered: July 2009

Senior Member

Max,

I'll look into creating a test case or getting you more data on the problem.

$cott

On 10/23/2014 1:17 PM, Maximilian Koegel wrote:
> Hi Scott,
>
> no this is definitely not expected. Can you provide an isolated test
> case maybe?
>
> Cheers,
> Maximilian
>
> Am 22.10.2014 01:44, schrieb scott@xxxxxxxx:
>> I noticed that the CPU was pegged when applying large EMFStore change
>> packages operations (updating), and since my application is somewhat
>> performance-sensitive, I decide to profile the execution of the EMFStore
>> code that does this work.
>>
>> Applying 100 or so changes takes 25 seconds my quad core machine. The
>> attached output breaks down where the CPU time was spent by method call.
>>
>> Hand-written EMF code that performs that adds the same elements to the
>> same lists takes less than a second. I expected that EMFStore would be
>> 2x or 3x slower than the equivalent EMF hand-written code, but not two
>> orders of magnitude slower.
>>
>> Is this expected?
>
>

Report message to a moderator

Previous Topic:	[EMFStore] [ECP] Add user via UI gives exception in 1.4
Next Topic:	[EMFForms] EKeys on EReference validation constrain only fired on initial AddCommand

Goto Forum:

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

]

Current Time: Tue Apr 23 12:46:10 GMT 2024

.:: Contact :: Home ::.

Breadcrumbs

Sign up to our Newsletter