|
| Re: The time of Eclipse loading [message #250344 is a reply to message #249785] |
Tue, 08 June 2004 07:17   |
Eclipse User |
|
|
|
Originally posted by: dcolombo.etad.com
Just start Eclipse with the -F option (fast load)... ;-))
Seriously:
When you first start a program (eclipse or any other) it has to be loaded
from your hard disk to the main memory before the processor can execute it.
If you restart the same program again while it is still residing in the main
memory this step is skipped and the program starts faster.
To load your program faster you have to find a way to load it as a
background job (perhaps after booting). But in my opinion it would be easier
to upgrade your computer....
Wish you a happy day
Diana
"Alex Chapiro" <achapiro@qnx.com> schrieb im Newsbeitrag
news:c9qpog$g81$1@eclipse.org...
> I was wondering why there so big difference in time of Eclipse loading
> the first time after computer re-booting and next loadings? For instance
> I tested it for medium size workspace:
>
> Eclipse 2.1.2:
> 1-st load - 40c
> 2-nd load - 20c
>
> Eclipse 3.0RC1:
> 1-st load - 75c
> 2-nd load - 35c
>
> I performed tests on W2K machine with 512Mb memory, -Xmx400m.
>
> Why so big difference and is there any way to improve the first load case?
>
>
> Thanks.
|
|
|
|
| Re: The time of Eclipse loading [message #250507 is a reply to message #249785] |
Tue, 08 June 2004 14:24 |
| Eclipse User |
|
|
|
Originally posted by: thomas_maeder.ch.ibm.com
I can start eclipse with a reasonably large workspace in about 10-15
secs (2.6 GHz Pentium 4, -Xmx 256M). Things to try would be:
1) make sure neither eclipse nor the workspace are loaded from a network
drive
2) use a slightly smaller -Xmx value (you might be taking away too much
memory from the system). I've never needed more than 256M.
3) How fast does an empty workspace come up? What's different (do you
have many perspectives open, many editors, etc)?
Close various perspectives/editors/whatevers to find out who's
responsible. If you find a culprit, file a bug report.
It would be interesting to know how your workspace is set up. How many
classes, how big are the files, how is the classpath set up, etc.
Thomas
Alex Chapiro wrote:
> I was wondering why there so big difference in time of Eclipse loading
> the first time after computer re-booting and next loadings? For instance
> I tested it for medium size workspace:
>
> Eclipse 2.1.2:
> 1-st load - 40c
> 2-nd load - 20c
>
> Eclipse 3.0RC1:
> 1-st load - 75c
> 2-nd load - 35c
>
> I performed tests on W2K machine with 512Mb memory, -Xmx400m.
>
> Why so big difference and is there any way to improve the first load case?
>
>
> Thanks.
|
|
|
| Re: The time of Eclipse loading [message #250516 is a reply to message #250478] |
Tue, 08 June 2004 14:29 |
| Eclipse User |
|
|
|
Originally posted by: user.domain.invalid
Performance is one of the things they are currently addressing for the
final 3.0. Remember that you are using beta s/w. RC1 has been a big
improvement for me in overall performance in general. Still not there,
but let them finish their work before worrying.
Alex Chapiro wrote:
> No, there is something else. While experimenting, I tried to push it out
> of RAM just starting a lot of time-consuming applications and checking
> physical memory consumption through PM. It didn't change the situation.
>
> Anyway, I agree that my complain sounds ridiculous. Unfortunately, one
> of our valued customers is not very happy because their dual Pentium-4
> 3Ghz machines with 2 Gb RAM and ultra-fast SCSI HDs spend every morning
> more than a minute for initial Eclipse loading. :-( That's why I'm
> trying to find any reasonable explanation (or one that at least sounds
> reasonable) for such a stupid question.
>
> Thanks.
>
> Diana Colombo wrote:
>
>> Just start Eclipse with the -F option (fast load)... ;-))
>>
>>
>> Seriously:
>> When you first start a program (eclipse or any other) it has to be loaded
>> from your hard disk to the main memory before the processor can
>> execute it.
>> If you restart the same program again while it is still residing in
>> the main
>> memory this step is skipped and the program starts faster.
>>
>> To load your program faster you have to find a way to load it as a
>> background job (perhaps after booting). But in my opinion it would be
>> easier
>> to upgrade your computer....
>>
>> Wish you a happy day
>>
>> Diana
>>
>>
>>
>> "Alex Chapiro" <achapiro@qnx.com> schrieb im Newsbeitrag
>> news:c9qpog$g81$1@eclipse.org...
>>
>>> I was wondering why there so big difference in time of Eclipse loading
>>> the first time after computer re-booting and next loadings? For instance
>>> I tested it for medium size workspace:
>>>
>>> Eclipse 2.1.2:
>>> 1-st load - 40c
>>> 2-nd load - 20c
>>>
>>> Eclipse 3.0RC1:
>>> 1-st load - 75c
>>> 2-nd load - 35c
>>>
>>> I performed tests on W2K machine with 512Mb memory, -Xmx400m.
>>>
>>> Why so big difference and is there any way to improve the first load
>>> case?
>>>
>>>
>>> Thanks.
|
|
|
| Re: The time of Eclipse loading [message #250602 is a reply to message #250507] |
Tue, 08 June 2004 17:52 |
 Alex Chapiro
Messages: 82
Registered: July 2009 |
Member |
|
|
You are right. In my case the cause of slow startup is the number of
opened files. It let me to assume that in my customer case the cause is
not in Eclipse kernel, but in background CDT activity (maybe, parsing or
indexing).
Below is a brief description about customer's configuration (Win XP).
"There are a lot of projects (this is C++) in the workspace. There are
only 2 or 3 files open. After a reboot, it takes about 20-30 seconds
for the splashscreen to come up. Then about another 30 seconds for the
IDE to make its presence known. After Eclipse has started up at least
once, it will only take about 30 seconds to start up."
Still not clear why there is such a big difference between 1-st and 2-nd
start-up. The same effect I observe in my Java case as well (Win 2000).
Thanks
Thomas Mäder wrote:
> I can start eclipse with a reasonably large workspace in about 10-15
> secs (2.6 GHz Pentium 4, -Xmx 256M). Things to try would be:
>
> 1) make sure neither eclipse nor the workspace are loaded from a network
> drive
> 2) use a slightly smaller -Xmx value (you might be taking away too much
> memory from the system). I've never needed more than 256M.
> 3) How fast does an empty workspace come up? What's different (do you
> have many perspectives open, many editors, etc)?
> Close various perspectives/editors/whatevers to find out who's
> responsible. If you find a culprit, file a bug report.
>
> It would be interesting to know how your workspace is set up. How many
> classes, how big are the files, how is the classpath set up, etc.
>
> Thomas
>
> Alex Chapiro wrote:
>
>> I was wondering why there so big difference in time of Eclipse loading
>> the first time after computer re-booting and next loadings? For
>> instance I tested it for medium size workspace:
>>
>> Eclipse 2.1.2:
>> 1-st load - 40c
>> 2-nd load - 20c
>>
>> Eclipse 3.0RC1:
>> 1-st load - 75c
>> 2-nd load - 35c
>>
>> I performed tests on W2K machine with 512Mb memory, -Xmx400m.
>>
>> Why so big difference and is there any way to improve the first load
>> case?
>>
>>
>> Thanks.
|
|
|
| Re: The time of Eclipse loading [message #250657 is a reply to message #250602] |
Tue, 08 June 2004 21:31 |
| Eclipse User |
|
|
|
Originally posted by: ed.burnette.REMOVE.THIS.sas.com
You can try the -vmargs -Xms128m option ('s' means minimum startup size);
this will reduce some heap growing when you start up. Also I'll bet that
400m is too big, try playing around with that number to find an optimal one
for this application, for example, -Xmx256m. JDK 1.5 is supposed to make
this more automatic but we'll see.
Also, I recommend you look at Dan Kehn's articles on tuning Eclipse's
startup performance if you haven't already:
http://www-106.ibm.com/developerworks/opensource/library/os- ecspy1/
http://www-106.ibm.com/developerworks/opensource/library/os- ecspy2/
Report any CDT performance issues over on their newsgroup or in bugzilla,
and let us know what you find out so we can learn from your experiences.
--
Ed
"Alex Chapiro" <achapiro@qnx.com> wrote in message
news:ca4tl3$s7j$1@eclipse.org...
> You are right. In my case the cause of slow startup is the number of
> opened files. It let me to assume that in my customer case the cause is
> not in Eclipse kernel, but in background CDT activity (maybe, parsing or
> indexing).
|
|
|
|
Powered by
FUDForum. Page generated in 0.27608 seconds
] 
Search
Help
Register
Login
Home
