tomcat-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Adrian Gonzalez <adr_gonza...@yahoo.fr>
Subject Re : Re : Performance issue in simpleTags and tagfiles
Date Wed, 07 Sep 2011 20:17:17 GMT
One more test at home (usin Sun Jdk 1.6.23).

Results are in nanosecond and correspond to generated JSP pages (loop with 1000 iteration
is replaced with 1000 tags).

With this test I have simpleTag = 10 * classicTag.
An this ratio is the same in 7.0.21 or in trunk.

trunk
simple = 13461836 [http://localhost:8080/tagfilesperf/generated/edit-simpletag.jsp]
classic=    1491251 [http://localhost:8080/tagfilesperf/generated/edit-classictag.jsp]
tagfile=   13660309 [http://localhost:8080/tagfilesperf/generated/edit-tagfile.jsp]

7.0.21
simple = 12590453
classic =   1388338
tagfile =  15088080

More tomorrow, sorry to spam...

----- Mail original -----
De : Adrian Gonzalez <adr_gonzalez@yahoo.fr>
À : Tomcat Users List <users@tomcat.apache.org>
Cc : 
Envoyé le : Mercredi 7 Septembre 2011 21h36
Objet : Re : Re : Performance issue in simpleTags and tagfiles

Sorry, no real difference.

Even worse, I'm on my home computer and here I don't see anymore the *5 difference between
simple tags and classic tags.

I'll continue these tests tomorrow at work (perhaps there's some file scanning / reading in
action - and since I'm with AV software...).

 - Tests results - 

Downloaded tomcat trunk sources from http://svn.apache.org/repos/asf/tomcat/trunk/

Used sample project here http://forum.springsource.org/attachment.php?attachmentid=4227&d=1315397882
from this issue http://forum.springsource.org/showthread.php?114210-Roo-tagfiles-performance-issues.


There are 4 different jsps in the test (one using tagfile, one using classic tag, etc...).
Each one does 1000 loops and prints overall elapsed time.
//timer.start
<c:forEach begin="1" end="10000">
  <appfield:inputS label="personne.nom" field="personne.nom" id="personne.nom" value="THOMAS"/>
</c:forEach>
//sysout timer.elapsed

Tomcat trunk
 * tagfile = 730
 * classictag = 730
 * simpletag = 740
 * notag = 730

Tomcat 7.0.21
 * tagfile = 700
 * classictag = 720
 * simpletag = 700
 * notag = 750



----- Mail original -----
De : Adrian Gonzalez <adr_gonzalez@yahoo.fr>
À : Tomcat Users List <users@tomcat.apache.org>
Cc : 
Envoyé le : Mercredi 7 Septembre 2011 20h43
Objet : Re : Re : Performance issue in simpleTags and tagfiles

Ouch, that was fast ! Thanks very much !

Are you willing to build Tomcat 7.0.x from source and see if the new cache works for you?

-> Of course ! Donno if I'm doing it now or tomorrow morning though - I'm a bit ill at
the moment ;(
sources are in trunk, right ? (http://svn.apache.org/repos/asf/tomcat/trunk/) 


----- Mail original -----
De : Mark Thomas <markt@apache.org>
À : Tomcat Users List <users@tomcat.apache.org>
Cc : 
Envoyé le : Mercredi 7 Septembre 2011 20h36
Objet : Re: Re : Performance issue in simpleTags and tagfiles

On 07/09/2011 18:37, Adrian Gonzalez wrote:
> Thanks for taking your time to look this issue !

Adding a cache for annotation scanning (that was taking most of the
time) was relatively simple. Are you willing to build Tomcat 7.0.x from
source and see if the new cache works for you?

Mark

> ----- Mail original -----
> De : Mark Thomas <markt@apache.org>
> À : Tomcat Users List <users@tomcat.apache.org>
> Cc : 
> Envoyé le : Mercredi 7 Septembre 2011 19h13
> Objet : Re: Performance issue in simpleTags and tagfiles
> 
> On 07/09/2011 15:57, Remy Maucherat wrote:
>> On Wed, Sep 7, 2011 at 2:23 PM, Adrian Gonzalez <adr_gonzalez@yahoo.fr> wrote:
>>> Hello,
>>>
>>> I've noticed a performance difference between classic Tags and simple Tags in
Tomcat 7.0.21 (also tested it on 7.0.6 with the same results).
>>>
>>> Simple tags or tagfiles execution is at 5 times superior to classic tag execution.
>>
>> Simple tags (and derivatives) are easier to write, but are not
>> poolable. So regular tags are best if you're going to use them a lot.
> 
> +1
> 
> A look with a profiler shows that a fair amount of time is spend in the
> InstanceManager with the non-poolable tags. We could probably speed that
> up with some caching.
> 
> I'll take a look to see if the benefit is worth the complexity.
> 
> Mark
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
> For additional commands, e-mail: users-help@tomcat.apache.org
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
> For additional commands, e-mail: users-help@tomcat.apache.org
> 


---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
For additional commands, e-mail: users-help@tomcat.apache.org

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
For additional commands, e-mail: users-help@tomcat.apache.org  

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
For additional commands, e-mail: users-help@tomcat.apache.org

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
For additional commands, e-mail: users-help@tomcat.apache.org


Mime
View raw message