Mailing-List: contact user-help@ant.apache.org; run by ezmlm
Precedence: bulk
Reply-To: "Ant Users List" <user@ant.apache.org>
Received-SPF: pass (asf.osuosl.org: local policy)
Content-class: urn:content-classes:message
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
Subject: RE: Performance difference between 1.6.1 and 1.6.4 (in eclipse)
Date: Mon, 27 Jun 2005 16:41:56 +0200
Message-ID: 
 <761E4BE8F2661749BF0922125848D036126082@SEEDEBREXCHCL.seeburger.de>
Thread-Topic: [[**SPAM**]] - RE: Performance difference between 1.6.1 and
 1.6.4 (in eclipse) - Bayesian Filter detected spam
thread-index: AcV7JKfd40fhGoYCT8qHEF7VPHVvrAAAINSQ
From: =?iso-8859-1?Q?Poll=E4hne=2E_Ullrich?= <U.Pollaehne@seeburger.de>
To: "Ant Users List" <user@ant.apache.org>

Hi Dominique,

> -----Original Message-----
> From: Dominique Devienne [mailto:ddevienne@gmail.com]
> Sent: Monday, June 27, 2005 4:29 PM
> To: 'Ant Users List'
> Cc: 'Ant Developers List'
> Subject: [[**SPAM**]] - RE: Performance difference between 1.6.1 and
> 1.6.4 (in eclipse) - Bayesian Filter detected spam
>=20
>=20
> Ullrich was kind enough to provide me data gathered by
> my own performance listener, which shows timings for
> tasks as well as targets:
>=20
> Timing by tasks in Ant 1.6.2:        in Ant 1.6.5
> ----------- ------ ------ ---------- ----------- ------
> 19.578,0 ms 175,2%    24x antcall    62.235,0 ms 189,0%
>  5.638,0 ms  50,5%   434x see.get     6.308,0 ms  19,2%
>  1.957,0 ms  17,5%    25x import     14.238,0 ms  43,2%
>  1.269,0 ms  11,4% 15131x property   11.775,0 ms  35,8%
>    563,0 ms   5,0%    43x delete        608,0 ms   1,8%
>     93,0 ms   0,8%   101x dirname       221,0 ms   0,7%
>     47,0 ms   0,4%     1x tstamp         47,0 ms   0,1%
>     30,0 ms   0,3%    25x presetdef      16,0 ms   0,0%
>     15,0 ms   0,1%   430x basename      453,0 ms   1,4%
>      0,0 ms   0,0%    78x mkdir          64,0 ms   0,2%
>=20
> 11.172,0 ms 100,0%     1x TOTAL      32.922,0 ms 100,0%
>=20
> So this data seems to indicate a significant slow down of
> <import> and <property> more than <antcall>. The combined
> slow down of these 2 tasks amounts to 23.2 seconds which
> is pretty much the whole slow down. <antcall>'s own perf
> data is masked by that of the nested tasks, but most
> likely peanuts.
>=20
> Does anybody have any clue to why these two tasks got so
> much slower in between patch releases? The slow down of
> <import> is more dramatic than <property>, since for
> 25 calls 'only'.
>=20
> 15K <property> calls might be unusual a bit, but still.
>=20
> Thanks, --DD
>=20
> PS: the 9,987.65 are for a German locale, reverse of US convention.
> PPS: Percentages go above 100% because consider nested tasks too.
> PPPS: the task execution count was identical in 1.6.5 (not shown)


And always keep in mind that the import in my case imports a script that
contains 510 <property> definitions (I had not counted them until now).
Since the <antcall> creates a new project which <include>s anew this
means that 12750 of the 15131 <property> calls are from the imported =
script.

Looking at the source of the property task of 1.6.2 and 1.6.5 shows no
difference between the two. So the reason for the slowdown seems to be
more deeply hidden.

Ullrich.


---------------------------------------------------------------------
To unsubscribe, e-mail: user-unsubscribe@ant.apache.org
For additional commands, e-mail: user-help@ant.apache.org