You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@flex.apache.org by Frédéric THOMAS <we...@hotmail.com> on 2012/10/05 14:30:25 UTC
Performence issue
Hi,
Even after I read that
http://blogs.adobe.com/aharui/2011/04/migratory-foul-performance-problems-mi
grating-from-flex-3-x-to-flex-4-x.html I thought that being in flex
compatibility 3.0.0 would be enough to have the same performances than in
Flex 3, that's not the case at all, does someone can explain me why and if I
can do something to improve them ?
Frédéric THOMAS
RE: Performence issue
Posted by Frédéric THOMAS <we...@hotmail.com>.
Ok, thank's
-----Original Message-----
From: Carol Frampton [mailto:cframpto@adobe.com]
Sent: Monday, October 15, 2012 3:55 PM
To: flex-users@incubator.apache.org
Subject: Re: Performence issue
We are doing SDK work in
https://svn.apache.org/repos/asf/incubator/flex/sdk/branches/develop.
Carol
On 10/15/12 4 :46AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
>I have to base my patch on develop or trunk ?
>
>-----Original Message-----
>From: Frédéric THOMAS [mailto:webdoublefx@hotmail.com]
>Sent: Friday, October 12, 2012 7:05 PM
>To: flex-users@incubator.apache.org
>Subject: Re: Performence issue
>
>Ok, I'll do that next week, thank's for the tip.
>
>-----Message d'origine-----
>From: Alex Harui
>Sent: Friday, October 12, 2012 6:16 PM
>To: flex-users@incubator.apache.org
>Subject: Re: Performence issue
>
>Please file a bug (and a Patch) for that.
>
>BTW, the FlashBuilder profiler can profile SWFs that are built from Maven.
>Just go to the profiler screens and choose an external launch
>configuration.
>
>
>On 10/12/12 5:46 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
>
>> Unfortunally, I wasn't able to do that in flashbuilder, I've got to
>> many mavenized projects already under IDEA, too much work to
>> reconfigure them with flashbuilder.
>>
>> Anyway, I've been able to deterninate the problem, the charts in flex
>> 3 had a static initializer for the styles, it's not the case in flex
>> 4, at each addChild the styles are re-computed which seems to be a
>> wrong aproach to me, It should have one static initializer per
>> moduleFactory.
>>
>> Example on CartesianChart :
>>
>> instead of :
>> private var _moduleFactoryInitialized:Boolean = false;
>>
>> should be :
>> private static var _moduleFactoryInitialized:Dictionary= new
>> Dictionary(true);
>>
>>
>> instead of :
>>
>> override public function set
>> moduleFactory(factory:IFlexModuleFactory):void
>> {
>> super.moduleFactory = factory;
>>
>> if (_moduleFactoryInitialized)
>> return;
>>
>> _moduleFactoryInitialized = true;
>>
>>
>> // our style settings
>> initStyles();
>> styleManager.registerInheritingStyle("axisTitleStyleName");
>> }
>>
>> should be :
>>
>> override public function set
>> moduleFactory(factory:IFlexModuleFactory):void
>> {
>> super.moduleFactory = factory;
>>
>> if (_moduleFactoryInitialized[factory])
>> return;
>>
>> _moduleFactoryInitialized[factory] = true;
>>
>>
>> // our style settings
>> initStyles();
>> styleManager.registerInheritingStyle("axisTitleStyleName");
>> }
>>
>> Re-compiling the charts with this modification I've been able to get
>> back my 35% lost - on some heavy screens where I had a lot of charts
>> (ItemRenderer containing charts) - migrating from flex 3 to flex 4.
>>
>>
>> At the moment anyway, I'm not satisfy yet, I've still got a lot of
>> flickering I hadn't got in Flex 3 and at the moment, I don't
>> understand why, it seems to be relative to styles again, the
>> initProtoChain is called too often, the component are inexplicably
>> re-layed out.
>>
>> Fred.
>>
>>
>> -----Message d'origine-----
>> From: Alex Harui
>> Sent: Monday, October 08, 2012 5:03 PM
>> To: flex-users@incubator.apache.org
>> Subject: Re: Performence issue
>>
>> Get a copy of Flash Builder and use the profiler that comes with it.
>> It will break down the most heavily used functions.
>>
>>
>> On 10/8/12 5:35 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
>>
>>> Sorry for the firsts numbers, I did a "dir *.as" and a "dir *.mxml"
>>> /s but the sdk was in the sources and others projects too, well,
>>> finally, it's only
>>> 1928 .as and 447 .mxml, not the same numbers but anyway, the rest
>>> are good numbers.
>>>
>>> What do you mean by "What does the profiler say is going on?", we're
>>> benchmarking every dashboard we load (our functional parts), we
>>> haven't got any particular profiler.
>>>
>>> In Flex 3, we still have some of the invalidatedFlags set at true
>>>after the update complete is fired, so to measure the load time of
>>>our components, we check for those flags and report the loaded status
>>>to the container.
>>> protected function
>>>monitorUIComponentStatus(component:UIComponent):void {
>>> if (!m_pagelet.started) {
>>> return;
>>> }
>>> if (!component.mx_internal::invalidateSizeFlag &&
>>>!component.mx_internal::invalidatePropertiesFlag &&
>>> !component.mx_internal::invalidateDisplayListFlag) {
>>> updateSpinner();
>>> markAsLoaded();
>>> } else {
>>> // Not ready yet, try later
>>> component.callLater(monitorUIComponentStatus,
>>> [component]);
>>> }
>>> }
>>>
>>> In Flex 4, the same code gives about 35% of time in more but I
>>> noticed I received 2 update complete from the container if it
>>> contains DataGrid with a lot of charts in the ItemRenderer, one
>>> about 10% later, the other really long time after.
>>>
>>> The logs:
>>> 2012-10-08 13:39:34,699 [829134827@qtp-1377590487-103] INFO
>>> fx0000000b.dashboardsBundle.dashboardViewerMediator - Dashboard
>>> HVP.OperationalSupervisor.Intraday.AboveProfile has finished loading.
>>> Loading time: 11681 ms
>>> 2012-10-08 13:39:38,268 [497865763@qtp-1377590487-101] INFO
>>> fx0000000b.dashboardsBundle.dashboardsUIModule - Dashboard
>>> application.ShellUI37.innerScrollBarsContent.modulesContainer.viewst
>>> a ck.Canv as142.dashboardViewer has finished loading with
>>> updateComplete. Loading
>>> time: 15578 ms
>>> 2012-10-08 13:40:54,689 [823563825@qtp-1377590487-95] INFO
>>> fx0000000b.dashboardsBundle.dashboardsUIModule - Dashboard
>>> application.ShellUI37.innerScrollBarsContent.modulesContainer.viewst
>>> a ck.Canv as142.dashboardViewer has finished loading with
>>> updateComplete. Loading
>>> time: 91995 ms
>>>
>>> dashboardsUIModule is the container of the AboveProfile dashboard,
>>> it seems AboveProfile is processed twice, dashboardsUIModule
>>> contains AboveProfile and a toolbar.
>>>
>>> I noticed the process of updating the display list happened more
>>> than
>>> 1 time when I¹ve got lot of chart into the dashboard, which is the
>>> case of AboveProfile and LostAndStuck.
>>>
>>> The loading times: flex 4 / flex 3 for the dashboards containing a
>>> lot of charts AboveProfile : 10538/ 7785
>>> LostAndStuck: 4860 / 2704
>>>
>>> Frédéric THOMAS
>>>
>>> -----Original Message-----
>>> From: Alex Harui [mailto:aharui@adobe.com]
>>> Sent: Friday, October 05, 2012 6:42 PM
>>> To: flex-users@incubator.apache.org
>>> Subject: Re: Performence issue
>>>
>>> What does the profiler say is going on?
>>>
>>>
>>> On 10/5/12 9:10 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
>>>
>>>> Hi Alex,
>>>>
>>>> - The application is 6778 .as and 10315 .mxml (some of the .as are
>>>> visual components and optimized as noted in your document)
>>>> - The migration is from 3.6 to 4.8
>>>> - I migrated from mxml 2006 to 2009 as much as I could, only
>>>> modules, components which contains States are still in mxml 2006
>>>> - Almost all the styles are in .css
>>>> - Each functional part of the application is benchmarked.
>>>> - I ran both version (flex 3 / 4) on the same computer, with the
>>>> same data loaded, at the same time and with FP 11.4
>>>>
>>>> The Flex 4 version was in compatibility mode because otherwise,
>>>> with spark, I found much more gap between both versions.
>>>>
>>>> The result is:
>>>> I found a gap of 35% between flex 3 / 4 ( Flex 3 compatibility
>>>> mode) for the richest screens, 20% more or less for the others.
>>>>
>>>>
>>>> -----Original Message-----
>>>> From: Alex Harui [mailto:aharui@adobe.com]
>>>> Sent: Friday, October 05, 2012 5:41 PM
>>>> To: flex-users@incubator.apache.org
>>>> Subject: Re: Performence issue
>>>>
>>>> Are you doing dynamic add of display objects? If so, the latter
>>>> part of that article might pertain to you.
>>>>
>>>> Otherwise, please provide more information on your scenario and
>>>> what changed going from 3 to 4. Which version of 4?
>>>>
>>>>
>>>> On 10/5/12 5:30 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
>>>>
>>>>> Hi,
>>>>>
>>>>> Even after I read that
>>>>>
>>>>>http://blogs.adobe.com/aharui/2011/04/migratory-foul-performance-pr
>>>>> ob l ems-mi grating-from-flex-3-x-to-flex-4-x.html I thought that
>>>>>being in flex compatibility 3.0.0 would be enough to have the same
>>>>>performances than in Flex 3, that's not the case at all, does
>>>>>someone can explain me why and if I can do something to improve
>>>>>them ?
>>>>>
>>>>> Frédéric THOMAS
>>>>>
>>>>
>>>> --
>>>> Alex Harui
>>>> Flex SDK Team
>>>> Adobe Systems, Inc.
>>>> http://blogs.adobe.com/aharui
>>>>
>>>>
>>>
>>> --
>>> Alex Harui
>>> Flex SDK Team
>>> Adobe Systems, Inc.
>>> http://blogs.adobe.com/aharui
>>>
>>>
>
>--
>Alex Harui
>Flex SDK Team
>Adobe Systems, Inc.
>http://blogs.adobe.com/aharui
>
>
Re: Performence issue
Posted by Carol Frampton <cf...@adobe.com>.
We are doing SDK work in
https://svn.apache.org/repos/asf/incubator/flex/sdk/branches/develop.
Carol
On 10/15/12 4 :46AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
>I have to base my patch on develop or trunk ?
>
>-----Original Message-----
>From: Frédéric THOMAS [mailto:webdoublefx@hotmail.com]
>Sent: Friday, October 12, 2012 7:05 PM
>To: flex-users@incubator.apache.org
>Subject: Re: Performence issue
>
>Ok, I'll do that next week, thank's for the tip.
>
>-----Message d'origine-----
>From: Alex Harui
>Sent: Friday, October 12, 2012 6:16 PM
>To: flex-users@incubator.apache.org
>Subject: Re: Performence issue
>
>Please file a bug (and a Patch) for that.
>
>BTW, the FlashBuilder profiler can profile SWFs that are built from Maven.
>Just go to the profiler screens and choose an external launch
>configuration.
>
>
>On 10/12/12 5:46 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
>
>> Unfortunally, I wasn't able to do that in flashbuilder, I've got to
>> many mavenized projects already under IDEA, too much work to
>> reconfigure them with flashbuilder.
>>
>> Anyway, I've been able to deterninate the problem, the charts in flex
>> 3 had a static initializer for the styles, it's not the case in flex
>> 4, at each addChild the styles are re-computed which seems to be a
>> wrong aproach to me, It should have one static initializer per
>> moduleFactory.
>>
>> Example on CartesianChart :
>>
>> instead of :
>> private var _moduleFactoryInitialized:Boolean = false;
>>
>> should be :
>> private static var _moduleFactoryInitialized:Dictionary= new
>> Dictionary(true);
>>
>>
>> instead of :
>>
>> override public function set
>> moduleFactory(factory:IFlexModuleFactory):void
>> {
>> super.moduleFactory = factory;
>>
>> if (_moduleFactoryInitialized)
>> return;
>>
>> _moduleFactoryInitialized = true;
>>
>>
>> // our style settings
>> initStyles();
>> styleManager.registerInheritingStyle("axisTitleStyleName");
>> }
>>
>> should be :
>>
>> override public function set
>> moduleFactory(factory:IFlexModuleFactory):void
>> {
>> super.moduleFactory = factory;
>>
>> if (_moduleFactoryInitialized[factory])
>> return;
>>
>> _moduleFactoryInitialized[factory] = true;
>>
>>
>> // our style settings
>> initStyles();
>> styleManager.registerInheritingStyle("axisTitleStyleName");
>> }
>>
>> Re-compiling the charts with this modification I've been able to get
>> back my 35% lost - on some heavy screens where I had a lot of charts
>> (ItemRenderer containing charts) - migrating from flex 3 to flex 4.
>>
>>
>> At the moment anyway, I'm not satisfy yet, I've still got a lot of
>> flickering I hadn't got in Flex 3 and at the moment, I don't
>> understand why, it seems to be relative to styles again, the
>> initProtoChain is called too often, the component are inexplicably
>> re-layed out.
>>
>> Fred.
>>
>>
>> -----Message d'origine-----
>> From: Alex Harui
>> Sent: Monday, October 08, 2012 5:03 PM
>> To: flex-users@incubator.apache.org
>> Subject: Re: Performence issue
>>
>> Get a copy of Flash Builder and use the profiler that comes with it.
>> It will break down the most heavily used functions.
>>
>>
>> On 10/8/12 5:35 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
>>
>>> Sorry for the firsts numbers, I did a "dir *.as" and a "dir *.mxml"
>>> /s but the sdk was in the sources and others projects too, well,
>>> finally, it's only
>>> 1928 .as and 447 .mxml, not the same numbers but anyway, the rest are
>>> good numbers.
>>>
>>> What do you mean by "What does the profiler say is going on?", we're
>>> benchmarking every dashboard we load (our functional parts), we
>>> haven't got any particular profiler.
>>>
>>> In Flex 3, we still have some of the invalidatedFlags set at true
>>> after the update complete is fired, so to measure the load time of
>>> our components, we check for those flags and report the loaded status
>>> to the container.
>>> protected function
>>>monitorUIComponentStatus(component:UIComponent):void {
>>> if (!m_pagelet.started) {
>>> return;
>>> }
>>> if (!component.mx_internal::invalidateSizeFlag &&
>>> !component.mx_internal::invalidatePropertiesFlag &&
>>> !component.mx_internal::invalidateDisplayListFlag) {
>>> updateSpinner();
>>> markAsLoaded();
>>> } else {
>>> // Not ready yet, try later
>>> component.callLater(monitorUIComponentStatus,
>>> [component]);
>>> }
>>> }
>>>
>>> In Flex 4, the same code gives about 35% of time in more but I
>>> noticed I received 2 update complete from the container if it
>>> contains DataGrid with a lot of charts in the ItemRenderer, one about
>>> 10% later, the other really long time after.
>>>
>>> The logs:
>>> 2012-10-08 13:39:34,699 [829134827@qtp-1377590487-103] INFO
>>> fx0000000b.dashboardsBundle.dashboardViewerMediator - Dashboard
>>> HVP.OperationalSupervisor.Intraday.AboveProfile has finished loading.
>>> Loading time: 11681 ms
>>> 2012-10-08 13:39:38,268 [497865763@qtp-1377590487-101] INFO
>>> fx0000000b.dashboardsBundle.dashboardsUIModule - Dashboard
>>> application.ShellUI37.innerScrollBarsContent.modulesContainer.viewsta
>>> ck.Canv as142.dashboardViewer has finished loading with
>>> updateComplete. Loading
>>> time: 15578 ms
>>> 2012-10-08 13:40:54,689 [823563825@qtp-1377590487-95] INFO
>>> fx0000000b.dashboardsBundle.dashboardsUIModule - Dashboard
>>> application.ShellUI37.innerScrollBarsContent.modulesContainer.viewsta
>>> ck.Canv as142.dashboardViewer has finished loading with
>>> updateComplete. Loading
>>> time: 91995 ms
>>>
>>> dashboardsUIModule is the container of the AboveProfile dashboard, it
>>> seems AboveProfile is processed twice, dashboardsUIModule contains
>>> AboveProfile and a toolbar.
>>>
>>> I noticed the process of updating the display list happened more than
>>> 1 time when I¹ve got lot of chart into the dashboard, which is the
>>> case of AboveProfile and LostAndStuck.
>>>
>>> The loading times: flex 4 / flex 3 for the dashboards containing a
>>> lot of charts AboveProfile : 10538/ 7785
>>> LostAndStuck: 4860 / 2704
>>>
>>> Frédéric THOMAS
>>>
>>> -----Original Message-----
>>> From: Alex Harui [mailto:aharui@adobe.com]
>>> Sent: Friday, October 05, 2012 6:42 PM
>>> To: flex-users@incubator.apache.org
>>> Subject: Re: Performence issue
>>>
>>> What does the profiler say is going on?
>>>
>>>
>>> On 10/5/12 9:10 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
>>>
>>>> Hi Alex,
>>>>
>>>> - The application is 6778 .as and 10315 .mxml (some of the .as are
>>>> visual components and optimized as noted in your document)
>>>> - The migration is from 3.6 to 4.8
>>>> - I migrated from mxml 2006 to 2009 as much as I could, only
>>>> modules, components which contains States are still in mxml 2006
>>>> - Almost all the styles are in .css
>>>> - Each functional part of the application is benchmarked.
>>>> - I ran both version (flex 3 / 4) on the same computer, with the
>>>> same data loaded, at the same time and with FP 11.4
>>>>
>>>> The Flex 4 version was in compatibility mode because otherwise, with
>>>> spark, I found much more gap between both versions.
>>>>
>>>> The result is:
>>>> I found a gap of 35% between flex 3 / 4 ( Flex 3 compatibility mode)
>>>> for the richest screens, 20% more or less for the others.
>>>>
>>>>
>>>> -----Original Message-----
>>>> From: Alex Harui [mailto:aharui@adobe.com]
>>>> Sent: Friday, October 05, 2012 5:41 PM
>>>> To: flex-users@incubator.apache.org
>>>> Subject: Re: Performence issue
>>>>
>>>> Are you doing dynamic add of display objects? If so, the latter
>>>> part of that article might pertain to you.
>>>>
>>>> Otherwise, please provide more information on your scenario and what
>>>> changed going from 3 to 4. Which version of 4?
>>>>
>>>>
>>>> On 10/5/12 5:30 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
>>>>
>>>>> Hi,
>>>>>
>>>>> Even after I read that
>>>>> http://blogs.adobe.com/aharui/2011/04/migratory-foul-performance-pr
>>>>> ob l ems-mi grating-from-flex-3-x-to-flex-4-x.html I thought that
>>>>> being in flex compatibility 3.0.0 would be enough to have the same
>>>>> performances than in Flex 3, that's not the case at all, does
>>>>> someone can explain me why and if I can do something to improve them
>>>>>?
>>>>>
>>>>> Frédéric THOMAS
>>>>>
>>>>
>>>> --
>>>> Alex Harui
>>>> Flex SDK Team
>>>> Adobe Systems, Inc.
>>>> http://blogs.adobe.com/aharui
>>>>
>>>>
>>>
>>> --
>>> Alex Harui
>>> Flex SDK Team
>>> Adobe Systems, Inc.
>>> http://blogs.adobe.com/aharui
>>>
>>>
>
>--
>Alex Harui
>Flex SDK Team
>Adobe Systems, Inc.
>http://blogs.adobe.com/aharui
>
>
RE: Performence issue
Posted by Frédéric THOMAS <we...@hotmail.com>.
I have to base my patch on develop or trunk ?
-----Original Message-----
From: Frédéric THOMAS [mailto:webdoublefx@hotmail.com]
Sent: Friday, October 12, 2012 7:05 PM
To: flex-users@incubator.apache.org
Subject: Re: Performence issue
Ok, I'll do that next week, thank's for the tip.
-----Message d'origine-----
From: Alex Harui
Sent: Friday, October 12, 2012 6:16 PM
To: flex-users@incubator.apache.org
Subject: Re: Performence issue
Please file a bug (and a Patch) for that.
BTW, the FlashBuilder profiler can profile SWFs that are built from Maven.
Just go to the profiler screens and choose an external launch configuration.
On 10/12/12 5:46 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
> Unfortunally, I wasn't able to do that in flashbuilder, I've got to
> many mavenized projects already under IDEA, too much work to
> reconfigure them with flashbuilder.
>
> Anyway, I've been able to deterninate the problem, the charts in flex
> 3 had a static initializer for the styles, it's not the case in flex
> 4, at each addChild the styles are re-computed which seems to be a
> wrong aproach to me, It should have one static initializer per
> moduleFactory.
>
> Example on CartesianChart :
>
> instead of :
> private var _moduleFactoryInitialized:Boolean = false;
>
> should be :
> private static var _moduleFactoryInitialized:Dictionary= new
> Dictionary(true);
>
>
> instead of :
>
> override public function set
> moduleFactory(factory:IFlexModuleFactory):void
> {
> super.moduleFactory = factory;
>
> if (_moduleFactoryInitialized)
> return;
>
> _moduleFactoryInitialized = true;
>
>
> // our style settings
> initStyles();
> styleManager.registerInheritingStyle("axisTitleStyleName");
> }
>
> should be :
>
> override public function set
> moduleFactory(factory:IFlexModuleFactory):void
> {
> super.moduleFactory = factory;
>
> if (_moduleFactoryInitialized[factory])
> return;
>
> _moduleFactoryInitialized[factory] = true;
>
>
> // our style settings
> initStyles();
> styleManager.registerInheritingStyle("axisTitleStyleName");
> }
>
> Re-compiling the charts with this modification I've been able to get
> back my 35% lost - on some heavy screens where I had a lot of charts
> (ItemRenderer containing charts) - migrating from flex 3 to flex 4.
>
>
> At the moment anyway, I'm not satisfy yet, I've still got a lot of
> flickering I hadn't got in Flex 3 and at the moment, I don't
> understand why, it seems to be relative to styles again, the
> initProtoChain is called too often, the component are inexplicably
> re-layed out.
>
> Fred.
>
>
> -----Message d'origine-----
> From: Alex Harui
> Sent: Monday, October 08, 2012 5:03 PM
> To: flex-users@incubator.apache.org
> Subject: Re: Performence issue
>
> Get a copy of Flash Builder and use the profiler that comes with it.
> It will break down the most heavily used functions.
>
>
> On 10/8/12 5:35 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
>
>> Sorry for the firsts numbers, I did a "dir *.as" and a "dir *.mxml"
>> /s but the sdk was in the sources and others projects too, well,
>> finally, it's only
>> 1928 .as and 447 .mxml, not the same numbers but anyway, the rest are
>> good numbers.
>>
>> What do you mean by "What does the profiler say is going on?", we're
>> benchmarking every dashboard we load (our functional parts), we
>> haven't got any particular profiler.
>>
>> In Flex 3, we still have some of the invalidatedFlags set at true
>> after the update complete is fired, so to measure the load time of
>> our components, we check for those flags and report the loaded status
>> to the container.
>> protected function monitorUIComponentStatus(component:UIComponent):void {
>> if (!m_pagelet.started) {
>> return;
>> }
>> if (!component.mx_internal::invalidateSizeFlag &&
>> !component.mx_internal::invalidatePropertiesFlag &&
>> !component.mx_internal::invalidateDisplayListFlag) {
>> updateSpinner();
>> markAsLoaded();
>> } else {
>> // Not ready yet, try later
>> component.callLater(monitorUIComponentStatus,
>> [component]);
>> }
>> }
>>
>> In Flex 4, the same code gives about 35% of time in more but I
>> noticed I received 2 update complete from the container if it
>> contains DataGrid with a lot of charts in the ItemRenderer, one about
>> 10% later, the other really long time after.
>>
>> The logs:
>> 2012-10-08 13:39:34,699 [829134827@qtp-1377590487-103] INFO
>> fx0000000b.dashboardsBundle.dashboardViewerMediator - Dashboard
>> HVP.OperationalSupervisor.Intraday.AboveProfile has finished loading.
>> Loading time: 11681 ms
>> 2012-10-08 13:39:38,268 [497865763@qtp-1377590487-101] INFO
>> fx0000000b.dashboardsBundle.dashboardsUIModule - Dashboard
>> application.ShellUI37.innerScrollBarsContent.modulesContainer.viewsta
>> ck.Canv as142.dashboardViewer has finished loading with
>> updateComplete. Loading
>> time: 15578 ms
>> 2012-10-08 13:40:54,689 [823563825@qtp-1377590487-95] INFO
>> fx0000000b.dashboardsBundle.dashboardsUIModule - Dashboard
>> application.ShellUI37.innerScrollBarsContent.modulesContainer.viewsta
>> ck.Canv as142.dashboardViewer has finished loading with
>> updateComplete. Loading
>> time: 91995 ms
>>
>> dashboardsUIModule is the container of the AboveProfile dashboard, it
>> seems AboveProfile is processed twice, dashboardsUIModule contains
>> AboveProfile and a toolbar.
>>
>> I noticed the process of updating the display list happened more than
>> 1 time when I¹ve got lot of chart into the dashboard, which is the
>> case of AboveProfile and LostAndStuck.
>>
>> The loading times: flex 4 / flex 3 for the dashboards containing a
>> lot of charts AboveProfile : 10538/ 7785
>> LostAndStuck: 4860 / 2704
>>
>> Frédéric THOMAS
>>
>> -----Original Message-----
>> From: Alex Harui [mailto:aharui@adobe.com]
>> Sent: Friday, October 05, 2012 6:42 PM
>> To: flex-users@incubator.apache.org
>> Subject: Re: Performence issue
>>
>> What does the profiler say is going on?
>>
>>
>> On 10/5/12 9:10 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
>>
>>> Hi Alex,
>>>
>>> - The application is 6778 .as and 10315 .mxml (some of the .as are
>>> visual components and optimized as noted in your document)
>>> - The migration is from 3.6 to 4.8
>>> - I migrated from mxml 2006 to 2009 as much as I could, only
>>> modules, components which contains States are still in mxml 2006
>>> - Almost all the styles are in .css
>>> - Each functional part of the application is benchmarked.
>>> - I ran both version (flex 3 / 4) on the same computer, with the
>>> same data loaded, at the same time and with FP 11.4
>>>
>>> The Flex 4 version was in compatibility mode because otherwise, with
>>> spark, I found much more gap between both versions.
>>>
>>> The result is:
>>> I found a gap of 35% between flex 3 / 4 ( Flex 3 compatibility mode)
>>> for the richest screens, 20% more or less for the others.
>>>
>>>
>>> -----Original Message-----
>>> From: Alex Harui [mailto:aharui@adobe.com]
>>> Sent: Friday, October 05, 2012 5:41 PM
>>> To: flex-users@incubator.apache.org
>>> Subject: Re: Performence issue
>>>
>>> Are you doing dynamic add of display objects? If so, the latter
>>> part of that article might pertain to you.
>>>
>>> Otherwise, please provide more information on your scenario and what
>>> changed going from 3 to 4. Which version of 4?
>>>
>>>
>>> On 10/5/12 5:30 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
>>>
>>>> Hi,
>>>>
>>>> Even after I read that
>>>> http://blogs.adobe.com/aharui/2011/04/migratory-foul-performance-pr
>>>> ob l ems-mi grating-from-flex-3-x-to-flex-4-x.html I thought that
>>>> being in flex compatibility 3.0.0 would be enough to have the same
>>>> performances than in Flex 3, that's not the case at all, does
>>>> someone can explain me why and if I can do something to improve them ?
>>>>
>>>> Frédéric THOMAS
>>>>
>>>
>>> --
>>> Alex Harui
>>> Flex SDK Team
>>> Adobe Systems, Inc.
>>> http://blogs.adobe.com/aharui
>>>
>>>
>>
>> --
>> Alex Harui
>> Flex SDK Team
>> Adobe Systems, Inc.
>> http://blogs.adobe.com/aharui
>>
>>
--
Alex Harui
Flex SDK Team
Adobe Systems, Inc.
http://blogs.adobe.com/aharui
Re: Performence issue
Posted by Frédéric THOMAS <we...@hotmail.com>.
Ok, I'll do that next week, thank's for the tip.
-----Message d'origine-----
From: Alex Harui
Sent: Friday, October 12, 2012 6:16 PM
To: flex-users@incubator.apache.org
Subject: Re: Performence issue
Please file a bug (and a Patch) for that.
BTW, the FlashBuilder profiler can profile SWFs that are built from Maven.
Just go to the profiler screens and choose an external launch configuration.
On 10/12/12 5:46 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
> Unfortunally, I wasn't able to do that in flashbuilder, I've got to many
> mavenized projects already under IDEA, too much work to reconfigure them
> with flashbuilder.
>
> Anyway, I've been able to deterninate the problem, the charts in flex 3
> had
> a static initializer for the styles, it's not the case in flex 4, at each
> addChild the styles are re-computed which seems to be a wrong aproach to
> me,
> It should have one static initializer per moduleFactory.
>
> Example on CartesianChart :
>
> instead of :
> private var _moduleFactoryInitialized:Boolean = false;
>
> should be :
> private static var _moduleFactoryInitialized:Dictionary= new
> Dictionary(true);
>
>
> instead of :
>
> override public function set
> moduleFactory(factory:IFlexModuleFactory):void
> {
> super.moduleFactory = factory;
>
> if (_moduleFactoryInitialized)
> return;
>
> _moduleFactoryInitialized = true;
>
>
> // our style settings
> initStyles();
> styleManager.registerInheritingStyle("axisTitleStyleName");
> }
>
> should be :
>
> override public function set
> moduleFactory(factory:IFlexModuleFactory):void
> {
> super.moduleFactory = factory;
>
> if (_moduleFactoryInitialized[factory])
> return;
>
> _moduleFactoryInitialized[factory] = true;
>
>
> // our style settings
> initStyles();
> styleManager.registerInheritingStyle("axisTitleStyleName");
> }
>
> Re-compiling the charts with this modification I've been able to get back
> my
> 35% lost - on some heavy screens where I had a lot of charts (ItemRenderer
> containing charts) - migrating from flex 3 to flex 4.
>
>
> At the moment anyway, I'm not satisfy yet, I've still got a lot of
> flickering I hadn't got in Flex 3 and at the moment, I don't understand
> why,
> it seems to be relative to styles again, the initProtoChain is called too
> often, the component are inexplicably re-layed out.
>
> Fred.
>
>
> -----Message d'origine-----
> From: Alex Harui
> Sent: Monday, October 08, 2012 5:03 PM
> To: flex-users@incubator.apache.org
> Subject: Re: Performence issue
>
> Get a copy of Flash Builder and use the profiler that comes with it. It
> will break down the most heavily used functions.
>
>
> On 10/8/12 5:35 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
>
>> Sorry for the firsts numbers, I did a "dir *.as" and a "dir *.mxml" /s
>> but
>> the sdk was in the sources and others projects too, well, finally, it's
>> only
>> 1928 .as and 447 .mxml, not the same numbers but anyway, the rest are
>> good
>> numbers.
>>
>> What do you mean by "What does the profiler say is going on?", we're
>> benchmarking every dashboard we load (our functional parts), we haven't
>> got
>> any particular profiler.
>>
>> In Flex 3, we still have some of the invalidatedFlags set at true after
>> the
>> update complete is fired, so to measure the load time of our components,
>> we
>> check for those flags and report the loaded status to the container.
>> protected function monitorUIComponentStatus(component:UIComponent):void {
>> if (!m_pagelet.started) {
>> return;
>> }
>> if (!component.mx_internal::invalidateSizeFlag &&
>> !component.mx_internal::invalidatePropertiesFlag &&
>> !component.mx_internal::invalidateDisplayListFlag) {
>> updateSpinner();
>> markAsLoaded();
>> } else {
>> // Not ready yet, try later
>> component.callLater(monitorUIComponentStatus,
>> [component]);
>> }
>> }
>>
>> In Flex 4, the same code gives about 35% of time in more but I noticed I
>> received 2 update complete from the container if it contains DataGrid
>> with
>> a
>> lot of charts in the ItemRenderer, one about 10% later, the other really
>> long time after.
>>
>> The logs:
>> 2012-10-08 13:39:34,699 [829134827@qtp-1377590487-103] INFO
>> fx0000000b.dashboardsBundle.dashboardViewerMediator - Dashboard
>> HVP.OperationalSupervisor.Intraday.AboveProfile has finished loading.
>> Loading time: 11681 ms
>> 2012-10-08 13:39:38,268 [497865763@qtp-1377590487-101] INFO
>> fx0000000b.dashboardsBundle.dashboardsUIModule - Dashboard
>> application.ShellUI37.innerScrollBarsContent.modulesContainer.viewstack.Canv
>> as142.dashboardViewer has finished loading with updateComplete. Loading
>> time: 15578 ms
>> 2012-10-08 13:40:54,689 [823563825@qtp-1377590487-95] INFO
>> fx0000000b.dashboardsBundle.dashboardsUIModule - Dashboard
>> application.ShellUI37.innerScrollBarsContent.modulesContainer.viewstack.Canv
>> as142.dashboardViewer has finished loading with updateComplete. Loading
>> time: 91995 ms
>>
>> dashboardsUIModule is the container of the AboveProfile dashboard, it
>> seems
>> AboveProfile is processed twice, dashboardsUIModule contains AboveProfile
>> and a toolbar.
>>
>> I noticed the process of updating the display list happened more than 1
>> time
>> when I¹ve got lot of chart into the dashboard, which is the case of
>> AboveProfile and LostAndStuck.
>>
>> The loading times: flex 4 / flex 3 for the dashboards containing a lot of
>> charts
>> AboveProfile : 10538/ 7785
>> LostAndStuck: 4860 / 2704
>>
>> Frédéric THOMAS
>>
>> -----Original Message-----
>> From: Alex Harui [mailto:aharui@adobe.com]
>> Sent: Friday, October 05, 2012 6:42 PM
>> To: flex-users@incubator.apache.org
>> Subject: Re: Performence issue
>>
>> What does the profiler say is going on?
>>
>>
>> On 10/5/12 9:10 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
>>
>>> Hi Alex,
>>>
>>> - The application is 6778 .as and 10315 .mxml (some of the .as are
>>> visual components and optimized as noted in your document)
>>> - The migration is from 3.6 to 4.8
>>> - I migrated from mxml 2006 to 2009 as much as I could, only modules,
>>> components which contains States are still in mxml 2006
>>> - Almost all the styles are in .css
>>> - Each functional part of the application is benchmarked.
>>> - I ran both version (flex 3 / 4) on the same computer, with the same
>>> data loaded, at the same time and with FP 11.4
>>>
>>> The Flex 4 version was in compatibility mode because otherwise, with
>>> spark, I found much more gap between both versions.
>>>
>>> The result is:
>>> I found a gap of 35% between flex 3 / 4 ( Flex 3 compatibility mode)
>>> for the richest screens, 20% more or less for the others.
>>>
>>>
>>> -----Original Message-----
>>> From: Alex Harui [mailto:aharui@adobe.com]
>>> Sent: Friday, October 05, 2012 5:41 PM
>>> To: flex-users@incubator.apache.org
>>> Subject: Re: Performence issue
>>>
>>> Are you doing dynamic add of display objects? If so, the latter part
>>> of that article might pertain to you.
>>>
>>> Otherwise, please provide more information on your scenario and what
>>> changed going from 3 to 4. Which version of 4?
>>>
>>>
>>> On 10/5/12 5:30 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
>>>
>>>> Hi,
>>>>
>>>> Even after I read that
>>>> http://blogs.adobe.com/aharui/2011/04/migratory-foul-performance-prob
>>>> l ems-mi grating-from-flex-3-x-to-flex-4-x.html I thought that being
>>>> in flex compatibility 3.0.0 would be enough to have the same
>>>> performances than in Flex 3, that's not the case at all, does someone
>>>> can explain me why and if I can do something to improve them ?
>>>>
>>>> Frédéric THOMAS
>>>>
>>>
>>> --
>>> Alex Harui
>>> Flex SDK Team
>>> Adobe Systems, Inc.
>>> http://blogs.adobe.com/aharui
>>>
>>>
>>
>> --
>> Alex Harui
>> Flex SDK Team
>> Adobe Systems, Inc.
>> http://blogs.adobe.com/aharui
>>
>>
--
Alex Harui
Flex SDK Team
Adobe Systems, Inc.
http://blogs.adobe.com/aharui
Re: Performence issue
Posted by Alex Harui <ah...@adobe.com>.
Please file a bug (and a Patch) for that.
BTW, the FlashBuilder profiler can profile SWFs that are built from Maven.
Just go to the profiler screens and choose an external launch configuration.
On 10/12/12 5:46 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
> Unfortunally, I wasn't able to do that in flashbuilder, I've got to many
> mavenized projects already under IDEA, too much work to reconfigure them
> with flashbuilder.
>
> Anyway, I've been able to deterninate the problem, the charts in flex 3 had
> a static initializer for the styles, it's not the case in flex 4, at each
> addChild the styles are re-computed which seems to be a wrong aproach to me,
> It should have one static initializer per moduleFactory.
>
> Example on CartesianChart :
>
> instead of :
> private var _moduleFactoryInitialized:Boolean = false;
>
> should be :
> private static var _moduleFactoryInitialized:Dictionary= new
> Dictionary(true);
>
>
> instead of :
>
> override public function set moduleFactory(factory:IFlexModuleFactory):void
> {
> super.moduleFactory = factory;
>
> if (_moduleFactoryInitialized)
> return;
>
> _moduleFactoryInitialized = true;
>
>
> // our style settings
> initStyles();
> styleManager.registerInheritingStyle("axisTitleStyleName");
> }
>
> should be :
>
> override public function set moduleFactory(factory:IFlexModuleFactory):void
> {
> super.moduleFactory = factory;
>
> if (_moduleFactoryInitialized[factory])
> return;
>
> _moduleFactoryInitialized[factory] = true;
>
>
> // our style settings
> initStyles();
> styleManager.registerInheritingStyle("axisTitleStyleName");
> }
>
> Re-compiling the charts with this modification I've been able to get back my
> 35% lost - on some heavy screens where I had a lot of charts (ItemRenderer
> containing charts) - migrating from flex 3 to flex 4.
>
>
> At the moment anyway, I'm not satisfy yet, I've still got a lot of
> flickering I hadn't got in Flex 3 and at the moment, I don't understand why,
> it seems to be relative to styles again, the initProtoChain is called too
> often, the component are inexplicably re-layed out.
>
> Fred.
>
>
> -----Message d'origine-----
> From: Alex Harui
> Sent: Monday, October 08, 2012 5:03 PM
> To: flex-users@incubator.apache.org
> Subject: Re: Performence issue
>
> Get a copy of Flash Builder and use the profiler that comes with it. It
> will break down the most heavily used functions.
>
>
> On 10/8/12 5:35 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
>
>> Sorry for the firsts numbers, I did a "dir *.as" and a "dir *.mxml" /s but
>> the sdk was in the sources and others projects too, well, finally, it's
>> only
>> 1928 .as and 447 .mxml, not the same numbers but anyway, the rest are good
>> numbers.
>>
>> What do you mean by "What does the profiler say is going on?", we're
>> benchmarking every dashboard we load (our functional parts), we haven't
>> got
>> any particular profiler.
>>
>> In Flex 3, we still have some of the invalidatedFlags set at true after
>> the
>> update complete is fired, so to measure the load time of our components,
>> we
>> check for those flags and report the loaded status to the container.
>> protected function monitorUIComponentStatus(component:UIComponent):void {
>> if (!m_pagelet.started) {
>> return;
>> }
>> if (!component.mx_internal::invalidateSizeFlag &&
>> !component.mx_internal::invalidatePropertiesFlag &&
>> !component.mx_internal::invalidateDisplayListFlag) {
>> updateSpinner();
>> markAsLoaded();
>> } else {
>> // Not ready yet, try later
>> component.callLater(monitorUIComponentStatus,
>> [component]);
>> }
>> }
>>
>> In Flex 4, the same code gives about 35% of time in more but I noticed I
>> received 2 update complete from the container if it contains DataGrid with
>> a
>> lot of charts in the ItemRenderer, one about 10% later, the other really
>> long time after.
>>
>> The logs:
>> 2012-10-08 13:39:34,699 [829134827@qtp-1377590487-103] INFO
>> fx0000000b.dashboardsBundle.dashboardViewerMediator - Dashboard
>> HVP.OperationalSupervisor.Intraday.AboveProfile has finished loading.
>> Loading time: 11681 ms
>> 2012-10-08 13:39:38,268 [497865763@qtp-1377590487-101] INFO
>> fx0000000b.dashboardsBundle.dashboardsUIModule - Dashboard
>> application.ShellUI37.innerScrollBarsContent.modulesContainer.viewstack.Canv
>> as142.dashboardViewer has finished loading with updateComplete. Loading
>> time: 15578 ms
>> 2012-10-08 13:40:54,689 [823563825@qtp-1377590487-95] INFO
>> fx0000000b.dashboardsBundle.dashboardsUIModule - Dashboard
>> application.ShellUI37.innerScrollBarsContent.modulesContainer.viewstack.Canv
>> as142.dashboardViewer has finished loading with updateComplete. Loading
>> time: 91995 ms
>>
>> dashboardsUIModule is the container of the AboveProfile dashboard, it
>> seems
>> AboveProfile is processed twice, dashboardsUIModule contains AboveProfile
>> and a toolbar.
>>
>> I noticed the process of updating the display list happened more than 1
>> time
>> when I¹ve got lot of chart into the dashboard, which is the case of
>> AboveProfile and LostAndStuck.
>>
>> The loading times: flex 4 / flex 3 for the dashboards containing a lot of
>> charts
>> AboveProfile : 10538/ 7785
>> LostAndStuck: 4860 / 2704
>>
>> Frédéric THOMAS
>>
>> -----Original Message-----
>> From: Alex Harui [mailto:aharui@adobe.com]
>> Sent: Friday, October 05, 2012 6:42 PM
>> To: flex-users@incubator.apache.org
>> Subject: Re: Performence issue
>>
>> What does the profiler say is going on?
>>
>>
>> On 10/5/12 9:10 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
>>
>>> Hi Alex,
>>>
>>> - The application is 6778 .as and 10315 .mxml (some of the .as are
>>> visual components and optimized as noted in your document)
>>> - The migration is from 3.6 to 4.8
>>> - I migrated from mxml 2006 to 2009 as much as I could, only modules,
>>> components which contains States are still in mxml 2006
>>> - Almost all the styles are in .css
>>> - Each functional part of the application is benchmarked.
>>> - I ran both version (flex 3 / 4) on the same computer, with the same
>>> data loaded, at the same time and with FP 11.4
>>>
>>> The Flex 4 version was in compatibility mode because otherwise, with
>>> spark, I found much more gap between both versions.
>>>
>>> The result is:
>>> I found a gap of 35% between flex 3 / 4 ( Flex 3 compatibility mode)
>>> for the richest screens, 20% more or less for the others.
>>>
>>>
>>> -----Original Message-----
>>> From: Alex Harui [mailto:aharui@adobe.com]
>>> Sent: Friday, October 05, 2012 5:41 PM
>>> To: flex-users@incubator.apache.org
>>> Subject: Re: Performence issue
>>>
>>> Are you doing dynamic add of display objects? If so, the latter part
>>> of that article might pertain to you.
>>>
>>> Otherwise, please provide more information on your scenario and what
>>> changed going from 3 to 4. Which version of 4?
>>>
>>>
>>> On 10/5/12 5:30 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
>>>
>>>> Hi,
>>>>
>>>> Even after I read that
>>>> http://blogs.adobe.com/aharui/2011/04/migratory-foul-performance-prob
>>>> l ems-mi grating-from-flex-3-x-to-flex-4-x.html I thought that being
>>>> in flex compatibility 3.0.0 would be enough to have the same
>>>> performances than in Flex 3, that's not the case at all, does someone
>>>> can explain me why and if I can do something to improve them ?
>>>>
>>>> Frédéric THOMAS
>>>>
>>>
>>> --
>>> Alex Harui
>>> Flex SDK Team
>>> Adobe Systems, Inc.
>>> http://blogs.adobe.com/aharui
>>>
>>>
>>
>> --
>> Alex Harui
>> Flex SDK Team
>> Adobe Systems, Inc.
>> http://blogs.adobe.com/aharui
>>
>>
--
Alex Harui
Flex SDK Team
Adobe Systems, Inc.
http://blogs.adobe.com/aharui
Re: Performence issue
Posted by Frédéric THOMAS <we...@hotmail.com>.
Unfortunally, I wasn't able to do that in flashbuilder, I've got to many
mavenized projects already under IDEA, too much work to reconfigure them
with flashbuilder.
Anyway, I've been able to deterninate the problem, the charts in flex 3 had
a static initializer for the styles, it's not the case in flex 4, at each
addChild the styles are re-computed which seems to be a wrong aproach to me,
It should have one static initializer per moduleFactory.
Example on CartesianChart :
instead of :
private var _moduleFactoryInitialized:Boolean = false;
should be :
private static var _moduleFactoryInitialized:Dictionary= new
Dictionary(true);
instead of :
override public function set moduleFactory(factory:IFlexModuleFactory):void
{
super.moduleFactory = factory;
if (_moduleFactoryInitialized)
return;
_moduleFactoryInitialized = true;
// our style settings
initStyles();
styleManager.registerInheritingStyle("axisTitleStyleName");
}
should be :
override public function set moduleFactory(factory:IFlexModuleFactory):void
{
super.moduleFactory = factory;
if (_moduleFactoryInitialized[factory])
return;
_moduleFactoryInitialized[factory] = true;
// our style settings
initStyles();
styleManager.registerInheritingStyle("axisTitleStyleName");
}
Re-compiling the charts with this modification I've been able to get back my
35% lost - on some heavy screens where I had a lot of charts (ItemRenderer
containing charts) - migrating from flex 3 to flex 4.
At the moment anyway, I'm not satisfy yet, I've still got a lot of
flickering I hadn't got in Flex 3 and at the moment, I don't understand why,
it seems to be relative to styles again, the initProtoChain is called too
often, the component are inexplicably re-layed out.
Fred.
-----Message d'origine-----
From: Alex Harui
Sent: Monday, October 08, 2012 5:03 PM
To: flex-users@incubator.apache.org
Subject: Re: Performence issue
Get a copy of Flash Builder and use the profiler that comes with it. It
will break down the most heavily used functions.
On 10/8/12 5:35 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
> Sorry for the firsts numbers, I did a "dir *.as" and a "dir *.mxml" /s but
> the sdk was in the sources and others projects too, well, finally, it's
> only
> 1928 .as and 447 .mxml, not the same numbers but anyway, the rest are good
> numbers.
>
> What do you mean by "What does the profiler say is going on?", we're
> benchmarking every dashboard we load (our functional parts), we haven't
> got
> any particular profiler.
>
> In Flex 3, we still have some of the invalidatedFlags set at true after
> the
> update complete is fired, so to measure the load time of our components,
> we
> check for those flags and report the loaded status to the container.
> protected function monitorUIComponentStatus(component:UIComponent):void {
> if (!m_pagelet.started) {
> return;
> }
> if (!component.mx_internal::invalidateSizeFlag &&
> !component.mx_internal::invalidatePropertiesFlag &&
> !component.mx_internal::invalidateDisplayListFlag) {
> updateSpinner();
> markAsLoaded();
> } else {
> // Not ready yet, try later
> component.callLater(monitorUIComponentStatus,
> [component]);
> }
> }
>
> In Flex 4, the same code gives about 35% of time in more but I noticed I
> received 2 update complete from the container if it contains DataGrid with
> a
> lot of charts in the ItemRenderer, one about 10% later, the other really
> long time after.
>
> The logs:
> 2012-10-08 13:39:34,699 [829134827@qtp-1377590487-103] INFO
> fx0000000b.dashboardsBundle.dashboardViewerMediator - Dashboard
> HVP.OperationalSupervisor.Intraday.AboveProfile has finished loading.
> Loading time: 11681 ms
> 2012-10-08 13:39:38,268 [497865763@qtp-1377590487-101] INFO
> fx0000000b.dashboardsBundle.dashboardsUIModule - Dashboard
> application.ShellUI37.innerScrollBarsContent.modulesContainer.viewstack.Canv
> as142.dashboardViewer has finished loading with updateComplete. Loading
> time: 15578 ms
> 2012-10-08 13:40:54,689 [823563825@qtp-1377590487-95] INFO
> fx0000000b.dashboardsBundle.dashboardsUIModule - Dashboard
> application.ShellUI37.innerScrollBarsContent.modulesContainer.viewstack.Canv
> as142.dashboardViewer has finished loading with updateComplete. Loading
> time: 91995 ms
>
> dashboardsUIModule is the container of the AboveProfile dashboard, it
> seems
> AboveProfile is processed twice, dashboardsUIModule contains AboveProfile
> and a toolbar.
>
> I noticed the process of updating the display list happened more than 1
> time
> when I¹ve got lot of chart into the dashboard, which is the case of
> AboveProfile and LostAndStuck.
>
> The loading times: flex 4 / flex 3 for the dashboards containing a lot of
> charts
> AboveProfile : 10538/ 7785
> LostAndStuck: 4860 / 2704
>
> Frédéric THOMAS
>
> -----Original Message-----
> From: Alex Harui [mailto:aharui@adobe.com]
> Sent: Friday, October 05, 2012 6:42 PM
> To: flex-users@incubator.apache.org
> Subject: Re: Performence issue
>
> What does the profiler say is going on?
>
>
> On 10/5/12 9:10 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
>
>> Hi Alex,
>>
>> - The application is 6778 .as and 10315 .mxml (some of the .as are
>> visual components and optimized as noted in your document)
>> - The migration is from 3.6 to 4.8
>> - I migrated from mxml 2006 to 2009 as much as I could, only modules,
>> components which contains States are still in mxml 2006
>> - Almost all the styles are in .css
>> - Each functional part of the application is benchmarked.
>> - I ran both version (flex 3 / 4) on the same computer, with the same
>> data loaded, at the same time and with FP 11.4
>>
>> The Flex 4 version was in compatibility mode because otherwise, with
>> spark, I found much more gap between both versions.
>>
>> The result is:
>> I found a gap of 35% between flex 3 / 4 ( Flex 3 compatibility mode)
>> for the richest screens, 20% more or less for the others.
>>
>>
>> -----Original Message-----
>> From: Alex Harui [mailto:aharui@adobe.com]
>> Sent: Friday, October 05, 2012 5:41 PM
>> To: flex-users@incubator.apache.org
>> Subject: Re: Performence issue
>>
>> Are you doing dynamic add of display objects? If so, the latter part
>> of that article might pertain to you.
>>
>> Otherwise, please provide more information on your scenario and what
>> changed going from 3 to 4. Which version of 4?
>>
>>
>> On 10/5/12 5:30 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
>>
>>> Hi,
>>>
>>> Even after I read that
>>> http://blogs.adobe.com/aharui/2011/04/migratory-foul-performance-prob
>>> l ems-mi grating-from-flex-3-x-to-flex-4-x.html I thought that being
>>> in flex compatibility 3.0.0 would be enough to have the same
>>> performances than in Flex 3, that's not the case at all, does someone
>>> can explain me why and if I can do something to improve them ?
>>>
>>> Frédéric THOMAS
>>>
>>
>> --
>> Alex Harui
>> Flex SDK Team
>> Adobe Systems, Inc.
>> http://blogs.adobe.com/aharui
>>
>>
>
> --
> Alex Harui
> Flex SDK Team
> Adobe Systems, Inc.
> http://blogs.adobe.com/aharui
>
>
--
Alex Harui
Flex SDK Team
Adobe Systems, Inc.
http://blogs.adobe.com/aharui
Re: Performence issue
Posted by Alex Harui <ah...@adobe.com>.
Get a copy of Flash Builder and use the profiler that comes with it. It
will break down the most heavily used functions.
On 10/8/12 5:35 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
> Sorry for the firsts numbers, I did a "dir *.as" and a "dir *.mxml" /s but
> the sdk was in the sources and others projects too, well, finally, it's only
> 1928 .as and 447 .mxml, not the same numbers but anyway, the rest are good
> numbers.
>
> What do you mean by "What does the profiler say is going on?", we're
> benchmarking every dashboard we load (our functional parts), we haven't got
> any particular profiler.
>
> In Flex 3, we still have some of the invalidatedFlags set at true after the
> update complete is fired, so to measure the load time of our components, we
> check for those flags and report the loaded status to the container.
> protected function monitorUIComponentStatus(component:UIComponent):void {
> if (!m_pagelet.started) {
> return;
> }
> if (!component.mx_internal::invalidateSizeFlag &&
> !component.mx_internal::invalidatePropertiesFlag &&
> !component.mx_internal::invalidateDisplayListFlag) {
> updateSpinner();
> markAsLoaded();
> } else {
> // Not ready yet, try later
> component.callLater(monitorUIComponentStatus, [component]);
> }
> }
>
> In Flex 4, the same code gives about 35% of time in more but I noticed I
> received 2 update complete from the container if it contains DataGrid with a
> lot of charts in the ItemRenderer, one about 10% later, the other really
> long time after.
>
> The logs:
> 2012-10-08 13:39:34,699 [829134827@qtp-1377590487-103] INFO
> fx0000000b.dashboardsBundle.dashboardViewerMediator - Dashboard
> HVP.OperationalSupervisor.Intraday.AboveProfile has finished loading.
> Loading time: 11681 ms
> 2012-10-08 13:39:38,268 [497865763@qtp-1377590487-101] INFO
> fx0000000b.dashboardsBundle.dashboardsUIModule - Dashboard
> application.ShellUI37.innerScrollBarsContent.modulesContainer.viewstack.Canv
> as142.dashboardViewer has finished loading with updateComplete. Loading
> time: 15578 ms
> 2012-10-08 13:40:54,689 [823563825@qtp-1377590487-95] INFO
> fx0000000b.dashboardsBundle.dashboardsUIModule - Dashboard
> application.ShellUI37.innerScrollBarsContent.modulesContainer.viewstack.Canv
> as142.dashboardViewer has finished loading with updateComplete. Loading
> time: 91995 ms
>
> dashboardsUIModule is the container of the AboveProfile dashboard, it seems
> AboveProfile is processed twice, dashboardsUIModule contains AboveProfile
> and a toolbar.
>
> I noticed the process of updating the display list happened more than 1 time
> when I¹ve got lot of chart into the dashboard, which is the case of
> AboveProfile and LostAndStuck.
>
> The loading times: flex 4 / flex 3 for the dashboards containing a lot of
> charts
> AboveProfile : 10538/ 7785
> LostAndStuck: 4860 / 2704
>
> Frédéric THOMAS
>
> -----Original Message-----
> From: Alex Harui [mailto:aharui@adobe.com]
> Sent: Friday, October 05, 2012 6:42 PM
> To: flex-users@incubator.apache.org
> Subject: Re: Performence issue
>
> What does the profiler say is going on?
>
>
> On 10/5/12 9:10 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
>
>> Hi Alex,
>>
>> - The application is 6778 .as and 10315 .mxml (some of the .as are
>> visual components and optimized as noted in your document)
>> - The migration is from 3.6 to 4.8
>> - I migrated from mxml 2006 to 2009 as much as I could, only modules,
>> components which contains States are still in mxml 2006
>> - Almost all the styles are in .css
>> - Each functional part of the application is benchmarked.
>> - I ran both version (flex 3 / 4) on the same computer, with the same
>> data loaded, at the same time and with FP 11.4
>>
>> The Flex 4 version was in compatibility mode because otherwise, with
>> spark, I found much more gap between both versions.
>>
>> The result is:
>> I found a gap of 35% between flex 3 / 4 ( Flex 3 compatibility mode)
>> for the richest screens, 20% more or less for the others.
>>
>>
>> -----Original Message-----
>> From: Alex Harui [mailto:aharui@adobe.com]
>> Sent: Friday, October 05, 2012 5:41 PM
>> To: flex-users@incubator.apache.org
>> Subject: Re: Performence issue
>>
>> Are you doing dynamic add of display objects? If so, the latter part
>> of that article might pertain to you.
>>
>> Otherwise, please provide more information on your scenario and what
>> changed going from 3 to 4. Which version of 4?
>>
>>
>> On 10/5/12 5:30 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
>>
>>> Hi,
>>>
>>> Even after I read that
>>> http://blogs.adobe.com/aharui/2011/04/migratory-foul-performance-prob
>>> l ems-mi grating-from-flex-3-x-to-flex-4-x.html I thought that being
>>> in flex compatibility 3.0.0 would be enough to have the same
>>> performances than in Flex 3, that's not the case at all, does someone
>>> can explain me why and if I can do something to improve them ?
>>>
>>> Frédéric THOMAS
>>>
>>
>> --
>> Alex Harui
>> Flex SDK Team
>> Adobe Systems, Inc.
>> http://blogs.adobe.com/aharui
>>
>>
>
> --
> Alex Harui
> Flex SDK Team
> Adobe Systems, Inc.
> http://blogs.adobe.com/aharui
>
>
--
Alex Harui
Flex SDK Team
Adobe Systems, Inc.
http://blogs.adobe.com/aharui
RE: Performence issue
Posted by Frédéric THOMAS <we...@hotmail.com>.
Sorry for the firsts numbers, I did a "dir *.as" and a "dir *.mxml" /s but
the sdk was in the sources and others projects too, well, finally, it's only
1928 .as and 447 .mxml, not the same numbers but anyway, the rest are good
numbers.
What do you mean by "What does the profiler say is going on?", we're
benchmarking every dashboard we load (our functional parts), we haven't got
any particular profiler.
In Flex 3, we still have some of the invalidatedFlags set at true after the
update complete is fired, so to measure the load time of our components, we
check for those flags and report the loaded status to the container.
protected function monitorUIComponentStatus(component:UIComponent):void {
if (!m_pagelet.started) {
return;
}
if (!component.mx_internal::invalidateSizeFlag &&
!component.mx_internal::invalidatePropertiesFlag &&
!component.mx_internal::invalidateDisplayListFlag) {
updateSpinner();
markAsLoaded();
} else {
// Not ready yet, try later
component.callLater(monitorUIComponentStatus, [component]);
}
}
In Flex 4, the same code gives about 35% of time in more but I noticed I
received 2 update complete from the container if it contains DataGrid with a
lot of charts in the ItemRenderer, one about 10% later, the other really
long time after.
The logs:
2012-10-08 13:39:34,699 [829134827@qtp-1377590487-103] INFO
fx0000000b.dashboardsBundle.dashboardViewerMediator - Dashboard
HVP.OperationalSupervisor.Intraday.AboveProfile has finished loading.
Loading time: 11681 ms
2012-10-08 13:39:38,268 [497865763@qtp-1377590487-101] INFO
fx0000000b.dashboardsBundle.dashboardsUIModule - Dashboard
application.ShellUI37.innerScrollBarsContent.modulesContainer.viewstack.Canv
as142.dashboardViewer has finished loading with updateComplete. Loading
time: 15578 ms
2012-10-08 13:40:54,689 [823563825@qtp-1377590487-95] INFO
fx0000000b.dashboardsBundle.dashboardsUIModule - Dashboard
application.ShellUI37.innerScrollBarsContent.modulesContainer.viewstack.Canv
as142.dashboardViewer has finished loading with updateComplete. Loading
time: 91995 ms
dashboardsUIModule is the container of the AboveProfile dashboard, it seems
AboveProfile is processed twice, dashboardsUIModule contains AboveProfile
and a toolbar.
I noticed the process of updating the display list happened more than 1 time
when Ive got lot of chart into the dashboard, which is the case of
AboveProfile and LostAndStuck.
The loading times: flex 4 / flex 3 for the dashboards containing a lot of
charts
AboveProfile : 10538/ 7785
LostAndStuck: 4860 / 2704
Frédéric THOMAS
-----Original Message-----
From: Alex Harui [mailto:aharui@adobe.com]
Sent: Friday, October 05, 2012 6:42 PM
To: flex-users@incubator.apache.org
Subject: Re: Performence issue
What does the profiler say is going on?
On 10/5/12 9:10 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
> Hi Alex,
>
> - The application is 6778 .as and 10315 .mxml (some of the .as are
> visual components and optimized as noted in your document)
> - The migration is from 3.6 to 4.8
> - I migrated from mxml 2006 to 2009 as much as I could, only modules,
> components which contains States are still in mxml 2006
> - Almost all the styles are in .css
> - Each functional part of the application is benchmarked.
> - I ran both version (flex 3 / 4) on the same computer, with the same
> data loaded, at the same time and with FP 11.4
>
> The Flex 4 version was in compatibility mode because otherwise, with
> spark, I found much more gap between both versions.
>
> The result is:
> I found a gap of 35% between flex 3 / 4 ( Flex 3 compatibility mode)
> for the richest screens, 20% more or less for the others.
>
>
> -----Original Message-----
> From: Alex Harui [mailto:aharui@adobe.com]
> Sent: Friday, October 05, 2012 5:41 PM
> To: flex-users@incubator.apache.org
> Subject: Re: Performence issue
>
> Are you doing dynamic add of display objects? If so, the latter part
> of that article might pertain to you.
>
> Otherwise, please provide more information on your scenario and what
> changed going from 3 to 4. Which version of 4?
>
>
> On 10/5/12 5:30 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
>
>> Hi,
>>
>> Even after I read that
>> http://blogs.adobe.com/aharui/2011/04/migratory-foul-performance-prob
>> l ems-mi grating-from-flex-3-x-to-flex-4-x.html I thought that being
>> in flex compatibility 3.0.0 would be enough to have the same
>> performances than in Flex 3, that's not the case at all, does someone
>> can explain me why and if I can do something to improve them ?
>>
>> Frédéric THOMAS
>>
>
> --
> Alex Harui
> Flex SDK Team
> Adobe Systems, Inc.
> http://blogs.adobe.com/aharui
>
>
--
Alex Harui
Flex SDK Team
Adobe Systems, Inc.
http://blogs.adobe.com/aharui
Re: Performence issue
Posted by Alex Harui <ah...@adobe.com>.
What does the profiler say is going on?
On 10/5/12 9:10 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
> Hi Alex,
>
> - The application is 6778 .as and 10315 .mxml (some of the .as are visual
> components and optimized as noted in your document)
> - The migration is from 3.6 to 4.8
> - I migrated from mxml 2006 to 2009 as much as I could, only modules,
> components which contains States are still in mxml 2006
> - Almost all the styles are in .css
> - Each functional part of the application is benchmarked.
> - I ran both version (flex 3 / 4) on the same computer, with the same data
> loaded, at the same time and with FP 11.4
>
> The Flex 4 version was in compatibility mode because otherwise, with spark,
> I found much more gap between both versions.
>
> The result is:
> I found a gap of 35% between flex 3 / 4 ( Flex 3 compatibility mode) for the
> richest screens, 20% more or less for the others.
>
>
> -----Original Message-----
> From: Alex Harui [mailto:aharui@adobe.com]
> Sent: Friday, October 05, 2012 5:41 PM
> To: flex-users@incubator.apache.org
> Subject: Re: Performence issue
>
> Are you doing dynamic add of display objects? If so, the latter part of
> that article might pertain to you.
>
> Otherwise, please provide more information on your scenario and what changed
> going from 3 to 4. Which version of 4?
>
>
> On 10/5/12 5:30 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
>
>> Hi,
>>
>> Even after I read that
>> http://blogs.adobe.com/aharui/2011/04/migratory-foul-performance-probl
>> ems-mi grating-from-flex-3-x-to-flex-4-x.html I thought that being in
>> flex compatibility 3.0.0 would be enough to have the same performances
>> than in Flex 3, that's not the case at all, does someone can explain
>> me why and if I can do something to improve them ?
>>
>> Frédéric THOMAS
>>
>
> --
> Alex Harui
> Flex SDK Team
> Adobe Systems, Inc.
> http://blogs.adobe.com/aharui
>
>
--
Alex Harui
Flex SDK Team
Adobe Systems, Inc.
http://blogs.adobe.com/aharui
RE: Performence issue
Posted by Frédéric THOMAS <we...@hotmail.com>.
Hi Alex,
- The application is 6778 .as and 10315 .mxml (some of the .as are visual
components and optimized as noted in your document)
- The migration is from 3.6 to 4.8
- I migrated from mxml 2006 to 2009 as much as I could, only modules,
components which contains States are still in mxml 2006
- Almost all the styles are in .css
- Each functional part of the application is benchmarked.
- I ran both version (flex 3 / 4) on the same computer, with the same data
loaded, at the same time and with FP 11.4
The Flex 4 version was in compatibility mode because otherwise, with spark,
I found much more gap between both versions.
The result is:
I found a gap of 35% between flex 3 / 4 ( Flex 3 compatibility mode) for the
richest screens, 20% more or less for the others.
-----Original Message-----
From: Alex Harui [mailto:aharui@adobe.com]
Sent: Friday, October 05, 2012 5:41 PM
To: flex-users@incubator.apache.org
Subject: Re: Performence issue
Are you doing dynamic add of display objects? If so, the latter part of
that article might pertain to you.
Otherwise, please provide more information on your scenario and what changed
going from 3 to 4. Which version of 4?
On 10/5/12 5:30 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
> Hi,
>
> Even after I read that
> http://blogs.adobe.com/aharui/2011/04/migratory-foul-performance-probl
> ems-mi grating-from-flex-3-x-to-flex-4-x.html I thought that being in
> flex compatibility 3.0.0 would be enough to have the same performances
> than in Flex 3, that's not the case at all, does someone can explain
> me why and if I can do something to improve them ?
>
> Frédéric THOMAS
>
--
Alex Harui
Flex SDK Team
Adobe Systems, Inc.
http://blogs.adobe.com/aharui
Re: Performence issue
Posted by Alex Harui <ah...@adobe.com>.
Are you doing dynamic add of display objects? If so, the latter part of
that article might pertain to you.
Otherwise, please provide more information on your scenario and what changed
going from 3 to 4. Which version of 4?
On 10/5/12 5:30 AM, "Frédéric THOMAS" <we...@hotmail.com> wrote:
> Hi,
>
> Even after I read that
> http://blogs.adobe.com/aharui/2011/04/migratory-foul-performance-problems-mi
> grating-from-flex-3-x-to-flex-4-x.html I thought that being in flex
> compatibility 3.0.0 would be enough to have the same performances than in
> Flex 3, that's not the case at all, does someone can explain me why and if I
> can do something to improve them ?
>
> Frédéric THOMAS
>
--
Alex Harui
Flex SDK Team
Adobe Systems, Inc.
http://blogs.adobe.com/aharui