[Fix] Issue890 Mantis6535 SortedCollection loose sortBlock

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
14 messages Options
Reply | Threaded
Open this post in threaded view
|

[Fix] Issue890 Mantis6535 SortedCollection loose sortBlock

Nicolas Cellier
http://code.google.com/p/pharo/issues/list?updated=890 ready for inclusion
Use SLICES described in my last comment

Sorry I left some garbage in the PharoInbox, but it's quite unhandy to
manage incremental dependency (when order of changes DOES matter).
Wouldn't atomic loading be an easy solution for such problems?

I usually don't complain, but here are my feelings:
<rant>
Monticello is quite a drag, and pharo feels globally unresponsive
(even the core image).
I still experiment some lost keystroke and menus are sometimes gooey.
It's not rare I click on the wrong item.
If 3GHz Intel core2 duo 64 bits really becomes a requirement, I will
give Cuis a ride to get some fresh air ;)
To me UI feel is mitigating all the good job made so far.
</rant>

Nicolas

_______________________________________________
Pharo-project mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Reply | Threaded
Open this post in threaded view
|

Re: [Fix] Issue890 Mantis6535 SortedCollection loose sortBlock

Stéphane Ducasse

On Jun 16, 2009, at 10:10 PM, Nicolas Cellier wrote:

> http://code.google.com/p/pharo/issues/list?updated=890 ready for  
> inclusion
> Use SLICES described in my last comment
>
> Sorry I left some garbage in the PharoInbox, but it's quite unhandy to
> manage incremental dependency (when order of changes DOES matter).

I will clean it up. And yes this is not really satisfactory.

> Wouldn't atomic loading be an easy solution for such problems?

in theory it would and I would love to. Now I need a week without
students, emails, papers and I want to have a look at MC1.5 but so far  
I could not
find the time.

>
> I usually don't complain, but here are my feelings:
> <rant>
> Monticello is quite a drag,

I think that monticello is really cool compare to cs.
Now with atomic loading and other speed up it should be really better

> and pharo feels globally unresponsive
> (even the core image).

I'm planning to replace package info by some real packages but it  
takes times

> I still experiment some lost keystroke and menus are sometimes gooey.

this is strange because on mac we do not have that behavior

> It's not rare I click on the wrong item.
> If 3GHz Intel core2 duo 64 bits really becomes a requirement, I will
> give Cuis a ride to get some fresh air ;)

It should not.

But really we are really paying attention to go faster on the UI level
So I hope that the morphic cleaning + sensor will pay off.

> To me UI feel is mitigating all the good job made so far.

we are doing our best.
But we accept speed up proposal at the scrollbar level
If somebody wants to integrate the speed up made on cuis for the UI
then feel free.

I think that once 1.0 is out of the door the first thing we should do
as a community is to take time to rethink our tools.

Stef


_______________________________________________
Pharo-project mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Reply | Threaded
Open this post in threaded view
|

Re: [Fix] Issue890 Mantis6535 SortedCollection loose sortBlock

Adrian Lienhard
In reply to this post by Nicolas Cellier
Hi Nicolas,

On Jun 16, 2009, at 22:10 , Nicolas Cellier wrote:

[...]

> <rant>
> Monticello is quite a drag, and pharo feels globally unresponsive
> (even the core image).

Yep, responsiveness can be improved. The core feels quite fast for me  
(on a MacBook Pro), but still, for example opening a browser is not  
instantaneous. Also, if you type really fast, the characters start  
appearing late (I think Bill complained about that recently).

OB is another story. For me, its at the border of being usable because  
of its slowness.

[...]

> To me UI feel is mitigating all the good job made so far.

So we should try to improve the GUI performance of the core. However,  
I'm not sure who can help...

Cheers,
Adrian

_______________________________________________
Pharo-project mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Reply | Threaded
Open this post in threaded view
|

Re: [Fix] Issue890 Mantis6535 SortedCollection loose sortBlock

Lukas Renggli
> OB is another story. For me, its at the border of being usable because
> of its slowness.

Just for the record: OB included with the Seaside One-Click image is
only marginally slower than the standard system browser.

   Opening and closing OB system browser on the class WACounter: 636.15 ms
   Opening and closing SystemBrowser on the class WACounter: 555.00 ms

Lukas

-=--=--=--=--=--=--=--=--=--=--=--=--=--=-

" Omni Browser "
[ browsers := (1 to: 20)
                collect: [ :each | OBSystemBrowser onClass: WACounter ].
        browsers do: [ :each |
                each open.
                World doOneCycle ].
        browsers do: [ :each |
                each close.
                World doOneCycle ] ]
                        timeToRun / 20.0

" System Browser "
[ browsers := (1 to: 20)
                collect: [ :each | Browser new setClass: WACounter selector: nil ].
        browsers do: [ :each |
                Browser openBrowserView: (each openEditString: nil) label: 'System Browser'.
                World doOneCycle ].
        browsers do: [ :each |
                each changed: #close.
                World doOneCycle ] ]
                        timeToRun / 20.0

-=--=--=--=--=--=--=--=--=--=--=--=--=--=-

--
Lukas Renggli
http://www.lukas-renggli.ch

_______________________________________________
Pharo-project mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Reply | Threaded
Open this post in threaded view
|

Re: [Fix] Issue890 Mantis6535 SortedCollection loose sortBlock

Adrian Lienhard
Thanks, I like benchmarks! I've run this in the recent Pharo-core and  
Pharo images (on the class String):

Pharo core 10432 SystemBrowser: 686.6ms
Pharo0.1-10342dev09.06.3 SystemBrowser: 940.8ms
Pharo0.1-10342dev09.06.3 OBSystemBrowser: 1087.1ms

Similar to Lukas' results is that the OBSystemBrowser is not that much  
slower than the normal SystemBrowser (slowdown of 15%). However, there  
is a larger difference between the SystemBrowser in Pharo and in Pharo-
core (37%). Hence other features seem to contribute to the slowdown:  
freetype fonts, code highlighting, ...?

Adrian

On Jun 17, 2009, at 11:18 , Lukas Renggli wrote:

>> OB is another story. For me, its at the border of being usable  
>> because
>> of its slowness.
>
> Just for the record: OB included with the Seaside One-Click image is
> only marginally slower than the standard system browser.
>
>   Opening and closing OB system browser on the class WACounter:  
> 636.15 ms
>   Opening and closing SystemBrowser on the class WACounter: 555.00 ms
>
> Lukas
>
> -=--=--=--=--=--=--=--=--=--=--=--=--=--=-
>
> " Omni Browser "
> [ browsers := (1 to: 20)
> collect: [ :each | OBSystemBrowser onClass: WACounter ].
> browsers do: [ :each |
> each open.
> World doOneCycle ].
> browsers do: [ :each |
> each close.
> World doOneCycle ] ]
> timeToRun / 20.0
>
> " System Browser "
> [ browsers := (1 to: 20)
> collect: [ :each | Browser new setClass: WACounter selector: nil ].
> browsers do: [ :each |
> Browser openBrowserView: (each openEditString: nil) label: 'System  
> Browser'.
> World doOneCycle ].
> browsers do: [ :each |
> each changed: #close.
> World doOneCycle ] ]
> timeToRun / 20.0
>
> -=--=--=--=--=--=--=--=--=--=--=--=--=--=-
>
> --
> Lukas Renggli
> http://www.lukas-renggli.ch
>
> _______________________________________________
> Pharo-project mailing list
> [hidden email]
> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project


_______________________________________________
Pharo-project mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Reply | Threaded
Open this post in threaded view
|

Re: [Fix] Issue890 Mantis6535 SortedCollection loose sortBlock

Stéphane Ducasse
In reply to this post by Lukas Renggli
one of the problem is that package are slow since they rely on class  
category.

Stef

On Jun 17, 2009, at 11:18 AM, Lukas Renggli wrote:

>> OB is another story. For me, its at the border of being usable  
>> because
>> of its slowness.
>
> Just for the record: OB included with the Seaside One-Click image is
> only marginally slower than the standard system browser.
>
>   Opening and closing OB system browser on the class WACounter:  
> 636.15 ms
>   Opening and closing SystemBrowser on the class WACounter: 555.00 ms
>
> Lukas
>
> -=--=--=--=--=--=--=--=--=--=--=--=--=--=-
>
> " Omni Browser "
> [ browsers := (1 to: 20)
> collect: [ :each | OBSystemBrowser onClass: WACounter ].
> browsers do: [ :each |
> each open.
> World doOneCycle ].
> browsers do: [ :each |
> each close.
> World doOneCycle ] ]
> timeToRun / 20.0
>
> " System Browser "
> [ browsers := (1 to: 20)
> collect: [ :each | Browser new setClass: WACounter selector: nil ].
> browsers do: [ :each |
> Browser openBrowserView: (each openEditString: nil) label: 'System  
> Browser'.
> World doOneCycle ].
> browsers do: [ :each |
> each changed: #close.
> World doOneCycle ] ]
> timeToRun / 20.0
>
> -=--=--=--=--=--=--=--=--=--=--=--=--=--=-
>
> --
> Lukas Renggli
> http://www.lukas-renggli.ch
>
> _______________________________________________
> Pharo-project mailing list
> [hidden email]
> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project


_______________________________________________
Pharo-project mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Reply | Threaded
Open this post in threaded view
|

Re: [Fix] Issue890 Mantis6535 SortedCollection loose sortBlock

Stéphane Ducasse
In reply to this post by Adrian Lienhard
did you disable dynamic protocols?

On Jun 17, 2009, at 12:17 PM, Adrian Lienhard wrote:

> Thanks, I like benchmarks! I've run this in the recent Pharo-core and
> Pharo images (on the class String):
>
> Pharo core 10432 SystemBrowser: 686.6ms
> Pharo0.1-10342dev09.06.3 SystemBrowser: 940.8ms
> Pharo0.1-10342dev09.06.3 OBSystemBrowser: 1087.1ms
>
> Similar to Lukas' results is that the OBSystemBrowser is not that much
> slower than the normal SystemBrowser (slowdown of 15%). However, there
> is a larger difference between the SystemBrowser in Pharo and in  
> Pharo-
> core (37%). Hence other features seem to contribute to the slowdown:
> freetype fonts, code highlighting, ...?
>
> Adrian
>
> On Jun 17, 2009, at 11:18 , Lukas Renggli wrote:
>
>>> OB is another story. For me, its at the border of being usable
>>> because
>>> of its slowness.
>>
>> Just for the record: OB included with the Seaside One-Click image is
>> only marginally slower than the standard system browser.
>>
>>  Opening and closing OB system browser on the class WACounter:
>> 636.15 ms
>>  Opening and closing SystemBrowser on the class WACounter: 555.00 ms
>>
>> Lukas
>>
>> -=--=--=--=--=--=--=--=--=--=--=--=--=--=-
>>
>> " Omni Browser "
>> [ browsers := (1 to: 20)
>> collect: [ :each | OBSystemBrowser onClass: WACounter ].
>> browsers do: [ :each |
>> each open.
>> World doOneCycle ].
>> browsers do: [ :each |
>> each close.
>> World doOneCycle ] ]
>> timeToRun / 20.0
>>
>> " System Browser "
>> [ browsers := (1 to: 20)
>> collect: [ :each | Browser new setClass: WACounter selector: nil ].
>> browsers do: [ :each |
>> Browser openBrowserView: (each openEditString: nil) label: 'System
>> Browser'.
>> World doOneCycle ].
>> browsers do: [ :each |
>> each changed: #close.
>> World doOneCycle ] ]
>> timeToRun / 20.0
>>
>> -=--=--=--=--=--=--=--=--=--=--=--=--=--=-
>>
>> --
>> Lukas Renggli
>> http://www.lukas-renggli.ch
>>
>> _______________________________________________
>> Pharo-project mailing list
>> [hidden email]
>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
>
>
> _______________________________________________
> Pharo-project mailing list
> [hidden email]
> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project


_______________________________________________
Pharo-project mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Reply | Threaded
Open this post in threaded view
|

Re: [Fix] Issue890 Mantis6535 SortedCollection loose sortBlock

Damien Cassou
On Wed, Jun 17, 2009 at 1:36 PM, Stéphane
Ducasse<[hidden email]> wrote:
> did you disable dynamic protocols?

if dynamic protocols were the problem, they were only responsible of
the 15% difference between the SystemBrowser and OB-one.  The huge
difference is between the Core and Pharo which means dynamic protocols
are not responsible.

--
Damien Cassou
http://damiencassou.seasidehosting.st

"Lambdas are relegated to relative obscurity until Java makes them
popular by not having them." James Iry

_______________________________________________
Pharo-project mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Reply | Threaded
Open this post in threaded view
|

Re: [Fix] Issue890 Mantis6535 SortedCollection loose sortBlock

Stéphane Ducasse
In reply to this post by Adrian Lienhard
do we have an idea of the source of slow down?


>
> Pharo core 10432 SystemBrowser: 686.6ms
> Pharo0.1-10342dev09.06.3 SystemBrowser: 940.8ms


_______________________________________________
Pharo-project mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Reply | Threaded
Open this post in threaded view
|

Re: [Fix] Issue890 Mantis6535 SortedCollection loose sortBlock

Marcus Denker-3
In reply to this post by Adrian Lienhard

On 17.06.2009, at 04:39, Adrian Lienhard wrote:

> Hi Nicolas,
>
> On Jun 16, 2009, at 22:10 , Nicolas Cellier wrote:
>
> [...]
>
>> <rant>
>> Monticello is quite a drag, and pharo feels globally unresponsive
>> (even the core image).
>
> Yep, responsiveness can be improved. The core feels quite fast for me
> (on a MacBook Pro), but still, for example opening a browser is not
> instantaneous. Also, if you type really fast, the characters start
> appearing late (I think Bill complained about that recently).
>
> OB is another story. For me, its at the border of being usable because
> of its slowness.

I wonder if OB would not be much faster if the auto-category feature  
would
be turned off?

        Marcus

_______________________________________________
Pharo-project mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Reply | Threaded
Open this post in threaded view
|

Re: [Fix] Issue890 Mantis6535 SortedCollection loose sortBlock

Damien Cassou
On Wed, Jun 17, 2009 at 2:32 PM, Marcus Denker<[hidden email]> wrote:
> I wonder if OB would not be much faster if the auto-category feature
> would
> be turned off?

are you talking about the dynamic protocols (long methods,
override...)? They are much faster than before. You can disable them
to see if they are a source of slowdowns by going to the preference
browser, 'dynamic protocols' category.

--
Damien Cassou
http://damiencassou.seasidehosting.st

"Lambdas are relegated to relative obscurity until Java makes them
popular by not having them." James Iry

_______________________________________________
Pharo-project mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Reply | Threaded
Open this post in threaded view
|

Re: [Fix] Issue890 Mantis6535 SortedCollection loose sortBlock

Marcus Denker-3

On 17.06.2009, at 09:34, Damien Cassou wrote:

> On Wed, Jun 17, 2009 at 2:32 PM, Marcus Denker<[hidden email]> wrote:
>> I wonder if OB would not be much faster if the auto-category feature
>> would
>> be turned off?
>
> are you talking about the dynamic protocols (long methods,
> override...)? They are much faster than before. You can disable them
> to see if they are a source of slowdowns by going to the preference
> browser, 'dynamic protocols' category.
>

Helps a little... but it feels to me that displaying a method, the  
slowest
part is the Annotationpane. Maybe we can turn that off? It's off in  
the core
image, too.

        Marcus

_______________________________________________
Pharo-project mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Reply | Threaded
Open this post in threaded view
|

Re: [Fix] Issue890 Mantis6535 SortedCollection loose sortBlock

Damien Cassou
On Wed, Jun 17, 2009 at 5:44 PM, Marcus Denker<[hidden email]> wrote:
> Helps a little... but it feels to me that displaying a method, the
> slowest
> part is the Annotationpane. Maybe we can turn that off? It's off in
> the core
> image, too.

I have tried disabling it but haven't felt any difference.

--
Damien Cassou
http://damiencassou.seasidehosting.st

"Lambdas are relegated to relative obscurity until Java makes them
popular by not having them." James Iry

_______________________________________________
Pharo-project mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Reply | Threaded
Open this post in threaded view
|

Re: [Fix] Issue890 Mantis6535 SortedCollection loose sortBlock

Lukas Renggli
>> Helps a little... but it feels to me that displaying a method, the
>> slowest
>> part is the Annotationpane. Maybe we can turn that off? It's off in
>> the core
>> image, too.
>
> I have tried disabling it but haven't felt any difference.

Please check out the Seaside One-Click image. It is pretty fast,
especially when using the standard Squeak look.

Lukas

--
Lukas Renggli
http://www.lukas-renggli.ch

_______________________________________________
Pharo-project mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project