RE: Near completion of Color Pattern Toolkit: IDE, engine, and font problems (part two)

[MisterX]

I thought it might be the issue. I stand corrected. 

Have you tried destroying the graphic stuff and recreating it?
Doing garbage collection of sorts? Delete local arrays?

cheers
Xavier

> -Original Message-
> From: [EMAIL PROTECTED] 
> [mailto:[EMAIL PROTECTED] On Behalf Of 
> Wilhelm Sanke
> Sent: Friday, 25 November, 2005 22:54
> To: use-revolution@lists.runrev.com
> Subject: RE: Near completion of Color Pattern Toolkit: IDE, 
> engine, and font problems (part two)
> 
> Xavier,
> 
> I just followed your suggestion to try to use graphics 
> instead of fields
> - something I had also experimented with some time ago.
> 
> I substituted the 2700 fields of test stack "ScanTest2700" 
> with graphics (and again want to mention that my present 
> toolkit stack contains only
> *one* field now containing color information)
> 
> The result for three passes with the scan button:
> 
> Stack with graphics 2371, 7930, and 13565 milliseconds.
> Stack with fields 2667, 7611, 13109 milliseconds.
> 
> I would say there are no statistically significant 
> differences between the two stacks (we would have to continue 
> to get a possible significance on the basis of more data).
> 
> Cheers,
> 
> Wilhelm
> <http://www.sanke.org/MetaMedia>
> 
> 
> ___
> use-revolution mailing list
> use-revolution@lists.runrev.com
> Please visit this url to subscribe, unsubscribe and manage 
> your subscription preferences:
> http://lists.runrev.com/mailman/listinfo/use-revolution

___
use-revolution mailing list
use-revolution@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-revolution

RE: Near completion of Color Pattern Toolkit: IDE, engine, and font problems (part two)

[Wilhelm Sanke]

Xavier,

I just followed your suggestion to try to use graphics instead of fields 
- something I had also experimented with some time ago.


I substituted the 2700 fields of test stack "ScanTest2700" with graphics 
(and again want to mention that my present toolkit stack contains only 
*one* field now containing color information)


The result for three passes with the scan button:

Stack with graphics 2371, 7930, and 13565 milliseconds.
Stack with fields 2667, 7611, 13109 milliseconds.

I would say there are no statistically significant differences between 
the two stacks (we would have to continue to get a possible significance 
on the basis of more data).


Cheers,

Wilhelm



___
use-revolution mailing list
use-revolution@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-revolution

Re: Near completion of Color Pattern Toolkit: IDE, engine, and font problems (part two)

[Wilhelm Sanke]

An additional remark concerning the reported engine leak (Windows XP and 
MacOS):


The increments in script execution time occur probably in many scripts, 
if not in all, although for shorter scripts with no loops or limited 
loops this may be hardly noticeable.


One example which I just tested:

There are six buttons in the color pattern toolkit that can store the 
color values of a given image for further processing or later retrieving.


The following script stores the values in a one-dimensional array and 
customproperty.


"put 0 into counter
 repeat for each Char C in fld 1
   add 1 to counter
   put the backcolor of char counter of fld 1 into aColor[counter]
 end repeat
set the customproperties[oldColors] of me to aColor"

For the roughly 20.000 chars this takes 550 milliseconds when the stack 
has been newly opened on my XP computer.


For each subsequent running of the script an increment of 20 to 40 ms 
occurs, but this is an average increment, as it does not go up 
consistently. In between small decrements can happen, but overall the 
execution time goes up.


After a number of about 30 runs the execution time of the above script 
reaches 1500 milliseconds.


On the basis of this observation it is understandable that during a 
session with the toolkit - where all the time color values are put into 
arrays for faster processing or transforming inside the array or between 
two or more arrays - that performance gradually gets very slow and the 
engine may choke for a while.
But never before using the additional scan feature illustrated in the 
test stack  I have experienced 
such a quick and devastating breakdown like I - and subsequently Xavier 
- have described in our posts.



Regards,


Wilhelm Sanke




___
use-revolution mailing list
use-revolution@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-revolution

RE: Near completion of Color Pattern Toolkit: IDE, engine, and font problems (part two)

[MisterX]

 
Wilhelm

> Thanks for the confirmation of my findings.

>From one crazy painter to another, it's a pleasure to try your tools ;)

> The engine leak is there as before. Scanning an 100X100 area 
> of the image thumb brings down my computer immediately.

it does take 4 or 5 successive runs to make it much slower here. But the
effect is doubled each time...

> I just used fields as color units because that old stack of 
> mine had them.
> I do not think (did not try yet) that graphics would make a 
> difference.

My logic is that displaying a field takes more resources than just a
graphic... I could be wrong but needs testing...

> As I mentioned in the first part of my report (about IDE 
> problems of Nov
> 22) the present version of the toolkit uses only *one* single 
> field now, but with 19.800 color chars in it, which for a 
> "text" really  is not much. Thus we have about 20.000 fields less.

Text, no, but 1000 of anything controls wise will cost a lot more than 20K's
of text... This is why I went from graphics I could color to a single paint
control and the performance showed right away the benefits... Note that
painting was just as slow as creating the fields (though just an impression,
im sure the paint is much faster)... 

your code is not much different than mine btw... Mine is overly complicated
to allow differnet modes of drawing and types of colorimetric function
parametrizing but your does have the fields overhead and that's where things
go really slow. In my stack, there is not issues with speed like you had.
There wasn't in the old model with graphics, and there isn't in the new with
paint. 

My guess is that you are overloading something... Since the number of fields
is constant, consider what changes... I tried to isolate it but hand to do
some work and couldn't continue... 

cheers
Xavier
http://monsieurx.com/taoo

___
use-revolution mailing list
use-revolution@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-revolution

Re: Near completion of Color Pattern Toolkit: IDE, engine, and font problems (part two)

[Wilhelm Sanke]

On Fri, 25 Nov ,"MisterX" <[EMAIL PROTECTED]> wrote:


Wilhelm

i tried your stack. After each redraw it is 2X slower. When i tried to see
why, MC started showing really weird artifacts, menus bars gone wrong, etc
and then it crashed. (glad it was the only stack opened at the time).
(snip)
There IS a memory leak
(engine 2.6.6) and further another problem with the redraw of the window
in XP thereafter. Each iteration doubles the time it takes... After that,
MC was running on mollases (despite your stack closed)
(snip)
But the leak is well in the engine.
(snip)
One suggestion is to use "graphics" instead of fields. MoireX did that and
i dont think there was as much overhead...

Cheers
Xavier
-- http://monsieurx.com/taoo




Xavier,

Thanks for the confirmation of my findings.

I just used fields as color units because that old stack of mine had them.
I do not think (did not try yet) that graphics would make a difference.

As I mentioned in the first part of my report (about IDE problems of Nov 
22) the present version of the toolkit uses only *one* single field now, 
but with 19.800 color chars in it, which for a "text" really  is not 
much. Thus we have about 20.000 fields less.


The engine leak is there as before. Scanning an 100X100 area of the 
image thumb brings down my computer immediately.


Cheers,

Wilhelm




___
use-revolution mailing list
use-revolution@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-revolution

Re: Near completion of Color Pattern Toolkit: IDE, engine, and font problems (part two)

[MisterX]

Wilhelm

i tried your stack. After each redraw it is 2X slower. When i tried to see
why, MC started showing really weird artifacts, menus bars gone wrong, etc
and then it crashed. (glad it was the only stack opened at the time).

Second attempt:

The big problem is with the backdrop (recently damned by the list). This
creates all sort of hickups and artifacts and eventually the crash. Taking
a screenshot resulted also in an error...

After removing the backdrop, i could see better. There IS a memory leak
(engine 2.6.6) and further another problem with the redraw of the window
in XP thereafter. Each iteration doubles the time it takes... After that,
MC was running on mollases (despite your stack closed) and I had to quit a
third time - mc became unusable... Parts of the windows from other
programs would draw over the mctoolbar - i had this also happen in another
stack since 2.6.6 appeared.). Anyway,before i could quit or save or
anything MC crashed again...

But the leak is well in the engine. Opening mc takes 10MBs of RAM. I run
your script and each time MC eats 10MBs. Close the stack, memory is not
freed... I did click the stack memory buttons on close...

One suggestion is to use "graphics" instead of fields. MoireX did that and
i dont think there was as much overhead...

Last but not least, using classic windows window theme instead of XP style
helps a bit but not much. By far the worst things i found was the
backdrop, and the use of fields as coloring elements. I didn't try with
graphics...

all tests done on a W2K3 server...

Cheers
Xavier
-- 
http://monsieurx.com/taoo

___
use-revolution mailing list
use-revolution@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription 
preferences:
http://lists.runrev.com/mailman/listinfo/use-revolution