hello,

well, having 2 pd instance will create more problem that it will solve in this 
situation.
- pix_data will appear to use less cpu, but pix_data did not use cpu, it use 
bandwidth between cpu and gpu.
and this can't be share between cpu.

- the main bottleneck of this object is the rendering part. not really getting 
the data. so the pd~ / sharemem will not help.

one possible optimisation is using pix_resize/pix_dump to output all data as a 
list that can be put on a table.
pix_data will make as many call as pixel, but the pix_dump will make only 1 
call to the gpu.
it look like it work a bit better, but i did not really benchmark it. (since 
that's not where the bottleneck is)

i'm however disappointed that destroying / creating a new object is faster than 
changing an object colour.

cheers
c


Le 25/06/2013 11:16, Philippe Boisnard a écrit :
Hello

To continue this question, perhaps, this preview can be create between 2 
instances of puredata, with sharemem to transmit data colours. In this case, we 
ll no cpu problem for the first instance. and only preview-process in the 
second.
I try to create that this week to explore this solution. In this case, it will 
be possible to have better preview quality. isn't it ?

best

p


Le 25 juin 2013 à 10:58, Cyrille Henry a écrit :

hello,

thanks for the patch,
unfortunately, i think it's still too slow to be usable.

better than deleting everything and creating them back with new colour, i tried 
to change colours of the existing one.
i tried with both the canvas version and the data structure version.
in both case, things are lot's faster in the original patch than in my 
optimized version. that's not what i was expecting.

how come that it's faster to delete a canvas and create a new one than changing 
a canvas colour?
(same question for data structure).

cheers
c



Le 24/06/2013 20:31, Philippe Boisnard a écrit :

Le 24 juin 2013 à 20:20, Cyrille Henry a écrit :

i can see the discution, but where is the patch?
and is it fast enough to be usable?

cheers
c

Le 24/06/2013 19:31, Philippe Boisnard a écrit :
Yes

arnaud courcelles have made today with datastructures
you can find this on codelab.fr
on this thread
http://codelab.fr/4343


Le 24 juin 2013 à 18:25, Cyrille Henry a écrit :

hello,
it would really be useful to have an efficient pix_preview.
did anyone try using data structure? would it be faster?

i guess lluis pix_preview is the best option for now

cheers
c


Le 23/06/2013 22:31, Olivier Baudu a écrit :
I've received few messages asking me if it's a joke... :-°

Well...
It's a kind of joke because it is useless...
(I redraw the GEM stream in hundred of 2x2 canvas)

But it's not joke in the way that it really works on my laptop / Linux Mint DE 
/ Pd-extended 0.43.4 / GEM 0.93 at 4 fps...

Sorry if I've crashed your OS with my stuff...

Cheers...

01ivier

Le 23.06.2013 17:33, Olivier Baudu a écrit :
Hi list,

I want to share with you a big victory for me... :-p
I've succeeded in visualizing image/video GEM stream inside my patch
(at 4 fps !!).

Well...
You can't really do anything else when the automatic mode is on... :-D
... but I'm so proud, you know.

Patch attached.
Screenshot here : http://codelab.fr/4343#p23267

Cheers

01ivier


_______________________________________________
Pd-list@iem.at mailing list
UNSUBSCRIBE and account-management ->
http://lists.puredata.info/listinfo/pd-list


_______________________________________________
Pd-list@iem.at mailing list
UNSUBSCRIBE and account-management -> 
http://lists.puredata.info/listinfo/pd-list


_______________________________________________
Pd-list@iem.at mailing list
UNSUBSCRIBE and account-management -> 
http://lists.puredata.info/listinfo/pd-list




_______________________________________________
Pd-list@iem.at mailing list
UNSUBSCRIBE and account-management -> 
http://lists.puredata.info/listinfo/pd-list



_______________________________________________
Pd-list@iem.at mailing list
UNSUBSCRIBE and account-management -> 
http://lists.puredata.info/listinfo/pd-list


_______________________________________________
Pd-list@iem.at mailing list
UNSUBSCRIBE and account-management -> 
http://lists.puredata.info/listinfo/pd-list

Reply via email to