Feature Request (Rendering Engine)

The interface, modeling, 3d editing tools, import/export, feature requests, etc

Moderators: jesterKing, stiv

Caronte
Posts: 76
Joined: Wed Oct 16, 2002 12:53 am
Location: Valencia-Spain-Europe

Post by Caronte »

I think Ton is working now, I feel the power of the dark side :twisted:
Caronte.
"Some Day, All Will Be Digital"
http://www.nicodigital.com

kitsu
Posts: 0
Joined: Sun Jun 06, 2004 8:08 am

Post by kitsu »

Lol!

Ton is such a titan of a coder. I can't believe all the things he managed to get done during Orange. And more important than the number of additions is how general and stable they are! The mat and composite nodes are so great I shouldn't really ask for anything more... Damn greedy users :oops:

joeri
Posts: 96
Joined: Fri Jan 10, 2003 6:41 pm
Contact:

Post by joeri »

I don't think Ton will be working on this.
Because to be honest: human time is more 'expensive' then computer time.
Doing all the work to save some render time is just not very professional.
I understand there are some people out there who have more time then money (to buy cpu speed), but in most modern animations the camera is constantly moving so this pretty much makes the parts optimize useless.

Caronte
Posts: 76
Joined: Wed Oct 16, 2002 12:53 am
Location: Valencia-Spain-Europe

Post by Caronte »

joeri wrote:I don't think Ton will be working on this.
Because...
My latest comment was in response to the Kitsu post: "Has anyone else even touched the render code since Ton rewrote it for Orange?"

I'm only supposing (like you can suppose) because I read a Ton message from the developpers mailing list, talking about that :wink:
Caronte.
"Some Day, All Will Be Digital"
http://www.nicodigital.com

kitsu
Posts: 0
Joined: Sun Jun 06, 2004 8:08 am

Post by kitsu »

Well, this is quickly disolving into a useless thread.

Joeri: I agree Tons time is more valuable then the few seconds this would shave off of animation render times. I disagree about its general usefulness though. Adaptive subdivision would make this useful even in scenes with large amounts of movement, and there are plenty of times when you don't need the camera flying all over the place. But it is true, with even a litle camera move, every pixel will need to be updated for things to look right.

Anyway it sounds like an intersting problem if someone wants to try to solve it.

joeri
Posts: 96
Joined: Fri Jan 10, 2003 6:41 pm
Contact:

Post by joeri »

I'm not sure what you mean with adaptive subdivision.
Splitting up a render frame into different size parts?

I saw mental images do this years ago. Areas that are suspected to take high render times (lots of transparant materials, reflectiveness, etc) are cut into smaller parts. This does not help on sequences but does on multi processor single images. The user no longer has to wait untill the last part is finally done (by the slowest processor). Ofcourse this means rendering a single image on a multi computer/processor network, something blender cant do yet.

hold-and-copy would also be hard in render farms.
what if machine #3 starts on frame range 31-40 but needs to copy an unfinished 18... stuff like that.

If the camera is not moving then better split the scene up into "backdrop" and "front action". Render a still for the backdrop and then render the front action. The front action renderer will see lots of empty parts that render real quick and need no optimisation.
But yes it would be nice that when the camera does move a small segment of the animation it renders a new backdrop by itsself. It all comes down to the smartness of the (video)editor if you really want all those 'the same' images copied on your harddisk.

For arguments sake it would also be nice if blender would create a seq-edit sequence out of the none-moving backdrop (say layers 11-20) with alpha over front-action (say layers 1-19) images. This would not only save render times but also alot of diskspace.

layertimelaps mockup
Image

kitsu
Posts: 0
Joined: Sun Jun 06, 2004 8:08 am

Post by kitsu »

joeri: Right again! I was imagining using something like an octree to divide the render screen into the smallest regions where pixels would change.

You are also right that most of this optimization can already be done in the compositor/sequence editor. I guess I'm just lazy - I want the computer to do that work for me :roll:
Okay, you've convinced me, I will do my optimizing myself.

joeri
Posts: 96
Joined: Fri Jan 10, 2003 6:41 pm
Contact:

Post by joeri »

kitsu wrote:You are also right that most of this optimization can already be done in the compositor/sequence editor. I guess I'm just lazy - I want the computer to do that work for me :roll:
Okay, you've convinced me, I will do my optimizing myself.
It doesn't need to be or/or. I think a little patch can create the seq-edit strips for you. Let blender "run" trough the animation and if nothing moves then create an image clip and if something moves use an image-sequence clip.

I have no idea if creating clips is in python, i think not. Then it would be fairly easy to do.

Caronte
Posts: 76
Joined: Wed Oct 16, 2002 12:53 am
Location: Valencia-Spain-Europe

Post by Caronte »

joeri wrote:It doesn't need to be or/or. I think a little patch can create the seq-edit strips for you. Let blender "run" trough the animation and if nothing moves then create an image clip and if something moves use an image-sequence clip.
Nice idea!
Caronte.
"Some Day, All Will Be Digital"
http://www.nicodigital.com

kitsu
Posts: 0
Joined: Sun Jun 06, 2004 8:08 am

Post by kitsu »

Yeah, last time I check Python had no access to the sequencer.

advs89
Posts: 0
Joined: Wed Apr 26, 2006 10:22 pm
Location: Roanoke, VA, USA

Well...

Post by advs89 »

I sortof forgot about this thread... apparently the consensus is that it can be done by hand, so noone wants to waste the time coding it. I guess I would rather the developers work on something more important, such as an vector/svg renderer (i wish...), anyways.

Thanks for your thoughts,
Advs89

Post Reply