evaluation

gtoledo3's picture

I was contemplating some different evaluation models yesterday, and the discourse on network transmission brought these thoughts to mind again.

Doesn't it implicitly make sense that a pink patch on a lower layer could be able to output to a pink patch on a higher layer, and serve as a kind of texture transform, in a slightly different evaluation model?

In a model like that, it could make sense to actually have images piped to a bunch of sprites, say in different positions....imagine the classic pop art 4 sprites/different colors type of setup...

Then, instead of putting them in a 3D transform to move them as a group, you could select each sprite to have an output, and tie MULTIPLE spaghetti cables to one master "sum renderer" sprite, that would act as a 3D transform renderer, and the colors and image inputs on the master sprite would act in addition mode. Translation and rotation controls are pretty self explanatory. The blend mode becomes an image transform as well.

I know multiple cables to one input has been mentioned before, and it doesn't make sense with QC's model, but it would make sense in the scenario of having a kind of prerender stage that doesn't need "render in images", and flows the way I'm talking about.

The sprites could tie in to the master sprite as long as they were on lower layers. The "sum renderer" that renders any of the transforms connected. It could be implemented as a setting, and once the setting was enabled, the enable on the renderer would actually become an output port, and it would be turned into an image transform.

I don't see this happening in QC but I think it's a viable alternate concept of managing image processing in general.

dust's picture
Re: evaluation

you kind of lost me for a second there george. i think if i understand what you are trying to say it might be cool to be able to put sprites inside of sprites ???? that would be interesting and save space for sure i guess. i think maybe it might make more since to have like 4 sprites inside of each other for rgba or have that as an option with 4 outputs image and only one image in or maybe even 4 image inputs ?

gtoledo3's picture
Re: evaluation

Well, it would be for renders to be able to be switched to a kind of image transform. It would be somewhat like putting a renderer inside of a render in image, and feeding it to another renderer.

So, if you have something rendering on a sphere, and want to chain it to a cube, instead of putting the sphere in a render in image, you could select a mode to give it an output and have it turn into a transform patch. That way, you could see everything in-line on the same macro level.