hello ! surely a dummy thing, i m not arriving at all to spread the renderer output from rendertarget.fx.
i have reproduced the example but its not at all showing up a spread of texture from dx9. What is exaclty the trick ?
im thinking of it to spread ( vs queue) a texture
the multiple rendertargets.v4p in \girlpower didn’t really mention that. bad. the trick is that you need to manually spread one of the DX9Texture (EX9.Texture)'s inputs via inspektor to a maximum of NumSimultaneousRTs which is a device/graphiccard capability and if i am not mistaken it is usually 4.
note though that this may not be what you’re looking for. you now can only write different stuff out to those texture-slices from within the effects pixelshader.
what i was thinking is that we could output a spread of texture from a single renderer.
its what is somehow shown on render multiple targets, am i right ?
this would enable somehow feedbacking ( if i m not mistaking) fx creation.
and beside, answer to one of my main question:
how to copy/duplicate a video texture ( no duplicate node ) without troubles if the source is a videoIN ( sometimes some troubles with a videoIn> videoTexture) to different types of tetxure receiveres ( freeframe / fx / opencv ).
To come back to your answer ( and continuying to learn despite a surely off-the-topic), i m not at all figurating how i can manually spread the dx9 ( linear spread to X/Y res ?)
@karistouf: only way to spread texture for the moment is in MRT mode (eg, pixelshader outputs several colors). So for the moment you couldn’t for example apply the same effect to different textures.
I’m not sure how spreading renderer output outside mrt would help for feedback. One nice option for it would be things like panoramic/tile/cubemap generation.
@joreg: by the way does the mrt works in singlepass or is one pass done per rt?
Specially considering different bit depth is not possible in one pass, like one texture in R16G16B16A16 (color/normal) and one in R32 (depth) for example.
But if you use A8R8G8B8 and R32 as spread input then does it works single pass?
that would work but not make sense probably. you can just view the DX9Texture with Inspektor and there set the spreadcount of any of the input pins to 4. the slices can still all have the same values. its a strange manual thing you need to do for this special case, i know.
like vux i don’t see how this would be related to any kind of feedbacking.
when rendering you can specify multiple rendertargets at once so one drawcall then does it all. not sure what goes on internally though…
just did a quick test with the girlpowers rendertarget example. spreading the textureformat seemed to work for me. so yes: writing to multiple different bit depths at once from one effect with one draw call.