Segment rendering for different feature sets?

Yea, trying to explain this in the title isn’t work very well… I want to make an animation using fire, but fire isn’t the main portion of the animation. I think it’d be efficient to take advantage of GPU rendering for the one part of the animation and then use CPU rendering for the fire.

How can I render one part of a scene, then add it together with the second part of the scene rendered with a different feature set?

Also, if anyone knows if and when the foundation plans to implement opencl volume rendering, let me know.

You can make use of render layers.

https://www.blender.org/manual/render/post_process/layers.html

I doubt you’ll be able to render volumetrics with OpenCL anytime soon.

Use render layers or render the fire animation with transparency checked as a PNG(RGBA with Alpha) sequence and combine it later.