From browsing around, it seems like this isn’t supported, but is there a way to render to a target texture that is deeper than 8 bits per channel?
I realize it’s expensive memory/cache-thrashing-wise, but for an effect I’m working on, I could really use the precision…
Last time I checked, accumulation buffer was not actually supported by any hardware out there (and there’s no such concept in D3D at all).
Now, when it comes to floating point (or other higher than 8 bit/channel precision) render texture formats, then no, currently you can’t use them. I want to add them in some future release.