Performance optimizations for IVA mode?

[JonnyOThan]

After fixing a bunch of the VR resolution issues, the performance is notably worse (not that surprising). I need to do some profiler captures to see where it’s spending time. But I had a thought: most of the time in VR you are in IVA mode and very little of the screen is showing a planet’s atmosphere. Is there anything we can do specifically for this case?

Is it possible to maybe include the IVA model in the depth prepass (is there even such a thing?)? Or maybe draw the local and scaled space cameras at a lower resolution when in IVA mode?

I’m not sure if this issue really should be on EVE, scatterer, or deferred - or all 3? But volumetric clouds seem to be the most expensive thing; because I didn’t notice the perf hit when working with scatterer alone.

[LGhassen]

You'd need to render the IVA (or an iva depth pass) before the local camera, then make a copy/mask of its depth, merge it yourself into the main depth buffer setting all pixels where IVA is opaque to depth value 1.0, and pass that buffer to the clouds instead of the current depth buffer (there is a property that is set manually in the clouds renderer).
Personally I'm not interested in doing or supporting this as it needs a lot of changes but you should be able to do it if you want to using the above method.