Based on the discussion in the other thread http://forums.chaosgroup.com/showthr...218#post721218 I went over all render settings, and marked out those that I would expect to update automatically during IPR.
Now, I do realize some of these are not possible to be updated on the fly during IPR, as they require complete scene re-parse and/or restart of rendering session. And that's fine. In such case, I would expect equivalent of hitting refresh button... complete restart of IPR and re-translation of everything. But it should happen automatically, so that if I switch something, I am sure I see the effect it has on my scene (at the cost of waiting for scene to re-translate).
Reasoning behind some of the more controversial ones:
Light solver (adaptive/probabilistic lights):
These have some slight visual differences between each other. It would be good to have that reflected during IPR too. While slightly, they do affect look of the scene.
Show GI only:
Not sure about this one, but it could cause a confusion if someone flipped it on during IPR, and then nothing happened, but nasty surprise would come when doing production render. It would be nice to reflect the change immediately, even at the cost of complete IPR restart.
3ds Max photometric scale and legacy sun/sky/camera:
These both have visual impact on the scene
Use local subdivs:
I am not sure here... should we always force automatic subdivs for IPR? I don't know... they are mostly superior... I am not sure if there are cases where manual subdivs could significantly alter the look of the scene, aside from noise distribution. But then again, someone who still uses manual subdivs may want to tweak them in IPR... so I guess that should be reflected too. Balancing of subdivs is a prime example of repetitive F9 key mashing process.
Affect background:
I still wish this worked with progressive... but it doesnt... and IPR will probably never be bucket... But i still marked it, in case it'd ever start to work with progressive sampler.
Caustics:
Photon emission is often quite long, and photon mapping in this case can not be progressive (i think)... But not including it would break WYSIWYG... So this is a tough one... A choice between allowing something to slow down interactivity to the point of not being interactive at all, or preventing loss of interactivity at the cost of breaking WYSIWYG.
Dyn. mem limit and Geometry:
These are not that important, but it could be for example nice to be able to flip between auto and static on the fly and observe if there's performance difference. From my older memories, static geometry caused quite a boost in some types of scenes.
Optimized atmos. evaluation:
As far as I understand, this could, under some circumstances, have an effect on scene appearance... ?
Now, I do realize some of these are not possible to be updated on the fly during IPR, as they require complete scene re-parse and/or restart of rendering session. And that's fine. In such case, I would expect equivalent of hitting refresh button... complete restart of IPR and re-translation of everything. But it should happen automatically, so that if I switch something, I am sure I see the effect it has on my scene (at the cost of waiting for scene to re-translate).
Reasoning behind some of the more controversial ones:
Light solver (adaptive/probabilistic lights):
These have some slight visual differences between each other. It would be good to have that reflected during IPR too. While slightly, they do affect look of the scene.
Show GI only:
Not sure about this one, but it could cause a confusion if someone flipped it on during IPR, and then nothing happened, but nasty surprise would come when doing production render. It would be nice to reflect the change immediately, even at the cost of complete IPR restart.
3ds Max photometric scale and legacy sun/sky/camera:
These both have visual impact on the scene
Use local subdivs:
I am not sure here... should we always force automatic subdivs for IPR? I don't know... they are mostly superior... I am not sure if there are cases where manual subdivs could significantly alter the look of the scene, aside from noise distribution. But then again, someone who still uses manual subdivs may want to tweak them in IPR... so I guess that should be reflected too. Balancing of subdivs is a prime example of repetitive F9 key mashing process.
Affect background:
I still wish this worked with progressive... but it doesnt... and IPR will probably never be bucket... But i still marked it, in case it'd ever start to work with progressive sampler.
Caustics:
Photon emission is often quite long, and photon mapping in this case can not be progressive (i think)... But not including it would break WYSIWYG... So this is a tough one... A choice between allowing something to slow down interactivity to the point of not being interactive at all, or preventing loss of interactivity at the cost of breaking WYSIWYG.
Dyn. mem limit and Geometry:
These are not that important, but it could be for example nice to be able to flip between auto and static on the fly and observe if there's performance difference. From my older memories, static geometry caused quite a boost in some types of scenes.
Optimized atmos. evaluation:
As far as I understand, this could, under some circumstances, have an effect on scene appearance... ?
Comment