[Q] Proxies?

[valerostudio]

Does Skatter work with VRay proxies? Should we not proxy our components for skattering?

[jiminy-billy-bob]

With Vray you can use whatever you want. Proxies or not.
Though I think proxies will perform better.

[andybot]

I've used proxies in my tests. Doesn't seem to make a whole lot of difference in vray loading time though. I imagine that's an issue on vray's end. I see one of the changes in the recent vray 3.1 (for Max) update is that duplicate proxies are instanced/ treated as one, I forget the exact wording (I'm assuming it means it hasn't been the case thus far - including 2.0)

[jiminy-billy-bob]

I've used proxies in my tests. Doesn't seem to make a whole lot of difference in vray loading time though. I imagine that's an issue on vray's end.

Surely there is something weird with Vray's parsing time in general. When we look at the parsing time of Thea or Octane for the same scene, they take a tiny fraction of Vray's parsing time. So there is room for optimization.
I know Devin and his colleagues already have reduced it in the past, put I hope some day they'll be able to drastically decrease it. Maybe this needs to be completely rewritten, in which case it will take time!

I see one of the changes in the recent vray 3.1 (for Max) update is that duplicate proxies are instanced/ treated as one, I forget the exact wording (I'm assuming it means it hasn't been the case thus far - including 2.0)

I think this is actually related to max's instancing system.
When you copy an object you can choose to make a separate copy, or an instance (like SU's components). So what I understand here is that now Vray automatically recognize similar copies of proxy meshs and treat them as instances. This probably saves RAM.

[andybot]

I think this is actually related to max's instancing system.
When you copy an object you can choose to make a separate copy, or an instance (like SU's components). So what I understand here is that now Vray automatically recognize similar copies of proxy meshs and treat them as instances. This probably saves RAM.

Ah, so that relates to Max only. It [copy versus instance] was already correct in vray for Blender, so I was wondering about the change.

Yeah, I'm hoping for something drastically different for vfs 3.0 Parsing in vfs 2.0 is better, but still has a long way to go.

[valerostudio]

I gotta agree on the parsing time thing. I used Maxwell a little and parsing time was at least a tenth or less of that of V-Ray. From what I am told it has a lot to do with the nature of SketchUp and we shouldn't expect much to change in that department.

[valerostudio]

So i made a skatter group with a proxy, checked Render, it worked. Then I went to Skatter render list and hit Edit, modified, click render. blank.

[jiminy-billy-bob]

From what I am told it has a lot to do with the nature of SketchUp and we shouldn't expect much to change in that department.

This is also what I thought before. But a few weeks ago I tried benchmarking what happens when we launch a Vray render. This is what I found:

It seems that ~70% of the parsing time actually happens in the doParse method in archiveRead.py. I don't really understand what happens in there.
But it to me seems that reading the model and writing the xml is relatively fast, but parsing the xml file is slow. Which sound weird.

Another ~20% happens in addAssetsToRenderer in vrayrender.py

I don't know if you have already looked at all this to optimize the parsing time, but if you've not I think it could be worth taking a look.

So if this is indeed correct, the way Sketchup poorly exposes data is not the big issue here. If it was, why would the other engines be so fast at parsing the scene?


(I don't know if Devin follows this thread, it would be intersting to get his say on this)

---

So i made a skatter group with a proxy, checked Render, it worked. Then I went to Skatter render list and hit Edit, modified, click render. blank.

I think I fixed it in the next version.