A little age in developement

[Sirius]

Though, to be honest, if your computer is from the last 5 years or so, space trees will likely not help (or will make worse) the performance.

Ok, nevermind then.


I have another idea of what might cause lags, though...
Cyan's Ages are rarely laggy, even when Plasma has to draw a complex scene.
I did a few performance tests with my old computer (which was pretty good at lagging), comparing my materials to Cyan's, and found out setting "specularity" flag was a real performance-killer. Cyan almost never use this flag in their Ages, only for objects that are supposed to "reflect" ambient light (for instance, gems in the Laki'Ahn villa).

I had a look with PrpShop to Turtle Isle's PRP (I hope you don't mind), and found out most of your materials have this flag set. Maybe you could try to remove specularity from your materials and see if it is better ? (I think PyPRP disables it if you set "specularity" to 0. You can also try to do the same for "reflect" and "hard").

[dendwaler]

Yes, i remember i had this specularity setting when i builded turtle isle.
I never saw any difference with specularity enabled or not, i thought PyPRP1 ignored this setting.
(You can see the differece in Blenders internal renderview however.)
and never looked to it with PrpShop, but if you do so, thats ok with me.
We all only can benefit of everything that is discovered around what is causing lag.
But don't forget Turtle Isle is builded 2 years ago now.
You cannot longer compare what i builded then with what i do now.
I do more with less faces and build far more efficient.
For instance its only for a relative short period that i finally understand how to make a good UV map, without overlapping faces, and Uvmap those faces with the right size.
Its a matter of splitting objects on the right points, with seams marking the correct edges to unfold properly, and learning to see to unwrap from the right view.
Anyway specularity is not set on any object of the current age.
I think its just to many vertices with all the arches and flowers.
Despite everything is builded low poly.
I did not have time today, but I probably start with the visregions tomorrow.
I expect it will dramatically improve.

[Sirius]

Well, I just linked to Turtle Isle today, the specularity is not noticeable, except on the terrain (causing a light "shine" in some places). Anyway, if you didn't enable specularity, then that's ok on this side.
And I believe you did improve, Age modelling is a really good way to learn more and more...
Still, makes me wonder what else might slow down Uru... I don't know the exact number of polys of each Age but I think Ages as complex as Eder Kemo have roughly the same number of vertices and yet are displaying fine.

[Tsar Hoikas]

The specularity tip is probably good--that's runtime lighting, which may or may not be slow on some graphics cards (I don't know).

An old concern of mine is that Cyan's Max plugin marks all scene objects to have their persistent state thrown away (excluded) by default--PyPRP does not do this. It's possible we're seeing a bit of increased overhead from SDL. Though I somehow doubt it.

I suppose the real way to figure this out would be to convert problematic ages to MOUL and run a profiler (the in game tools and AMD Code Analyst) to see what's slow.

[dendwaler]

The terrain on turtle isle is not set to specularity
But..... the terrain is not set to shadeless! ...and it should be.
I have not done that , because there are many parts that needs some shading.
At that time i did not know how to do that.(set shaded vertex colors inside vertex paint)
I made a compromise there.

[Tsar Hoikas]

Just out of curiosity, what are the system specs of the user who is reporting lag in your new age? Your own specs are very close to my own desktop PC (ATI Radeon HD 4850, Core 2 Duo 3.0GHz, 4gb DDR2), so I expect we'll see basically the same performance.

[dendwaler]

A bit difficult to compare.

Its vm ware on a mac pro, with probably 512kb videomemory available.

[Tsar Hoikas]

Ouch. Yeah, I'd expect performance to be less than stellar on that (especially if the host system is an older Mac).

[dendwaler]

Yes,but hedid not have problems with my cathdral age ( stil unreleased) but at least twice as big!
But that one has well choosen visregions.

[Tsar Hoikas]

Right, in virtual machines, you can basically take what I said about modern graphics cards and throw it out the window. Running any game from a virtual machine is going to mean worse than usual graphics performance. You'll definitely see better performance in those cases by sacrificing some CPU time (this is going to be as fast as on the host machine) to save some GPU time. SpaceTrees would likely offer some benefit as well to ages that can be systematically subdivided. Perhaps backporting SpaceTree generation to PyPRP 1.6 would not be a wasted effort in this case.

I am somewhat hesitant to do that though since, in my profiling of CWE, I see that as the number of PRPs loaded increases, time spent on calculating visibility (with the SpaceTree) increases dramatically. As I said before, on modern PCs with Uru running directly on the main OS, the GPU can handle lots of data, so the time we're spending on the SpaceTree is essentially wasted.