High poly thing
-
What particular aspect of SU is it that makes it unable to handle models with lots of polys?
-
Inferencing - every edge must be kept track of. OpenGL limitations.
I still remember the time when running Osnaps (basically the same thing) caused similar problems with AutoCad, especially in the 286 and 386 hardware of the time. Then Autodesk found a way to rewrite thr snapping, and it started to work like a breeze.
Anssi
-
@anssi said:
Inferencing - every edge must be kept track of. OpenGL limitations...
Wow, I have never thought of that (still a newbie in the industry
)!In this case, a simple toggle inferencing on/off with a shortcut key could bve indeed useful (at least while trying to orbit large models).
-
Hmmm, thats interestign, like you gai id always thought it was just an inefficient method of drawing the geometry up that caused the problems, didnt think of the inferencing thing, although i suppose it makes sense as SU needs to find the end, mid and perpendicular of evey line on screen. Cheers for the quick answer anssi.
-
Would enabling SketchUp to fully utilise DirectX improve matters?
I have noticed that when using 3D Studio MAX that choosing to use DirectX as opposed to the other graphic modes available including OpenGL, the feedback, screen detail and overall perfomance seems superior.
I'm not a Microsoft fan boy by any means (see my other post complaing about animation codecs)and I understand people wanting to opt for "open source" software if possible.
But if Microsofts solution is better I would want to take advantage of that fact.Does anybody know which is best?
OpenGL or DirectX?Regards
Mr S -
Direct X definitely.
I don't know why OpenGL is still used. Probably to prevent riots from the Mac fanboys?
-
@kwistenbiebel said:
Direct X definitely.
I don't know why OpenGL is still used. Probably to prevent riots from the Mac fanboys?They can stick with OpenGL - and the Google should only develop the Windows version
-
What about the new openGL 2.1?
-
Don't know about that...but clearly the graphics is not the most important bottleneck in current SU I believe.
I think it also has to do with the constant updating of info spread all over the different panels.
For instance, didn't you ever notice the 'outliner' panel going bananas when editing ?
Also, the slowness of the material browser, etc....Somehow, some update processes should be unhooked from the viewport refresh imo.
Or else, spreading them over the different cores naturally (64 bit, multicore, etc..), the thing we wine about almost constantly... -
@unknownuser said:
Don't know about that...but clearly the graphics is not the most important bottleneck in current SU I believe.
I think you're broadly right. There seems to be a lot of work going on under the hood to get to the point of drawing (which incidentally is why the OpenGL/DX question is a bit of a red herring). BTW I can't believe inference is more than a few percent of frametime so I'd discount that too.
However, if you have complex models it does make a BIG difference in refresh rate to not draw shadows and not draw (thick) profile edges.
Adam
-
Adam, what makes you think inferencing does take much computing power? i would have tohugh its quite computationally intensive to find the mid, end and direction of very line in an SU model.
-
@remus said:
Adam, what makes you think inferencing does take much computing power? i would have tohugh its quite computationally intensive to find the mid, end and direction of very line in an SU model.
It would, so you don't do that!
Typically one creates some spatial structure that can narrow down the potential candidates very quickly. LightUp uses kd-trees to do it, I suspect Mr.Google will have something similar. The end result is you actually find mid, end and direction for only handful of lines.
Adam
Advertisement