Porting Pinball over to 0.9-38 / 2.7.3 - Frame Rate Problem

byte_wrench

08-01-2008 18:29:16

I finally bit the bullet and began porting over to the new NxOgre from the .4 version. I have most things working, but have a couple questions.

First, Is the best way to represent my pinball table collision model a TriangleMeshShape?
And, when I run my game in release mode, I can only get around 15fps whereas before it was normally around 55. I have scoured the forum but cannot find a solution.

Reckon I missing something silly?

betajaen

08-01-2008 18:59:28

Yep. It's a triangle mesh shape. As for your FPS issue, it is puzzling. I haven't experienced any sever differences in FPS between those generations of NxOgre.

Are you doing anything special?

byte_wrench

08-01-2008 19:06:03

Well, I am not sure if I am doing anything that would be out of the ordinary, I do notice that the cakebox demo does in fact get 60fps. So I will compare the differences and report back.

Thanks betajaen

byte_wrench

08-01-2008 19:55:25

I must be a dim-wit, but I cannot find any core differences in how the scene is created, etc.. I have even removed a bunch of the logic from the game, yet I still get around 17 fps. I have the Ogre profiler in place, but I cannot deduce anything from it. Anyone have any suggestions ?

betajaen

08-01-2008 20:36:35

Not of the top of my head. In theory 0.9 should be faster than 0.4, not only the code is neater; it uses a system of only updating the nodes of things that have moved, leaving out the ones who haven't. Plus some other speed improvements.

You said Cake gets 60fps, is that with both 0.9 and 0.4, or just 0.9?

byte_wrench

08-01-2008 23:41:44

That try was with .9, I have not opened up the .4 in a while, but the game was running good in .4 on a descent pc. I will try more in the morning. Guess I will start from scratch, adding to the Cake until it explodes or something like that. thanks again...

weak

09-01-2008 12:46:18

I have the Ogre profiler in place, but I cannot deduce anything from it.

that's the most promising approach imo. the 0.9 code is still blocking, so you should easily get the time needed for the simulation (and other parts of your app).
so try to find out which part of your app is eating the frame time and go from there...