Projected shadow performance (not physics)

[shadmar]

I was working on my joint editor and built a simple car to test. (it can go back and forward using the UP/DOWN keys, WASD for camera)

 

When wheels hit the ground I go from 400 to 18 fps, which seem terrible.

When the wheels are at rest it goes up to 400 ish again.

 

Here is my full test project including the car :

[Rick]

When I try to import I get no project file found. Are you doing Project Manager->Export when you make these project files because that will be the easiest way to share them.

 

[EDIT] NVM got it loaded. Looks like we can't load the zip file but have to select the werk file. Lame! Will see what I get.

 

On my macbook air I get about 80 fps when I'm moving and 160 fps when idle.

 

I still think the physics in LE 3 is much worse performance wise than LE 2, but it's so hard to prove this to Josh. I've never had physics kill performance in LE 2.

[shadmar]

Ah sorry here is it exported version (added to top post)

[YouGroove]

From 70 to 90 or 95 FPS on my big PC tower (hard to see as numbers are too much varying, you should siplay one FPS number each 0.5 second) The scene is empty, FPS could be better no ?

 

For the physics, the car is not running , but running in place and jumping,at final walking slowly.

Would need real physic car system instead.

 

Anyway great design and model car.

[beo6]

can confirm it. It goes from around 900 to around 80 fps when i have vertical sync completely deactivated in my driver settings.

 

When i set it to adaptive it stays at 60. (which is obvious)

[YouGroove]

Bullet physic Kickstarter program ?

[beo6]

i would be with you if it would be sure that any different physic engine helps. I however think the issue must be somwhere else since there are other games using newton as physic engine that do not seem to suffer from this.

 

//edit:

just looked into my game.

I get 1000 fps on my release compiled game (looks like the max limit that it displays).

 

depending on how fast i move with my physic controlled ball it drops to under 100 fps. (around 500-300 if i move slow)

 

I am not sure how extreme the speed of physic enabled objects should affect the framerate but it looks like quite a big drop.

 

//edit2: just to be sure. i set world->SetPhysicsDetail(1); in my game just like shadmar does. and 1 should be the least expensive setting here i think.

Edited September 18, 2013 by beo6

[Josh]

Will check it out, thanks.

[YouGroove]

Is shadmar code using NEwton car physics or it's own car physics system ? this can make all the difference.

I have 45 FPS running Nexwton 3 playground game and lauching the race track demo : 1 car, 1 circuit

[Josh]

It isn't slow due to physics. It's slow due to the projected shadow rendering. Though that still seems excessive.

[shadmar]

Ah physics performance was much better without shadows, thanks. The shadow performance is quite an impact then.

[Josh]

Like I said, it seems rather high.

[neseir]

Had the same experience with some simple Box2D tests. Suspected the physics part but it turned out to be the shadow performance that slowed down the simulation.

 

//Eirik

[beo6]

indeed. It is the dynamic shadow. It gets a lot faster on android too when i disable it.

[Admin]

Since we are focusing on dynamic lighting in the future this report will be closed. Thank you.