Hi! We're running bullet in our game and have experienced a strange performance issue. After a minute or so of running our app, the following happens with no apparent reason:
RayTest:0.155746 [ms]
RayTest:0.153939
RayTest:0.156107
The thread 'Win32 Thread' (0x179c) has exited with code 0 (0x0).
RayTest:0.010118
RayTest:0.010118
RayTest:0.010479
These are results of a unique raytest each frame. All of a sudden, a thread exits and the raytest increases its performance radically.
What's going on?
Thanks!
RayTest strange performance change
-
Nene
- Posts: 4
- Joined: Mon Apr 26, 2010 11:08 pm
Re: RayTest strange performance change
I found out that the culprit of the performance sudden change is the btAlignedObjectAray::resize that is performed each frame inside raytest, although I still don't understand the thread issue.. I've seen bullet overloads the new operator, but I haven't gone into detail yet..
-
B_old
- Posts: 79
- Joined: Tue Sep 26, 2006 1:10 pm
Re: RayTest strange performance change
Hi,
I just want to mention, that I experienced similar behavior. In my case I was doing a couple of hundred raytests per frame, and the performance went down considerably after a while. Waiting a little longer might improve performance again.
Also, it would be really cool it the raytests could be threadsafe. Currently I made a small change to how compound shapes are being tested because testing against such shapes concurrently can lead to crashes sometimes.
I just want to mention, that I experienced similar behavior. In my case I was doing a couple of hundred raytests per frame, and the performance went down considerably after a while. Waiting a little longer might improve performance again.
Also, it would be really cool it the raytests could be threadsafe. Currently I made a small change to how compound shapes are being tested because testing against such shapes concurrently can lead to crashes sometimes.