Hi,
I’m loading a model with about 50.000 or more triangles into my JOGL- application. Where under Windows this is no problem, under Linux the memory usage of the JVM grows and grows (up to 500 MB). I’ve set the heap size of the JVM to 128MB, so it seems that the memory leak is in the native part of JOGL. Any ideas what this could be or how to trace it ?
I’m useing RedHat 7.2 with Mesa 5.0.1 on 1 GHz / 512 MB RAM machine without any graphics hardware acceleration.
Best Regards,
gerd
No ideas. JOGL doesn’t do any management of storage for data that is passed down to OpenGL, so unless there is a memory leak in the window system binding (which is possible, actually there are a couple of known memory leaks, but they are tiny) I think it’s unlikely to be in JOGL itself. You may want to try some of the commercial profilers which can give you at least a Java heap view and possibly even a native heap view. I think NuMega / CompuWare used to have pretty good C heap analysis tools.
Blowfish, which JVM are you using on Linux? I’m running RedHat 9 and generally don’t run across much trouble - but then again I don’t do much in Linux.
I remembered last night that the last time I saw a problem like this it was because I had assigned the GL object to a data member instead of passing it around on the stack and was accidentally calling OpenGL functions from the AWT event queue thread during key presses. Check your program to make sure the multithreaded behavior seems correct. See the user guide for some more details.
Sorry, it was my failure. Because of an error in my scene loader there I had not only 50.000 triangles but 500.000 up to 1.000.000 !