The problem is that, randomly, some data in a VAR is corrupt. I use a
bunch of VAOs so I can have a bunch on VARs.
This problem has only been found to occur on dual CPU G4s. The first
is a 1.25G with a Radeon 8500 running 10.2.8. If a GeForce3 is swapped
in, the problem don't happen. However, the second is a 1G with a
GeForce4 and 10.3.8. I believe a Radeon was tried in there as well. If
the second CPU is disabled in the firmware, the problem does not occur.
This has never happened on my dual G5.
I do not explicitly spawn a thread anywhere during these tests.
It seems that the closer towards the end of the loading sequence an
object is, the more likely it is to be corrupted.
A workaround has been found. If I delay enabling the VAR until after
the first render, the problem has not yet occurred. This implies that
my data is not corrupted, but something is corrupting it on upload to
VRAM.
Anyone have any clue what could be causing a (seemingly)
multiprocessing VAR upload problem in a single-threaded app?
Thanks,
Jeff
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Mac-opengl mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/mac-opengl/email@hidden
