Hello,
I couldn’t find any information about if there’s a post-transformation (vertex) cache in the MBX Lite chip (that one is used in the iPhone, as far as I know), and (if exists) about the size of the cache.
Is there any public documentation about it, that I didn’t find?
Or, alternatively, is there a way to measure the cache size?
Thanks,
Remage.
The reason that there is no documentation concerning this is because there isn’t a post-transformation cache in the MBX Lite architecture.
That makes sense, then… 
So, generally, to get the best performance, one should optimize the mesh for the shortest vertex stream, right?
Yes, that is one factor in optimization. Also, the MBX processess vertices in triangle strips which is why we recommend using meshes with their vertices in strip order.
In the case of the iPhone reducing vertex data size seems to make even more difference to performance and there are some threads on the Apple developer forum about this that I would suggest you look at (if you have access to them).Gordon2009-06-01 09:44:21
Remage wrote:
That makes sense, then... :)
So, generally, to get the best performance, one should optimize the mesh for the shortest vertex stream, right?
So, generally, to get the best performance, one should optimize the mesh for the shortest vertex stream, right?
Yeah ... In the case of the iPhone it is partially handled on the CPU side so any problems with vertex processing will first show up as a CPU bottleneck.
