I was just testing PGMM (or shall I say geoMipTerrain), and when testing the auto-flattening function I stumbled upon this analyze() message:
264 vertices are unreferenced by any GeomPrimitives.
1007 GeomPrimitive arrays are redundant, wasting 141K.
I’ve checked the first line (indeed at some places it creates vertices where it does not need to. But, regarding my current way of calculating vertex connections, this is almost impossible to fix without adding lots of lots of extra calculations.)
Is this serious? Will this slow down my terrain and do I need to add a loop which fixes that, or will Panda take care of it?
And the second error, about the redundant arrays, is that directly related to the first, or has this a different cause?
Thanks in advance,
i dont thinkg that wasting 264 vertices is a problem on todays hardware.
about the second error i dont kow.
but neither one is falged as “warning” or “error” . so i assume it’s ok to leave it the way it is.
Right, neither of these is a major issue worth losing sleep over. I’ll explain what they mean, though.
In the first problem, you are sending useless vertices to the graphics card. This consumes bandwidth and wastes framebuffer memory. Some graphics cards will needlessly transform the unused vertices, wasting GPU time; others won’t. Of course, you only have 264 wasted vertices, which is nothing at all. If you had 264,000 wasted vertices, you would need to deal with it.
In the second problem, Panda noticed that you have 1007 identical GeomPrimitive objects, with different pointers. Instead of creating 1007 identical GeomPrimitive objects, you could have just created one, and used it 1007 times. Again, this consumes bandwidth and wastes framebuffer memory–but only a tiny bit, so it’s probably not worth fixing. (On the other hand, if you’re creating all of these in Python, you might save some significant Python time by eliminating 1006 addVertex loops.)
Thanks for your replies. I now know what’s causing it. I’ll just can’t split it up in multiple GeomPrimitives because the chunks are all different Geoms.
It’s C++, so it wouldn’t be a major speed issue I guess.