maz0r wrote:I didnt focus on T-Junctions (or what they were called) in this example and PVS isnt important either. So I wonder which solution will result in less tris count.
What.
Tris count is directly affected by both PVS and t-junctions (although there are more variables involved in the latter--e.g. whether faces are or can be merged, etc.), so they're both important to consider when looking for lower tris.
[/nitpicking]
That being said, I would almost always use the first method over the second (and miter the corners) simply because of ease of working with the resultant brushes. Whenever you have a single brush that contributes multiple faces to multiple walls/rooms/hallways it can quickly become a clumsy process when making alterations or doing any sort of detailing.
One simple example: Suppose you wanted to chop up the hallway on z and get some layering of textures in the hallway only; do that in your second example, and you would end up slicing up the wall of the room as well, or you would have to miter the brush. Once you start mitering up those hallway walls, you'll probably end up with about the same number of brushes as your first example anyway.
With triscounts quickly climbing into the tens or hundreds of thousands for a typical scene in recent games, fussiness regarding very precise and exacting construction techniques
purely as a nod to minor triscount reduction is becoming more and more irrelevent, and construction with focus on flexibility and easy modification (and thus less rebuilding and less wasted time) is more important.
The wasted time factor isn't necessarily a big one if you're just doing it for a hobby and you don't really care how long it takes, but even when that's the case, it's still nice to do something in two nights instead of three.
I beat the internet; the end guy is hard.
