max leaves

so i had that issue with map_max_faces. I still do, but when i compile the map with one room visgroup'd out (so its under the limit), the faces are only at 30%. Its the leaves, being at 98 (and going to about 105% when full) that are giving that error.

Are there any ways to lower leaf counts? I usually func_wall my detailings, but i found that those give way higher clipnode counts, which is what i was originally trying to lower by a lot. I went from about 140% clips to 98% my making all my func_wall etc into worldbrushes, sometimes also giving them better clips.

But now my leaves are over by 5%. I've alraedy removed tons of detail, i dont think i can handle any more. I've tried hints; i know their theory, use, etc, and have used pyramid hints in places where they absolutely reduce r_speeds by tons (using gl_wireframe 2). But, they almost always add about 2% leaves (each room goes from about 5% to about 7%). So now i'm sacrificing r_speeds to make the leaf max. Which is still at 5% over.

Is there any other way to do this?

I pretty much use hints to manually sculpt my vis leaves and often this can be dramatically more efficient in keeping the leaf count down than leaving the matter up to VBSP. Try using them to design the leaf layout almost completely - there are always going to be additional cuts by the compile tools but often one or two hint brushes can save half a dozen or so that the compilers would do otherwise. Do this throughout your (presumably large) level and it could save you hundreds. If you are using angled hints to cut visibility, consider just placing straight hints perpendicular to the walls at the furthest extents of the old angled one - won't be quite as effective for performance (though the area you miss out on improving is normally only reduced insignificantly in size) but keeping your leaf cuts along the 3 main axis wherever possible should help simplify matters and in turn keep that number down.

Other than that? Well it looks like your just gonna have to tread a fine line between too many func_details for your clipnode limit, and too many world brushes for your leaf limit. Choose carefully which are more effective at simplifying the BSP tree and which are more effective at simplifying the collision hull.

Good luck with it, sounds like this map is being a bit of a nightmare!

---

[img]http://card.mygamercard.net/sig/Default/reno84.png[/img]
Designer @ Haiku Interactive | ReNo-vation.net

It's Half-Life 1, Reno. I'd say the principles are the same, but in practice... :smile:

If you don't see any possibility of HINT brushes doing anything, I usually convert func_illusionary or func_wall details to world brushes to lower the leaf count. If you have any complex brush entity geometry in high traffic areas that would span multiple leaves, those are good candidates for converting to world. Keep in mind an old trick: raising it off the ground by 1 unit or moving it 1 unit away from a wall. 99% of players won't notice or won't care.

k hold the phone.

I thought that details we're supposed to be turned into func_walls etc to LOWER the leaf count, since they dont break up the bsp. A single empty room would presumably be 1 leaf, but the same room with a table in it (thats not a func_wall) would suddenly have to make leaves all around that... or so i thought. So shouldnt complex geometry be func'd to save tons of leaves?

and i thought the 1 unit trick was only for r_speeds. Wouldnt raising things by 1 unit actually raise the leaves by just a bit?

(Oops, 18 days ago!...)

When I hit the leaves limit, it's usually from having large func_walls in large highly visible areas - like a giant cliffside, for example. In that case, I find it usually helps to reduce the amount of func_walls and convert the brushes back to world, or delete the brushes altogether. Of course, maybe my situation/tactics should be the exception rather typical practice.

ive gotten past the leaves limit through simplification of brushwork. But, in all my dealings, i seem to have found that entities still cut leaves just like everything else. Test it out for yourself if you want.