While compiling with vvis on it causes vrad to crash, but if I turn off vvis during compile vrad will work correctly. Anyone familiar with this problem?
This is an optimization issue, not an issue with VVIS. When you don’t run VVIS, no vis information is created, so VRAD does direct lighting only. There should be a line in VRAD that lets you know about this.
When you do run VVIS, vis information is created, and therefore, vis lighting calculation may be done. This includes bounces and BuildVisLeafs and bounces. So when you do run BuildVisLeafs, it will sometimes run out of RAM and crash, the reason being that your lightmap scales are too low. So I suggest first optimizing your visleafs so that you won’t have any reason to not run VVIS, then optimizing your lightmap scales by raising them on faces with little to no shadow detail.
tl;dr - optimize
So there’s a memory issue… Does Hammer, vvis.exe, vrad.exe etc. have 64bit support?
It should never, ever hit your RAM limit… but if you do have an extremely detailed map that takes up the entire grid, there is a way to make VRAD LAA so that it can use all your RAM on a 64-bit OS. It would be pointless to do this for VVIS or VBSP, or hammer, etc.
It’s not a RAM issue, it’s an optimization issue. Stop trying to shy away from optimization…
Wow, that looks mad. What is that map exactly?
WHOA, that map is huge. Remember that large areas where every leaf can see every other leaf can be func_viscluster’d without any performance loss.