map breaks engine

map breaks engine

Re: map breaks engine Posted by haymaker on Wed Sep 17th 2008 at 11:25pm
haymaker
439 posts
Posted 2008-09-17 11:25pm
haymaker
member
439 posts 921 snarkmarks Registered: Apr 1st 2007 Location: CAN
I figured I'd make a last-minute contribution to the comp map effort, and take inspiration from a component of the Hadron contraption. Here is a shot of the Atlas magnetic device

http://www.phy.duke.edu/research/hep/images/atlas_magnet_600x391.jpg

OK looks a pile of repetition more than anything. Here is an early Hammer shot

[img=http://img522.imageshack.us/img522/8984/colllt5.th.jpg]

Going to compile, it's broken utterly. Yes everything is func_detail, no leak

[img=http://img522.imageshack.us/img522/7013/collider20000db2.th.jpg]

There are more shots but it's basically an HOM nightmare. The only way I could get it to run was by eliminating most of the detail contents:

[img=http://img237.imageshack.us/img237/2759/collider0000mc3.th.jpg]

I have to conclude that there are just too many vertices too close together in the same visleaf...they are all on grid...I've seen many maps more detailed than this run fine :(
Re: map breaks engine Posted by Le Chief on Wed Sep 17th 2008 at 11:55pm
Le Chief
2605 posts
Posted 2008-09-17 11:55pm
Le Chief
member
2605 posts 937 snarkmarks Registered: Jul 28th 2006 Location: Sydney, Australia
Whoa, nice stuff man.

Anyways, if your getting the HOM effect, you have some sort of error in the map. As far as I know, the only effect of having alot of vertices is lag. I can take a look at the vmf file if you like.

Anyway, there is a few trial and error tests you can do such as hide all the pipes, compile and retrexture the whole map temporarily to rule out a texture issue causing the effect test to rule out problems.

Oh and when you say everything is func_detail, I hope that dosen't include the maps hull :P.
Aaron's Stuff
Re: map breaks engine Posted by haymaker on Thu Sep 18th 2008 at 1:50pm
haymaker
439 posts
Posted 2008-09-18 1:50pm
haymaker
member
439 posts 921 snarkmarks Registered: Apr 1st 2007 Location: CAN
Thanks Aaron, yeah I've tried all the tests I could think of...converting to dev tex's makes no difference, the only thing that works is eliminating at least 75% of those rings lol.

The HOM in the shot was taken without VRAD compile, but it does the same thing, except that during full compile VRAD took about 45 minutes on building visleafs. It was choking on something, so on that old compile log I looked and saw that max_map vertices was about 17%, as was brush count.

I'm gonna try hinting each ring-set off by itself, but I think this idea is toast for the comp because of how much of max_everything it uses.