fibanocci on 22/7/2013 at 16:21
You're pretty much near the cell-limit (32760). Errors like that occur, if you're close to the limit. You could use area brushes to optimize parts of the map.
Recently I made a test-FM with 32752 cells. Portalizing was fine, optimize failed as well.
BTW, I had to remove some cylinders to get to that point.
R Soul on 22/7/2013 at 17:27
I've had that error too. It was caused by lots of pointy roofs. Every mission has its own issues, so if you don't have enough pyramids to eliminate, focus on cylinders, dodecahedrons and complex terrain.
fibanocci on 4/8/2013 at 13:23
It's hard to tell, why and when optimization doesn't work any more.
I recently tested the limits. In a map with 20133 WR-cells, optimization was fine. It has hundreds of dodecahedrons, cylinders and stuff. Some more dodecahedrons, and optimization failed. Same message as you got. Portalize was fine until I reached the limit.
So, 15809 WR-cells isn't the limit for optimization. You could try to replace some dodecahedrons with objects (this will save many cells) or optimize parts of the map using area brushes. Some brushes might be bad ones, I don't know.
EmperorSteele on 4/8/2013 at 15:24
Have you tried "hovering" any terrain? It will look weird if anyone gets too close, so save this for background or far-off elements, but basically there's an old trick where if you have, say, a column that goes from floor to ceiling, you can shave off like .5 or .25 from the total size so it doesn't quite touch the ceiling and/or floor. This, for some reason*, reduces poly/cell counts and scene complexity.
*iirc, if you have a blank rectangular room, Thief will render a couple polygons for each surface (let's say its 2 polys per surface, for a total of 12 (only 10 which the player can see at any given time). Conventional thinking would suggest that a column wouldn't do anything besides add a few polys of its own. This isn't the case. When two planes meet, the Dark Engine uses extra polys to render the brushes by sub-dividing the brush into several smaller brushes. But if you hover the column, the dark engine will render the original 12 polys, plus whatever the column is made up of, without adding extra brushes. It's hard to explain in words, but there should be an old design tech document in your dromed docs folder explaining the situation.
Sensut on 4/8/2013 at 17:36
I know what you say, but unfortunately I have no any places where I can apply this trick, since it's a city mission only with houses and rooms inside. The most complex part was a small cave with those dodecahedrons, and I have some air filled archways, that's all. The architecture is just suitable for me, only annoying that I see nothing there in order to justify this error.
Soul Tear on 14/8/2013 at 12:57
More than two weeks I am fighting with this problem but I have no progress.
I've had more than 65000 cells, and I got the message "too many cells." I deleted a giant air brush (together with all buildings) - and the number of cells decreased to 62 000. Optimization is OK.
Then I loaded the source file and I began to delete each brush successively. So I could see how the cell limit is reduced.
However, when I reached 62000 - optimization is not working! I reduced cell limit to 61 - NOT worked, 60 - as well. Then I deleted a big building - it reduced the limit to 59000 and optimization is OK.
I loaded the source file again and began to delete each brush. I got to 59000 and then to 58000 cells - but optimization does not work. WHAT IS IT? Dark Engine plays tricks on me...
I checked in other place with the different brushes.
I deleted the stair at once (located at an angle not a multiple of 90 or 22.5 degrees) - Optimizing is OK. I deleted the stair successively (in the end all the stairs deleted) - does not work.
I deleted all the pyramids, cylinders, and dodecahedrons which I added recently - not worked. I even deleted all of that I built up to the moment when the full optimization is worked - NOT worked...
I just hope my broken English will not be able to prevent you understand my problem. I have no idea what I to do.
Apparently my mission is too complicated or improperly constructed. Until I see only one way: carve up into two parts. :(
Sensut on 14/8/2013 at 15:48
Same here, but I don't know what's going on. My amount of cells is ridiculously low, less than in old dark (!), but still I have this damn error when I try to add 1-2 brushes of any types. I need to delete some not too important brushes from somewhere in order to be able to create new ones, although I have only 3600 brushes total. Finally I gave up the architecture, but one thing for sure: this is a strange bug again and not a cell or any other limit problem, at least in my case :tsktsk: .
fibanocci on 15/8/2013 at 11:07
I tested some more and compared FMs.
This is a test mission. It's full of cylinders and dodecahedrons. Optimize is fine
Code:
--------------- csgmerge ---------------
Epsilon is 1.e-004
File has 19813 cells.
Allocations: 19814
Min: -1541.1672 -1929.3213 -921.875
Max: 1205.7419 1836.1686 520.25
Input polys: 63609
Unique planes: 9194
Unique render aspects: 7849
Unique vertices: 78641
Allocations: 146908
Merges made: 32989
Merges with internal colinear points: 0
0 colinear points deleted
Allocations: 90124
Discarded degenerate polys: 62
Current allocations: 19960
Max allocations: 146909
Releasing remaining allocs: 19960
------------- csgmerge end -------------
I can add stuff until I get to this size. Optimise is still fine
Code:
--------------- csgmerge ---------------
Epsilon is 1.e-004
File has 20214 cells.
Allocations: 20215
Min: -1745.4189 -1623.1268 -921.875
Max: 1192.9919 1836.1686 520.25
Input polys: 64746
Unique planes: 9334
Unique render aspects: 7992
Unique vertices: 80151
Allocations: 149579
Merges made: 33276
Merges with internal colinear points: 0
0 colinear points deleted
Allocations: 92361
Totally coplanar... ick.
Discarded degenerate polys: 64
Current allocations: 21330
Max allocations: 149580
Releasing remaining allocs: 21330
Adding one more brush, and I get this
Code:
--------------- csgmerge ---------------
Epsilon is 1.e-004
File has 20511 cells.
Allocations: 20512
Min: -1745.4189 -1866.1963 -921.875
Max: 1192.9919 1836.1686 520.25
ERROR: Increase MAX_GLOBAL_RENDER
Current allocations: 144456
Max allocations: 144456
Releasing remaining allocs: 144456
------------- csgmerge end -------------
WARNING: Optimization failed, level remains unoptimized!
Looks like, I hit the limit. Which limit? It's not the WR-cell limit. I think, it's another one.
I loaded Beltzers "The Favour" into DromEd. It's a pretty big mission.
Beltzer: The Favour
Code:
--------------- csgmerge ---------------
Epsilon is 1.e-004
File has 23100 cells.
Allocations: 23101
Min: -104.5 -206.25 352.5
Max: 670.75 683.625 559.01843
Input polys: 68923
Unique planes: 5220
Unique render aspects: 2755
Unique vertices: 84861
Allocations: 160847
Merges made: 40178
Merges with internal colinear points: 0
0 colinear points deleted
Allocations: 85711
I've put tons of multibrushes (cylinders, dodecahedrons...) into Beltzers map until I reached this
Code:
--------------- csgmerge ---------------
Epsilon is 1.e-004
File has 30608 cells.
Allocations: 30609
Min: -905.63165 -976.36102 352.5
Max: 922.14026 889.33569 689.78076
Input polys: 98029
Unique planes: 10900
Unique render aspects: 7349
Unique vertices: 121984
Allocations: 226531
Merges made: 53587
Merges with internal colinear points: 0
0 colinear points deleted
Allocations: 130257
Totally coplanar... ick.
Totally coplanar... ick.
Totally coplanar... ick.
Totally coplanar... ick.
Discarded degenerate polys: 68
Current allocations: 19172
Max allocations: 226532
Releasing remaining allocs: 19172
optimise is still fine. This is strange. Comparing the data, I can see one big difference to my testing map. Beltzers map only has 2755 Unique render aspects. And I can put tons of stuff into Beltzers map. I hit the limit at 7992 in the testing map. I can't add a single brush without messing up optimization. So I guess, this is the limit for MAX_GLOBAL_RENDER.
I can even put more stuff in Beltzers map, but optimize doesn't work anymore
Code:
--------------- csgmerge ---------------
Epsilon is 1.e-004
File has 30851 cells.
Allocations: 30852
Min: -905.63165 -976.36102 352.5
Max: 932.89026 889.33569 689.78076
Input polys: 98984
Unique planes: 11083
Unique render aspects: 7533
Unique vertices: 123233
Allocations: 228682
Merges made: 54023
Merges with internal colinear points: 0
0 colinear points deleted
Allocations: 131719
Totally coplanar... ick.
Totally coplanar... ick.
Totally coplanar... ick.
Totally coplanar... ick.
ERROR: Increase MAX_GLOBAL_POINTSCurrent allocations: 55505
Max allocations: 228683
Releasing remaining allocs: 55505
------------- csgmerge end -------------
WARNING: Optimization failed, level remains unoptimized!
Uh, I hit another limit. This time MAX_GLOBAL_POINTS. But I did not hit the limit for MAX_GLOBAL_RENDER.
How did Beltzer built this mission? I guess, this is the clue for creating a very big mission.
We could compare some more "csmerges", so we can see, whats going on.