TTLG|Jukebox|Thief|Bioshock|System Shock|Deus Ex|Mobile
Results 1 to 12 of 12

Thread: Saving Cells

  1. #1
    Member
    Registered: Dec 2001
    Location: OldDark Detox Clinic

    Saving Cells

    When I first was playing around with this mission back in the mid 2000's, I figured I was just about finished for a few reasons.

    The biggest reason was I was having to select several area brushes and activate them in order to be able to compute pathfinding. I was making area brushes just above AI's heads, and making sure I was only activating anywhere an AI should be able to go. Otherwise it wouldn't work, and no AI would move.

    I've worked on it sporadically here and there since then, when I'm not busy with wildlife photography, as the mood strikes me. Nowadays, NewDark can pathfind everything without fail, even though I've double the size of the mission at least, since then.

    I've used Christine's trick of renumbering larger brushes to earlier time stamps, and that has dropped the cell count. Even lately, after more building, that has worked to some degree once again.

    But now, even with the time stamp tricks, I am at 21801 cells. I am only at 6020 terrain brushes, so I have more breathing room there than the impending cell count.

    Okay that was a long setup, but here is the question.

    Is it worth the work to go back into the first part (somewhere between third and half) of what I built between 2006 and 2009, and change the brushes around. Namely, those that I built the old way: a room with mouldings and wainscoting being built with 3 or four air brushes.

    Will I save a considerable number of cells going back and making those rooms one air brush, and putting 0.25 thick wainscotings flush with the walls, but a later timestamp?

    There would be quite a few such rooms to go back to, but if it would help lower the cell count to allow me to continue to expand the world, I'd be willing to do it.

  2. #2
    Member
    Registered: Apr 2011
    first: you still have a smidgen over 10,000 cells of headroom, so it seems a little premature to be worrying.

    second: the rooms you describe, with (presumably protruding) mouldings and wainscoting, will not save any cells from rebuilding them with a different brush arrangement; such a room is a minimum of 3 cells already (one for each different width).

    as a general rule, cube brushes that are unrotated or rotated only in multiples of 90° will not produce excess cells, because all six of their faces are “doing work”, i.e. defining actual geometry. excess cells come from two sources:

    (a) brush faces cutting other distant cells up as a result of the innate work that portalizing/optimizing involves; the advice to adjust brush timing to put larger brushes first can help with this, but a fair amount of excess cells is inevitable.

    (b) brush faces that arent doing any work at all, e.g. the classic example of the bottom half of an air cylinder that is making an arch: the bottom faces are air cutting air, and not defining any geometry. excess cells from these can be eliminated by eliminating such faces, such as by rebuilding with wedges so only the top half of the arch is defined.

    note that a brush face that is exactly parallel with another brush face that is doing work, but that is not itself doing work, is benign; they share the same “cutting plane”, so no additional cuts are made.

  3. #3
    Member
    Registered: Dec 2001
    Location: OldDark Detox Clinic
    Quote Originally Posted by vfig View Post
    first: you still have a smidgen over 10,000 cells of headroom, so it seems a little premature to be worrying.

    second: the rooms you describe, with (presumably protruding) mouldings and wainscoting, will not save any cells from rebuilding them with a different brush arrangement; such a room is a minimum of 3 cells already (one for each different width).

    So, for example, if I take an old room and make all three air brushes the same depth and width, the cell count would naturally go down? But, would making one air brush and then four solids for wainscoting which do not protude, and then four more solid brushes for ceiling mouldings, which also don't protrude, would both methods result in the same cell count?

    as a general rule, cube brushes that are unrotated or rotated only in multiples of 90° will not produce excess cells, because all six of their faces are “doing work”, i.e. defining actual geometry. excess cells come from two sources:

    (a) brush faces cutting other distant cells up as a result of the innate work that portalizing/optimizing involves; the advice to adjust brush timing to put larger brushes first can help with this, but a fair amount of excess cells is inevitable.

    Yes, I've changed timing on larger air brushes a few times, and lowered cell count each time, although I think I've probably milked this method just about dry by now.

    (b) brush faces that arent doing any work at all, e.g. the classic example of the bottom half of an air cylinder that is making an arch: the bottom faces are air cutting air, and not defining any geometry. excess cells from these can be eliminated by eliminating such faces, such as by rebuilding with wedges so only the top half of the arch is defined.

    I have quite great number of sewers that should be taller air brushes then, and then progressive wedges. Is it worth the work and all those extra solid brushes. I also have some basement areas with intersecting criss-cross arches, which as you have described, probably chew up the cells pretty ravenously. What is the drop in percent of cells used, roughly? I'll spend the time for sure, if it's significant enough of a drop in cells... If it helps determine whether it's better to change those basements and sewers, my terrain brush count right now is 6057.

    note that a brush face that is exactly parallel with another brush face that is doing work, but that is not itself doing work, is benign; they share the same “cutting plane”, so no additional cuts are made.

    Just the few additions I've made tonight have jumped the cell count to 21915, so I'm really interested in going back, and trying to see if I can really back off the cell count, so I can add more areas with better detail and/or make existing areas less "cubey".
    Sorry for all the perhaps patently obvious questions on this matter, but I haven't even looked at Dromed since before Southquarter stopped working (as far as I can remember). I'm around fifteen years behind.

  4. #4
    Member
    Registered: Apr 2011
    Quote Originally Posted by uncadonego View Post
    So, for example, if I take an old room and make all three air brushes the same depth and width, the cell count would naturally go down?
    that depends on how you build the room. if you have three air brushes stacked like pancakes, they still make 3 cells pre-csgmerge. but if you have one large air brush for the room, and then two more air brushes later in time, of the same width and depth, that are there to retexture the walls for wainscoting and moulding, that is only 1 cell pre-csgmerge, for the large brush; the two later brushes are fully contained within it, and dont add to the cells.

    you can see this for yourself by building these different structures in a new file and optimizing, and comparing the cell counts.

    however, this is in laboratory conditions. in a real level, due to the interaction of many brushes, you may not see exactly the reduction you expect from such a simplification; the actual reduction in practice may be less.

    Quote Originally Posted by uncadonego View Post
    But, would making one air brush and then four solids for wainscoting which do not protude, and then four more solid brushes for ceiling mouldings, which also don't protrude, would both methods result in the same cell count?
    just like above where (in laboratory conditions) the later air brushes fully contained in an earlier air brush do not contribute to cells, these solid brushes fully contained in solid (void) also do not create cells.

    Quote Originally Posted by uncadonego View Post
    I have quite great number of sewers that should be taller air brushes then, and then progressive wedges. Is it worth the work and all those extra solid brushes. I also have some basement areas with intersecting criss-cross arches, which as you have described, probably chew up the cells pretty ravenously. What is the drop in percent of cells used, roughly? I'll spend the time for sure, if it's significant enough of a drop in cells... If it helps determine whether it's better to change those basements and sewers, my terrain brush count right now is 6057.
    it cant be estimated as a percentage. you really have do a before and after comparison with the original geometry and rebuilt geometry to see what savings you get.

    but honestly, i am the wrong person to talk to about cell counts. my philosophy is: if i am getting close to the limit, then i am building the mission unnecessarily big and/or unnecessarily detailed, and in that case my time is better spent in deciding what to cut entirely to make the mission more focused, than in rebuilding geometry.

  5. #5
    Member
    Registered: Dec 2001
    Location: OldDark Detox Clinic
    Thanks vfig. If I may stretch everyone's patience on this line of questioning just one more time...

    I vaguely remember a thread where some dromeders were talking about an author who actually went to the trouble of building intersecting arches with wedges and, I think... upside-down corner-apex pyramids for the intesections.

    Can anyone remember what author that was, and the title of one of their missions? I'd like to open one of their missions up in dromed so I can "monkey-see-monkey-do" a few of my sewers, just to see what effect it has on the cell count.

    Thanks, and sorry for bothering people for old-hat I may have even asked about a decade ago, but no longer remember...

  6. #6
    Member
    Registered: Apr 2011
    i believe skacky is the foremost proponent of the wedges-for-arches style. The Sound of a Burrick in a Room is one mission showing the technique.

    but here is a quick sample of the technique for a barrel vault. each different brush shape is textured in a different colour for clarity.

    to mimic the 4-sided vault shape you get from an 8-sided align-by-vertices cylinder, you need two sets of wedges and corner-apex pyramids (4-sided, align by sides) that overlap:

    A. Vaulted ceiling with two 8-sided cylinders (i.e. the shape we want to achieve):


    B. The first set of wedges (stucco) and corner-apex pyramids (brick):


    C. The second set of wedges (wood) and corner apex pyramids (green brick):


    D. The completed vault, made by overlapping the brushes from B and C.


    and here is the .mis:

    wedgevault.zip on dropbox

  7. #7
    Member
    Registered: Dec 2001
    Location: OldDark Detox Clinic
    Ah! Thanks so much vfig. Quite simple to do really. It just takes a little time to place all the brushes, but it's quite straightforward.

    I only did a small section of my sewers...maybe 8 tunnels, but I'm optimizing right now, and it's already dropped by 202 cells.

    It might save more than 1000 cells by the time they are all done. Plus the arched basement areas.

    Muchas Gracias!

  8. #8
    Member
    Registered: Dec 2001
    Location: OldDark Detox Clinic
    For some close criss-crossed air brushes, I had to use upside-down 4 sided pyramids instead of wedges, but it was pretty the same process. Anyway it took a bit of time, but it dropped the cell count by 1621 cells.

    Thanks for that little .mis file vfig, seeing was understanding. You went the extra mile.

  9. #9
    Member
    Registered: May 2002
    Location: Texas
    Whenever I put an arch in my mission by using a cylinder, I place it first before placing the subsequent rectangular air brush so that the rectangular air brush wipes out the bottom half of the cylinder. I'll have to test this to see which way has the least cells.

  10. #10
    Member
    Registered: Dec 2001
    Location: OldDark Detox Clinic
    I'd be interested in your results if you post them here...

  11. #11
    Member
    Registered: Jan 2001
    Location: 27430 Cells
    Another way to save cells is to look for things like staircases and make sure the space underneath has been filled in (unless you want it to be accessible, and in that case you could consider whether it adds enough to the gameplay to justify leaving it open).

    Any complex terrain you have, in addition to using up a lot of cells, could well have some pockets of unused empty spaces where things don't quite meet.
    There could also be spaces between the edges of that terrain and the outer faces of that part of the mission.
    You could save quite a few cells by filling in those spaces before you start spending time simplifying what the player can see.

  12. #12
    Member
    Registered: May 2002
    Location: Texas
    Quote Originally Posted by uncadonego View Post
    I'd be interested in your results if you post them here...
    I made a small test mission, first trying it with the cylinder timing first and then the rectangle second, then compared that result to having the timing reversed. The number of cells was equal either way.

    Thanks Robin! I have a lot of complex terrain and one stair case that has a room underneath. Both can be cleaned up using your suggestion.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •