• Content count

  • Joined

  • Last visited

Community Reputation

124 Neutral

About Liguria

  • Rank
  1. Quadtrees, Octrees, BSP Trees, etc. are commonly known as 'Spatial Partitioning' or 'Spatial Subdivision' methods.
  2. Self-hosting the Epoch Compiler: Day Three

    Man, these 'captain's log book' style of entries are among the most interesting I have read so far.   Even though it seems you're not getting a whole lot of feedback/comments on the Day One/Two/Three/Etc... entries, please DO keep posting them.    Very, very entertaining!!   Oh, and good luck fixing and perfecting your compiler-compiler of course :)
  3.   I understand that, which is why I tried to point out that in your box-art it's spelled wrongly as 'Jericheo'.
  4. Looks nice overall.   I have one point of critique though: There's a spelling error in the logo (most prominently visible in the box-art). Unless you misspelled the protagonist's name wrong everywhere else (which seems unlikely), I don't think it should be 'Jericheo'.   Otherwise, good luck with the project!
  5. Plith's gameplay video

    Your game looks very interesting. It makes me think of a very nice twist between Tetris and Four in a Row. Keep up the good work and finish strong! Liguria
  6. How many people have bailed on

    [quote name='Waaayoff' timestamp='1299586057' post='4783078'] [quote name='SuperVGA' timestamp='1299585003' post='4783076'] I visited several times a day before the makeover. Now it's once or twice a week. The new layout simply annoys me. Can't really put a finger on why, but feels more dead now. Wish it would be changed back. [/quote] I know what you mean.. i loved the old design.. this one is just plain weird. It makes me feel like i'm browsing a social network website. [/quote] Same here. It doesn't feel like anymore. It doesn't work smoothly enough to really enjoy visiting. That plus the fact that the last few active threads are not shown at the top of the page is just a shame.
  7. A quick topic to say thanks...

    I congratulate you sincerely. Well done mate! So you must be the guy on the right of the screen: [url=""]BAFTA Interview[/url]
  8. Still looking for game engine

    Sorry for ignoring the DirectX11 requirement for now, but why would you want to pay for an engine? There's plenty of freely available engines/game-kits available (Unity, Unreal) that will allow you to do anything you can think of, if you have the creativity in you (that you would need even if you pay for an engine). I suggest you give it a try with one of those free engines first, before paying a lot of money and get disappointed soon. Do you have experience in game making? (you say you're retired, but not what from..)
  9. polar01, an octree is one of the most useful spatial subdivison method that I ever implemented. Even/especially with millions of polygons you get good results, depending on your configuration. (disclaimer: I'm not an expert on spatial subdivision algorithms!) Also moving entities are rather easily moved between nodes in an octree. There's are lot of information available all around the net about octrees. I'd suggest you have a look at that. You may even come up with a hybrid solution, mixing several methods. Don't restrict yourself to a single point of view! (that's the good thing being a programmer; you've got all the freedom :) )
  10. polar01, you were just a little to fast for me with your reply :) What kind of rendering are you doing? Are you implementing a software renderer? Some extra information might help to determine what could be the quickest solution for you.
  11. Just to clarify (I think we might have a slight misunderstanding due to terminology; which is my fault I'm pretty sure): In the second solution I proposed, I'm talking about casting a 3D ray. When I wrote screen pixel I actually meant a point on the Z==0 plane which relates to a screen pixel. When you have your calculations right and, more important, consistent, inaccuracy shouldn't be a problem, not even for objects that are far away. I never had trouble with it, as far as I can remember. The first solution is definitely the most straight forward and easiest to implement solution if you're only interested in picking top-most objects/entities. That is, if you don't mind having an extra buffer eating up some memory and losing some cycles during rendering (since this would be for an editor-type of application, I guess that shouldn't be an issue).
  12. Two relatively easy solutions: 1) The easiest and fastest way to do it is keeping a buffer of identifiers, which has the same dimensions as your projection buffer. This buffer would be filled with 'identifiers' (pointers (i.e. addresses), for example) by your rendering algorithm, depending on the 'selection mode' (vertex, edge, face, etc). Now you can simply read out the 'identifier' under the mouse cursor to know which 'object' was selected/picked. 2) Another option is to cast a ray from your viewpoint through a 'pixel' on screen and use some intersection math (again, depending on the 'selection mode') to determine what (if any) was selected/picked. I implemented both solutions several times thoughout the years in my software rasterizers. They are really quite simple to get working.
  13. Very well, glad you figured it out! More than welcome :)
  14. About the collision problems: Make sure you process all nodes that your character intersects between the start position of your move and the requested target position. About your performance problems I'm not sure. You might want to tweak the minimum size of your octree nodes a bit to get better performance. Making them too small might just choke performance. Play with your variables a bit (minimum node size vs. max content of node) Small addition: maybe your collision code wasn't so bad in the first place. Did you run a profiler to see if a lot of time is actually spent in the collision routines? Also, for better comparison of the numbers you get, try a bigger and more complicated map. [Edited by - Liguria on December 6, 2010 10:24:43 AM]
  15. Implement some kind of spatial subdivision. An octree would work well for your kind of geometry. Like that, you'll cut down the amount of geometry to test collision against A LOT (if implemented/configured well, that is) Just use your internet search-engine of choice to find information about spatial subdivision and octrees in particular. There's a hell of a lot written about it :)