ViLiO

Members
  • Content count

    2014
  • Joined

  • Last visited

Community Reputation

1326 Excellent

About ViLiO

  • Rank
    Contributor
  1. Back in Classic Black

    This is so very awesome!
  2. HL2 BSP Lightmap Coordinates

    The reason for the padding was to stop the texels from neighbouring packed lightmaps bleeding into each other. To do this I add a border around each packed lightmap (so a 16x16 lightmap becomes 18x18 when packed) and the padding gets filled with the nearest texel on the source lightmap. The texture coords are then adjusted to only address the 16x16 area inside the packed lightmap. There is quite an old article on flipcode that explains some of the reasons bleeding occurs http://www.flipcode.com/archives/Light_Mapping_Theory_and_Implementation.shtml There are a few other ways you could solve your texture seams, padding with a 1 texel border is just the one I favour :) All the best, ViLiO
  3. HL2 BSP Lightmap Coordinates

    Assuming you are creating a texture for each individual lightmap, this looks like it could be solved by setting the AddressU and AddressV render states to Clamp (or CLAMP_TO_EDGE_EXT in OpenGL). In my own hl2 bsp renderer I packed lightmaps into a fewer larger textures using a technique similar to http://www.blackpawn.com/texts/lightmaps/default.html and padded each lightmap with a 1 pixel border to avoid seams. Regards, ViLiO
  4. HL2 BSP Lightmap Coordinates

    My apologies if anything I am about to say is complete nonsense as it has been over 2 years since I last looked at my hl2 bsp source code :) The only thing I can see that you have omitted, is not subtracting the LightmapTextureMinsInLuxels from the lu/lv before dividing by the corresponding LightmapTextureSizeInLuxels. So I think your code should read... float lu = TextureInfo[Faces[faceId].texinfo].lightVecsU.dotProduct(Vertexes[vertexIndice]) + TextureInfo[Faces[faceId].texinfo].lightOffsetU - Faces[faceId].LightmapTextureMinsInLuxels[0]; float lv = TextureInfo[Faces[faceId].texinfo].lightVecsV.dotProduct(Vertexes[vertexIndice]) + TextureInfo[Faces[faceId].texinfo].lightOffsetV - Faces[faceId].LightmapTextureMinsInLuxels[1]; lu /= (Faces[faceId].LightmapTextureSizeInLuxels[0] + 1); lv /= (Faces[faceId].LightmapTextureSizeInLuxels[1] + 1); Regards, ViLiO
  5. Both Programmer and modeler

    Quote:Original post by dhulli I mean, I don't want to waste so much time just to make test models. Can someone give me an example of someone who's both a professional games programmer as well as a professional 3d modeler? From what I recall, Mark Healey did both on Rag Doll Kung Fu ...among many other things [smile] Regards, ViLiO
  6. G prefixed API

    This is a bit of a guess as I've never worked with it, but Scaleform seems to be common to all the games you have listed. Regards, ViLiO
  7. Recommend me a 24" Monitor

    Probably a bit late on this, but I will quite happily swear by the HP LP2475w [smile] I currently only have one of them, but strongly considering getting a second to replace my ageing Dell 1908fp. Regards, ViLiO
  8. Slow web page scrolling only on this site

    I've seen terrible page performance caused by certain flash adverts (an O2 one in particular) but not on this site as I only get GDNet+ ads [smile] (This was in latest Firefox with latest Shockwave) So it might be worth noting if any banner ad in particular is common to the times when site performance is poor and reporting it to GDNet Staff. Regards, ViLiO
  9. Hi, Just for a comparison, using my own .obj loader in a Release build C# application takes about 1.3 seconds to load the Sponza mesh, which contains roughly 65000 polygons. The general algorithm is as follows: string line = null; while ((line = streamReader.ReadLine()) != null) { string[] tokens = line.Split(new char[]{' '}, StringSplitOptions.RemoveEmptyEntries); if (tokens.Length > 0) { if (tokens[0].Equals("#")) // comment { continue; } else { if (tokens[0].Equals("mtllib")) // material library { continue; // ignoring these for now } else if (tokens[0].Equals("g")) // group { continue; // also ignoring these for now } else if (tokens[0].Equals("v")) // vertex position { AddVertexPosition(tokens); } else if (tokens[0].Equals("vt")) // vertex texture coordinate { AddTextureCoordinate(tokens); } else if (tokens[0].Equals("vn")) // vertex normal { AddVertexNormal(tokens); } else if (tokens[0].Equals("usemtl")) { SetMaterial(tokens); } else if (tokens[0].Equals("f")) { AddPolygon(tokens); } } } } Adding a polygon itself constructs vertexes using the various position/normal/texcoord indexes and tries to add them to a set of unique vertexes. This returns an index into the vertex list where it was added (or found from a previous add). Creating this list of unique vertexes adds roughly 0.6 seconds onto my total load time for the .obj but it needs doing as .obj is a rather crappy format (so the actually reading and parsing of those ~65000 polys takes about 0.7 seconds) [smile] I also then calculate tangent space for the polygons and batch them up using their material. So all-in-all ...the complete time to read Sponza (including materials) and mold it into something that I can actually render takes roughly 2 seconds (which is still really slow) [smile] Regards, ViLiO
  10. BEWARE of Edge Games!

    If he is hiring just to port Bobby Bearing to iPhone then it doesn't sound like a great opportunity for a budding game developer. However, if the racers game he claims is in development is actually being put on consoles such as PS3 and 360 (and he has the licenses and kits to do it) then I reckon it is definitely worth applying if you are on the lookout for a job. Chances are he is indeed just porting old games to iPhone to make some quick cash, and there is zero work being done on consoles, but that isn't a reason to not apply and potentially get an interview. Then you could see for yourself what he has in development and make an informed decision instead of avoiding him like the plague based on the hatred he receives on the internet. All the best, ViLiO
  11. Service to distribute small demo?

    Assuming it is the demo in your IOTD (which is very nice btw), there is devimg.net which allows you to post a screenshot and also upload your zip (or so I believe as I've never actually used the site I just know it exists [cool]) Also, I hope it goes well in your job search [smile] Regards, ViLiO
  12. AABB for animated model

    The yellow box in your diagram was indeed what I had said "definitely won't result in one that is too small", so massive fail on my part [grin] So yea, if having ragdoll or hitbox info for your animated mesh isn't something you are going to be doing and naturally calculating an AABB from the transformed verts is serious overkill (I mean it is just an AABB, it doesn't need to be perfect), you could just grow your AABBs to something that "works" or even just calculate a single worst case AABB offline (as Sneftel suggested) that will definitely give no false negatives in whatever bounding volume tests you are going to use it for [smile] All the best, ViLiO [Edited by - ViLiO on October 22, 2009 5:32:06 PM]
  13. Either those are some of the most futuristic vacuum cleaners ever made or you meant "Hover" [smile] All the best, ViLiO
  14. AABB for animated model

    If you have the AABBs of the keyframes you are interpolating, just min/max those AABBs to create an AABB of the AABBs (won't necessarily result in a perfect AABB but definitely won't result in one that is too small) [smile] Alternatively, (and this makes a bit more sense for ragdolls etc. where you can't really precompute an AABB) you create OOBs (hitboxes) for all major bones in your skeleton and transform those when you have calculated the resulting pose for the current blended animations. These can then used to create an overall AABB for visibility culling etc. and also for speeding up your ray intersection tests (and other gameplay related things like headshots etc.) Here is an example with a HL2 .mdl (hard to make out the hitboxes in the thumbnail) Regards, ViLiO
  15. Portal Vids

    Quote:Original post by Demosthenes That looks really, really good in motion. :) The only thing is, maybe the portal could be bigger, in the last part of the video it seems hard to get Squishy in. Agreed, maybe even a slight change of shape would be enough, like a square with rounded corners or something [smile]