OCTOBER 1997 GAME DEVELOPER MAGAZINE SAYS YOU More Caviar? probably uses a variation of A* with a lim- Design Documents ited search horizon, possibly using some- thing similar to the cut-locked path read Dave Sieks’ interesting report method for avoiding other units. I guess just finished Tzvi Freeman’s article I about Caviar technology (“Dawn of they also had some look-up tables in I (“Creating A Great Design the 3D Pixel Sprite,” pp.58-61, Game there to decide which paths were impossi- Document,” pp.58-66, Game Developer, June 1997). I am a program- ble (the way their sea transport boarding Developer, August 1997). Many mer with a restricted budget. I want to algorithm works suggests this). While I thanks! The article provided many experiment with something like this think the original COMMAND & CONQUER insights for me and several ways that technology. I guess it’s Animatek’s used a variation of the wall-tracing algo- I can improve my next game develop- secret, but would you guide me further? rithm, RED ALERT probably uses A*. Unit ment cycle. How could I get more information about avoidance there is almost definitely step- I am a producer in the game indus- the technique? If you know of some lit- based (the way the units deal with try, and my last project had me writ- erature references it would stimulate my blocked bridges points in that direction), ing the design document. Many of imagination. but some interunit communication might the foul-ups that can happen, did (it Ahmet Engin Karahan also be present. Obviously, RED ALERT gave me a good chuckle to read that also has code in it that allows several they happen to others too!). I espe- CONTRIBUTING EDITOR DAVE SIEKS units to follow a main path, which is easy cially liked your acknowledgment of RESPONDS: Yes, Animatek has patented their to do if you work with step-based unit the “soul” of a game, and how it can Caviar technology. As mentioned in the arti- collision avoidance. focus those wildly creative members 7 cle, you can download freeware Caviar view- of the team. I also agree ers from their web site (http://www.ani- with Freeman’s matek.com/VoxelCh.htm), which will at least Rant thoughts on afford you a glimpse of the process in action. Hey Folks! Drop us a line at how to allow a Also available for licensing by game devel- [email protected]. Or write to Game game design opers is another implementation of voxel would like a moment of your Developer, 600 Harrison, San document to sprite technology from Attention To Detail. Itime to let you know that I guide the You can find details and a downloadable like the “new” look, Francisco, CA 94107. development demo online (http://www.atd.co.uk/ but I am a little without restrict- voxel.html). And though it’s not voxel fearful that the ing the creativity and sprites, you might be interested in checking content will allowing “sparks” of brilliance to out a list of 3D voxel landscape engines drop to adver- improve the overall product when maintained by Karsten Isakovic at the tising like appropriate. Technical University of Berlin. The list cur- other “used-to- Anyway, I just wanted to say rently boasts 33 different engines, with avail- be-good” maga- thanks. Keep up the good work! able source code and documentation - zines. Chris Longpre (http://cg.cs.tu-berlin.de/~ki/3de_land.html). I own every Good luck. issue of Game Y Developer. I enjoy it that much. The problem is that I have Secure Them Pants! Algorithms for Warcraft, C&C already noticed a drop in content; Game Developer spends more and more time with “talk-to-this person” stuff, our editorial in the July 1997 hat was a great article on multiple “political” issues (in the form of “what Y issue of Game Developer made me T object pathfinding in the June I think about the industry”) and laugh out loud when I visualized 1997 Game Developer ("Real-Time “advertising” in the form of new game “security breeches.” Yeah, I guess that Pathfinding for Multiple Objects," pp explanations. confirms that I’m a geek. 36-44). What type of pathfinding algo- Yes, I like to know when the games Anyway, I know that you know rithms do popular real-time strategy come out and if they’re worth looking what you meant. games such as WARCRAFT II and into, but I don’t want to see the impor- Dan Kollmorgen COMMAND & CONQUER: RED ALERT use tant things such as how-tos and exam- when moving large numbers of vehi- ples disappear to this “talk-mag” style EDITOR ALEX DUNNE RESPONDS: That is cles? It's great to learn how to do (examples past and present include really funny. You’re the first person to things the right way, but as a game Commodore and Byte magazines). point that out to us, and I nearly died developer, I'm also interested in learn- I am sorry that I felt compelled to laughing when I reread that particular ing how the other guy did it. write a not-so-positive review, but I just sentence. Dylan Miller don’t want to lose the valuable infor- Therein lies the main problem with spell mation I glean from articles in Game checkers — there’s no way to check the AUTHOR SWEN VINCKE RESPONDS: I have Developer. Thank you for listening. context of the sentence with them. to warn you, this is pure speculation, and Anyway, perhaps security breeches are a I'm doing this from memory. WARCRAFT II Gregory L. Miller good idea too, in certain situations! http://www.gdmag.com OCTOBER 1997 GAME DEVELOPER GRAPHIC CONTENT The Zenith of Hardware Accelerated Rasterization and Beyond… ast month, I talked about what I’d like to see as the baseline of features for 3D accelerators in the 1998 timeframe. This month, I’m going to get into the future of 3D acceleration, including the apex of hardware accelerated rasterization L and moving beyond the triangle model of representing and rendering data. So I’d like to finish up our coverage some noticeable quantization error 10 The Peak of Rasterization of hardware accelerators and then talk with 16-bit frame buffers. Significant or the next few years, we’re still a bit about what the future of hardware errors can be introduced as early as the F going to be using hardware accel- acceleration and 3D graphics might third pass. With 24-bit frame buffers, erators that represent objects as tri- look like when it comes to games. however, quantization errors due to angles. This model is still viable for a 24-BIT AND 32-BIT FRAME BUFFERS. Twenty– multipass rendering are greatly reduced. few more years, but it’s getting a little four-bit (RGB) and thirty-two bit Currently, hardware vendors are long in the tooth for a lot of applica- (RGBA) frame buffers will become shying away from deep frame buffers tions, and with games pushing the increasingly important in the future as because of their higher bandwidth envelope far harder than “real” appli- fill rates increase and developers use requirements (deeper frame buffers cations, I wouldn’t be surprised to see that fill rate for multipass rendering consume more refresh time and also games that represent data in radical techniques. As I’ve bemoaned before, require more bandwidth for fill opera- new ways. multipass rendering can introduce tions) and their greater memory foot- GAME DEVELOPER OCTOBER 1997 http://www.gdmag.com by Brian Hook print. But with memory speeds getting ly useful, especially if there is no perfor- back to very high resolution, 8-bit per faster and memory parts getting mance penalty for enabling it. component texture maps. cheaper, I’m hoping that deeper frame When it comes to filtering, I guess it 24-BIT AND 32-BIT Z-BUFFERS. The limita- buffers become a reality. boils down to this: the less aliasing and tions of 16-bit Z-buffers are becoming A 32-bit frame buffer with a destina- visual artifacts, the better. If a hardware obvious even with today’s games — tion alpha channel would be even bet- ter — if you give game developers more buffers, they will figure out something Currently, hardware vendors are shying cool to do with them (such as alpha saturate antialiasing). away from deep frame buffers because of PER-PIXEL MIP MAPPING AND BETTER FILTERING. As of today, the only accelerators that their higher bandwidth requirements and I’ve seen with usable per-pixel MIP mapping have been the PowerVR PCX2 greater memory footprint. and the 3Dfx Interactive Voodoo. I used to think it wasn’t a vital feature, accelerator does this by traditional tri- GLQUAKE exhibits some Z-buffer alias- but after seeing GLQUAKE’S horrible linear-filtered MIP mapping, anisotrop- ing artifacts on 3Dfx Interactive’s aliasing artifacts on non-MIP-mapping ic filtering, or by contracting the Voodoo accelerator’s 16-bit Z-buffer. A hardware, I’ve revised my opinion. Per- antialiasing fairy to sprinkle magic filter move to a 24-bit Z-buffer format would pixel MIP mapping really is going to be dust on every chip that gets stuck on a help with Z-aliasing problems greatly. A 11 a necessity. board, I don’t care — just make texture 32-bit Z-buffer might be overkill, how- While I don’t consider trilinear filter- aliasing go away, hopefully without ever, but I imagine in a few years it will ing completely necessary, anyone who blurring everything into the Hell of be the norm.