home *** CD-ROM | disk | FTP | other *** search
- Hi,
-
- > I don't understand totally what you mean here ...
- > What do you mean by the full version with large table ?
-
- The original DVIEW code always used a couple of really large arrays to
- keep information about what horizontal and vertical lines to draw. After
- everything had been calculated it then walked through the arrays drawing
- first all celings, then all floors and finally all walls.
-
- The arrays I'm talking about are mainly:
- short intersections[50][320]; /* 16000 16 bit integers */
- wall_run walls[8000]; /* 24000 16 bit integers */
- floor_run floorlist[200][40]; /* 32000 16 bit integers */
-
- As you can see they are _large_. If we're going to put the drawing engine
- in the DSP something drastic has to be done about them.
-
- > What do you mean by my optimized one ?
-
- The tables I mentioned above were almost eliminated when I changed DVIEW
- to draw everything as soon as it was calculated. This mostly works very
- nicely (take a look at the drawing in single step mode), but the checks
- against previously drawn walls sometimes fails.
- I believe the failures should be easy to fix, but since I've spent several
- hours on it without success, I might be wrong.
-
- > I mean : Which one is the best for the game ? ( with no bugs, speed,...),
-
- At the moment the speed is about the same, but the memory requirements are
- much smaller for the direct drawing approach. I doubt the other one will
- be possible to implement on the DSP, but on the other hand I also doubt
- that the sprite drawing can be accomplished without at least most of those
- tables...
- Anyway, this is only the parts that do the actual drawing and the step
- immediately above that. All BSP tree searching, clipping and 3D calculations
- are the same, so those would be quite safe to convert to assembly.
-
- > and I don't want to recode everything everytime, so I would like to see
- > the code you fell the most interesting for the game.
-
- The changes are not really very great since the optimized version simply
- uses some information directly instead of putting it in a table.
- As for what's most interesting for the game, I'm not sure. The direct
- approach has definite advantages if you only need to draw the environment
- but it might not be enough to include the sprites.
- I discussed some of these things with Bertrand yesterday and I'll try to
- make those letters available on the mailinglist.
-
- > Anything else, if you can, tell me which functions do you thing that won't
- > be modified a lot and which ones will be changed frequently, I'll code the
- > first ones before.
-
- In view.c, DrawNode and DrawSSector are unlikely to change. You should
- probably try to get rid of the recursion in DrawNode, though. These two
- functions, together with those called in the if-statements in DrawNode
- are the BSP traversal itself. The algorithm can probably be improved. I
- even think some suggestions are in the code somewhere.
- Also in view.c, LoadSeg does all the 3D calculations and clipping, which
- is unlikely to change. Again the algorithms used can probably be improved.
-
- > Do you know where I can fing Gcc ?
-
- On every ftp site I've seen, but I don't know about BBSes, especially not
- in France. ;-)
-
- > By clip the wall, I mean impossibility to walk through them.
-
- OK, I'll try to fix that. As I said it'll be done completely outside the
- drawing engine since it's only a matter of not allowing the player to
- move everywhere.
-
- > I'd like to see what dview would look like with textured walls.
-
- Who wouldn't. ;-)
-
- > The last problem is that I've tried twice to subscribe the mailing list,
- > but it didn't work, I received twice a message like unknown adress...
-
- This is very strange. It worked when I tried it yesterday, but today I
- also get some kind of 'unknown mailer error 1'.
- I'll try to fix it ASAP and let you know when it seems to work again.
-
- > Is 'bad_mood-request@rand.thn.htu.se' the correct adress ?
-
- Yes.
-
- > Other explanation ?
-
- Not yet...
-
- Regards,
- Johan
-
-
- --
- Chalmers University | Why are these | e-mail: d8klojo@dtek.chalmers.se
- of Technology | .signatures | rand@cd.chalmers.se
- | so hard to do | www/ftp: rand.thn.htu.se
- Gothenburg, Sweden | well? | (MGIFv5 and QLem)
-
-
-