P3D - project status update

Hello clickers,

codename - P3D is a little project for a 3d engine in fusion I'm working on since about four months,
P may stay for Perspective, for Projection, for Pseudo-, for Project ...or even Pinhole Camera.

Since I talked about this a few weeks ago, and I said I was about to finish by the end of july,
I tought someone could've been interested in a little update on the progress.

Well... >:) august is already here, and there's still much to do!

Fact is I'm expanding the thing a lot (and thus encountering many complications)
I've thrown in some optics for a better control over vertices positions, and got also some bonuses from this setup,
i.e. now we have focal lenght and zoom power (may be useful for a sniper game? who knows!)
and most of all we got rotations along other axis than camera Z (added camera X axis by now, you can look up/down. I'll probably test Y axis too.)

Here's a really quick video while "flying" over a simple test area, with the background not following by now:

https://www.youtube.com/watch?v=IBMCOSzy-aA

Too bad this screen capture couldn't catch the buttons aligned in the top grey bar,
you would've noticed switching between map/render view, the "resampling" option ticked/unticked, and most of all the multiple camera selector
(you can see the two cameras being switched soon after the map view, in the video).
I'll do better videos in the future.

But you may also see so many things still to fix XD (:()
Oddly, many easy things turned out being complicated, while some complex things turned out being easy!

I'm still struggling to find the best clipping (or no-clipping) method - the process of intersecting polygons with the view pane to remove segments behind the view angle, before they get too close and distorted by proximity.
This is by far the hardest thing ('till now I've spent 2 months doing clipping, 2 months all the rest!)
Clipping gives cleaner results against my other solution, the "backward projection",
but unfortunately, under some circumstances, it can't be applied (quadrilateral should become a pentagon.. no clean way to do this and keep textures consistent)
so I have to go with a somewhat mixed approach.

I did some further testing on performances.
At 60 FPS (like the video above) we can use about 300 "faces" (quadrilaterals)
At 30 FPS we have three times the number of faces, thus about 1000 faces on screen at the same time, without noticeable differences.
But on those numbers there's a starting "loading lag" of about one second.
Tested on an average machine.

For those interested, and missing previous episodes XD,
the thing will be really easy to use (drag and drop, add to qualifier, write some values),
extensionless (but not cross platform, since uses Muddymole's superb pixel shader)
thus being fully customizable (it's Fusion!),
and -if results will be overall satisfactory- payment, don't know how much but for sure below $15.

Buildings can already have many more faces automatically joined,
will be moveable (probably) and their faces can be animated (for sure)
a reenginering I've already done will grant advanced control to build "non-linear" geometries, like "skewed" towers,
and many more things being still in working...

I realize I can't make a believable esteem on ending date,
but I'm commited to finish asap, since I've already spent twice the time I thought would have been fine for this project.

...well, to the next episode! ;D

Speechless Shcrodinger. That's quite amazing!!!

Amazing :O

Looking great! Keep up the good work! ;)

Thanks everyone, I'll keep up working and keep you updated!

This looks great!

If it can be used to rotate the polygons enough, this could probably be used to make something that even looks on-part with some N64 games using this, and if not then at least something that looks like Minecraft!

Also, 1000 currently-displayed polygons at 30FPS=enough for at least a few objects using half-decent 3D models (probably). :)

P.S. What texture size did you use in your test with 1000 polygons at 30FPS, and did the objects have alpha channels or not?

P.P.S. Out of curiosity, will this be able to create and destroy objects at runtime, or not really?

Thank you HGF!

I still have to do rotation on polygons
(you mean like "rotating the buildings" and not "rotating the camera around the buildings", right?)
I will for sure try to pull out this feature, I'm thinking on adding a default "center of rotation" in the middle of the object that could be altered by advanced users.

Main problem as always lies in clipping,
if I won't be able to ultimate a satisfying clipping algorithm working under every situation... well... I will be very sad :(
(and it is devilish and distressful that such a limited thing is delaying and complicating this thing so much.. I mean... closer polygons are just a very limited percentage of what you will see on screen, that on further distances jsut works without any annoyance!)

P.S.: textures I used were small, all "tiles" you see are 64x64
Actually, the system works twice as fast with twice sized tiles (=less objects on screen for the same surface covered!)
but again, clipping doesn't XD
I'm mainly working on this now.

EDIT: sorry, I just realized you were asking about my 1000 objects test (video was shot with about 400 objects at 60FPS)
I'm not 100% sure, but I guess I used same dimensions for textures, no alpha channels (it is doable however).

P.P.S.: right now I've implemented this (automated) function only for 2D elements (sprites).
This would be doable for 3D elements, and could probably grant some performance benefit (like creating objects on a pre-selected proximity),
I will probably test this, but would require a good bunch of extra code.

Anyway, if and when this thing will be complete, being done in Fusion, well commented and "open",
you'll also have chance to code your own custom "engine workings" in case of some particular need!

That's all great to hear! 64x64 texture size is actually pretty good as far as I'm concerned. :)

Also, 400 objects at 60FPS is great as well. :)

I'm really looking forwards to this, I may actually have enough money to buy it when it's released as long as I don't decide to spend said money on something else (which I probably will resist the urge to do as much as possible), with the real question being whether or not my parents will help me out with buying it (under the age of 18 and therefore don't have a credit card and such, which is what they would need to help me with) when it is, in fact, released. :)

You could possibly be over the age of 18 when I'll be completing this XDXDXD

Really, I'm investing my holidays on this (when not taken in other occupations - like emptying the basement this morning :(..)
and since I'm dreaming slope-intercepts and arctangents at night I really want to bring this project to its completion.
But this can't be approximated, any little unconsistance can become big and noticeable, so I realize it still requires a good amount of work.

As for pricing, I tought that could be a decent tag, given the amount of work it required (and will require)
and also the -possibily- wide range of applications it could be useful for.
But I'll be thinking on it, and decide upon completion!

Hope this will turn out being solid, easy to use and useful in the end :)

I hope the same, since this seems like we're finally getting a good ("good" being an important part of this), easy-to-use 3D engine in CF2.5 that isn't fake 3D and that doesn't require crazy dll files and such, and that doesn't lag like crazy with even only a few objects. :D