1 yes for flash but you can do whatever you want with RAM
2 CE cause ez80 :p
3 usually m'y calc tend to be a bit damaged by scratches and such but they are usually fine

Implemented per vertex lightning, so I have an almost rock solid vertex shader now  I also have a key transparency for texture set.

It should indeed. Well as long as framerate is somewhere between 6 to 10 fps it feel responsive enough. Still more polygons would be welcome 

I think the code I am most proud of is the vertex shader which is quite efficient and especially the perspective divide,  who doesnt even use a true division. It is about 2500 cycles max per vertex.

Really nice  Do you plan to support more feature as animated tiles or other portal than screen bounds ?

Well I don't have a 80 tris model in my hand right now and I have a old and buggy (and pretty much unoptimized) glib version on this computer; but :



212 tris, 155 verts. I would expect a minimum 30% speed boost against the current code I have.

EDIT : take a look at this screen, though :



Nearly 1400 tris, 1000 verts. And it run at about 280ms/frame. Given that rendering time scale quite well number of tris and vertex, I *think* 80 poly model should run under 40ms.

Well not really by hand since I think you can do your geometry and then select all triangle you want to asign the material too, and set it directly. You can start asign at the beging and all triangle created after that will use the same material than their parent too.

You need to create mutliple material in blender, which have their own texture, specular, etc... and then when you design your landscape, asign proprely each triangle to the correct material, and then do your uv.
When you render the landscape, you usually group rendering per group of triangle sharing the same material (usually linked with the same shader too), and you can keep the material created in blender.

Quote from: gameblabla on April 13, 2017, 08:52:55 PM

Quote from: DJ Omnimaga on April 13, 2017, 05:16:38 PM
I was planning to get one last year, but then I heard that the d-pad controls were kinda wonky. I wish they added a proper d-pad or something for easier gameplay.

Not only that, i heard that first production builds would stop working after a few weeks. (with no visible damages or anything)

I guess we can say that the chip is cheap.

Nah, it is RAM limited 

I tried some 4000+ polygones model (the engine limit being 4096), and it look good. 









Still got some work, and speed issues, but yeah, at least it as a pretty much (even if slow) steady progress.

Raycasting should definitly be faster than polygon engine, if done right. Just totally different use 

yeah, there are quite funny part 

Anyway :

Got a 115 but yeah I am bad at matching pattern and I had to do it quickly (10min instead of '30min). Anyway, yeah, it only test one aspect of IQ, it is hard to get a good general test. Usually those last at least 2 or 3 hours and are a collection of multiple stuff. Quite more accurate I'll say. (I did one actually when I was young and I am supposed to be >130). But I know for a fact that I am much more a space visualisation guy than a 2D pattern matching machine :p

I did a very small C library for fixed 8.8 point format, https://github.com/TheMachine02/FixedPoint8/blob/master/fixedpoint.h . Should be quite fast (at least almost as integer and much better than float) although I guess it could be better and could  support 16.8 too 
There isn't all math functions either, so that will need to be coded