Ahah awesome :)

It's going to be very very hard to not fall to 20fps but I'm looking at ways to precompute the player, sprites etc

Also, for the road and background, to speed up 3d clipping and projection it is always facing Z. I use skewing instead of rotation. But all transformations are 3d, and the ship is properly rotated, so it's hard to notice. The camera feels like a proper 3d camera

It's the same idea as my 1bpp demo github.com/anael-seghez.... For 2bpp you only need 2 palette entries. It's then possible to fill twice the pixels in one write but bit-masking hell to handle the tips of the scanline

Wow thanks ! :)

Thanks :) I should replace the ship with sprite and try adding 3d shapes in the background. I hope it can maintain 30fps. I also want to try double buffering, it takes more ram but it's possible to gain some cpu %

If a mainstream solution like Ares is already out there, maybe you can free yourself from answering to the average user and focus on what you enjoy the most with blastem. I know it's easier said than done.

What is the most challenging thing for you to keep it going ?

First I want to get the driving "feel" right, speed, feedbacks etc. Then I'll add as much as I can so it doesn't go below 30fps.

I have a lot a free vram so it could work. But I'm thinking I could use a sprite in highlight mode on top of the triangle ship for the propulsion and fx (to add another color). And sprite highlight doesn't work on top of another sprite.

Well, for a slow paced game (slow movement, adventure) it might be ok if the framerate drops at 20 or 15fps. For a fast paced game to keep at 30fps it will probably require line doubling or reducing the 3d view in height.

Thanks!

It took me a long time to find an efficient way to handle the scanline tips. The logical way to do it is "output = (input and mask) or (output and (not mask))". But in asm it is faster to do "output = output xor ((input xor output) and mask); Maybe it is widely known but I struggled...

Thanks to the 2 colors constraint it is possible to write 32 pixels with one 32bit write. The frame buffer is also 1/4 the size so it can be send in one vblank using DMA. The difficulty if to do bit masking fast and accurately. Lot of asm and lookup tricks. github.com/anael-seghez... #SGDK

Merci ! <3