It's been a year since I started working on the picogame game engine https://picogame.makerclass.cz/.
In my opinion, it is ready for the next step. Over the last two months I haven't had to change anything in its core API except adding new primitives for the 3D helpers, so I consider it stable - and by now dozens of users have run it on their own boards, some have built their own games with it. I'd like to discuss integrating the C module into CircuitPython.
I also need help deciding a few details about how to finalize the engine, and what requirements would Adafruit have for it.
Two features still need a direction:
-
Core1: on SPI boards this adds about 30 % performance for pseudo-3D scenes; on the Fruit Jam without USB host it generally adds about 50–70 % for full-motion games. In reality this is primarily a benefit for 3D scenes, which picogame wasn't originally designed for. The question is whether it makes sense to keep improving this aspect.
-
ROMFS asset partition: for games with large assets, a small flash partition (e.g. 32–96 kB carved from the CIRCUITPY drive) that the engine can blit from directly - assets then cost zero RAM, which unlocks much better-looking games. It requires a partition layout change, so I'd like to hear how you'd prefer that to be handled.
I'd also like to discuss distribution. I would enable picogame on the PicoPad board I maintain (https://circuitpython.org/board/pajenicko_picopad/), and I think it makes sense enabled on other gaming boards and the Fruit Jam. It's useful on a bare Pico too, but there it probably shouldn't be on by default - for those maybe a separate build variant, similar to how the Zephyr builds, should be nice.
A few details: the whole C module costs about 37 kB of flash on a PicoPad with optimized compile flags (~6 kB of that came with 3D). The C module is fully usable on its own - see https://picogame.makerclass.cz/concepts/engine-only/ - the rest of the engine is 30+ Python libraries that would stay a normal library bundle outside the tree: they abstract the hardware (GPIO buttons, USB gamepads/keyboards and I2C pads all behave the same), audio and other conveniences, configured through settings.toml so boards and peripherals can be tuned without touching game code. The engine is covered by a desktop simulator, a browser (WASM) build and a benchmark suite, and 40+ existing games and demos serve as a regression corpus - all of which I intend to keep maintaining.
What should be the next step?
It's been a year since I started working on the picogame game engine https://picogame.makerclass.cz/.
In my opinion, it is ready for the next step. Over the last two months I haven't had to change anything in its core API except adding new primitives for the 3D helpers, so I consider it stable - and by now dozens of users have run it on their own boards, some have built their own games with it. I'd like to discuss integrating the C module into CircuitPython.
I also need help deciding a few details about how to finalize the engine, and what requirements would Adafruit have for it.
Two features still need a direction:
Core1: on SPI boards this adds about 30 % performance for pseudo-3D scenes; on the Fruit Jam without USB host it generally adds about 50–70 % for full-motion games. In reality this is primarily a benefit for 3D scenes, which picogame wasn't originally designed for. The question is whether it makes sense to keep improving this aspect.
ROMFS asset partition: for games with large assets, a small flash partition (e.g. 32–96 kB carved from the CIRCUITPY drive) that the engine can blit from directly - assets then cost zero RAM, which unlocks much better-looking games. It requires a partition layout change, so I'd like to hear how you'd prefer that to be handled.
I'd also like to discuss distribution. I would enable picogame on the PicoPad board I maintain (https://circuitpython.org/board/pajenicko_picopad/), and I think it makes sense enabled on other gaming boards and the Fruit Jam. It's useful on a bare Pico too, but there it probably shouldn't be on by default - for those maybe a separate build variant, similar to how the Zephyr builds, should be nice.
A few details: the whole C module costs about 37 kB of flash on a PicoPad with optimized compile flags (~6 kB of that came with 3D). The C module is fully usable on its own - see https://picogame.makerclass.cz/concepts/engine-only/ - the rest of the engine is 30+ Python libraries that would stay a normal library bundle outside the tree: they abstract the hardware (GPIO buttons, USB gamepads/keyboards and I2C pads all behave the same), audio and other conveniences, configured through settings.toml so boards and peripherals can be tuned without touching game code. The engine is covered by a desktop simulator, a browser (WASM) build and a benchmark suite, and 40+ existing games and demos serve as a regression corpus - all of which I intend to keep maintaining.
What should be the next step?