Problem to solve

Now the SDL renderer is working but it's not fully functional. The concept is that it renders the draw tasks into SDL textures which are cached. Later if a texture is cached it can be rendered by SDL directly.

Success criteria

1. Support all LVGL drawing tasks
2. Fix caching issues. (e.g. strings are compared by address now, and not strcmp)
3. Use LVGL's image decoders instead of SDL_Image
4. Use SDL features directly where possible instead of caching (e.g. for fill)
5. Have at least double performance compared to SW rendering.

Solution outline

It's mainly finalizing the already started work. No new solution is required.

The only conceptual question is shall we use SDL to transform images (textures), or we shall cache the LVGL transformed images in the textures.

Rabbit holes

None.

Testing

It would be great to add unit tests to verify the output, but it's not clear if we can do it in GitHub CI, and if we can get back the result from the SDL Rendered.

Teaching

Docs needs to be added.

Considerations

No response