TODO

TODO list:
- Create CPU implementation
  Currently project cannot be properly verified on github.actions because GPU available runners are paid
  - Fix CI after CPU algorithm is implemented
  Also, CPU implementation will allow to have a build without CUDA
  - Create build flag to build the project without CUDA
- Add more tests
- Don't show console along with application launch
  Currently console displays some debug info, but in general it's redundant
  - Create debug window that can be optionally opened in runtime
- Implement image debugger
  Sometimes Image needs to be visualized, but there is no any way except saving to disk
  - Implement a mechanism that will allow to display image at runtime
  Idea: debug window with textbox that will accept pointer to Image object
- Create better serialization mechanism
  There are several json files that store application config, rendering palettes, etc...
  Current serialization mechanism needs to be improved to easily serialize all datatypes
- Implement smooth zoom 
- Exception window
  In case of exception, message box with problem details should be displayed
- Remove extra coloring modes
  Some of the coloring mode are experimental and don't look pretty
- Create Carousel class
  There are several flows that slide between elements, changing:
  - Coloring mode
  - Palette
  - Bookmark
  Make them use new Carousel class
- Implement states/event listener mechanisms
  Currently entire logic is located in application callbacks itself
  - Create some mechanism to separate parts of the code that are responsible for bookmarks, coloring mode, etc...
- Improve GPUMemoryPool
- Improve Logger
- Create application icon
- Treat warnings as errors
- Allow to resize window in windowed mode
- Write documentation
- Make install script
- Add precision switch (to use floats instead of doubles)
- Render image at higher resolution that display

Ideas to try:
- Improve performance
  Current GPU implementation is fast enough even on high resolution
  But if CPU algorithm will be implemented, performance may be really bad
  General optimizations:
  - Calculate fractal at lower resolution first
  CPU optimizations:
  - Use AVX2 instructions
  - Use parallelization
- Implement arbitrary precision float
  Although the fractal is infinite, the current zoom's capabilities are not
  The accuracy of the double ends very quickly and at some point further increase of the fractal becomes impossible
  - Use some CPU-based arbitrary precision calculations library to increase maximum zoom level at the cost of performance
- Use precompiled 3rd-party
  Almost all 3rd-party are built from sources. It doesn't take a lot of time, but using precompiled binaries may be a nit improvement
- Implement Julia set support
  At any point of Mandelbrot set you can calculate corresponding Julia set
- Make rendering on separate thread
  Currently there is not a lot of sense of it, since there are no any background calculations, but may be useful in the future
- Create application factory
  Implement in case of support of different platforms
- Update bookmarks.json and config.json on application shutdown
- Add WASM platform support
- Try Vulkan