Skip to main content

December

Sunday 1st​

Monday 2nd​

  • Added UniTask to project
  • Using UniTask in solar system loader
  • Rewritten loader to use scenes to load planets
    • Reference the scene, instead of the prefab
    • Async loading
    • Doesn't load all of the planet content ahead of time (which it would with prefabs)
    • It's like addressables, but less buggy
  • Updated scenes to use new loading system
  • Messing with realtime super-resolution ideas (to make enormous planetary textures usable)
    • Idea:
      • Train a (very) small neural network to upscale one specific image, overfit as much as possible. Train it on the high res textures.
      • Use that NN in the shaders, sampling the texture and upscaling it on the fly.
  • Started setting up new NFGO menu/lobby/game scene, using Myriad this time

Tuesday 3rd​

  • NFGO lobby connection approval
  • Lobby sync (names, ready state, starting game etc)

Wednesday 4th​

  • Started on game systems for networked pong (in Myriad)
  • Improving method docs for Myriad queries
  • Exposed GetCachedQuery in public Myriad API
  • Experimenting with message/event processing for network spawning of entities

Thursday 5th​

  • Spawning objects (the ball) with a "GameObject lead" system:
    • Spawn a GameObject
    • GO creates the relevant Entity locally
    • RPCs for this specific Entity can be sent through that GameObject
  • Wrestling with sync issues (servers spawns ball in game scene, but client hasn't loaded that scene yet)
    • Can't find a good way around this that isn't messy
    • Consider a totally different approach (custom messaging)

Friday 6th​

  • Setting up Mirror test project
  • Setup very basic Mirror lobby

Monday 9th​

  • Updating Myriad integration packages to handle some breaking changes made over the weekend (removing BaseSystem)
  • Updated Ephemeris for the same breaking changes
  • Mirror lobby using code from ECOM
  • Setting up lobby with a Myriad world, to hold per-player data (e.g. ready state, name)
    • This can be synced just like normal game data

Tuesday 10th​

  • Spawning player objects, putting them into ECS
  • Setup local only paddle movement

Wednesday 11th​

  • Synchronising object spawning using Mirror
    • Developed a pattern to spawn things with Myriad entities and GameObjects

Thursday 12th​

  • Cleaning up sim setup across clients and server
  • Setting up some pong gameplay systems

Friday 13th​

  • Fixed SolarSystem test scene (some changes made in the docking sim development broke it)
  • Developing editor visualisers for some ECS components:
    • ScenePosition
    • UnityTransform
    • WorldVelocity
  • Reviewing existing code to see where [BurstCompile] can be added

Monday 16th​

  • Updating query caching system in Myriad.ECS
  • Designing debris clouds
    • Modelling debris from an impact as a single entity, with a position and spread that expands over time. This way debris can be fully nbody simulated without simulating thousands of objects.
    • This has problems:
      • What's it for? only for spherical expanding clouds of debris. That doesn't actually fit very much. e.g. an impact on a ship creates a cone of debris not a sphere!
  • Designing bullet chains
    • Modelling a lot of bullets as a "chain" of entities with full nbody simulation and bullets evenly distributed along the links in the chain. At the cost of just 2 entities in the integrator this can simulate lots of bullets.
    • Store a bitfield of which bullets are missing, to account for bullets hitting things and being removed.
    • Calculate a random "divergence" for each bullet, away from the perfect centre line to account for spread when firing.

Tuesday 17th​

Wednesday 18th​

  • Considering curves for bullet chains
    • Experimented with cubic bezier, this is easy to implement but not particularly accurate
  • Setup a test scene to test bullets at true orbital scale
  • Cleaning up some spawning code used in previous test scenes