Hey folks, I’m building a small pixel-art game and I’m trying to keep the animation feeling smooth while still shipping features weekly. Right now the “easy” path is adding more layers, particles, and per-sprite effects, but the failure mode is death-by-a-thousand-cuts where one new enemy pushes frame time over budget and everything stutters.
How do you actually pick and enforce a performance budget for canvas/WebGL sprite animation (draw calls, overdraw, texture swaps, update costs) in a way that catches regressions early without turning every change into a profiling project?
The “one new enemy and everything stutters” bit is exactly why I’d budget around a couple of worst-case scenes you can replay, not abstract draw-call numbers. I’m not sure what engine you’re on, but can you build a tiny “perf zoo” room that spawns your top three nasties (max particles, max layers, max simultaneous sprites), then have the game print a rolling 95th-percentile frame time in the corner so you can spot regressions without opening a profiler?
Draw-call budgets are a trap in pixel art because the thing that kills you is usually “somebody added per-sprite work” and the renderer just gets blamed for it.
I’d do exactly what @ArthurDent described, but make it boringly deterministic: fixed RNG seed, fixed camera path, fixed spawn order, same 30–60s run every time. Then budget in milliseconds (split update vs render if you can), and treat even ~1ms drift in that specific zoo scene as a regression—otherwise you end up tuning to noise from particle chaos and feeling good about it.
