Performance Tuning

Jump to bottom Edit New page

Isthimius edited this page Jun 22, 2026 · 1 revision

Gondwana already gives you a lot of leverage if you follow its intended model.

The biggest performance win is simple: preserve the dirty-region model whenever possible.

Good performance habits

prefer small targeted world invalidations over full-scene refreshes
avoid unnecessary camera or viewport churn
keep parallax layers cleanly separated
use bitmap backbuffers when dirty-region redraw is the goal
use GPU backbuffers when full-viewport redraw is acceptable and presentation overhead matters more

Things that force more work

These commonly trigger full refresh behavior:

camera motion
viewport resize
viewport zoom changes
layer z-order changes
layer parallax changes
layer origin changes
visibility toggles
debug overlays like grid lines or collision boxes

Watch the right metrics

If you are tuning, pay attention to:

how often FullRefreshNeeded is set
how large refresh queues get
how large the backbuffer dirty rectangle becomes
CPS/FPS sampling results

Practical advice

If performance is poor, first ask:

am I invalidating too much?
am I moving the camera every cycle?
am I on the right backbuffer type?
am I drawing too many overlays as view-level direct drawings?

Where to read next

Gondwana/Rendering/RefreshQueue.cs
Gondwana/Rendering/RenderSurfaceHost.cs
Gondwana/Rendering/Backbuffers/BackbufferBase.cs
Gondwana/Engine.cs