Compose performance

Core principle

Measure one user-visible transition, identify the runtime axis that causes the work, then apply the smallest correction at the phase or boundary where that axis begins.

Procedure

1. Reproduce one concrete transition and capture the observable evidence: recomposition counts, compiler reports, profiler data, or a clear trace.
2. Classify the primary axis: parameter stability and skipping, State read phase, or snapshot state written back into an earlier phase.
3. Check for a false lead: a real data change, a correctness defect, or an unchanged lazy item that is expected to recompose.
4. Read the corresponding focused reference before proposing a change.
5. For review work, pair every finding with the smallest evidence-supported repair. For a false stability promise, explicitly say to replace mutable non-snapshot properties with immutable data or snapshot-observable state, verify that contract, and only then decide whether an annotation remains needed. For a phase problem, name the layout or draw consumer where the changing read or calculation should move. A cross-phase write proves invalidation or an extra pass; do not claim oscillation or an infinite loop unless evidence shows values repeatedly changing. Do not stop at diagnosis. When only source code is available, separate the proven feedback path from its unmeasured runtime cost. Give a conditional repair direction based on the intended measurement boundary; do not prescribe a callback swap, equality guard, or content-layout rewrite as a performance fix without evidence that it changes the observed problem. For an Int held in mutableStateOf with the default structural equality policy, an unchanged value already avoids invalidation.
6. Change one axis at a time and re-measure the same transition.
7. Finish when the evidence improves at the observed boundary without hiding state changes, caching stale values, or moving work to a less correct owner.

Topic router