Skip to content

JetBrains/intellij-community / .claude/skills/compose-ui

compose-ui

Verified

Write or audit Compose Desktop UI state, effects, and boundaries.

FileKindSize
OWNERSHIPother20 B
SKILL.mdskill md637 B
architecture/GUIDE.mdmarkdown5.7 KB
architecture/evals/evals.jsonconfig6.5 KB
architecture/references/animation.mdmarkdown11.3 KB
architecture/references/component-api-lists-performance.mdmarkdown4.1 KB
architecture/references/desktop-seams-review-output.mdmarkdown1.5 KB
architecture/references/effects.mdmarkdown1.7 KB
architecture/references/state-boundaries.mdmarkdown5.4 KB
architecture/skill-source.jsonconfig2.0 KB
audit/GUIDE.mdmarkdown14.1 KB
audit/evals/evals.jsonconfig18.1 KB
audit/references/api-hygiene.mdmarkdown8.8 KB
audit/references/audit-search-patterns.mdmarkdown3.9 KB
audit/references/boundary-violations.mdmarkdown3.3 KB
audit/references/effects/derived-state-of-guide.mdmarkdown2.2 KB
audit/references/effects/effect-selection-guide.mdmarkdown2.7 KB
audit/references/effects/infinite-loop-patterns.mdmarkdown2.6 KB
audit/references/effects/overview.mdmarkdown7.7 KB
audit/references/lazy-layouts/animate-item.mdmarkdown2.0 KB
audit/references/lazy-layouts/cache-window-and-prefetch.mdmarkdown3.2 KB
audit/references/lazy-layouts/desktop-scroll-measurement.mdmarkdown2.0 KB
audit/references/lazy-layouts/keys-and-content-type.mdmarkdown2.2 KB
audit/references/lazy-layouts/nested-lazy-layouts.mdmarkdown1.6 KB
audit/references/lazy-layouts/overview.mdmarkdown7.5 KB
audit/references/measurement/benchmarking-methodology.mdmarkdown1.9 KB
audit/references/measurement/overview.mdmarkdown5.5 KB
audit/references/measurement/profiler-setup.mdmarkdown2.5 KB
audit/references/modifiers/custom-modifier-node.mdmarkdown3.4 KB
audit/references/modifiers/modifier-order.mdmarkdown2.1 KB
audit/references/modifiers/overview.mdmarkdown7.4 KB
audit/references/modifiers/phase-deferral.mdmarkdown2.6 KB
audit/references/reporting-guidance.mdmarkdown2.3 KB
audit/references/stability/ci-stability-gate.mdmarkdown2.0 KB
audit/references/stability/desktop-measurement.mdmarkdown2.2 KB
audit/references/stability/overview.mdmarkdown7.6 KB
audit/references/stability/reading-classes-txt.mdmarkdown2.1 KB
audit/references/stability/reading-composables-txt.mdmarkdown1.9 KB
audit/references/stability/stabilising-types.mdmarkdown3.1 KB
audit/references/stability/stability-config-syntax.mdmarkdown1.6 KB
audit/references/stability/stability-inference-algorithm.mdmarkdown3.4 KB
audit/references/stability/strong-skipping-escape-hatches.mdmarkdown2.6 KB
audit/references/subcomposition/desktop-subcomposition-measurement.mdmarkdown1.7 KB
audit/references/subcomposition/overview.mdmarkdown7.2 KB
audit/references/subcomposition/subcomposition-decision-tree.mdmarkdown1.8 KB
audit/references/subcomposition/swing-panel-subcomposition.mdmarkdown1.8 KB
audit/references/tracing-setup.mdmarkdown5.3 KB
audit/references/tracing-usage.mdmarkdown6.3 KB
audit/skill-source.jsonconfig2.5 KB

audit/references/measurement/benchmarking-methodology.md

View on GitHub
# Benchmarking Methodology

## Sample size

Run the scenario **at least 5 times** and discard the first run (JVM warm-up). Use the median,
not the mean, to reduce outlier impact.

| Metric | Minimum runs | Discard first? |
|--------|-------------|----------------|
| Recomposition count | 5 | Yes |
| CPU time % | 5 | Yes |
| Memory retained | 3 | Yes (force GC before each) |
| Frame time | 10 | Yes |

## Variance and significance

If the standard deviation is >20% of the mean, the scenario is not reproducible. Fix the
scenario (same data, same steps, same duration) before comparing numbers.

A change is "significant" if:

- The median after is >10% better than the median before.
- The after range (median ± stddev) does not overlap the before range.

## Common pitfalls

### Cold-start bias

The first run after JVM start is always slower. Always discard it.

### Debug build bias

Debug builds disable optimisations and enable Live Literals. They can be 2-5× slower than
release. Always verify the final number on a release build.

### GC noise

Memory measurements vary with GC timing. Force GC before each measurement:

```kotlin
// Call from a background coroutine, not from the main/EDT thread
suspend fun forceGc() = withContext(Dispatchers.IO) {
    System.gc()
    delay(100) // let GC settle
}
```

### Observer effect

Attaching a profiler (especially YourKit with allocation recording) slows the app. Use
low-overhead tools (JFR, async-profiler) for the final verification.

### Combining changes

Never A/B test multiple optimisations at once. If A + B together show 15% improvement but A
alone shows 12% and B alone shows -3%, you would incorrectly keep B.

## Documentation template

```markdown
## Performance fix: <brief description>

### Scenario
1. <step 1>
2. <step 2>

### Baseline (median of 5 runs)
- Metric: <value>

### After (median of 5 runs)
- Metric: <value>

### Delta
- <percentage>% improvement

### Tool
- <JFR / YourKit / logging>
```
rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first.Stars: 194kUpdated Claude CodeCodexVerifiedInstall