perfetto-trace-analysis

name	perfetto-trace-analysis
description	Analyze Android performance traces using Perfetto/TraceProcessor. Supports Systrace, Perfetto traces, and Android Studio CPU/Memory profiles. Provides SQL-based trace analysis, flamegraph generation, and performance bottleneck identification.

Perfetto Trace Analysis Skill

Overview

This skill enables deep analysis of Android performance traces using the Perfetto open-source trace analysis framework. It supports:

• Perfetto traces (.perfetto-trace, .ctrace)
• Systrace (.html, .json)
• Android Studio CPU profiles (.trace, .cpuprofile)
• Ftrace / atrace kernel traces

Core Concepts

TraceProcessor SQL Interface

Perfetto's TraceProcessor allows SQL queries against trace data:

-- Top CPU consumers
SELECT process.name, thread.name, sum(dur)/1e9 as cpu_sec
FROM sched JOIN thread USING (utid) JOIN process USING (upid)
GROUP BY upid ORDER BY cpu_sec DESC LIMIT 20;

-- UI thread jank (frame misses)
SELECT slice.name, ts, dur/1e6 as dur_ms
FROM slice JOIN track ON slice.track_id = track.id
WHERE track.name GLOB '*ui*'
AND dur > 16e6
ORDER BY dur DESC;

-- Lock contention
SELECT slice.name, count(*) as cnt, sum(dur)/1e6 as total_ms
FROM slice
WHERE slice.name LIKE '%lock%'
GROUP BY slice.name ORDER BY total_ms DESC;

-- Memory allocation hotspots
SELECT slice.name, count(*) as allocs, sum(dur)/1e6 as total_ms
FROM slice
WHERE slice.name LIKE '%alloc%' OR slice.name LIKE '%malloc%'
GROUP BY slice.name ORDER BY total_ms DESC;

Key Tables

Table	Description
sched	CPU scheduling events
slice	Trace events (begin/end)
counter	Counter values (CPU freq, memory)
thread_track	Thread-level tracks
process_track	Process-level tracks
actual_frame	Frame timing
memory_snapshot	Heap snapshots
heap_graph	Object graphs

Analysis Dimensions

1. CPU Performance

• Per-thread CPU usage
• Scheduling latency (time from runnable to running)
• CPU frequency analysis
• Wakeup analysis

2. UI Performance (Jank Analysis)

• Frame timeline (expected vs actual)
• Choreographer frame drops
• Inflate/layout/measure/draw breakdown
• RecyclerView bind times

3. Memory Analysis

• Java heap growth patterns
• Native allocation patterns
• GC event frequency and duration
• OOM risk assessment
• Memory leak indicators

4. I/O Performance

• File read/write latency
• Disk I/O patterns
• SharedPreferences commit times
• Database query times

5. Network Performance

• Request latency breakdown
• Connection setup time
• TLS handshake duration
• Data transfer rates

6. Power/Battery

• Wakelock analysis
• Alarm frequency
• GPS usage patterns
• Network radio state transitions

Tools & Integration

trace_processor_shell

# Interactive SQL queries
trace_processor_shell trace.perfetto-trace

# Batch query
trace_processor_shell -q query.sql trace.perfetto-trace

# JSON output
trace_processor_shell -q query.sql --json trace.perfetto-trace

Perfetto UI (ui.perfetto.dev)

• Upload traces for visual analysis
• Flamegraph views
• Counter tracks
• Slice analysis

Integration with AppSmartInspector

• Parse trace files and extract key metrics
• Generate structured analysis reports
• Identify top N bottlenecks
• Compare traces before/after optimization
• Automated regression detection

Analysis Workflow

1. Collect trace - adb shell perfetto -c - -o /data/misc/perfetto-traces/trace.pb
2. Pull trace - adb pull /data/misc/perfetto-traces/trace.pb
3. Load into TraceProcessor - Run SQL queries
4. Identify bottlenecks - Sort by impact (duration × frequency)
5. Generate report - Structured markdown with metrics and recommendations
6. Compare baselines - Track regressions across builds

Spec Document Template

When generating analysis specs, include:

1. Executive Summary (top findings)
2. Methodology (trace collection config)
3. Detailed Analysis (per dimension)
4. SQL Queries Used
5. Metrics & Thresholds
6. Recommendations (priority-sorted)
7. Baseline for Regression Detection