Skip to content

SOURCE.md

Latest commit

 

History

History
197 lines (140 loc) · 4.95 KB

SOURCE.md

File metadata and controls

197 lines (140 loc) · 4.95 KB
Raw

Performance Monitoring System

Overview of Changes

The performance monitoring system has been completely rewritten to provide more accurate and detailed metrics.

Key Improvements

1. Statistical Analysis

Instead of one-time measurements, the system now collects data from multiple runs and provides:

  • Minimum value
  • Maximum value
  • Mean value
  • Median value
  • Standard Deviation (shows variability)
  • Number of measurements

2. Detailed Measurements

Each module is now measured separately:

  • AppPreload
  • App
  • Auth
  • User
  • Dictionaries
  • Words
  • Notification
  • Updater
  • AppVersion

3. Automatic Logging

All metrics are automatically saved to files:

  • Location: ~/Library/Application Support/electron-dictionaries/performance-logs/
  • Format: performance-YYYY-MM-DDTHH-MM-SS.log
  • Includes: timestamps, all measurements, statistics

4. Memory Tracking

Detailed information about memory usage:

  • Heap Used/Total
  • RSS (Resident Set Size)
  • External memory
  • Memory delta (changes between before/after)

Usage

Running with the New Monitoring System

npm run dev:electron

After starting, you will see:

================================================================================
🚀 Starting Application Bootstrap with Performance Monitoring
================================================================================

[MEMORY] app-start: heap: X/X MB, rss: X MB, ext: X MB
[PERFORMANCE] module-AppPreload: X.XXms
[PERFORMANCE] module-App: X.XXms
...

================================================================================
PERFORMANCE STATISTICS
================================================================================

📊 module-App:
  Runs:    X
  Mean:    X.XXms
  Median:  X.XXms
  Min:     X.XXms
  Max:     X.XXms
  StdDev:  X.XXms
  Memory Δ: heap X.XXMB, rss X.XXMB

...

================================================================================
📁 Full log saved to: /Users/.../performance-logs/performance-XXX.log
================================================================================

✅ Application ready!

Automatic Testing (Recommended)

For the most accurate results, use the test script:

./scripts/performance-test.sh

This script:

  1. Runs the application 5 times
  2. Collects all metrics
  3. Saves results in ./performance-results/
  4. Allows you to analyze aggregated data

You can change the number of runs by editing the RUNS variable in the script.

Why Are Metrics Always Different?

Variations in metrics are completely normal and occur due to:

  1. Garbage Collection (GC): V8 can trigger GC at any moment
  2. JIT Compilation: Code is compiled "on the fly" and cached
  3. OS Caching: Files may or may not be in the cache
  4. Background Processes: The OS performs other tasks
  5. Cold vs Warm Start: The first run is always slower
  6. CPU Throttling: The processor may reduce frequency to save energy

How to Interpret Results

Standard Deviation (StdDev)

  • < 10ms: Very stable performance ✅
  • 10-30ms: Normal variability ✅
  • > 30ms: High variability ⚠️ (may indicate an issue)

Mean vs Median

  • If Mean and Median are close - data is stable
  • If Mean > Median significantly - there are outliers, possibly due to GC or background processes

Example Interpretation

📊 total-bootstrap:
  Runs:    10
  Mean:    245.32ms
  Median:  240.15ms
  Min:     178.77ms
  Max:     316.73ms
  StdDev:  38.21ms

Analysis:

  • Average time: ~245ms
  • Best time: ~179ms (likely a warm start)
  • Worst time: ~317ms (possibly a cold start or GC)
  • StdDev 38ms - moderate variability, normal for an Electron app

Tips for More Accurate Measurements

  1. Close other applications before testing
  2. Plug in your Mac (disables CPU throttling)
  3. Run 10-20 times for statistically significant data
  4. Ignore the first 2-3 runs (warmup period)
  5. Test under consistent conditions (temperature, system load)

API Performance Monitor

Basic Usage

const monitor = new PerformanceMonitor();

// Simple measurement
monitor.startMeasure("my-operation");
await doSomething();
monitor.endMeasure("my-operation");

// Measurement with memory
monitor.startMeasure("my-operation", true); // captureMemory = true
await doSomething();
monitor.endMeasure("my-operation", true);

// Show statistics
monitor.printStatistics();

// Show statistics for a specific operation
monitor.printStatistics("my-operation");

// Clear measurements
monitor.clearMeasurements(); // Clear all
monitor.clearMeasurements("my-operation"); // Clear specific

// Get log file path
const logPath = monitor.getLogFilePath();

Project Files

  • src/main/performance-monitor.ts - PerformanceMonitor class
  • src/main/app.ts - Integration of monitoring into bootstrap
  • scripts/performance-test.sh - Script for automatic testing
  • Logs: ~/Library/Application Support/electron-dictionaries/performance-logs/