Profiling measures and analyzes execution behavior and resource usage to make performance problems visible before code ships. It complements performance budgets: budgets define targets, while profiling shows whether a change exceeds them. Integrating profiling into development and review keeps fixes small, clear, and accountable. section 7.0
Use a repeatable profiling loop:
- Profile after writing or modifying code against the relevant performance budget. section 7.0
- Investigate the measured hotspot, resource growth, or blocking work. section 7.0 section 7.4
- Fix the cause, profile again, and move the change to review only when it meets its budget. section 7.0
- Repeat measurements across the device tiers users actually have, including the cheapest supported device. section 7.3.5
Table 7.1 Web profiling concerns, tools, and diagnostic patterns
|
Concern
|
Primary tool
|
What to look for
|
|
Loading performance
|
Network panel
|
Large bundles (over 200KB compressed), long dependency
chains, render-blocking resources, and missing cache headers
|
|
Runtime behavior
|
Performance panel
|
Long tasks (over 50ms), layout thrashing, expensive
paints, and poor Core Web Vitals scores
|
|
Memory stability
|
Memory panel
|
Growing heap size across navigations, detached DOM nodes,
and event listeners that accumulate
|
Use browser profiling panels to separate loading, runtime, and memory problems: inspect network size and dependencies, long tasks and rendering, then heap growth and retained DOM or event listeners. section 7.1
A realistic desktop workload
A photo editor's batch export works in under 30 seconds for 10 images, but profiling a realistic 200-image queue exposes behavior that small tests can hide. Choose tools according to the architecture: platform profilers for WPF, AppKit, or GTK; Qt Creator alongside platform tools for Qt; and browser plus host-process profilers for Electron or Tauri. section 7.4














