jlfwong/speedscope
Speedscope
Speedscope is a web-based performance profiler that visualizes profiling data through interactive flamegraphs and timeline views. It ingests performance profiles from a wide range of sources, including Chrome, Firefox, Safari, Node.js, .NET Core, Instruments, Hermes, GHC, and Ruby, normalizing them into a common schema for unified analysis.
The tool distinguishes itself with a canvas-based rendering engine that draws flamegraphs without DOM nodes for each frame, and a WebAssembly-based rendering pipeline for high-performance drawing. It offers left-heavy stack sorting to surface the most time-consuming call paths first, and a single-pass stack aggregation that groups identical stacks by weight. A time-ordered timeline view reveals how application behavior changes during execution, while a sortable function table with linked flamegraphs enables drill-down into caller and callee relationships.
Beyond the browser, speedscope provides a command-line interface for downloading mobile traces and opening saved profiles offline. Its import pipeline supports a documented custom JSON format, the Trace Event format, and a collapsed stack text format, making it adaptable to custom profiling tools. The tool also accepts profiles via drag-and-drop or file picker, and can import aggregate statistics from Instruments.app through deep copy.
Features
- Flamegraph Renderers - An interactive web tool for visualizing and analyzing performance profiles from multiple profiler formats using flamegraphs and timeline views.
- Call Stack Timelines - Renders call stacks chronologically along a horizontal timeline to reveal temporal patterns in execution.
- Profiling Data Normalizers - Normalizes profiling data from multiple profiler formats into a shared JSON schema for unified analysis.
- Profile Data Importers - Accepts profile data by drag-and-drop or file browser, supporting multiple profiler formats for analysis.
- Multi-Format Importers - Loads performance profiles from Chrome, Firefox, Safari, Node.js, .NET, and other profilers into a single viewer.
- Multi-Format Importers - Ingests profiling data from a wide range of profilers across multiple programming languages and environments.
- Sortable Timing Tables - Inspects a sortable table of functions with self and total times, plus caller/callee flamegraphs.
- Weighted Stack Sorters - Aggregates identical call stacks and sorts them by heaviest first to pinpoint where time is spent across interleaved calls.
- Application Performance Profiling - Provides interactive flamegraph visualization for identifying performance bottlenecks in application code.
- Performance Profilers - Opens JSON profiles exported from Firefox's developer tools performance tab for interactive analysis.
- Browser Profile Importers - Opens JSON profiles exported from Chrome DevTools to inspect performance data interactively.
- Performance Profiling - Loads and visualizes profiling data from multiple file formats to help identify performance bottlenecks.
- Heaviest Path Sorters - Groups identical stacks and sorts by total time to pinpoint the most time-consuming execution paths.
- Profiling Data Tables - Provides a sortable function timing table with linked flamegraphs for caller and callee exploration.
- Canvas Rendering Compilers - Compiles C++ layout logic to WebAssembly for high-performance canvas drawing of flamegraphs and timelines.
- Single-Pass Stack Aggregators - Aggregates identical call stacks into weighted groups in a single linear pass over raw sample data.
- Debugging and Inspection - Generate flame graphs from Time Profiler data.
- Performance Analysis Tools - Interactive web-based viewer for performance profiles.
