flamegraph を Homebrew, dnf, MacPorts, Nix でインストール

プロジェクトリンク

• ホームページ
• 上流ドキュメント
• パッケージマネージャページ

コマンドとエイリアス

• difffolded.pl
• files.pl
• flamegraph.pl
• pkgsplit-perf.pl
• range-perf.pl
• stackcollapse-aix.pl
• stackcollapse-elfutils.pl
• stackcollapse-gdb.pl
• stackcollapse-go.pl
• stackcollapse-instruments.pl
• stackcollapse-jstack.pl
• stackcollapse-ljp.awk
• stackcollapse-perf-sched.awk
• stackcollapse-perf.pl
• stackcollapse-pmc.pl
• stackcollapse-recursive.pl
• stackcollapse-sample.awk
• stackcollapse-stap.pl
• stackcollapse-vsprof.pl
• stackcollapse-vtune.pl
• stackcollapse.pl

プロジェクトの歴史

The official Flame Graphs page describes flame graphs as a visualization Brendan Gregg created to identify the most frequent code paths in profiled software. The GitHub project packages the open source programs used to create interactive SVGs.

The repository documents a three-step workflow: capture stacks, fold stacks, and render with `flamegraph.pl`. Its converter scripts cover profiler outputs from Linux perf, DTrace, SystemTap, FreeBSD pmcstat, Xcode Instruments, Intel VTune, Java jstack, Go pprof, gdb, and other sources.

採用の歴史

The visualization spread from Gregg's performance-engineering work into talks, articles, operating-system profilers, open source tools, and commercial profilers. The official site says there are now over 100 other implementations and that flame graphs are available in most commercial profilers.

A major adoption path was the simplicity of the file format: profilers only need stack samples that can be collapsed into semicolon-separated folded stacks, after which the reference Perl script emits a portable interactive SVG.

使われ方

The common package workflow is `perf record` or another profiler to capture stacks, `stackcollapse-*` to produce folded stacks, and `flamegraph.pl` to create an SVG. The README shows examples for Linux perf and DTrace and notes that folded input can be filtered with tools such as `grep` before rendering.

The scripts expose options for title, width, frame height, colors, palettes, inverted icicle graphs, flame charts, differential colors, and consistent palette maps.

パッケージ好きにとっての重要性

FlameGraph is one of the canonical tiny-tools packages: mostly scripts, little ceremony, and huge influence. Package managers distribute it because one checkout provides the reference renderer plus a toolbox of converters for many profiling ecosystems.

Its significance is also cultural. The package gave systems programmers a common visual language for stack-profile data, and the folded-stack interchange format made it easy for unrelated profilers and language runtimes to interoperate.

タイムライン

• 2013: Brendan Gregg gave an early LISA plenary talk on flame graphs.
• 2016: The Flame Graph article was published in ACM Queue and Communications of the ACM.
• 2017: Gregg presented an updated USENIX ATC talk, Visualizing Performance with Flame Graphs.
• 2020s: Official site reports over 100 implementations and broad support in commercial profilers.

Related projects

• d3-flame-graph is identified by the official site as a widely used interactive d3 implementation.
• FlameScope is linked from the official site for time-range-oriented flame graph analysis.
• Linux perf, DTrace, SystemTap, Xcode Instruments, VTune, jstack, and Go pprof are among the profiler ecosystems supported by converter scripts.

ソース

• https://github.com/brendangregg/FlameGraph
• https://raw.githubusercontent.com/brendangregg/FlameGraph/master/README.md
• https://www.brendangregg.com/flamegraphs.html