Flamegraph

Flamegraph is a visualization technique and related tooling used to analyze the performance of software systems by depicting profiled call stacks as a flame-shaped graph. The approach highlights hot paths where most time is spent, aiding developers in identifying optimization opportunities.

Typically, flamegraphs are generated from CPU profiling data obtained by sampling profilers such as perf, DTrace, or eBPF-based tools. Each sampled call stack is recorded, and identical stacks are collapsed into a single representation (a folded stack) with counts aggregated. The resulting flame graph is a horizontal stacked diagram where the width of each block is proportional to the time spent in that function, and vertical position indicates the depth in the call stack.

Interpreting a flamegraph involves looking for wide blocks, which signify functions that consume significant time either directly or through their callees. A hot function appearing across multiple levels of the stack helps trace how callers contribute to the overall cost. The visualization makes it easier to compare different runs or configurations to observe how changes affect performance.

History and usage: Flamegraphs were popularized by Brendan Gregg in the early 2010s and have since become a standard tool in performance analysis. The FlameGraph suite of scripts can process collapsed stacks from many sources and produce interactive or static SVG flame graphs, supporting a variety of languages and platforms.

Limitations: Flamegraphs rely on sampling and thus may miss short-lived events; very large or deeply recursive call graphs can become dense and hard to interpret. They illustrate time distribution but do not show absolute numerical timing without proper normalization. They are best used alongside other profiling methods.