Menu

Individual skill

linux-perf

Use when collecting sampling profiles with perf record, reading perf report or perf annotate, measuring PMU counters with perf stat -e, or using PAPI instrumentation to interpret IPC, miss rates, MPKI, or memory bandwidth on Linux. Not for SVG rendering: use flamegraphs.

.devin/skills/linux-perf/SKILL.md

Read pinned source

Source commit b0e8ce89a19fac880251dc3ea1babfeb4503a4fe

What was checked

Matches the observed repository HEAD
Yes
Scanner rules are current
Yes
Repository scan complete
No
Skill scan complete
No
Observed behavior
Network or broad shell
Last checked
Nov 4, 2026

These are observations at the recorded commit, not a guarantee of safety. Review the source before running it.

Install this skill

Choose an agent in your Stack or copy a command below. Commands fetch the pinned commit and copy only this skill directory. Existing destinations are refused. Copying a command does not install anything.

An install command is unavailable: current rules, complete repository and skill scans, an eligible behavior level and a permitted license are required.

Repository context

Repository stars (not skill ratings)
55
Repository last updated
Sep 28, 2026
License
Apache-2.0

Observed agent integration: No specific integration evidence. Standard SKILL.md installation does not imply verified compatibility.

Files and excerpt

File metadata and permitted, redacted excerpts come from the recorded scan. Source links are pinned to that commit.

Browse files
Read excerpt

# Linux perf and PMU counters

`perf` ships in the kernel source tree under `tools/perf`, so its feature set follows the running kernel line (mainline 7.2, LTS 6.18). This skill covers the three jobs perf does for a CPU-bound program: count hardware events with `perf stat`, sample where cycles go with `perf record` and `perf report`, and attribute samples to instructions with `perf annotate`. It also covers PAPI region instrumentation, counter-derived metrics, memory-bandwidth counters, and raw PMU event lookup.

## Contract

| Field | Bound contract |
|---|---|
| Trigger | A Linux program is slow or CPU-bound and the user wants hot functions, counter totals, or instruction-level attribution from `perf`, or asks for PAPI instrumentation, IPC, miss rates, MPKI, or counter-based memory bandwidth. Also fires on `[unknown]` frames, `Permission denied` from perf, or a request to feed perf data to a flamegraph pipeline. |
| Authority | Reversible local: writes only `perf.data` files (or the `-o` target), compiled or PAPI-instrumented test binaries, and PMU tool checkouts in the working directory; rollback is deleting those files. `sysctl` and MSR module changes are proposed to the user, never applied. No remote mutation. |
| Side effect | Profile and counter data on disk. Counting mode (`perf stat`) adds negligible overhead; PAPI adds instrumentation overhead around the measured region; sampling mode slows the program by a small factor that depends on sample frequency and call-graph mode. |
| Done | The report names the hot symbols with their sample share, every counter metric carries its formula, raw counts, workload, microarchitecture, and comparison it supports, and any unresolved frame is explained with its fix. |

## Inputs