Perf Runtime Device

Benchmark the hardware performance of all scene tests under a runtime (tests/st/<arch>/<runtime>/, $ARGUMENTS). Use when the user asks to benchmark all examples of a runtime on hardware.

49 stars 102 forksTesting & QA

Install

skills CLI (Claude Code, Cursor, Codex and more)
npx skills add hw-native-sys/simpler --skill perf-runtime-device
Manual install
git clone --depth 1 https://github.com/hw-native-sys/simpler.git /tmp/simpler
mkdir -p ~/.claude/skills
cp -r /tmp/simpler/.claude/skills/perf-runtime-device ~/.claude/skills/perf-runtime-device

Personal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.

From the SKILL.md

# Benchmark the hardware performance of all scene tests under `tests/st/<arch>/<runtime>/`

If `$ARGUMENTS` is provided, use it as the runtime name. Otherwise default to
`tensormap_and_ringbuffer`. Reference `tools/benchmark_rounds.sh` for the full
implementation pattern (device log resolution, timing parsing, reporting
format). Detection / isolation procedures live in
[`../../lib/onboard-detection.md`](../../lib/onboard-detection.md).

1. Validate the runtime is one of: `host_build_graph`,
   `tensormap_and_ringbuffer`. If not, list valid runtimes and stop.
2. If `command -v npu-smi` is not found, tell the user this requires hardware
   and stop.
3. **Precheck + detect platform** (§A) — gate on real silicon, then read the
   detected arch into `PLATFORM`.
4. **Select a single idle device** (§C) — or let `task-submit` pick via
   `--device auto --device-num 1`. Hold one lock for the whole sweep (let
   `task-submit` own the loop) rather than re-locking per example.
5. Enumerate all subdirectories under `tests/st/<platform>/$ARGUMENTS/` that
   contain a `test_*.py` directly beneath them.
6. **Run through `task-submit`** (§E). For each example, run the same
   `run_bench()` pattern from `tools/benchmark_rounds.sh`: snapshot logs, run
   `python <example>/test_<name>.py -p <platform> -d $TASK_DEVICE --rounds 10 --skip-golden`,
   find the new log, parse timing, report results.
7. Print only metrics backed by captured markers: TMR normally has Host /
Read the full skill on GitHub

Details

Skill name
perf-runtime-device
Path
.claude/skills/perf-runtime-device/SKILL.md
Repo updated
18 November 2026

FAQ

What does the Perf Runtime Device skill do?

Benchmark the hardware performance of all scene tests under a runtime (tests/st/<arch>/<runtime>/, $ARGUMENTS). Use when the user asks to benchmark all examples of a runtime on hardware.

How do I install the Perf Runtime Device skill?

skills CLI (Claude Code, Cursor, Codex and more): npx skills add hw-native-sys/simpler --skill perf-runtime-device. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.

Is Perf Runtime Device popular?

It lives in hw-native-sys/simpler, which has 49 GitHub stars and 102 forks.

Related skills

Benchmark

49

hw-native-sys/simpler

Benchmark runtime performance on hardware. If the current branch has commits ahead of upstream/main or uncommitted changes, compares against the fork point (merge-base). Otherwise benchmarks current state only. Use when the user asks to benchmark, measure performance, or compare latency.

Git & Code Review102

Checkout PR

49

hw-native-sys/simpler

Add a remote for a given PR's head repository and checkout a local pr-xxx-work branch. Use when the user wants to work on someone else's PR locally.

Git & Code Review102

Clean Branches

49

hw-native-sys/simpler

Remove stale local and remote git branches that have been merged into main. Detects regular merges and squash merges. Cleans up fork remote branches and prunes tracking refs. Use when the user asks to clean up branches, remove merged branches, or tidy up git.

Git & Code Review102

Plan Audit

49

lllyys/vreader

Audit an implementation against a plan (dev-docs/plans/*). Use when a user asks to check for gaps, logic errors, or missing tests relative to a plan or Work Items.

Testing & QA21

Onboard Arch Precheck

49

hw-native-sys/simpler

Detect the host's actual Ascend silicon and refuse mismatched onboard hardware tests. On shared hosts, acquire a task-submit device allocation first and run the precheck and hardware test in that same job. Skip for a2a3sim and a5sim.

Testing & QA102

Test All Device

49

hw-native-sys/simpler

Run the full hardware (onboard) CI pipeline (examples + tests/st) with automatic platform and device detection. Use when the user asks to run all device/hardware tests or the full onboard CI locally.

Testing & QA102

Browse the most popular Claude skills, all testing & qa skills, or write your own with the skills generator.