Run any Skill in Manus with one click

Get Started

benchmark-first

Stars1

Forks0

UpdatedApril 17, 2026 at 16:32

Benchmark converted code against the frozen upstream implementation before making performance claims.

Installation

Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.

Run Skill in Manus

Source

strmt7

strmt7/ome-zarr-C

View GitHub Repository View Creator Repositories

Download

Run Skill in Manus

Related occupationsSOC

Based on SOC occupation classification

Software Quality Assurance Analysts and TestersComputer and Mathematical Occupations·SOC 15-1253

SKILL.md

readonly

name	benchmark-first
description	Benchmark converted code against the frozen upstream implementation before making performance claims.
origin	repo-local

Benchmark First

Use this skill after parity has been demonstrated for a converted surface.

Rules

Compare the upstream implementation and the converted implementation on the same inputs and same machine class.
Verify parity on the exact benchmark input before timing it.
Prefer representative real data. If real data is unavailable, document the synthetic fallback clearly.
Use the narrowest suite, group, and match filters that isolate the edited hot path before widening to larger reports.
Run repeated measurements and report medians or geometric means, not one-off wall-clock numbers.
Report time terms first: Python time, native C++ time, time saved per operation, and native C++ time reduction. If a ratio is included, label it as native C++ speedup over Python (python_time / native_cpp_time).
Report version and environment details with the results.
If the benchmark points to boundary overhead instead of native arithmetic, load cpp-performance-optimization before changing code.

More from this repository

same repository

browser-fallback

strmt7/ome-zarr-C

Use a deterministic browser workflow only when direct fetch or extraction is insufficient for a public-web task.

2026-04-171

verification-loop

strmt7/ome-zarr-C

Run the smallest correct verification lane and report it precisely.

2026-04-171

compliance-and-rate-limit

strmt7/ome-zarr-C

Keep public-web research compliant, bounded, and efficient through rule checks, caching, and backoff instead of evasive scraping.

2026-04-171

cpp-performance-optimization

strmt7/ome-zarr-C

Optimize a parity-proven ome-zarr native C++ surface for measurable speedups by improving data layout, copies, cache behavior, and boundary design without changing behavior.

2026-04-171

cpp-parity-porting

strmt7/ome-zarr-C

Port one upstream ome-zarr-py surface to standalone native C++ without touching the frozen snapshot.

2026-04-161

pybind11-runtime-parity

strmt7/ome-zarr-C

Legacy reference only. Current main does not use pybind11 ports; use this only to understand old debt before removing it.

2026-04-161

name	benchmark-first
description	Benchmark converted code against the frozen upstream implementation before making performance claims.
origin	repo-local

Benchmark First

Use this skill after parity has been demonstrated for a converted surface.

Rules

Compare the upstream implementation and the converted implementation on the same inputs and same machine class.
Verify parity on the exact benchmark input before timing it.
Prefer representative real data. If real data is unavailable, document the synthetic fallback clearly.
Use the narrowest suite, group, and match filters that isolate the edited hot path before widening to larger reports.
Run repeated measurements and report medians or geometric means, not one-off wall-clock numbers.
Report time terms first: Python time, native C++ time, time saved per operation, and native C++ time reduction. If a ratio is included, label it as native C++ speedup over Python (python_time / native_cpp_time).
Report version and environment details with the results.
If the benchmark points to boundary overhead instead of native arithmetic, load cpp-performance-optimization before changing code.