# Performance Parallelism

> Use Jeff Dean and Sanjay Ghemawat's canonical parallelization and synchronization guidance. Use when tuning threads, locks, contention, critical sections, sharding, SIMD, false sharing, context switches, pipelines, channels, or lock-free designs.

- Skill: `boshjerns/performance-parallelism` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add boshjerns/performance-parallelism`
- Raw SKILL.md: https://api.skillmd.com/api/skills/boshjerns/performance-parallelism/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: boshjerns (https://skillmd.com/u/boshjerns)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/boshjerns/performance-parallelism

---


# Parallelization and synchronization

Read the exact current source section before applying it:

```bash
python3 ~/.codex/skills/performance-hints/scripts/fetch_section.py parallelization-and-synchronization --output text
```

Canonical section: [Parallelization and synchronization](https://abseil.io/fast/hints#parallelization-and-synchronization)

Attribute the guidance to Jeff Dean and Sanjay Ghemawat. Do not assume every hint applies: inspect the code, identify the relevant cost, and measure or estimate before recommending a change.

This section includes:
- Exploit parallelism
- Amortize lock acquisition
- Keep critical sections short
- Reduce contention by sharding
- SIMD Instructions
- Reduce false sharing
- Reduce frequency of context switches
- Use buffered channels for pipelining
- Consider lock-free approaches