False Sharing Lab

This project demonstrates the phenomenon of false sharing in C by benchmarking a concurrent counter increment with and without padding to cache lines. It includes a Python agent using LangChain that autonomously tests thread counts and writes an analysis.

Prerequisites

• Windows Subsystem for Linux (WSL) with gcc, make, nproc.
• Python 3.9+
• Anthropic API Key

Setup

1. Create a .env file in the agent/ directory:

   ANTHROPIC_API_KEY=your_key_here

2. Install Python dependencies:

   cd agent
   pip install -r requirements.txt

Running the Lab

You can execute the entire flow by simply running the python script:

cd agent
python agent.py

This script will:

1. Detect your available WSL cores.
2. Direct the AI agent to build and run the C benchmark across multiple thread counts (unpadded & padded).
3. Record data in results/benchmark_results.csv.
4. Command the AI to review the CSV and write a final report to reports/analysis.md.