NA-Race Detector

🧠 Overview

This project detects non-atomic data races (NA-races) in execution traces of C++ concurrent programs. It parses a trace of program execution, builds a happens-before (HB) graph using synchronization events and program order, and identifies potential data races where at least one access is non-atomic.

---

🚀 How It Works

1. Parse Execution Trace
   Reads the execution trace (e.g., signal_wait.txt) and converts it into structured trace events.

2. Build Happens-Before (HB) Graph
   Constructs a graph capturing the HB relations between events, based on:

   • Program order (sequential execution within a thread)
   • Synchronization primitives (e.g., signal/wait)

3. Detect NA-Races
   Identifies pairs of conflicting memory accesses (i.e., accesses to the same variable where at least one is a write) that are not ordered by the HB relation — and where at least one is non-atomic.

📌 Usage

1. Place your execution trace file in the benchmark/ directory.

2. Run the main script:

python main.py

3. Output will list all detected non-atomic data races, or indicate that none were found.

---

🧪 Example Output

--Output for execution trace 10
Data race detected between
ThreadAction(id=13, thread_id=3, action_type='nonatomic write', memory_order='relaxed', location='0000000000404088', value='0x2a')
ThreadAction(id=7, thread_id=2, action_type='atomic read', memory_order='acquire', location='0000000000404088', value='0', read_from=0)

Data race detected between
ThreadAction(id=8, thread_id=2, action_type='nonatomic write', memory_order='relaxed', location='0000000000404084', value='0')
ThreadAction(id=12, thread_id=3, action_type='atomic read', memory_order='acquire', location='0000000000404084', value='0', read_from=7)

Or:

No data races found in this program

---