OSFINAL: C vs Rust Thread Performance

Project Question

How does the performance of threads in C compare to Rust for parallel workloads?

This project compares C pthread programs against Rust std::thread programs using the same thread count and increasing workload sizes. The workloads are:

• Monte Carlo estimation of pi
• Matrix multiplication
• DNS hostname resolution

Folder Layout

• c/: C pthread implementations and Makefile
• rust/: Rust implementations using std::thread
• scripts/: benchmark runner, result analyzer, and graph generator
• data/dns/: static DNS hostname inputs, including fixed per-trial files
• docs/: final report PDF/Markdown, graph assets, Obsidian styling, and AI citation log
• results/: raw and summarized benchmark CSV results

Recommended Experiment

Use one thread count for all runs and vary only the amount of work. The current collected results use 12 worker threads:

• Monte Carlo: 1,000,000; 5,000,000; 10,000,000 points
• Matrix multiplication: 128x128; 256x256; 512x512 matrices
• DNS: 50; 200; 500 unique hostnames

Run each case many times. The assignment suggests that 30 or more runs per case is common for meaningful timing results.

Build and Run

These commands are intended for Ubuntu, WSL, or another Linux environment with gcc, make, python3, and Rust installed.

cd OSFINAL
make -C c
cargo build --release --manifest-path rust/Cargo.toml

Run the full benchmark suite:

THREADS=12 TRIALS=50 DNS_SIZES=50,200,500 python3 scripts/run_benchmarks.py
python3 scripts/analyze_results.py results/benchmark_raw.csv results/benchmark_summary.csv
python3 scripts/make_graphs.py results/benchmark_summary.csv docs/assets/graphs

For quick smoke tests:

THREADS=12 TRIALS=2 DNS_SIZES=50,200,500 python3 scripts/run_benchmarks.py
python3 scripts/analyze_results.py results/benchmark_raw.csv results/benchmark_summary.csv
python3 scripts/make_graphs.py results/benchmark_summary.csv docs/assets/graphs

From Windows PowerShell on this machine, use WSL:

wsl
cd ~/code/OSFINAL
THREADS=12 TRIALS=50 DNS_SIZES=50,200,500 python3 scripts/run_benchmarks.py
python3 scripts/analyze_results.py results/benchmark_raw.csv results/benchmark_summary.csv
python3 scripts/make_graphs.py results/benchmark_summary.csv docs/assets/graphs

The final report is in docs/final-report.md, and the PDF export is docs/final-report.pdf. The Markdown report embeds the SVG graphs from docs/assets/graphs/.

The DNS inputs are static checked-in files. The simple submission files live in data/dns/names_50.txt, data/dns/names_200.txt, and data/dns/names_500.txt. The actual benchmark uses the fixed per-trial/per-language files in data/dns/trials/, which keeps hostnames unique across the benchmark run and reduces repeated-name cache effects.

To open the report cleanly in Obsidian, open docs/ as the vault instead of opening the whole repository.