Chess HPC Project

This folder is the cleaned project bundle for the parallel search used in chess engine. It is organized to be easy to share, review, and upload as a standalone package.

Package Structure

• code/
  C++ source files, headers, Makefile, and usage notes for the benchmark implementation.

• data/
  Input dataset used for the project, including the Tal PGN benchmark source.

• results/
  Benchmark outputs, merged CSV files, representative run logs, and analysis-ready experiment results.

• plots/
  Final report plots and supporting figures used to summarize scaling behavior and profiling findings.

• slurm_scripts/
  Slurm job scripts for full runs, array sharding, CSV merge, and weak-scaling experiments.

What This Project Demonstrates

This project studies the scalability of sequential and OpenMP-parallel Alpha-Beta chess search on shared-memory CPU systems.

Key outcomes:

• Benchmarked 2,431 chess positions from the Tal PGN dataset
• Evaluated depths 4, 5, and 6
• Generated 43,758 benchmark measurements in the final study
• Added search optimizations including:
  • transposition-table caching
  • root alpha-beta window reuse
  • improved parallel bound sharing
  • dynamic thread-count handling
• Observed best overall efficiency-performance behavior around 8 CPU threads

Recommended Review Order

If someone is reviewing the project for the first time, the best order is:

1. code/README.md
2. results/Tal_final_results.csv
3. results/tal_full_results.csv
4. plots/report_plots/strong_scaling_runtime.svg
5. plots/report_plots/strong_scaling_speedup.svg
6. plots/report_plots/strong_scaling_efficiency.svg
7. plots/report_plots/weak_scaling_runtime.svg
8. plots/report_plots/weak_scaling_efficiency.svg