v1.0.2

v1.0.2 Release

FEATURES

• Add --solve-all option to find all solutions (default limit: 100)
• Add --max-solutions N to set solution limit (0 = unlimited)
• Add name and difficulty metadata support for JSON puzzles
• Display puzzle metadata (name, difficulty) when present
• Add throughput display for single-threaded benchmarks

BUG FIXES

• Fix invalid 16x16 built-in puzzle (was unsolvable)
• Fix unlimited solution search bug (maxSolutions=0)
• Remove redundant APP_VERSION macro (defined in CMakeLists.txt)

CI IMPROVEMENTS

• Skip CI builds for non-code changes (*.md, docs/, LICENSE)
• Add single-threaded benchmark test (-b 100 -t 25)

BENCHMARK RESULTS (Single-threaded, 25x25 puzzle)

• WSL2 9950X3D: 57.43 puzzles/sec
• Windows 9950X3D: 51.90 puzzles/sec
• Android SD8G3: 39.27 puzzles/sec
• macOS M1 (VM): 31.39 puzzles/sec

NEW CLI OPTIONS
--solve-all Find all solutions
--max-solutions N Limit solutions (0 = unlimited)

EXAMPLE USAGE
SudokuSolver -t 9 --solve-all
SudokuSolver -i puzzle.json --solve-all --max-solutions 10
SudokuSolver -i puzzle.json --solve-all -o solutions.json

v1.0.1

v1.0.1 - Cross-Platform Support & Enhanced System Detection

🚀 Multi-Platform Build Support

• CMakePresets.json for Windows, Linux, macOS, Android
• Conditional Tesseract OCR compilation (HAS_TESSERACT)
• vcpkg integration with platform-specific triplets
• Static linking for portable binaries

🤖 GitHub Actions CI/CD

• Automated builds for Windows, Linux, macOS
• Cross-platform testing with built-in puzzles (-t 9/16/25)
• Release artifact publishing

📱 Enhanced System Detection

• Android: SoC detection (Snapdragon, Dimensity, Exynos, Tensor)
• Android: ARM CPU Part ID lookup (Cortex-X4, A720, A520)
• macOS: Apple Silicon detection with reference clock
• macOS: Version codename mapping (Sequoia, Sonoma, etc.)
• Linux: Full /proc/cpuinfo parsing

🔧 Build Fixes

• CMake target ordering for cross-platform compatibility
• Added #include for GCC/Clang
• Fixed MSVC type conversion warnings
• Optimized OpenCV features to avoid fontconfig dependency
• Built-in test puzzles for CI benchmarks

📊 Benchmark Highlights

Platform	CPU	DLX/s	Efficiency
Windows	Ryzen 9950X3D	885.59/s	55.25%
WSL2	Ryzen 9950X3D	903.59/s	49.92% ⭐
macOS CI	Apple M1 (VM)	94.92/s	98.75% ⭐
Android	SD 8 Gen 3	105.31/s	34.23%

🔬 Key Findings

• WSL2 DLX outperforms native Windows by 2%
• GCC excels at DLX (pointer ops), MSVC at Backtracking (recursion)
• Apple Silicon achieves 98.75% efficiency (UMA advantage)
• Compiler optimization > OS virtualization overhead

Platform Support:

Platform	Solver	Benchmark	OCR
Windows x64	✅	✅	✅
Linux x64	✅	✅	✅
macOS (Intel/ARM)	✅	✅	✅
Android ARM64	✅	✅	❌

v1.0.0 - Initial Release

🎉 Initial Release

High-performance Sudoku Solver with multiple algorithms and OCR support.

✨ Features

• Algorithms

  • Dancing Links (DLX) - optimal for large puzzles & multi-threading
  • Backtracking with Constraint Propagation - fast for 9x9

• OCR Support - Extract puzzles from images (requires eng.traineddata)

• Benchmarking - Multi-threaded performance analysis with system info detection

• Puzzle Support - 9x9 and 16x16 grids

📊 Benchmark Highlights

Puzzle	Algorithm	Throughput (Ryzen 9 9950X3D)
16x16	DLX	23,075 solves/sec
9x9	Backtracking	204,557 solves/sec

📥 Downloads

File	Description
SudokuSolver-v1.0.0-win-x64.zip	Windows x64 binary (Release build)

📋 Requirements

• Windows 10/11 x64
• For OCR: Download eng.traineddata → tessdata/

🚀 Quick Start

# Solve puzzle
SudokuSolver.exe -i puzzle.json

# Benchmark
SudokuSolver.exe -i puzzle.json -b 1000 -w 0

# Show help
SudokuSolver.exe --help

# 📁 Binary File
SudokuSolver-v1.0.0-win-x64.zip

Contents of ZIP file ：

SudokuSolver-v1.0.0-win-x64/
├── SudokuSolver.exe
├── tessdata/           (Empty folder; prompt user to place eng.traineddata)
└── examples/