Fast Rectangle Collision Detection (WebAssembly)

Due to the highly simplified new algorithm, the speed benefit of SIMD is now minimal (honestly, almost negligible). We'll stick with it for the time being :).

A high-performance WebAssembly module for rectangle collision detection, optimized with SIMD instructions.

Features

• Fast AABB (Axis-Aligned Bounding Box) and OBB (Oriented Bounding Box) collision detection (OBB and OBB will be automatically converted to AABB and OBB)
• SIMD-optimized calculations using WebAssembly SIMD128
• Support for both f32 (single precision) and f64 (double precision) floating-point calculations
• Efficient memory usage and computation

Exported Functions

The module exports four variants of the collision detection function:

• rectCollisionf: 32-bit floating-point collision detection
• rectCollision: 64-bit floating-point collision detection

Function Parameters

Each function takes the following parameters for two rectangles:

(x1, y1, width1, height1, rotation1, x2, y2, width2, height2, rotation2)

• x, y: Center position of the rectangle
• width, height: Dimensions of the rectangle
• rotation: Rotation angle in radians

Returns 1 if the rectangles are colliding, 0 otherwise.

Implementation Details

• Uses SAT (Separating Axis Theorem) for OBB collision detection
• Implements efficient SIMD operations for vector calculations

Building

1. Please ensure that you have installed the Emscripten SDK and added the SDK root directory to your environment variables.
2. Run the compilation script:

powershell -File "./compile.ps1"

The script will:

1. Compile separate object files for each variant (f32/f64)
2. Link them into a single WebAssembly module
3. Clean up temporary object files

Usage Example

// Load the WebAssembly module
const wasmModule = await WebAssembly.instantiateStreaming(fetch('rectCollision.wasm'));
const { rectCollisionf } = wasmModule.instance.exports;

// Check collision between two rectangles
const isColliding = rectCollisionf(
    0, 0, 100, 50, Math.PI/4,  // First rectangle: centered at (0,0), 100x50, rotated 45°
    50, 50, 80, 60, 0          // Second rectangle: centered at (50,50), 80x60, no rotation
);

Performance Considerations

• F32 variants are generally faster but less precise
• SIMD operations provide significant speedup for vector calculations

License

Apache-2.0 License - See LICENSE file for details