Crate cliffy_gpu 

Cliffy GPU

WebGPU compute shaders and SIMD-optimized CPU operations for geometric algebra.

This crate provides hardware-accelerated geometric algebra operations using:

• WebGPU: Massive parallel computation on GPU
• SIMD: CPU-optimized operations using portable SIMD intrinsics

Overview

Every browser becomes a compute node with cliffy-gpu:

use cliffy_gpu::{GpuContext, GpuMultivector, AutoDispatcher};
use cliffy_core::GA3;

// Initialize auto dispatcher (chooses GPU or SIMD-CPU based on batch size)
let dispatcher = AutoDispatcher::new().await;

// Batch geometric products - automatically dispatched
let a_batch: Vec<GA3> = vec![...];
let b_batch: Vec<GA3> = vec![...];
let results = dispatcher.geometric_product(&a_batch, &b_batch).await?;

Features

• Batch Operations: Process thousands of multivectors in parallel
• Auto Dispatch: Automatic CPU/GPU selection based on batch size
• SIMD Fallback: Optimized CPU operations when GPU unavailable
• WASM Support: Works in browsers with WebGPU
• Compute Shaders: WGSL shaders for geometric product, sandwich, exp, slerp

Re-exports

pub use simd::addition_simd;

pub use simd::geometric_product_simd;

pub use simd::sandwich_simd;

pub use simd::SimdBatch;

Modules

simd

SIMD-optimized CPU operations for geometric algebra.

Structs

AutoDispatcher

Automatic dispatcher that chooses CPU (SIMD) or GPU based on batch size.

GpuContext

GPU compute context for geometric algebra operations.

GpuMultivector

GPU-compatible multivector representation.

Enums

GpuError

Errors that can occur during GPU operations.

Constants

GPU_DISPATCH_THRESHOLD

Threshold for automatic GPU dispatch. Below this count, CPU is often faster due to GPU overhead.