ix-gpu

name	ix-gpu
description	GPU compute via WGPU — cosine similarity, matrix multiply, batch vector search
disable-model-invocation	true

GPU Compute

Cross-platform GPU computation via WGPU (Vulkan/DX12/Metal).

When to Use

When the user needs GPU-accelerated operations: batch cosine similarity search, matrix multiplication, distance computation, or batch quaternion/sedenion operations on large datasets.

Capabilities

• Cosine Similarity — GPU batch cosine similarity search across vector collections
• Matrix Multiply — WGPU shader-based matrix multiplication (f32)
• Distance Matrix — Pairwise distance computation on GPU
• Batch Operations — Quaternion transforms, sedenion multiplication, BSP kNN queries
• Rips Complex — GPU-accelerated Vietoris-Rips complex construction

Method Selection

• Small data (<1000 vectors) — CPU is faster (no GPU dispatch overhead)
• Large batches (>10000) — GPU provides significant speedup
• No GPU available — All operations have CPU fallbacks

Programmatic Usage

use ix_gpu::similarity::GpuCosineSimilarity;
use ix_gpu::matmul::GpuMatMul;
use ix_gpu::distance::GpuDistanceMatrix;
use ix_gpu::context::GpuContext;

MCP Tool Reference

Not yet available as a dedicated MCP tool (requires GPU hardware). Use the Rust API directly.