hypercube-gpu-core (v6.0.6)

Scientific Audit Node for WebGPU Compute & Coupled Physics.

A deterministic, high-concurrency architecture designed for Lattice Boltzmann (LBM), FDTD, and Poisson-Boltzmann numerical methods. The core implements a Zero-Stall memory model and a v9 Direct-Read Manifest Architecture to maximize effective VRAM bandwidth.

v6.0.6 Highlight: Enterprise DX & Coupled Physics

The v6.0.6 release stabilizes the Multi-Cube Support for coupled physics domains sharing a single contiguous physical GPUBuffer. It introduces Semantic Aliasing (SimulationParams), Custom EntryPoints, and Transient Execution for one-shot physical impulses (e.g., heat injection, pressure bursts) with zero parity overhead.

Technical Specifications

• Synchronous MasterBuffer Layout: Host-mirrored VRAM partitioning for low-latency state synchronization.
• Batched Command Orchestration: Minimal command encoder overhead via consolidated rule dispatching.
• Zero-Stall Pipeline : Asynchronous uniform updates and multi-rule compute passes.
• Direct-Read Manifest (v9) : Read individual global variables (Cd, Cl) without host shadowing.
• Scientific Macro Engine (v6.0.6) : Automated WGSL read_NAME_Now(x, y) and write_NAME_Next(x, y, val) helpers with Idiomatic Semantic Aliasing via SimulationParams.
• Transient Dispatch (New) : executeTransient() for one-shot compute passes without temporal parity rotation.
• Dynamic Insertion : injectData() for mid-simulation data manipulation without host read-back.

Numerical Validation (v6.0 Certified)

• Spatial Order: Verified second-order spatial convergence ($O(\Delta x^2)$).
• Physical Accuracy: Drag coefficient ($C_D$) validated within 1.2% error on the Schäfer & Turek (1996) benchmark.
• Computational Throughput: 1042.12 MLUPS (D3Q19) recorded on modern GPU architecture.

Reproducibility Suite

Absolute throughput and numerical precision can be verified through the integrated audit suite:

1. Initialize: npm run dev
2. Access the Hub: http://localhost:5173/docs/index.html

Multi-Cube Usage (New in v6.0.1)

import { createSimulation, linkSimulation, SharedMasterBuffer } from 'hypercube-gpu-core';

// 1. Create a large Shared Buffer (1GB)
const shared = new SharedMasterBuffer(1024 * 1024 * 1024);

// 2. Link Multiple Engines to the same physical memory
const fluidEngine = await linkSimulation(lbmManifest, shared);
const thermalEngine = await linkSimulation(heatManifest, shared);

// Both engines now share the same GPUBuffer using internal offsets.
// Kernels in 'thermalEngine' can read 'fluidEngine' data faces directly.

Scientific Solver taxonomy

| Discipline | Model | Status | | :--- | :--- | :--- | | Fluid Dynamics | LBM 2D/3D | v6.0 CERTIFIED | | Electromagnetics | FDTD Maxwell | v6.0 CERTIFIED | | Potential Fields | Poisson Solver | v6.0 CERTIFIED | | Signal Physics | Wave Equation | v6.0 CERTIFIED | | AI Agents | Agent Guidelines | ACTIVE |

🤖 AI Collaborator Support

AI agents developing with Hypercube should initialize by reading agents/README.md and parsing docs/index.html for technical specifications of the Manifest DSL and WGSL Macro system.

License

MIT — Hypercube GPU Core v6.0.6 (Helron/Hypercube)