Expand description
GPU-accelerated compute infrastructure for quantum chemistry and visualization.
This module provides a production-quality GPU pipeline for:
- Context — wgpu device initialization with explicit CPU fallback
- Backend report — Every execution reports which backend was used and why
- Orbital grid — Evaluate ψ_i(r) on 3D grids (GPU or CPU)
- Marching cubes — Isosurface extraction from scalar fields
- Isosurface — High-level orbital mesh generation (dual-lobe, normal smoothing)
- Memory budget — WebGPU/WASM memory limit enforcement and pre-flight checks
- Shader registry — Centralized WGSL shader catalogue with tier classification
- Pipeline coordinator — Multi-kernel dispatch coordination with memory-aware scheduling
- Two-electron GPU — O(N⁴) ERI computation on GPU (Tier 1)
- Fock build GPU — Fock matrix F = H + G(P) construction on GPU (Tier 1)
- One-electron GPU — Overlap/kinetic/nuclear matrix build on GPU (Tier 2)
- Density grid GPU — Electron density ρ(r) on 3D grids (Tier 1)
- ESP grid GPU — Point-charge ESP evaluation on 3D grids (Tier 1)
- D4 GPU — Pairwise two-body D4 dispersion accumulation (Tier 3)
- EEQ GPU — Damped Coulomb matrix build for electronegativity equilibration (Tier 3)
- CPM GPU — Coulomb matrix build for constant-potential charges (Tier 3)
§Feature Gates
experimental-gpu-rendering— Enables#[repr(C)]aligned types (bytemuck)experimental-gpu— Full wgpu runtime (Vulkan/Metal/DX12), includes rendering
When experimental-gpu is disabled, all functions fall back to CPU
and the backend report makes the fallback explicit.
Modules§
- aligned_
types - GPU-aligned data structures for zero-copy transfer to VRAM.
- alpb_
born_ gpu - GPU-accelerated ALPB Born-radii evaluation.
- ani_
aev_ gpu - GPU-accelerated ANI Atomic Environment Vector (AEV) computation.
- backend_
report - Backend execution reporting for GPU/CPU compute dispatch.
- buffer_
manager - GPU buffer lifecycle management for quantum chemistry workloads.
- context
- GPU compute context — wgpu device initialization with CPU fallback.
- d4_
dispersion_ gpu - GPU-accelerated two-body D4 dispersion accumulation.
- density_
grid_ gpu - GPU electron-density grid evaluation.
- eeq_gpu
- GPU-assisted EEQ charges: GPU damped Coulomb matrix + CPU linear solve.
- esp_
grid_ gpu - GPU point-charge ESP grid evaluation.
- fock_
build_ gpu - GPU-accelerated Fock matrix construction.
- gamma_
matrix_ gpu - GPU-accelerated SCC-DFTB gamma-matrix construction.
- hct_
born_ gpu - GPU-accelerated HCT Born-radii evaluation for Generalized Born solvation.
- isosurface
- Isosurface mesh generation for molecular orbital visualization.
- marching_
cubes - Marching cubes isosurface extraction from 3D scalar fields.
- memory_
budget - GPU Memory Budget — WebGPU/WASM limit enforcement.
- mmff94_
gpu - GPU-accelerated MMFF94 force field evaluation.
- one_
electron_ gpu - GPU-accelerated one-electron matrix construction.
- orbital_
grid - Molecular orbital evaluation on 3D grids (GPU + CPU).
- pipeline_
coordinator - GPU Pipeline Coordinator — multi-kernel dispatch with memory-aware scheduling.
- shader_
registry - Centralized WGSL shader registry.
- two_
electron_ gpu - GPU-accelerated two-electron repulsion integrals (ERIs).