Colfax and MIT speed up lattice QCD staple networks 700–1200x on H200 with CuTe DSL
snowclipsed · x · 2026-11-21
Colfax Research, collaborating with MIT physicists, published a blog detailing optimizations to the staple network—a core component of normalizing flows for lattice QCD Monte Carlo sampling. On an NVIDIA H200 NVL, their CuTe DSL kernels accelerate forward-pass hidden layers by 700–1200x over a reference PyTorch implementation and 9–11x over a previous fused Triton version, and 2.0–3.4x over a Triton port of their optimizations. The workload—many small 3x3 complex double-precision matrix multiplies in a stencil pattern on a 4D lattice—differs sharply from typical LLM operators; the post covers the physics background, per-layer math, kernel design, and further optimizations for more channels.
More from Infra
- Sarvam resells DeepSeek in India, drawing backlash over AI sovereignty stance — teortaxesTex · 2026-11-21
- Ilya Sutskever warns weak neocloud security could let rogue agents hijack compute — npinto · 2026-11-21
- 'Computers Have Broken Everything': NYT on America's growing tech fragility — nordicinst · 2026-11-21
- Can foundries build photonic processors with unclonable fingerprints? — jwt0625 · 2026-11-21
- Optical waveguide crossing lets light pass through while keeping circuits electrically isolated — jwt0625 · 2026-11-21
- Inkling 2 Mini quantized to 13.9GB runs at ~60 t/s decode on a single RTX 4090 — indrasmirror · 2026-11-21