Skip to content
← Attention Kernels / 16
hardNewNVIDIAAMDOpenAI

A decode-attention kernel wins at 128K context. What evidence would justify enabling it for other shapes?

Review a shape-by-shape timing fixture, catch a ragged-cache correctness failure, and calculate whether selective dispatch pays for itself. Includes a small reference you can run without a GPU.

Updated Oct 2026 · Worked answers on inference mechanisms, with calculations, implementation details, and follow-up questions.

Review a shape-by-shape timing fixture, catch a ragged-cache correctness failure, and calculate whether selective dispatch pays for itself. Includes a small reference you can run without a GPU.

10 answers per topic instead of 5, plus saved progress and bookmarks · costs nothingor unlock all 136 remaining answers · ₹2,000 / $25

Every worked answer and derivation in all 12 topics, every concept page, and the full practice-test bank.

6 months · One payment · No auto-renewal

UP NEXT ON YOUR JOURNEY
Popular right nowWhy is batch-1 decode on a 70B model bandwidth-bound, and what is its latency floor on two H100s?Popular right nowDerive the FLOPs per token for a dense transformer from its config. Where does the 2N rule stop being accurate?Popular right nowWhy does prefill turn compute-bound after a few hundred tokens on an H100, and how long does a 4K-token prompt take on a 70B model?
FEDITOR'S NOTE

The rollout unit is a supported shape and execution path. A successful long-context timing cannot approve a different head layout, mask convention or graph-capture path.

DISCUSSION · 0

Nobody has commented on this answer. Add a correction or another way to work it.