Weight-Stationary Deep Dive: Review Checklist

Review checklist

Review Checklist for Weight-Stationary Deep Dive is anchored on Off-array weight bandwidth per inference and achieved weight reuse across batch and sequence profiles.. Convert measurements into mechanism-backed decisions with clear owner accountability.

• Workload scope and SLA targets are explicit.

• Environment metadata is locked and reproducible.

• Mechanism classification is evidence-backed.

• Owner, rollback trigger, and validation matrix are documented.

• Owners signed: accelerator architect, compiler mapping owner, SRAM subsystem owner, runtime scheduler owner.

AI accelerator deep dive

Dataflow choices are durable architecture decisions that shape memory and scheduling cost.

Concept diagram

DATAFLOW DECISION

input/output/weight stationary
  -> locality pattern
  -> movement cost
  -> throughput and power

Metric graph

DATAFLOW COST MIX

activation traffic   ███████
weight traffic       █████
partial-sum traffic  ██████

Metrics and artifacts to collect

• reuse factor map

• buffer pressure profile

• NoC traffic mix

• shape sensitivity analysis

Mini case study

A dataflow that won for convolution lost on attention-heavy batches due to activation movement pressure.

Debug branches

• Segment by model family

• Compare reuse vs movement

• Re-check mapping assumptions under batch variance

Senior review question

Ask: which first-principles bottleneck class explains the symptom, and what artifact proves it reproducibly?

Key takeaways

• Tie every accelerator claim to a reproducible workload slice and one primary metric trend.

• Prefer bounded fixes with clear owner and rollback boundary over broad tuning bundles.

Common pitfalls

• Optimizing synthetic kernels without production-shape validation.

• Reading average latency while ignoring p95 and p99 behavior.

• Declaring sparse or precision wins without fallback and quality evidence.

Review checklist explanation

Checklist quality determines whether teams close on proof or on optimism.

Minimum packet: metric trend (Off-array weight bandwidth per inference and achieved weight reuse across batch and sequence profiles.), artifact set (Reuse-and-bandwidth report mapping layer classes to weight residency and feeder utilization.), bottleneck class, owner fix, rollback trigger, and validation matrix.

If precision changes are involved, include quality guardrail evidence for each deployment slice.