Snoop & Cache Maintenance: Debug Playbook

Debug playbook

Debug Playbook for Snoop & Cache Maintenance focuses on cache maintenance latency, invalidation count, stale data escapes. The goal is to connect the observable symptom to protocol mechanism, ownership, and regression risk.

Protocol debug is a search for the FIRST deviation, not the loudest symptom. Timeouts and error flags are usually many cycles downstream of the real cause.

Root-cause tree

ROOT-CAUSE TREE — Snoop & Cache Maintenance

cache maintenance latency, invalidation count, stale data escapes looks wrong
        |
   reproducible?
     /        \
   no          yes
   |            |
 flaky env   same first transaction every time?
 / seed         /            \
              yes             no
               |               |
        protocol rule     timing/reset/PVT
        or config bug     or load-dependent

1. Freeze the failing seed, firmware tag, and spec revision.

2. Find the first bad transaction, not the loudest downstream timeout.

3. Map the transaction to signals and VIP monitor events.

4. Classify the failure: protocol rule, integration config, timing/reset, or performance pressure.

5. Prove the mechanism with one reduced sequence.

6. Patch the smallest owner-controlled boundary and rerun compliance plus workload traffic.

Review memo template

STAFF PROTOCOL REVIEW MEMO — Coherence Fabrics (ACE / CHI) / Snoop & Cache Maintenance

1. Symptom
   - Watched metric: cache maintenance latency, invalidation count, stale data escapes
   - Failing interface: <master/slave/endpoint/controller/PHY>
   - Transaction identity: <ID/tag/address/endpoint/lane>
   - Repro setup: <sim/emulation/FPGA/silicon + firmware tag>

2. Mechanism hypothesis
   - Primary mechanism: snoops and maintenance operations move cache lines between valid sharing states and clean stale visibility.
   - Competing hypothesis: <timing, reset, bridge, ordering, firmware, or VIP issue>
   - Missing evidence: <waveform, analyzer trace, counter, spec clause, or log>

3. Proposed action
   - Minimal reversible change: <RTL, register setting, bridge config, scheduler, VIP check>
   - Expected metric movement: <delta and workload>
   - Regression risk: ordering, compatibility, performance, power, area, or timing

4. Signoff
   - Re-run artifact: line-state timeline, maintenance operation trace, software flush sequence
   - Required owners: software owner, cache RTL owner, system verification owner
   - Final decision: fix, waive, document limitation, or escalate to architecture

Protocol deep dive

Coherence extends memory transactions with snoop and state — traffic multiplies when software shares cache lines.

Concept diagram

COHERENCE TRAFFIC FLOW

RN issues coherent read
   -> HN looks up directory
   -> snoops to sharers
   -> data + state update returned

False sharing: different variables, same cache line -> coherence storm.

Metric graph

COHERENCY TRAFFIC STACK

data fetch        ████████
snoop responses   ██████████████
writebacks        ██████
maintenance ops   ████

High snoop stack with good IPC -> suspect line sharing before faster NoC.

Metrics and artifacts to collect

• snoop rate

• intervention latency

• coherency transaction mix

• false sharing indicators

Mini case study

Benchmark IPC looked fine but system power spiked: per-core counters were on one cache line. Padding counters fixed coherency traffic without any NoC change.

Debug branches

• If snoop latency high, check home node placement and directory policy.

• If ordering bug, run litmus sequences before microarch changes.

• If traffic storm, profile cache line sharing in software layout.

Senior review question

Ask: what is the first transaction that deviates, and which spec rule does it test?

Key takeaways

• Connect every protocol claim to a transaction identity and measurable metric.

• Store the artifact (waveform, log, counter) next to every signoff decision.

Common pitfalls

• Debugging timeouts without finding the first bad transaction.

• Quoting peak bus width without payload efficiency and retry overhead.

• Treating VIP compliance as a substitute for system integration replay.

Principal review addendum

Re-read Snoop & Cache Maintenance against one concrete product workload, not a synthetic directed test.

snoops and maintenance operations move cache lines between valid sharing states and clean stale visibility.