SA260324R-0098_r1

SA260324R-0098

① SA Source

• Source: 開啟完整 SA 文章
• Section: GPU and LPU Integration: Attention FFN Disaggregation (AFD)
• Line hint: 105

Context Before

As we explained in our InferenceX article, LLM inference involves two phases: prefill and decode. Prefill processes the full input context: It is compute-intensive, which is suitable for GPUs.

Evidence

On the other hand, decode predicts new tokens and is memory-bounded.

Context After

Decode is latency-sensitive because the model predicts new tokens one by one, and LPU’s high SRAM bandwidth and low-latency capabilities can help accelerate this iterative process.

② Atomic Claim

另一方面，decode 會預測新的 token，且屬於 memory-bounded 工作。

• Epistemic Mode: ASSERTED
• Mapping Status: COMPLETE

③ Semantic Frame

{
  "attribute": "WORKLOAD_CHARACTERISTICS",
  "context_nodes": [],
  "entity": {
    "id": "04_knowledge_base/Decode",
    "label": "decode"
  },
  "frame_type": "ATTRIBUTE",
  "qualifiers": {
    "condition_text": null,
    "numeric_mentions": [],
    "temporal_mentions": []
  },
  "value": {
    "numeric_mentions": [],
    "value_text": "預測新的 token，且屬於 memory-bounded 工作。"
  }
}

④ Canonical Entity Mapping

Role	Surface Label	Canonical Target
entity	decode	Decode

⑤ Human Review

請在 Properties 逐項確認：

• 原文 → Atomic Claim 是否忠實
• Atomic Claim → Semantic Frame 是否忠實
• Canonical Entity mapping 是否正確
• Epistemic mode 是否保留原文語氣
• 最後選擇 review_action

Review state

Markdown 內文不是正式 approval。只有 Apply bridge 寫入的 Decision Ledger event 才是正式決策。