MOCAP proposes MBKR and LBCP techniques for chunked pipelining on wafer-scale chips to reduce memory imbalance and latency skew in prefill LLM inference, reporting 76.4% lower latency and 3.24x throughput vs GPipe.
arXiv preprint arXiv:2406.19580 (2024)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AR 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
MOCAP: Wafer-Scale-Chip-Oriented Memory-Orchestrated Chunked Pipelining Framework for Prefill-Only LLM Inference
MOCAP proposes MBKR and LBCP techniques for chunked pipelining on wafer-scale chips to reduce memory imbalance and latency skew in prefill LLM inference, reporting 76.4% lower latency and 3.24x throughput vs GPipe.