MMA routes host-GPU transfers over multiple available paths to deliver 4.62x higher peak bandwidth and lower latencies in LLM serving without hardware or driver changes.
Scalable and efficient intra-and inter-node in- terconnection networks for post-exascale supercomput- ers and data centers
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
fields
cs.DC 2verdicts
UNVERDICTED 2representative citing papers
A new tabular abstraction for pipeline schedules shows communication can reverse rankings from bubble analysis alone, with GPipe and 1F1B runtime-equivalent but 1F1B lower in activation memory.
citing papers explorer
-
MultiPath Memory Access: Breaking Host-GPU Bandwidth Bottlenecks in LLM Services
MMA routes host-GPU transfers over multiple available paths to deliver 4.62x higher peak bandwidth and lower latencies in LLM serving without hardware or driver changes.
-
A Tabular Schedule Abstraction for Communication-Aware Evaluation of Pipeline-Parallel LLM Training
A new tabular abstraction for pipeline schedules shows communication can reverse rankings from bubble analysis alone, with GPipe and 1F1B runtime-equivalent but 1F1B lower in activation memory.