BWhat changed
Introduced a loop-based communication IR treating remote GPU memory as schedulable storage, with parallelize, shift, shard and replicate transformations and autotuned multi-GPU operator schedules.
Scientific scope
Introduced a loop-based communication IR treating remote GPU memory as schedulable storage, with parallelize, shift, shard and replicate transformations and autotuned multi-GPU operator schedules. The available anonymous technical preprint and the named authorship source are separate documents. Equivalence with the final proceedings version and its exact benchmark details is unconfirmed. Reuses established ring/Ulysses strategies; no all-operator coverage or global optimum is claimed.
Evidence
The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. The available anonymous technical preprint and the named authorship source are separate documents. Equivalence with the final proceedings version and its exact benchmark details is unconfirmed. Reuses established ring/Ulysses strategies; no all-operator coverage or global optimum is claimed.
Contributors
- Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.
- Mercury: Unlocking Multi-GPU Operator Optimization for LLMs via Remote Memory Scheduling — author-linked anonymous preprint
- Keren Zhou — Mercury publication and complete byline
Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.
Full scientific record · Browse contribution grades · Request a correction