Metraipolis › archive › dc_b24b331dd4fb

KOSMOS: Self-scheduling Persistent Kernels for Communication-Compute Overlap — arXiv:2611.07477

Source: arXiv (en)
Original: https://arxiv.org/abs/2611.07477
License (as archived): CC0-1.0
Archived: 2026-11-06T14:42:24.269Z
SHA-256 of the archived text: 19fdd0ed64c3411b10eb4d5a5b9c72edc763e942911acd2ba5bab663ab6b00ee
Stable URL of this page: https://metraipolis.com/documento/dc_b24b331dd4fb

How to cite

KOSMOS: Self-scheduling Persistent Kernels for Communication-Compute Overlap — arXiv:2611.07477 — arXiv (en), https://arxiv.org/abs/2611.07477 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_b24b331dd4fb (Metraipolis archive dc_b24b331dd4fb, sha256 19fdd0ed64c3411b10eb4d5a5b9c72edc763e942911acd2ba5bab663ab6b00ee)

BibTeX

@misc{dcb24b331dd4fb,
  title = {KOSMOS: Self-scheduling Persistent Kernels for Communication-Compute Overlap — arXiv:2611.07477},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.07477},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_b24b331dd4fb; sha256 19fdd0ed64c3411b10eb4d5a5b9c72edc763e942911acd2ba5bab663ab6b00ee},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.07477. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Tensor and sequence parallelism (TP+SP) and expert parallelism (EP) put collectives on the critical path of transformer layers. State-of-the-art fused kernels mitigate this with fixed roles or schedules, but this rigidity leaves large idle windows. Predicting a communication-computation split does not remove them either; a predetermined split cannot follow data readiness. We present KOSMOS, a library for self-scheduling persistent kernels whose workgroups use queues ordered by consumer demand to switch between communication and computation roles as data become ready, with no scheduler, per-call cost model, or tuned split. KOSMOS fuses every TP+SP collective of a transformer layer with its GEMM, and runs an entire expert layer, from router to fold, in one launch per direction. Evaluated on AMD Instinct MI350X and AMD Instinct MI355X GPUs, KOSMOS speeds up TP+SP end-to-end training in Megatron-LM by 1.10-1.28x in BF16 and 1.10-1.25x in MXFP8. KOSMOS accelerates end-to-end training of DeepSeek-V3 and Qwen3-235B proxies by 1.29-1.83x over DeepEP.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Distributed, Parallel, and Cluster Computing (cs.DC) ; Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Performance (cs.PF).
Pagina del lavoro: https://arxiv.org/abs/2611.07477 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-11-06T14:18:42.676Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).