Backfilling explains most scheduling gains, while prediction benefits are Trace-Dependent: A cluster-size sweep and cross-trace bootstrap of forecast-driven GPU scheduling in the LLM era
Array · 2026-12 · 32 · 101374
DOI: 10.1016/j.array.2026.101374 ↗At a glance
Journal article in Array, 2026-12. Written by Don Harl C. Malabanan. 17 references. The full text is freely available (open access, CC BY 4.0). Funded by: Republic of the Philippines Department of Science and Technology.
Abstract (original language)
Kokkuvõte pole saadaval.
Funding
- Republic of the Philippines Department of Science and Technology
Cite
APA
Malabanan, D. H. C. (2026). Backfilling explains most scheduling gains, while prediction benefits are Trace-Dependent: A cluster-size sweep and cross-trace bootstrap of forecast-driven GPU scheduling in the LLM era. Array, 32, 101374. https://doi.org/10.1016/j.array.2026.101374
Vancouver
Malabanan DHC. Backfilling explains most scheduling gains, while prediction benefits are Trace-Dependent: A cluster-size sweep and cross-trace bootstrap of forecast-driven GPU scheduling in the LLM era. Array. 2026;32:101374. doi:10.1016/j.array.2026.101374
BibTeX
@article{Malabanan2026,
author = {Malabanan, Don Harl C.},
title = {Backfilling explains most scheduling gains, while prediction benefits are Trace-Dependent: A cluster-size sweep and cross-trace bootstrap of forecast-driven GPU scheduling in the LLM era},
journal = {Array},
year = {2026},
volume = {32},
pages = {101374},
doi = {10.1016/j.array.2026.101374}
}
EUVIMED