2026-10-23T02:36:40Z https://drops.dagstuhl.de/oai

oai:drops-oai.dagstuhl.de:23599 2025-10-27T09:21:57Z ddc:004 open_access

Hardware Compute Partitioning on NVIDIA GPUs for Composable Systems Bakita, Joshua Anderson, James H. Real-time systems composable systems graphics processing units CUDA As GPU-using tasks become more common in embedded, safety-critical systems, efficiency demands necessitate sharing a single GPU among multiple tasks. Unfortunately, existing ways to schedule multiple tasks onto a GPU often either result in a loss of ability to meet deadlines, or a loss of efficiency. In this work, we develop a system-level spatial compute partitioning mechanism for NVIDIA GPUs and demonstrate that it can be used to execute tasks efficiently without compromising timing predictability. Our tool, called nvtaskset, supports composable systems by not requiring task, driver, or hardware modifications. In our evaluation, we demonstrate sub-1-μs overheads, stronger partition enforcement, and finer-granularity partitioning when using our mechanism instead of NVIDIA’s Multi-Process Service (MPS) or Multi-instance GPU (MiG) features. Schloss Dagstuhl – Leibniz-Zentrum für Informatik Joshua Bakita and James H. Anderson 2025 Is Part Of LIPIcs, Volume 335, 37th Euromicro Conference on Real-Time Systems (ECRTS 2025) InProceedings Text doc-type:ResearchArticle publishedVersion application/pdf doi:10.4230/LIPIcs.ECRTS.2025.21 urn:nbn:de:0030-drops-235998 https://drops.dagstuhl.de/entities/document/10.4230/LIPIcs.ECRTS.2025.21 eng https://creativecommons.org/licenses/by/4.0/legalcode