SET is a new CUDA runtime framework that combines event-chaining, work-stealing, and per-stream buffers in graph-based pipelines to deliver 1.15-1.44X speedups and 18-54% lower scheduling overhead versus prior CUDA graph methods.
In: 56th Annual IEEE/ACM International Symposium on Microarchitecture
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.DC 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
SET: Stream-Event-Triggered Scheduling for Efficient CUDA Graph Pipelines
SET is a new CUDA runtime framework that combines event-chaining, work-stealing, and per-stream buffers in graph-based pipelines to deliver 1.15-1.44X speedups and 18-54% lower scheduling overhead versus prior CUDA graph methods.