Answers are grounded in this paper and cite their sources.
Empirical study introduces a framework to measure execution variability in computer systems, suggesting more precise evaluations.
This work introduces a schema-based measurement framework for observing execution-level variability and redundant computation in computer systems. Conventional system evaluation commonly relies on aggregate metrics such as latency, throughput, and utilization. While useful, these metrics do not fully capture changes in realized execution structure or the presence of additional computation across repeated executions. The proposed framework defines execution-level concepts including execution path variability, reuse failure, redundant computation, observed work, reference work, work inflation, and recompute ratio as measurable and reproducible observable properties of system behavior. The experiments are implemented using Argus, an execution-level observation and reporting framework, together with execution-instability-bench, a reproducible workload and benchmarking framework for execution observation experiments. The included experiments are conducted in a controlled repeated-execution environment on a local CPU-based macOS system. The reported observations demonstrate measurable differences in execution path behavior, repeated operations, reuse failures, and work inflation across repeated execution conditions. The purpose of this work is not to establish universal causal claims about execution behavior, but to provide a reproducible measurement framework for observing and characterizing execution-level variability and redundant computation across computing environments.
No takes yet. Share an insight, caveat, or question.
Hakjun Kim (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: