CS Paper Radar

Daily listing

2026-11-17 · cs.PF

4 papers first submitted this day (UTC) in cs.PF — Performance. arXiv announces a day's submissions the next evening, so the newest day fills up over ~24 h.

Score = weighted mean ofCitations 20%Upvotes 10%Code stars 10%Venue 15%Code/data 10%Novelty 15%Momentum 15%Substance 5%how it works

Systems & Networking 3

On Traffic Characterization in Time-Sensitive Networking and Deterministic Networking

cs.NI cs.PF ·2026-11-17 ·Yuming Jiang
Abstract

The traffic of an audio/video stream in IEEE Time-Sensitive Networking (TSN) or an asynchronous traffic flow in IETF Deterministic Networking is specified by three parameters: an observation time interval, a maximum packet size, and a maximum number of packets in the interval. This traffic specification (TSpec) triggers three fundamental questions for TSN and DetNet: (1) if and how the traffic of a flow can be specified by these parameters after it traverses a system, (2) how to specify the traffic of a flow aggregate using the three parameters, and (3) how to use it in performance guarantee analysis. Among them, the literature has focused on (3), leaving the other two little investigated and largely open. This paper is intended to bridge the gap. Specifically, we introduce an equivalent traffic characterization for TSN TSpec, which is simple in representation and simplifies analytical development. Based on this result, we establish an output characterization property for (1) and show that the output of a TSN flow traversing a bounded-delay system can be characterized by the TSN TSpec with updated parameters. For (2) we study the superposition of TSN flows and derive TSN traffic characterizations for the aggregate of multiple TSN flows. For (3), based on a relationship between the TSN traffic characterization and the well-known token-bucket traffic characterization, backlog and delay bounds are derived. Together, these results form a calculus for TSN traffic characterization.

0.26

Critical heavy traffic and diffusion limits for shortest-of-d dynamic routing

cs.PF math.PR ·2026-11-17 ·Elmira Yu. Kalimulina
Abstract

We study shortest-of-$d$ routing, with fixed $d\ge2$, in $N$ parallel queues with independent finite-state Markov service environments switching at rate $κ_N$. At the critical load $ρ_N=1-1/N$, the empirical measure of the centered queue lengths, under stationarity, converges in distribution to a random probability measure determined by its mean. We identify the mean's density and the conditional independence of fixed finite collections of centered queues in the limit. On time scales of order $N$, the centered empirical mean converges to a diffusion with explicit drift and constant diffusion coefficient. These limits hold when $κ_N/N\to\infty$, along subsequences with convergent lattice phase. The diffusion limit extends to centered initial empirical measures tight in the first Wasserstein topology, with means converging in law. A short initial layer prepares the limiting family while preserving the limiting mean. Empirical approximation holds in time-integrated Wasserstein distance.

0.24

How well does graph superoptimization transfer? Adapting tensor-graph superoptimizer to a Triton-based NPU

cs.PF cs.ARcs.PL ·2026-11-17 ·Weronika Hryniewska-Guzik, Mateusz Comblik, Daniel Murawski, Jakub Schmiegel, Maciej Świetlik +2 more
published
Abstract · Accepted for FlagOS Forum at MICRO 2026 - OASIS-AI workshop: Open Compute Architecture and Software Infrastructure for Heterogeneous AI Syst…

For automatic optimization systems, performance portability across increasingly diverse AI accelerators remains a challenge. We investigate how well graph-level superoptimization transfers across architectures by adapting Mirage from its CUDA-oriented design to the Huawei Ascend 910B3 through a Triton-based compilation stack. We evaluate the adapted system on transformer-relevant workloads, including matrix multiplication, softmax, RMSNorm+MatMul, gated MLP, FNN, and attention. The results show that graph-level optimization can successfully transfer and outperform available reference implementations by up to 87.0% on fused workloads. However, performance is highly dependent on tensor shape and workload structure, with substantial regressions for other configurations. We identify target-specific tensor partitioning, execution-resource modeling, backend lowering and fusion, and search and verification costs as key limitations. Our findings demonstrate that a portable graph representation and programming model do not imply a portable optimization space, motivating modular target-specific mechanisms within otherwise reusable compiler infrastructure.

0.20

Software & Programming 1

RyBench: Runtime Verification Benchmarking Framework over Publish/Subscribe Networks

cs.LO cs.PFcs.SE ·2026-11-17 ·Mehmet Hilmi Dündar, Doğan Ulus
publishedbenchmark
Abstract · In Proceedings FMAS 2026, arXiv:2611.21375

Autonomous systems built as distributed, node-based systems communicate via publish/subscribe middleware. Runtime Verification (RV) tools are used to monitor those systems against formal specifications, receiving monitored events through the pub/sub channel. An RV tool's practical value depends on its monitoring throughput capacity and verdict latency, besides its correctness, so a systematic benchmarking setup is needed to evaluate how efficiently these tools run. This paper presents RyBench, a pub/sub-based benchmarking framework that acts as a synthetic event publisher in place of a monitored system, providing a common benchmarking setup. RyBench separates benchmark generation, serialization, and transport into distinct components, enabling the combination of different event-generation strategies, data formats, and transport layers. We also evaluate the framework's throughput capacity across these configurations using Open Simulation Interface (OSI) schemas and Timescales properties, and illustrate the end-to-end benchmarking flow through an application to Reelay and an integration with MonPoly.

0.51