Searcharxiv⌕ Search
Home/Archive/arXiv

arXiv · 2610.23383

Data-Driven Prediction of Optimal MPI Rank Placement and Node Count for Memory-Bound CFD on AMD EPYC Clusters

Marcin Lawenda·Łukasz Szustak·Aleksandra Krasicka
View original source ↗Read source document ↗Export metadata ↓

Abstract

Users who run a memory-bound OpenFOAM case on an AMD EPYC cluster must decide how many nodes to use and how many MPI ranks to place on each core complex. The fastest choice is often not the cheapest one. This paper revisits our earlier analytical model of this performance and energy trade-off with 560 measured configurations of the motorBike case on five Zen 4 and Zen 5 EPYC processors, including socket power from processor telemetry. On up to four nodes, small meshes are both fastest and cheapest on a single node on most processors. For the largest mesh the largest node count is both the fastest and the cheapest option, because its speedup is superlinear. A genuine trade-off arises only for medium meshes, where full speed costs a median of 20 to 97% more node hours than the cheapest option. The L3 miss rate and the halo traffic per cell are consistent with this pattern. For a new mesh of the same case, the configuration that was optimal on the nearest measured mesh is near optimal in 75 to 85% of the decisions. Running the two configurations it recommends for speed and for cost raises this share to 85 to 90% for runtime and node hours and determines the trade-off with median errors of 3 to 5 percentage points. Neither gradient boosted trees trained on the same data nor the earlier polynomial model improves on this simple procedure.

Explore related subjects

cs.PFcs.LG

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Marcin Lawenda, Łukasz Szustak, Aleksandra Krasicka. 2026-10-16. Data-Driven Prediction of Optimal MPI Rank Placement and Node Count for Memory-Bound CFD on AMD EPYC Clusters. https://arxiv.org/abs/2610.23383

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

PairSight: Characterizing and Predicting Operator-Level Interference in GPU Sharing for DNN Inference

Serving DNN inference is costly, yet a single task seldom saturates both the compute units and the memory bandwidth of a GPU. GPU sharing raises utilization by co-locating several tasks on one GPU, at the cost of interference among them. Existing systems estimate this interference with binary compute-memory rules, standalone scaling curves, or model-level predictors. However, these estimates rest on a limited understanding of interference. They leave the software and hardware properties of co-located operators largely unexplored and give no slowdown prediction that is both interpretable and accurate. We conduct an operator-level study of interference under SM partitioning and use hardware counters to trace it to contention on the shared memory path, where the slowdown depends jointly on both co-located operators. Based on these findings, we propose PairSight, a hardware-aware interference predictor that learns how the hardware behaviors of co-located tasks interact and predicts the latency of each task before co-location. On six DNN models including LLMs, PairSight predicts latency under sharing with 4.3% mean error, against 19.0% for a compute-memory intensity rule, and adapts to unseen models and inputs with a few samples.

cs.PF↗
arXiv2026

The Optimal Ranking Tradeoffs Between Two Ranking Scores in Two-Class Classification

The accuracy, true negative rate (negative class recall), true positive rate (positive class recall), negative predictive value (negative class precision), positive predictive value (positive class precision), class-specific intersections over unions, class-specific F-scores, Cohen's kappa, and balanced accuracy are scores that one intuitively wants to maximize when developing machine learning methods. However, they lead to different rankings of such methods. So, how can we establish a ranking that is an optimal tradeoff between any pair of these scores? It turns out that all these scores are special cases of ranking scores: a continuum of scores parameterized by the relative importance values assigned to the true negative, false positive, false negative, and true positive outcomes. Since all these scores are known to induce meaningful rankings, we seek the optimal tradeoffs among them. We consider the tradeoff between two scores to be optimal when it minimizes the Fréchet variance regarding the Kendall distance (also known as the bubble-sort distance) between rankings. For any given set of classifiers to rank, the optimal tradeoff lies at the center of the shortest (geodesic) path on the manifold of all rankings induced by the ranking scores. However, in the parameter space of the ranking scores, it is rarely achieved by averaging the importance values corresponding to the initial scores. In this paper, we present an exact analytical solution for determining a ranking score that is an optimal tradeoff for any given pair of initial scores, in the common case where the class priors are the same for all compared performances. We also experimentally investigate what happens when performance follows some distributions with fixed priors.

cs.PF↗
arXiv2026

Phase-Scoped Query Layouts for Reference-Preserving Speculative Decoding

Speculative verification changes how a target model batches its queries. In a quantized deployment, that change can alter the generated tokens even when the verification algorithm is mathematically valid. We present a phase-scoped query-to-batch (QB) transformation that preserves an existing greedy reference while retaining batched projections and feed-forward computation. QB represents independent attention queries through matched query and mask views, shares K/V storage, and makes the selected layout part of graph reuse. In a six-session experiment spanning two dates, it reduces speculative request latency relative to explicit per-query attention by 3.06%, 6.17% and 2.83% for Qwen3, SmolLM2 and Llama, with complete reference-output agreement on the fixed panel. A separate 72-context CUDA replay covers three layer positions, both execution phases and query widths one to four; QB and per-query attention agree bitwise at QK, softmax and AV. Controlled Vulkan experiments identify two limits to transferring this local repair. Earlier computation can create consequential state differences, and endpoint-equivalent states need not remain equivalent under component replacement. Independently, the lifted query axis fails a necessary operator-support condition in the frozen Vulkan backend. Together, the results establish a concrete local implementation benefit and distinguish the representation, state and backend conditions that a reference-preserving deployment must satisfy.

cs.PF↗