arXiv subjects

Paul H J Kelly

Publications and source records attributed to Paul H J Kelly.

4 recordsLinked to original sources

arXiv2026

PG-MDP: Profile-Guided Memory Dependence Prediction for Area-Constrained Cores

Memory Dependence Prediction (MDP) is a speculative technique to predict which stores, if any, a given load will depend on. Area-constrained cores are increasingly relevant in various applications such as energy-efficient or edge systems, and often have limited space for MDP tables. This leads to a high rate of false dependencies as memory independent loads alias with unrelated predictor entries, causing unnecessary stalls in the processor pipeline. The conventional way to address this problem is with greater predictor size or complexity, but this is unattractive on area-constrained cores. This paper demonstrates that targeting the predictor working set delivers the majority of available performance without scaling any hardware structures. We achieve this with profile-guided memory dependence prediction (PG-MDP), a hardware-software co-design to label consistently memory independent loads via their opcode and remove them from the MDP working set. These loads bypass querying the MDP and always issue as soon as possible. In the event that a labeled load incorrectly passes a store to the same address, a rollback is triggered as usual but no new MDP entry is created. Across the SPECspeed 2017 suites, PG-MDP reduces MDP load queries by 80%, false dependencies by 84%, and improves geomean IPC for a small (ROB=128) simulated core by 4.6% (to within 1.2% of the IPC when using 8x more predictor entries), with no area cost and no additional instruction bandwidth.

cs.PL↗

arXiv2020

Scalable Uncertainty for Computer Vision with Functional Variational Inference

As Deep Learning continues to yield successful applications in Computer Vision, the ability to quantify all forms of uncertainty is a paramount requirement for its safe and reliable deployment in the real-world. In this work, we leverage the formulation of variational inference in function space, where we associate Gaussian Processes (GPs) to both Bayesian CNN priors and variational family. Since GPs are fully determined by their mean and covariance functions, we are able to obtain predictive uncertainty estimates at the cost of a single forward pass through any chosen CNN architecture and for any supervised learning task. By leveraging the structure of the induced covariance matrices, we propose numerically efficient algorithms which enable fast training in the context of high-dimensional tasks such as depth estimation and semantic segmentation. Additionally, we provide sufficient conditions for constructing regression loss functions whose probabilistic counterparts are compatible with aleatoric uncertainty quantification.

cs.CV↗

arXiv2015

A Fast and Scalable Graph Coloring Algorithm for Multi-core and Many-core Architectures

Irregular computations on unstructured data are an important class of problems for parallel programming. Graph coloring is often an important preprocessing step, e.g. as a way to perform dependency analysis for safe parallel execution. The total run time of a coloring algorithm adds to the overall parallel overhead of the application whereas the number of colors used determines the amount of exposed parallelism. A fast and scalable coloring algorithm using as few colors as possible is vital for the overall parallel performance and scalability of many irregular applications that depend upon runtime dependency analysis. Catalyurek et al. have proposed a graph coloring algorithm which relies on speculative, local assignment of colors. In this paper we present an improved version which runs even more optimistically with less thread synchronization and reduced number of conflicts compared to Catalyurek et al.'s algorithm. We show that the new technique scales better on multi-core and many-core systems and performs up to 1.5x faster than its predecessor on graphs with high-degree vertices, while keeping the number of colors at the same near-optimal levels.

cs.DC↗

arXiv2014

Multidisciplinary Engineering Models: Methodology and Case Study in Spreadsheet Analytics

This paper demonstrates a methodology to help practitioners maximise the utility of complex multidisciplinary engineering models implemented as spreadsheets, an area presenting unique challenges. As motivation we investigate the expanding use of Integrated Resource Management(IRM) models which assess the sustainability of urban masterplan designs. IRM models reflect the inherent complexity of multidisciplinary sustainability analysis by integrating models from many disciplines. This complexity makes their use time-consuming and reduces their adoption. We present a methodology and toolkit for analysing multidisciplinary engineering models implemented as spreadsheets to alleviate such problems and increase their adoption. For a given output a relevant slice of the model is extracted, visualised and analysed by computing model and interdisciplinary metrics. A sensitivity analysis of the extracted model supports engineers in their optimisation efforts. These methods expose, manage and reduce model complexity and risk whilst giving practitioners insight into multidisciplinary model composition. We report application of the methodology to several generations of an industrial IRM model and detail the insight generated, particularly considering model evolution.

cs.SE↗