June 29, 2026 | Randles Lab
Ayman Yousef presented this work at the 2026 Platform for Advanced Scientific Computing (PASC) Conference. The paper was co-authored by Aristotle Martin and Amanda Randles.
Randles Lab PhD student Ayman Yousef presented new research at the 2026 Platform for Advanced Scientific Computing (PASC) Conference on making analysis of large-scale scientific simulations more flexible and efficient.
The paper, “GPU Halo Replay: Lossless Twin Simulations for Flexible In Situ Analysis of Stencil-Based Solvers,” by Ayman Yousef, Aristotle Martin, and Amanda Randles, introduces GPU Halo Replay, a framework that enables lossless reconstruction of simulation data for analysis by dedicated computational twin simulations.
Large-scale simulations can generate enormous amounts of data, creating a fundamental challenge: storing every output for later analysis is often impractical, while performing analysis during the simulation can consume computational resources and constrain what researchers can examine. GPU Halo Replay provides an alternative by capturing the information needed to reproduce simulation behavior and allowing separate computational twins to replay and analyze the simulation without slowing the primary run or sacrificing fidelity.
The approach creates new flexibility for in situ analysis, scientific visualization, and large-scale computational workflows. It also supports the Randles Lab’s broader efforts to develop scalable computational infrastructure for simulations that increasingly span massive spatial and temporal scales.