Search arXiv⌕ Search
Home/Archive/arXiv

arXiv · 2507.09337

ORCHA: A Performance Portability System for Extreme Heterogeneity

Youngjun Lee·Klaus Weide·Wesley Kwiecinski·Jared O'Neal·Johann Rudi·Anshu Dubey
View original source ↗Read source document ↗Export metadata ↓

Abstract

Heterogeneity is the prevalent trend in the rapidly evolving high-performance computing (HPC) landscape in both hardware and application software. The diversity in hardware platforms, currently comprising various accelerators and a future possibility of specializable chiplets, poses a significant challenge for scientific software developers aiming to harness optimal performance across different computing platforms while maintaining the quality of solutions when their applications are simultaneously growing more complex. Code synthesis and code generation can provide mechanisms to mitigate this challenge. We have developed a divide-and-conquer approach where different aspects of performance are handled by different stand-alone tools that are interfaced with the application through generated code. This system, ORCHA, enables users to configure and orchestrate their computations among available resources on a platform by specifying a high-level recipe, thereby permitting a many-to-many paradigm where each recipe results in a different variant of the application. The core design goal is to let users decide the application's hardware mapping and orchestration by editing only the high-level recipe. Tools in ORCHA distribution are: CG-Kit for translating the recipe into an execution graph; Milhoja to execute the graph by orchestrating data and task among hardware resources; and Macroprocessor that enables users to define their own code-shorthand for higher composability and easier management of code variants. In this paper, we describe the design of ORCHA and the role that code-generation plays in isolating applications from tools. We demonstrate the breadth of configurations ORCHA enables with a case study in which an application configuration is realized on three distinct hardware mappings---a GPU-centric, a CPU/GPU balanced, and a CPU/GPU concurrent layouts by using different recipes.

Explore related subjects

math.NAcs.NA

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Youngjun Lee, Klaus Weide, Wesley Kwiecinski, Jared O'Neal, Johann Rudi, Anshu Dubey. 2026-09-09. ORCHA: A Performance Portability System for Extreme Heterogeneity. https://doi.org/10.1177/10943420261465929

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

KEEP EXPLORING

Related papers

arXiv2026

A Stabilized Finite Element Method for a Morpho-Visco-Poroelastic Model

Studying the structure of soft tissues is important and relevant in biology, particularly in some diseases, such as tumor growth and dermal contraction after burn injury. Based on the complicated characteristics of the tissue and for the sake of a better understanding of the underlying biomechanics, we propose a mathematical model that combines elastic, viscous, and porous effects with growth or shrinkage due to microstructural changes. The framework is referred to as morpho-visco-poroelasticity. Although the existence results of the solution to the problem are not given in this study, we assess the stability of the equilibria for both the continuous and semi-discrete versions of the model, and the key features of this modelling framework have been discussed. To obtain reliable numerical solutions, a stabilized finite element (FE) scheme is proposed for the morpho-visco-poroelasticity equations to avoid spurious oscillations in the pressure profile; the success of this FE scheme is verified by numerical simulations and convergence investigation in both spatial and temporal aspects. For a more quantitative assessment, the total variation of the pressure profile is evaluated as a function of the stabilization parameter.

math.NA↗
arXiv2026

Efficient third-order iterative algorithms for computing zeros of special functions

This manuscript presents a novel and reliable third-order iterative procedure for computing the zeros of solutions to second-order ordinary differential equations. By approximating the solution of the related Riccati differential equation using the trapezoidal rule, this study has derived the proposed third-order method. This work establishes sufficient conditions to ensure the theoretical non-local convergence of the proposed method. This study provides suitable initial guesses for the proposed third-order iterative procedure to compute all zeros in a given interval of the solutions to second-order ordinary differential equations. The orthogonal polynomials like Legendre and Hermite, as well as the special functions like Bessel, Coulomb wave, confluent hypergeometric, and cylinder functions, satisfy the proposed conditions for convergence. Numerical simulations demonstrate the effectiveness of the proposed theory. This work also presents a comparative analysis with recent studies.

math.NA↗
arXiv2026

Machine-Learning-Enhanced Discretize-then-Project Reduced-Order Modeling of Turbulent Flows on Collocated Grids

This study presents a hybrid reduced-order modeling (ROM) framework for incompressible flows on collocated finite-volume grids, combining a discretize-then-project consistent-flux formulation for velocity and pressure with a non-intrusive neural-network closure for turbulent viscosity. The intrusive formulation preserves discrete mass conservation and pressure-velocity coupling, while a reduced pressure reference-cell constraint fixes pressure gauge ambiguity. We evaluate Multilayer Perceptron (MLP), Transformer, and Long Short-Term Memory (LSTM) closures. For a three-dimensional lid-driven cavity at $Re=100$, the LSTM-based ROM achieves relative errors of 0.7% in velocity and 4% in turbulent viscosity. At $Re=3200$, a mode-sensitivity study identifies $N=15$ POD modes as the best overall configuration, balancing accuracy, dimension, robustness, and cost. It yields a final relative velocity error of approximately 12.3% and an online wall-clock speedup of approximately $50\times$ over the full-order model; energy and enstrophy errors remain below 11% for all three architectures. This regime requires case-specific neural-network retraining and pressure reference-cell parameter retuning. In a time-extrapolation test trained on $t\in[0,3]$,s and rolled out to $t=6$,s, the ROM remains bounded, although velocity and pressure errors increase beyond the training window. The LSTM turbulent-viscosity closure remains robust, identifying long-horizon pressure accuracy as the main limitation. These results demonstrate the potential of consistent projection-based modeling combined with data-driven turbulence closure for efficient reduced-order simulation.

math.NA↗