Ascend to Science: Exploration of AI Chips for Scientific Computing

DOI10.48550/arXiv.2607.20120MaRDI QIDQ7284387arXivFDO

Authors Yongxiang Liu, Xiang-Lin Liu, Yonghong Tian, Weicheng Xue, Kai Yang, Baisong Xu, Dengdong Fan, Xianglin Liu, Pengxiang Xu

Publication date 22 July 2026

Published in arXiv

Tldr We explore the feasibility of deploying scientific computing applications on AI-native NPUs and present strategies for performance improvement through heterogeneous execution, mixed-precision numerical formulations, precision emulation, hierarchical memory orchestration, and communication--computation overlap.

Full work available at URL https://doi.org/10.48550/arXiv.2607.20120

arXiv classification

cs.DC

Cites work

• Error-free transformations of matrix multiplication by using fast routines of matrix multiplication and its applications
• Finding structure with randomness: probabilistic algorithms for constructing approximate matrix decompositions
• GPU accelerated Monte Carlo simulation of the 2D and 3D Ising model
• Hardware Trends Impacting Floating-Point Computations In Scientific Applications
• High performance implementations of the 2D Ising model on GPUs
• HPL-MxP Benchmark: Mixed-Precision Algorithms, Iterative Refinement, and Scalable Data Generation
• JAX-fluids 2.0: towards HPC for differentiable CFD of compressible two-phase flows
• JAX-fluids: a fully-differentiable high-order computational fluid dynamics solver for compressible two-phase flows
• Quantum computation and quantum information
• Streaming low-rank matrix approximation with an application to scientific simulation
• The LINPACK Benchmark: past, present and future

This page was built for publication: Ascend to Science: Exploration of AI Chips for Scientific Computing

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7284387)