Leveraging Hardware-Aware Computation in Mixed-Precision Matrix Multiply: A Tile-Centric Approach

2025/08/20 by Qiao Zhang, Zhang, Qiao, Rabab Alomairy +7
Computer Science · #Cellular Automata and Applications #Distributed #FOS: Computer and information sciences #Interconnection Networks and Systems #Parallel #Parallel Computing and Optimization Techniques #and Cluster Computing (cs.DC)

paper · pdf · doi:10.48550/arxiv.2508.14848

openalex publication_date 2025/08/20 · openalex created_date 2025/10/09 · openalex updated_date 2026/07/28

Abstract

General Matrix Multiplication (GEMM) is a critical operation underpinning a wide range of applications in high-performance computing (HPC) and artificial intelligence (AI). The emergence of hardware optimized for low-precision arithmetic necessitates a reevaluation of numerical algorithms to leverage mixed-precision computations, achieving improved performance and energy efficiency. This research introduces an adaptive mixed-precision GEMM framework that supports different precision formats at fine-grained tile/block levels. We utilize the PaRSEC runtime system to balance workloads across various architectures. The performance scales well on ARM CPU-based Fugaku supercomputer, Nvidia GPU-based A100 DGX, and AMD GPU-based Frontier supercomputer. This research aims to enhance computational efficiency and accuracy by bridging algorithmic advancements and hardware innovations, driving transformative progress in various applications.

Citations

• Toward Capturing Genetic Epistasis From Multivariate Genome-Wide Association Studies Using Mixed-Precision Kernel Ridge Regression
• Boosting Earth System Model Outputs And Saving PetaBytes in their Storage Using Exascale Climate Emulators
• Mixed-Precision Computing in the GRIST Dynamical Core for Weather and Climate Modelling
• On floating point precision in computational fluid dynamics using OpenFOAM
• Pushing the limit of molecular dynamics with ab initio accuracy to 100 million atoms with machine learning
• Mixed Precision Solver Scalable to 16000 MPI Processes for Lattice Quantum Chromodynamics Simulations on the Oakforest-PACS System
• Attention Is All You Need

Related

• Cascading GEMM: High Precision from Low Precision
• The Cambrian Explosion of Mixed-Precision Matrix Multiplication for Quantized Deep Learning Inference
• Optimizing Irregular-Shaped Matrix-Matrix Multiplication on Multi-Core DSPs
• A Flexible Instruction Set Architecture for Efficient GEMMs
• Open-Source GEMM Hardware Kernels Generator: Toward Numerically-Tailored Computations