Developing a BLAS library for the AMD AI Engine

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Laan, Tristan, De Matteis, Tiziano
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866909331757203456
author Laan, Tristan
De Matteis, Tiziano
author_facet Laan, Tristan
De Matteis, Tiziano
contents Spatial (dataflow) computer architectures can mitigate the control and performance overhead of classical von Neumann architectures such as traditional CPUs. Driven by the popularity of Machine Learning (ML) workloads, spatial devices are being marketed as ML inference accelerators. Despite providing a rich software ecosystem for ML practitioners, their adoption in other scientific domains is hindered by the steep learning curve and lack of reusable software, which makes them inaccessible to non-experts. We present our ongoing project AIEBLAS, an open-source, expandable implementation of Basic Linear Algebra Routines (BLAS) for the AMD AI Engine. Numerical routines are designed to be easily reusable, customized, and composed in dataflow programs, leveraging the characteristics of the targeted device without requiring the user to deeply understand the underlying hardware and programming model.
format Preprint
id arxiv_https___arxiv_org_abs_2410_00825
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Developing a BLAS library for the AMD AI Engine
Laan, Tristan
De Matteis, Tiziano
Distributed, Parallel, and Cluster Computing
Emerging Technologies
Spatial (dataflow) computer architectures can mitigate the control and performance overhead of classical von Neumann architectures such as traditional CPUs. Driven by the popularity of Machine Learning (ML) workloads, spatial devices are being marketed as ML inference accelerators. Despite providing a rich software ecosystem for ML practitioners, their adoption in other scientific domains is hindered by the steep learning curve and lack of reusable software, which makes them inaccessible to non-experts. We present our ongoing project AIEBLAS, an open-source, expandable implementation of Basic Linear Algebra Routines (BLAS) for the AMD AI Engine. Numerical routines are designed to be easily reusable, customized, and composed in dataflow programs, leveraging the characteristics of the targeted device without requiring the user to deeply understand the underlying hardware and programming model.
title Developing a BLAS library for the AMD AI Engine
topic Distributed, Parallel, and Cluster Computing
Emerging Technologies
url https://arxiv.org/abs/2410.00825