Publication:
AXI4MLIR: User-Driven automatic host code generation for custom AXI-Based accelerators

Simple item page
dc.contributor.authorBohm Agostini, Nicolas
dc.contributor.authorHaris, Jude
dc.contributor.authorGibson, Perry
dc.contributor.authorJayaweera, Malith
dc.contributor.authorRubin, Norm
dc.contributor.authorTumeo, Antonino
dc.contributor.authorAbellán, José L.
dc.contributor.authorCano, José
dc.contributor.authorKaeli, David
dc.contributor.departmentIngeniería y Tecnología de Computadores
dc.date.accessioned2024-03-14T08:53:30Z
dc.date.available2024-03-14T08:53:30Z
dc.date.issued2023-12-22
dc.descriptionThis document is a PrePrint . You can find it also in arXiv.org, with DOI: https://doi.org/10.48550/arXiv.2312.14821es
dc.description.abstractThis paper addresses the need for automatic and efficient generation of host driver code for arbitrary custom AXI-based accelerators targeting linear algebra algorithms, an important workload in various applications, including machine learning and scientific computing. While existing tools have focused on automating accelerator prototyping, little attention has been paid to the host-accelerator interaction. This paper introduces AXI4MLIR, an extension of the MLIR compiler framework designed to facilitate the automated generation of host-accelerator driver code. With new MLIR attributes and transformations, AXI4MLIR empowers users to specify accelerator features (including their instructions) and communication patterns and exploit the host memory hierarchy. We demonstrate AXI4MLIR's versatility across different types of accelerators and problems, showcasing significant CPU cache reference reductions (up to 56%) and up to a 1.65x speedup compared to manually optimized driver code implementations. AXI4MLIR implementation is open-source and available at: t: https://github.com/AXI4MLIR/axi4mlires
dc.formatapplication/pdfes
dc.format.extent13es
dc.identifier.doihttps://doi.org/10.48550/arXiv.2312.14821
dc.identifier.urihttp://hdl.handle.net/10201/140191
dc.languageenges
dc.relationDMC Initiative, the AT SCALE Initiative, and the Compiler Frameworks and Hardware Generators to Support Innovative US Government Designs project at Pacific Northwest National Laboratory; Engineering and Physical Sciences Research Council (grant EP/R513222/1); the grant RYC2021-031966-I funded by MCIN/AEI/10.13039/50110es
dc.rightsinfo:eu-repo/semantics/openAccesses
dc.rights.urihttp://creativecommons.org/licenses/by-nc-nd/4.0/*
dc.titleAXI4MLIR: User-Driven automatic host code generation for custom AXI-Based acceleratorses
dc.typeinfo:eu-repo/semantics/articlees
dspace.entity.typePublicationes
Files
Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
AXI4MLIR.pdf
Size:
1.11 MB
Format:
Adobe Portable Document Format
Description:
Download
License bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
license.txt
Size:
2.26 KB
Format:
Item-specific license agreed upon to submission
Description:
Download
Collections
Artículos