A functional pattern-based language in MLIR

Martin Lücke; Michel Steuwer; Aaron Smith

A functional pattern-based language in MLIR

Martin Lücke, Michel Steuwer, Aaron Smith

Research output: Contribution to conference › Paper › peer-review

Abstract

Machine learning systems are stuck in a rut. Paul Barham and Michael Isard, two of the original authors of TensorFlow, come to this conclusion in their recent HotOS paper. They argue that while TensorFlow and similar frameworks have enabled great advances in machine learning, their current design and implementations focus on a fixed set of monolithic and inflexible kernels. They continueto say that “this reliance on high performance but inflexible kernels reinforces the dominant style of programming model” and argue that “these programming abstractions lack expressiveness, maintainability, and modularity; all of which hinders research progress”.

Original language	English
Number of pages	6
Publication status	Published - 31 May 2020
Event	2nd Workshop on Accelerated Machine Learning @ ISCA 2020 - Virtual workshop Duration: 31 May 2020 → 31 May 2020 http://workshops.inf.ed.ac.uk/accml/

Workshop

Workshop	2nd Workshop on Accelerated Machine Learning @ ISCA 2020
Abbreviated title	AccML 2020
City	Virtual workshop
Period	31/05/20 → 31/05/20
Internet address	http://workshops.inf.ed.ac.uk/accml/

Access to Document

A functional_LUCKE_DOA20052020_AFVAccepted author manuscript, 563 KBLicence: All Rights Reserved

Cite this

@conference{98d574fcf471473898a887a125db43cc,

title = "A functional pattern-based language in MLIR",

abstract = "Machine learning systems are stuck in a rut. Paul Barham and Michael Isard, two of the original authors of TensorFlow, come to this conclusion in their recent HotOS paper. They argue that while TensorFlow and similar frameworks have enabled great advances in machine learning, their current design and implementations focus on a fixed set of monolithic and inflexible kernels. They continueto say that “this reliance on high performance but inflexible kernels reinforces the dominant style of programming model” and argue that “these programming abstractions lack expressiveness, maintainability, and modularity; all of which hinders research progress”.",

author = "Martin L{\"u}cke and Michel Steuwer and Aaron Smith",

year = "2020",

month = may,

day = "31",

language = "English",

note = "2nd Workshop on Accelerated Machine Learning @ ISCA 2020, AccML 2020 ; Conference date: 31-05-2020 Through 31-05-2020",

url = "http://workshops.inf.ed.ac.uk/accml/",

}

TY - CONF

T1 - A functional pattern-based language in MLIR

AU - Lücke, Martin

AU - Steuwer, Michel

AU - Smith, Aaron

PY - 2020/5/31

Y1 - 2020/5/31

N2 - Machine learning systems are stuck in a rut. Paul Barham and Michael Isard, two of the original authors of TensorFlow, come to this conclusion in their recent HotOS paper. They argue that while TensorFlow and similar frameworks have enabled great advances in machine learning, their current design and implementations focus on a fixed set of monolithic and inflexible kernels. They continueto say that “this reliance on high performance but inflexible kernels reinforces the dominant style of programming model” and argue that “these programming abstractions lack expressiveness, maintainability, and modularity; all of which hinders research progress”.

AB - Machine learning systems are stuck in a rut. Paul Barham and Michael Isard, two of the original authors of TensorFlow, come to this conclusion in their recent HotOS paper. They argue that while TensorFlow and similar frameworks have enabled great advances in machine learning, their current design and implementations focus on a fixed set of monolithic and inflexible kernels. They continueto say that “this reliance on high performance but inflexible kernels reinforces the dominant style of programming model” and argue that “these programming abstractions lack expressiveness, maintainability, and modularity; all of which hinders research progress”.

M3 - Paper

T2 - 2nd Workshop on Accelerated Machine Learning @ ISCA 2020

Y2 - 31 May 2020 through 31 May 2020

ER -

A functional pattern-based language in MLIR

Abstract

Workshop

Access to Document

Fingerprint

Cite this