Generating Efficient FFT GPU Code with Lift

Bastian Köpcke; Michel Steuwer; Sergei Gorlatch

doi:10.1145/3331553.3342613

Generating Efficient FFT GPU Code with Lift

Bastian Köpcke
, Michel Steuwer
, Sergei Gorlatch

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

The Fast Fourier Transform is a well-known algorithm used in many high-performance applications, ranging from signal processing to convolutional neural networks.

In this paper, we encode FFTs by building high-level abstractions based on a set of functional parallel patterns in the Lift language. Abstractions are derived from and closely resemble mathematical definitions for FFTs. We leverage the Lift performance-portable code generator to generate high performing GPU code for FFTs. No FFT-specific patterns are required for this, showing the expressive power of the generic parallel patterns used in Lift.

Our experimental results show that our approach achieves performance better than AMD's OpenCL implementation clFFT on an Nvidia GPU. Nvidia's highly optimized cuFFT implementation still performs better on their GPUs.

Original language	English
Title of host publication	Proceedings of the 8th ACM SIGPLAN International Workshop on Functional High-Performance and Numerical Computing
Place of Publication	New York, NY, USA
Publisher	ACM Association for Computing Machinery
Pages	1–13
Number of pages	13
ISBN (Print)	9781450368148
DOIs	https://doi.org/10.1145/3331553.3342613
Publication status	Published - 18 Aug 2019
Event	FHPNC 2019 Functional High-Performance and Numerical Computing : ICFP 19 Berlin - Berlin, Germany Duration: 18 Aug 2019 → 18 Aug 2019 https://icfp19.sigplan.org/home/FHPNC-2019#About

Workshop

Workshop	FHPNC 2019 Functional High-Performance and Numerical Computing
Abbreviated title	FHPNC
Country/Territory	Germany
City	Berlin
Period	18/08/19 → 18/08/19
Internet address	https://icfp19.sigplan.org/home/FHPNC-2019#About

Keywords / Materials (for Non-textual outputs)

Code Generation
GPU
FFT
Lift

Access to Document

10.1145/3331553.3342613Licence: All Rights Reserved

https://dl.acm.org/doi/abs/10.1145/3331553.3342613Licence: All Rights Reserved

Cite this

@inproceedings{1c8df810cbaa4473a8f3fab591f95675,

title = "Generating Efficient FFT GPU Code with Lift",

abstract = "The Fast Fourier Transform is a well-known algorithm used in many high-performance applications, ranging from signal processing to convolutional neural networks. In this paper, we encode FFTs by building high-level abstractions based on a set of functional parallel patterns in the Lift language. Abstractions are derived from and closely resemble mathematical definitions for FFTs. We leverage the Lift performance-portable code generator to generate high performing GPU code for FFTs. No FFT-specific patterns are required for this, showing the expressive power of the generic parallel patterns used in Lift. Our experimental results show that our approach achieves performance better than AMD's OpenCL implementation clFFT on an Nvidia GPU. Nvidia's highly optimized cuFFT implementation still performs better on their GPUs.",

keywords = "Code Generation, GPU, FFT, Lift",

author = "Bastian K{\"o}pcke and Michel Steuwer and Sergei Gorlatch",

year = "2019",

month = aug,

day = "18",

doi = "10.1145/3331553.3342613",

language = "English",

isbn = "9781450368148",

pages = "1–13",

booktitle = "Proceedings of the 8th ACM SIGPLAN International Workshop on Functional High-Performance and Numerical Computing",

publisher = "ACM Association for Computing Machinery",

note = "FHPNC 2019 Functional High-Performance and Numerical Computing<br/> : ICFP 19 Berlin, FHPNC ; Conference date: 18-08-2019 Through 18-08-2019",

url = "https://icfp19.sigplan.org/home/FHPNC-2019\#About",

}

Köpcke, B, Steuwer, M & Gorlatch, S 2019, Generating Efficient FFT GPU Code with Lift. in Proceedings of the 8th ACM SIGPLAN International Workshop on Functional High-Performance and Numerical Computing. ACM Association for Computing Machinery, New York, NY, USA, pp. 1–13, FHPNC 2019 Functional High-Performance and Numerical Computing
, Berlin, Germany, 18/08/19. https://doi.org/10.1145/3331553.3342613

Generating Efficient FFT GPU Code with Lift. / Köpcke, Bastian; Steuwer, Michel; Gorlatch, Sergei.
Proceedings of the 8th ACM SIGPLAN International Workshop on Functional High-Performance and Numerical Computing. New York, NY, USA: ACM Association for Computing Machinery, 2019. p. 1–13.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - Generating Efficient FFT GPU Code with Lift

AU - Köpcke, Bastian

AU - Steuwer, Michel

AU - Gorlatch, Sergei

PY - 2019/8/18

Y1 - 2019/8/18

N2 - The Fast Fourier Transform is a well-known algorithm used in many high-performance applications, ranging from signal processing to convolutional neural networks. In this paper, we encode FFTs by building high-level abstractions based on a set of functional parallel patterns in the Lift language. Abstractions are derived from and closely resemble mathematical definitions for FFTs. We leverage the Lift performance-portable code generator to generate high performing GPU code for FFTs. No FFT-specific patterns are required for this, showing the expressive power of the generic parallel patterns used in Lift. Our experimental results show that our approach achieves performance better than AMD's OpenCL implementation clFFT on an Nvidia GPU. Nvidia's highly optimized cuFFT implementation still performs better on their GPUs.

AB - The Fast Fourier Transform is a well-known algorithm used in many high-performance applications, ranging from signal processing to convolutional neural networks. In this paper, we encode FFTs by building high-level abstractions based on a set of functional parallel patterns in the Lift language. Abstractions are derived from and closely resemble mathematical definitions for FFTs. We leverage the Lift performance-portable code generator to generate high performing GPU code for FFTs. No FFT-specific patterns are required for this, showing the expressive power of the generic parallel patterns used in Lift. Our experimental results show that our approach achieves performance better than AMD's OpenCL implementation clFFT on an Nvidia GPU. Nvidia's highly optimized cuFFT implementation still performs better on their GPUs.

KW - Code Generation

KW - GPU

KW - FFT

KW - Lift

U2 - 10.1145/3331553.3342613

DO - 10.1145/3331553.3342613

M3 - Conference contribution

SN - 9781450368148

SP - 1

EP - 13

BT - Proceedings of the 8th ACM SIGPLAN International Workshop on Functional High-Performance and Numerical Computing

PB - ACM Association for Computing Machinery

CY - New York, NY, USA

T2 - FHPNC 2019 Functional High-Performance and Numerical Computing<br/>

Y2 - 18 August 2019 through 18 August 2019

ER -

Generating Efficient FFT GPU Code with Lift

Abstract

Workshop

Keywords / Materials (for Non-textual outputs)

Access to Document

Fingerprint

Cite this