OpenCL JIT Compilation for Dynamic Programming Languages

Juan  Fumero; Michel Steuwer; Lukas Stadler; Christophe Dubach

OpenCL JIT Compilation for Dynamic Programming Languages

Juan Fumero, Michel Steuwer, Lukas Stadler, Christophe Dubach

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

Abstract

Graphics Processor Units (GPUs) are powerful hardware to parallelize and speed-up applications. However, programming these devices is too complex for most users and the existing standards for GPU programming are available only for low-level languages such as C.
Dynamic programming languages offer higher abstractions and functionality for many users. GPU programming is possible for dynamic languages through external libraries or via wrappers in which the GPU code is normally written in C. Either way, programmers have to rely on third-party libraries with a limited number of operations or program the GPU kernels themselves.
In this work we present a technique to automatically offload parts of the input program written in a dynamic language into OpenCL without any changes in the original source code. Our preliminary results show we achieve speedups of up to 150x when using the GPU.

Original language	English
Number of pages	2
Publication status	Published - 3 Apr 2017
Event	Workshop on Modern Language Runtimes, Ecosystems, and VMs @ <Programming> 2017 - Brussels, Belgium Duration: 3 Apr 2017 → 3 Apr 2017 https://2017.programming-conference.org/track/MoreVMs-2017-papers

Conference

Conference	Workshop on Modern Language Runtimes, Ecosystems, and VMs @ <Programming> 2017
Abbreviated title	MoreVMs 2017
Country/Territory	Belgium
City	Brussels
Period	3/04/17 → 3/04/17
Internet address	https://2017.programming-conference.org/track/MoreVMs-2017-papers

Access to Document

OpenCL JIT_FUMERO_DOA01032017_AFVFinal published version, 164 KBLicence: Creative Commons: Attribution (CC-BY)

Cite this

@conference{1d1e5248103c47e49bdffc5ca4284f98,

title = "OpenCL JIT Compilation for Dynamic Programming Languages",

abstract = "Graphics Processor Units (GPUs) are powerful hardware to parallelize and speed-up applications. However, programming these devices is too complex for most users and the existing standards for GPU programming are available only for low-level languages such as C. Dynamic programming languages offer higher abstractions and functionality for many users. GPU programming is possible for dynamic languages through external libraries or via wrappers in which the GPU code is normally written in C. Either way, programmers have to rely on third-party libraries with a limited number of operations or program the GPU kernels themselves. In this work we present a technique to automatically offload parts of the input program written in a dynamic language into OpenCL without any changes in the original source code. Our preliminary results show we achieve speedups of up to 150x when using the GPU.",

author = "Juan Fumero and Michel Steuwer and Lukas Stadler and Christophe Dubach",

year = "2017",

month = apr,

day = "3",

language = "English",

note = "Workshop on Modern Language Runtimes, Ecosystems, and VMs @ \<Programming\> 2017, MoreVMs 2017 ; Conference date: 03-04-2017 Through 03-04-2017",

url = "https://2017.programming-conference.org/track/MoreVMs-2017-papers",

}

TY - CONF

T1 - OpenCL JIT Compilation for Dynamic Programming Languages

AU - Fumero, Juan

AU - Steuwer, Michel

AU - Stadler, Lukas

AU - Dubach, Christophe

PY - 2017/4/3

Y1 - 2017/4/3

N2 - Graphics Processor Units (GPUs) are powerful hardware to parallelize and speed-up applications. However, programming these devices is too complex for most users and the existing standards for GPU programming are available only for low-level languages such as C. Dynamic programming languages offer higher abstractions and functionality for many users. GPU programming is possible for dynamic languages through external libraries or via wrappers in which the GPU code is normally written in C. Either way, programmers have to rely on third-party libraries with a limited number of operations or program the GPU kernels themselves. In this work we present a technique to automatically offload parts of the input program written in a dynamic language into OpenCL without any changes in the original source code. Our preliminary results show we achieve speedups of up to 150x when using the GPU.

AB - Graphics Processor Units (GPUs) are powerful hardware to parallelize and speed-up applications. However, programming these devices is too complex for most users and the existing standards for GPU programming are available only for low-level languages such as C. Dynamic programming languages offer higher abstractions and functionality for many users. GPU programming is possible for dynamic languages through external libraries or via wrappers in which the GPU code is normally written in C. Either way, programmers have to rely on third-party libraries with a limited number of operations or program the GPU kernels themselves. In this work we present a technique to automatically offload parts of the input program written in a dynamic language into OpenCL without any changes in the original source code. Our preliminary results show we achieve speedups of up to 150x when using the GPU.

M3 - Paper

T2 - Workshop on Modern Language Runtimes, Ecosystems, and VMs @ <Programming> 2017

Y2 - 3 April 2017 through 3 April 2017

ER -

OpenCL JIT Compilation for Dynamic Programming Languages

Abstract

Conference

Access to Document

Fingerprint

Cite this