An embedded DSL for stochastic processes

Michael Flænø Werk; Joakim Ahnfelt-Rønne; Ken Friis Larsen

doi:10.1145/2364474.2364488

An embedded DSL for stochastic processes

Michael Flænø Werk, Joakim Ahnfelt-Rønne, Ken Friis Larsen

Datalogisk Institut

4 Citationer (Scopus)

Abstract

We present a domain specific language embedded in Haskell for specifying stochastic processes, called SPL;. It is designed with the goal of matching the notation used in mathematical finance, where the price of a financial contract is specified using stochastic processes and distributions. SPL; is declarative in the sense that it is agnostic of the choice of discretization and of the computational model. We provide an implementation of SPL; that performs Monte Carlo simulation using GPGPU, and we present data indicating that this gives a 100x speedup compared to hand-written sequential C, and that the speedup scales linearly with the number of available cores.

Originalsprog	Engelsk
Titel	Proceedings of the 1st ACM SIGPLAN Workshop on Functional High-Performance Computing
Antal sider	9
Forlag	Association for Computing Machinery
Publikationsdato	2012
Sider	93-101
ISBN (Trykt)	978-1-4503-1577-7
DOI	https://doi.org/10.1145/2364474.2364488
Status	Udgivet - 2012
Begivenhed	1st ACM SIGPLAN Workshop on Functional High-Performance Computing - København, Danmark Varighed: 15 sep. 2012 → 15 sep. 2012 Konferencens nummer: 1

Konference

Konference	1st ACM SIGPLAN Workshop on Functional High-Performance Computing
Nummer	1
Land/Område	Danmark
By	København
Periode	15/09/2012 → 15/09/2012

Emneord

code generation, embedded domain specific language, gpgpu, haskell, monte-carlo simulation, opencl, stochastic processes

Adgang til dokumentet

10.1145/2364474.2364488

Citationsformater

@inproceedings{7a28b0052d35403c9520868fa3b79266,

title = "An embedded DSL for stochastic processes",

abstract = "We present a domain specific language embedded in Haskell for specifying stochastic processes, called SPL;. It is designed with the goal of matching the notation used in mathematical finance, where the price of a financial contract is specified using stochastic processes and distributions. SPL; is declarative in the sense that it is agnostic of the choice of discretization and of the computational model. We provide an implementation of SPL; that performs Monte Carlo simulation using GPGPU, and we present data indicating that this gives a 100x speedup compared to hand-written sequential C, and that the speedup scales linearly with the number of available cores.",

keywords = "code generation, embedded domain specific language, gpgpu, haskell, monte-carlo simulation, opencl, stochastic processes",

author = "Werk, {Michael Fl{\ae}n{\o}} and Joakim Ahnfelt-R{\o}nne and Larsen, {Ken Friis}",

year = "2012",

doi = "10.1145/2364474.2364488",

language = "English",

isbn = "978-1-4503-1577-7",

pages = "93--101",

booktitle = "Proceedings of the 1st ACM SIGPLAN Workshop on Functional High-Performance Computing",

publisher = "Association for Computing Machinery",

note = "1st ACM SIGPLAN Workshop on Functional High-Performance Computing, FHPC '12 ; Conference date: 15-09-2012 Through 15-09-2012",

}

TY - GEN

T1 - An embedded DSL for stochastic processes

AU - Werk, Michael Flænø

AU - Ahnfelt-Rønne, Joakim

AU - Larsen, Ken Friis

N1 - Conference code: 1

PY - 2012

Y1 - 2012

N2 - We present a domain specific language embedded in Haskell for specifying stochastic processes, called SPL;. It is designed with the goal of matching the notation used in mathematical finance, where the price of a financial contract is specified using stochastic processes and distributions. SPL; is declarative in the sense that it is agnostic of the choice of discretization and of the computational model. We provide an implementation of SPL; that performs Monte Carlo simulation using GPGPU, and we present data indicating that this gives a 100x speedup compared to hand-written sequential C, and that the speedup scales linearly with the number of available cores.

AB - We present a domain specific language embedded in Haskell for specifying stochastic processes, called SPL;. It is designed with the goal of matching the notation used in mathematical finance, where the price of a financial contract is specified using stochastic processes and distributions. SPL; is declarative in the sense that it is agnostic of the choice of discretization and of the computational model. We provide an implementation of SPL; that performs Monte Carlo simulation using GPGPU, and we present data indicating that this gives a 100x speedup compared to hand-written sequential C, and that the speedup scales linearly with the number of available cores.

KW - code generation, embedded domain specific language, gpgpu, haskell, monte-carlo simulation, opencl, stochastic processes

U2 - 10.1145/2364474.2364488

DO - 10.1145/2364474.2364488

M3 - Article in proceedings

SN - 978-1-4503-1577-7

SP - 93

EP - 101

BT - Proceedings of the 1st ACM SIGPLAN Workshop on Functional High-Performance Computing

PB - Association for Computing Machinery

T2 - 1st ACM SIGPLAN Workshop on Functional High-Performance Computing

Y2 - 15 September 2012 through 15 September 2012

ER -

An embedded DSL for stochastic processes

Abstract

Konference

Emneord

Adgang til dokumentet

Fingeraftryk

Citationsformater