Exploiting data parallelism and population Monte Carlo on massively-parallel architectures for geoacoustic inversion

Jan Dettmer, Stan E. Dosso, Charles Holland

Research output: Contribution to journalConference article

Abstract

Bayesian inference algorithms in geoacoustic inversion have high computational requirements on multiple computational scales. Predicting (modeling) data to match observations represents fine-grained computations which often cannot be implemented efficiently on CPU clusters since high latency and communication overhead outweigh parallelization gains. However, GPUs, which operate efficiently on 100,000s of parallel treads with low latency and high bandwidth, can provide signficant performance gains. Bayesian sampling schemes are generally coarse-grained, and can be implemented efficiently in parallel on multi-core/cluster architectures. For example, population Monte Carlo methods simulate many Markov chains in parallel, with chains running independently between interactions (at predefined intervals) which exchange information throughout the population, substantially increasing sampling efficiency. This paper combines fine- and coarse-grained parallelization to profoundly improve the efficiency of geoacoustic inversion of seabed reflection data. Spherical-wave reflection-coefficient predictions, which require solving the Sommerfeld integral for a large number of grazing angles and frequencies, constitute fine-grained, data-parallel computations which are implemented efficiently on a GPU. Sampling is based on population Monte Carlo simulation with chain interactions as exchange and crossover moves. The algorithm is applied to data from the Malta Plateau to study the frequency dependence of sound velocity and attenuation in marine sediments.

Original languageEnglish (US)
Article number070094
JournalProceedings of Meetings on Acoustics
Volume19
DOIs
StatePublished - Jun 19 2013
Event21st International Congress on Acoustics, ICA 2013 - 165th Meeting of the Acoustical Society of America - Montreal, QC, Canada
Duration: Jun 2 2013Jun 7 2013

Fingerprint

sampling
inversions
Malta
treads
spherical waves
wave reflection
Markov chains
grazing
acoustic velocity
inference
Monte Carlo method
plateaus
crossovers
sediments
attenuation
communication
interactions
bandwidth
intervals
reflectance

All Science Journal Classification (ASJC) codes

  • Acoustics and Ultrasonics

Cite this

@article{1993cda2cded4a9481ffc39182c0c82a,
title = "Exploiting data parallelism and population Monte Carlo on massively-parallel architectures for geoacoustic inversion",
abstract = "Bayesian inference algorithms in geoacoustic inversion have high computational requirements on multiple computational scales. Predicting (modeling) data to match observations represents fine-grained computations which often cannot be implemented efficiently on CPU clusters since high latency and communication overhead outweigh parallelization gains. However, GPUs, which operate efficiently on 100,000s of parallel treads with low latency and high bandwidth, can provide signficant performance gains. Bayesian sampling schemes are generally coarse-grained, and can be implemented efficiently in parallel on multi-core/cluster architectures. For example, population Monte Carlo methods simulate many Markov chains in parallel, with chains running independently between interactions (at predefined intervals) which exchange information throughout the population, substantially increasing sampling efficiency. This paper combines fine- and coarse-grained parallelization to profoundly improve the efficiency of geoacoustic inversion of seabed reflection data. Spherical-wave reflection-coefficient predictions, which require solving the Sommerfeld integral for a large number of grazing angles and frequencies, constitute fine-grained, data-parallel computations which are implemented efficiently on a GPU. Sampling is based on population Monte Carlo simulation with chain interactions as exchange and crossover moves. The algorithm is applied to data from the Malta Plateau to study the frequency dependence of sound velocity and attenuation in marine sediments.",
author = "Jan Dettmer and Dosso, {Stan E.} and Charles Holland",
year = "2013",
month = "6",
day = "19",
doi = "10.1121/1.4799784",
language = "English (US)",
volume = "19",
journal = "Proceedings of Meetings on Acoustics",
issn = "1939-800X",
publisher = "Acoustical Society of America",

}

Exploiting data parallelism and population Monte Carlo on massively-parallel architectures for geoacoustic inversion. / Dettmer, Jan; Dosso, Stan E.; Holland, Charles.

In: Proceedings of Meetings on Acoustics, Vol. 19, 070094, 19.06.2013.

Research output: Contribution to journalConference article

TY - JOUR

T1 - Exploiting data parallelism and population Monte Carlo on massively-parallel architectures for geoacoustic inversion

AU - Dettmer, Jan

AU - Dosso, Stan E.

AU - Holland, Charles

PY - 2013/6/19

Y1 - 2013/6/19

N2 - Bayesian inference algorithms in geoacoustic inversion have high computational requirements on multiple computational scales. Predicting (modeling) data to match observations represents fine-grained computations which often cannot be implemented efficiently on CPU clusters since high latency and communication overhead outweigh parallelization gains. However, GPUs, which operate efficiently on 100,000s of parallel treads with low latency and high bandwidth, can provide signficant performance gains. Bayesian sampling schemes are generally coarse-grained, and can be implemented efficiently in parallel on multi-core/cluster architectures. For example, population Monte Carlo methods simulate many Markov chains in parallel, with chains running independently between interactions (at predefined intervals) which exchange information throughout the population, substantially increasing sampling efficiency. This paper combines fine- and coarse-grained parallelization to profoundly improve the efficiency of geoacoustic inversion of seabed reflection data. Spherical-wave reflection-coefficient predictions, which require solving the Sommerfeld integral for a large number of grazing angles and frequencies, constitute fine-grained, data-parallel computations which are implemented efficiently on a GPU. Sampling is based on population Monte Carlo simulation with chain interactions as exchange and crossover moves. The algorithm is applied to data from the Malta Plateau to study the frequency dependence of sound velocity and attenuation in marine sediments.

AB - Bayesian inference algorithms in geoacoustic inversion have high computational requirements on multiple computational scales. Predicting (modeling) data to match observations represents fine-grained computations which often cannot be implemented efficiently on CPU clusters since high latency and communication overhead outweigh parallelization gains. However, GPUs, which operate efficiently on 100,000s of parallel treads with low latency and high bandwidth, can provide signficant performance gains. Bayesian sampling schemes are generally coarse-grained, and can be implemented efficiently in parallel on multi-core/cluster architectures. For example, population Monte Carlo methods simulate many Markov chains in parallel, with chains running independently between interactions (at predefined intervals) which exchange information throughout the population, substantially increasing sampling efficiency. This paper combines fine- and coarse-grained parallelization to profoundly improve the efficiency of geoacoustic inversion of seabed reflection data. Spherical-wave reflection-coefficient predictions, which require solving the Sommerfeld integral for a large number of grazing angles and frequencies, constitute fine-grained, data-parallel computations which are implemented efficiently on a GPU. Sampling is based on population Monte Carlo simulation with chain interactions as exchange and crossover moves. The algorithm is applied to data from the Malta Plateau to study the frequency dependence of sound velocity and attenuation in marine sediments.

UR - http://www.scopus.com/inward/record.url?scp=84878971659&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84878971659&partnerID=8YFLogxK

U2 - 10.1121/1.4799784

DO - 10.1121/1.4799784

M3 - Conference article

VL - 19

JO - Proceedings of Meetings on Acoustics

JF - Proceedings of Meetings on Acoustics

SN - 1939-800X

M1 - 070094

ER -