Data parallel multi-time-step time integration strategies for computational heat transfer

Raju R. Namburu; Farzad Rostam-Abadi; Ram V. Mohan; Kumar K Tamma

Data parallel multi-time-step time integration strategies for computational heat transfer

Raju R. Namburu, Farzad Rostam-Abadi, Ram V. Mohan, Kumar K Tamma

Mechanical Engineering

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

2 Scopus citations

Abstract

This paper explores two strategies for implementing multi-time-step explicit time integration procedure for computational heat transfer in data parallel mode on massively parallel computer, namely, the Connection Machine CM-5. Data layout, amount of communication, communication routing and load balancing are the main issues which hinder the efficiency of data parallel implementation. Among these timings, more than 70% of total time is spent performing communications for explicit finite element method. To reduce the amount of communication time and to improve the overall efficiency, two multi-time-step implementation strategies, namely, masking and grouping are discussed in this paper.

Original language	English (US)
State	Published - Jan 1 1993
Event	AIAA 28th Thermophysics Conference, 1993 - Orlando, United States Duration: Jul 6 1993 → Jul 9 1993

Other

Other	AIAA 28th Thermophysics Conference, 1993
Country/Territory	United States
City	Orlando
Period	7/6/93 → 7/9/93

Find It

Find It at U of M Libraries

Cite this

@conference{1cf7469491494a9ca117091a9c7549d9,

title = "Data parallel multi-time-step time integration strategies for computational heat transfer",

abstract = "This paper explores two strategies for implementing multi-time-step explicit time integration procedure for computational heat transfer in data parallel mode on massively parallel computer, namely, the Connection Machine CM-5. Data layout, amount of communication, communication routing and load balancing are the main issues which hinder the efficiency of data parallel implementation. Among these timings, more than 70% of total time is spent performing communications for explicit finite element method. To reduce the amount of communication time and to improve the overall efficiency, two multi-time-step implementation strategies, namely, masking and grouping are discussed in this paper.",

author = "Namburu, {Raju R.} and Farzad Rostam-Abadi and Mohan, {Ram V.} and Tamma, {Kumar K}",

year = "1993",

month = jan,

day = "1",

language = "English (US)",

note = "AIAA 28th Thermophysics Conference, 1993 ; Conference date: 06-07-1993 Through 09-07-1993",

}

TY - CONF

T1 - Data parallel multi-time-step time integration strategies for computational heat transfer

AU - Namburu, Raju R.

AU - Rostam-Abadi, Farzad

AU - Mohan, Ram V.

AU - Tamma, Kumar K

PY - 1993/1/1

Y1 - 1993/1/1

N2 - This paper explores two strategies for implementing multi-time-step explicit time integration procedure for computational heat transfer in data parallel mode on massively parallel computer, namely, the Connection Machine CM-5. Data layout, amount of communication, communication routing and load balancing are the main issues which hinder the efficiency of data parallel implementation. Among these timings, more than 70% of total time is spent performing communications for explicit finite element method. To reduce the amount of communication time and to improve the overall efficiency, two multi-time-step implementation strategies, namely, masking and grouping are discussed in this paper.

AB - This paper explores two strategies for implementing multi-time-step explicit time integration procedure for computational heat transfer in data parallel mode on massively parallel computer, namely, the Connection Machine CM-5. Data layout, amount of communication, communication routing and load balancing are the main issues which hinder the efficiency of data parallel implementation. Among these timings, more than 70% of total time is spent performing communications for explicit finite element method. To reduce the amount of communication time and to improve the overall efficiency, two multi-time-step implementation strategies, namely, masking and grouping are discussed in this paper.

UR - http://www.scopus.com/inward/record.url?scp=85001949407&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85001949407&partnerID=8YFLogxK

M3 - Paper

AN - SCOPUS:85001949407

T2 - AIAA 28th Thermophysics Conference, 1993

Y2 - 6 July 1993 through 9 July 1993

ER -

Data parallel multi-time-step time integration strategies for computational heat transfer

Abstract

Other

Find It

Other files and links

Fingerprint

Cite this