Progress towards accelerating the unified model on hybrid multi-core systems

Wei Zhang; Min Xu; Katherine Evans; Matthew Norman; Mario Morales-Hernandez; Salil Mahajan; Adrian Hill; James Manners; Ben Shipway; Maynard Christopher

doi:10.1145/3468267.3470612

Progress towards accelerating the unified model on hybrid multi-core systems

Wei Zhang, Min Xu, Katherine Evans, Matthew Norman, Mario Morales-Hernandez, Salil Mahajan, Adrian Hill, James Manners, Ben Shipway, Maynard Christopher

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

Abstract

The cloud microphysics scheme, CASIM, and the radiation scheme, SOCRATES, are two computationally intensive parts within the Met Office's Unified Model (UM). This study enables CASIM and SOCRATES to use accelerated multi-core systems for optimal computational performance of the UM. Using profiling to guide our efforts, we refactored the code for optimal threading and kernel arrangement and implemented OpenACC directives manually or through the CLAW source-to-source translator. Initial porting results achieved 10.02x and 9.25x speedup in CASIM and SOCRATES respectively on 1 GPU compared with 1 CPU core. A granular performance analysis of the strategy and bottlenecks are discussed. These improvements will enable UM to run on heterogeneous computers and a path forward for further improvements is provided.

Original language	English
Title of host publication	Proceedings of the Platform for Advanced Scientific Computing Conference, PASC 2021
Publisher	Association for Computing Machinery, Inc
ISBN (Electronic)	9781450385633
DOIs	https://doi.org/10.1145/3468267.3470612
State	Published - Jul 5 2021
Event	2021 Platform for Advanced Scientific Computing Conference, PASC 2021 - Virtual, Online, Switzerland Duration: Jul 5 2021 → Jul 9 2021

Publication series

Name	Proceedings of the Platform for Advanced Scientific Computing Conference, PASC 2021

Conference

Conference	2021 Platform for Advanced Scientific Computing Conference, PASC 2021
Country/Territory	Switzerland
City	Virtual, Online
Period	07/5/21 → 07/9/21

Funding

This research was supported by the U.S. Air Force LCMC collaboration with Oak Ridge National Laboratory (ORNL). The computational resources on Summit are provided by the Oak Ridge Leadership Computing Facility (OLCF) Director’s Discretion Project ATM112. The OLCF at Oak Ridge National Laboratory (ORNL) is supported by the Office of Science of the U.S. Department of Energy under Contract No.DE-AC05-00OR22725. Furthermore, we would like to acknowledge the contributions of Youngsung Kim at ORNL for the insightful suggestions on porting algorithm development and performance bottleneck detection. We also appreciate the great help on CLAW implementation from Valentim Clement at ORNL. This research was supported by the U.S. Air Force LCMC collaboration with Oak Ridge National Laboratory (ORNL). The computational resources on Summit are provided by the Oak Ridge Leadership Computing Facility (OLCF) Director's Discretion Project ATM112. The OLCF at Oak Ridge National Laboratory (ORNL) is supported by the Office of Science of the U.S. Department of Energy under Contract No.DE-AC05-00OR22725. Furthermore, we would like to acknowledge the contributions of Youngsung Kim at ORNL for the insightful suggestions on porting algorithm development and performance bottleneck detection.We also appreciate the great help on CLAW implementation from Valentim Clement at ORNL.

Funders	Funder number
Oak Ridge National Laboratory	ATM112
U.S. Department of Energy
Office of Science
Oak Ridge National Laboratory
U.S. Air Force

Keywords

CASIM
GPU porting
OpenACC
SOCRATES
Unified model

Access to Document

10.1145/3468267.3470612

Cite this

Zhang, W., Xu, M., Evans, K., Norman, M., Morales-Hernandez, M., Mahajan, S., Hill, A., Manners, J., Shipway, B., & Christopher, M. (2021). Progress towards accelerating the unified model on hybrid multi-core systems. In Proceedings of the Platform for Advanced Scientific Computing Conference, PASC 2021 (Proceedings of the Platform for Advanced Scientific Computing Conference, PASC 2021). Association for Computing Machinery, Inc. https://doi.org/10.1145/3468267.3470612

@inproceedings{1e78dce42ec84c4985c58dec95fad778,

title = "Progress towards accelerating the unified model on hybrid multi-core systems",

abstract = "The cloud microphysics scheme, CASIM, and the radiation scheme, SOCRATES, are two computationally intensive parts within the Met Office's Unified Model (UM). This study enables CASIM and SOCRATES to use accelerated multi-core systems for optimal computational performance of the UM. Using profiling to guide our efforts, we refactored the code for optimal threading and kernel arrangement and implemented OpenACC directives manually or through the CLAW source-to-source translator. Initial porting results achieved 10.02x and 9.25x speedup in CASIM and SOCRATES respectively on 1 GPU compared with 1 CPU core. A granular performance analysis of the strategy and bottlenecks are discussed. These improvements will enable UM to run on heterogeneous computers and a path forward for further improvements is provided.",

keywords = "CASIM, GPU porting, OpenACC, SOCRATES, Unified model",

author = "Wei Zhang and Min Xu and Katherine Evans and Matthew Norman and Mario Morales-Hernandez and Salil Mahajan and Adrian Hill and James Manners and Ben Shipway and Maynard Christopher",

note = "Publisher Copyright: {\textcopyright} 2021 Owner/Author.; 2021 Platform for Advanced Scientific Computing Conference, PASC 2021 ; Conference date: 05-07-2021 Through 09-07-2021",

year = "2021",

month = jul,

day = "5",

doi = "10.1145/3468267.3470612",

language = "English",

series = "Proceedings of the Platform for Advanced Scientific Computing Conference, PASC 2021",

publisher = "Association for Computing Machinery, Inc",

booktitle = "Proceedings of the Platform for Advanced Scientific Computing Conference, PASC 2021",

}

Zhang, W, Xu, M, Evans, K , Norman, M, Morales-Hernandez, M, Mahajan, S, Hill, A, Manners, J, Shipway, B & Christopher, M 2021, Progress towards accelerating the unified model on hybrid multi-core systems. in Proceedings of the Platform for Advanced Scientific Computing Conference, PASC 2021. Proceedings of the Platform for Advanced Scientific Computing Conference, PASC 2021, Association for Computing Machinery, Inc, 2021 Platform for Advanced Scientific Computing Conference, PASC 2021, Virtual, Online, Switzerland, 07/5/21. https://doi.org/10.1145/3468267.3470612

Progress towards accelerating the unified model on hybrid multi-core systems. / Zhang, Wei; Xu, Min; Evans, Katherine et al.
Proceedings of the Platform for Advanced Scientific Computing Conference, PASC 2021. Association for Computing Machinery, Inc, 2021. (Proceedings of the Platform for Advanced Scientific Computing Conference, PASC 2021).

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

TY - GEN

T1 - Progress towards accelerating the unified model on hybrid multi-core systems

AU - Zhang, Wei

AU - Xu, Min

AU - Evans, Katherine

AU - Norman, Matthew

AU - Morales-Hernandez, Mario

AU - Mahajan, Salil

AU - Hill, Adrian

AU - Manners, James

AU - Shipway, Ben

AU - Christopher, Maynard

PY - 2021/7/5

Y1 - 2021/7/5

N2 - The cloud microphysics scheme, CASIM, and the radiation scheme, SOCRATES, are two computationally intensive parts within the Met Office's Unified Model (UM). This study enables CASIM and SOCRATES to use accelerated multi-core systems for optimal computational performance of the UM. Using profiling to guide our efforts, we refactored the code for optimal threading and kernel arrangement and implemented OpenACC directives manually or through the CLAW source-to-source translator. Initial porting results achieved 10.02x and 9.25x speedup in CASIM and SOCRATES respectively on 1 GPU compared with 1 CPU core. A granular performance analysis of the strategy and bottlenecks are discussed. These improvements will enable UM to run on heterogeneous computers and a path forward for further improvements is provided.

AB - The cloud microphysics scheme, CASIM, and the radiation scheme, SOCRATES, are two computationally intensive parts within the Met Office's Unified Model (UM). This study enables CASIM and SOCRATES to use accelerated multi-core systems for optimal computational performance of the UM. Using profiling to guide our efforts, we refactored the code for optimal threading and kernel arrangement and implemented OpenACC directives manually or through the CLAW source-to-source translator. Initial porting results achieved 10.02x and 9.25x speedup in CASIM and SOCRATES respectively on 1 GPU compared with 1 CPU core. A granular performance analysis of the strategy and bottlenecks are discussed. These improvements will enable UM to run on heterogeneous computers and a path forward for further improvements is provided.

KW - CASIM

KW - GPU porting

KW - OpenACC

KW - SOCRATES

KW - Unified model

UR - http://www.scopus.com/inward/record.url?scp=85114312324&partnerID=8YFLogxK

U2 - 10.1145/3468267.3470612

DO - 10.1145/3468267.3470612

M3 - Conference contribution

AN - SCOPUS:85114312324

T3 - Proceedings of the Platform for Advanced Scientific Computing Conference, PASC 2021

BT - Proceedings of the Platform for Advanced Scientific Computing Conference, PASC 2021

PB - Association for Computing Machinery, Inc

T2 - 2021 Platform for Advanced Scientific Computing Conference, PASC 2021

Y2 - 5 July 2021 through 9 July 2021

ER -

Zhang W, Xu M, Evans K , Norman M, Morales-Hernandez M, Mahajan S et al. Progress towards accelerating the unified model on hybrid multi-core systems. In Proceedings of the Platform for Advanced Scientific Computing Conference, PASC 2021. Association for Computing Machinery, Inc. 2021. (Proceedings of the Platform for Advanced Scientific Computing Conference, PASC 2021). doi: 10.1145/3468267.3470612

Progress towards accelerating the unified model on hybrid multi-core systems

Abstract

Publication series

Conference

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this