Towards Use of Mixed Precision in ECP Math Libraries

Anzt, Hartwig; Boman, Eric G.; Gates, Marj; Kruger, Scott; Li, Sherry; Loe, Jennifer; Osei-Kuffuor, Daniel; Tomov, Stan; Tsai, Yaohung M.; Yang, Ulrike Meier

doi:10.2172/1762902

Title: Towards Use of Mixed Precision in ECP Math Libraries

Full Record
Other Related Research

Abstract

The use of multiple types of precision in mathematical software has the potential to increase its performance on new heterogeneous architectures. The xSDK project focuses both on the investigation and development of multiprecision algorithms as well as their inclusion into xSDK member libraries. This report summarizes current efforts on including and/or using mixed precision capabilities in the math libraries Ginkgo, heFFTe, hypre, MAGMA, PETSc/TAO, SLATE, SuperLU, and Trilinos, including KokkosKernels. It contains both numerical results from libraries that already provide mixed precision capabilities, as well as descriptions of the strategies to incorporate multiprecision into established libraries.

Authors:

Anzt, Hartwig ^[1]; Boman, Eric G. ^[2]; Gates, Marj ^[3]; Kruger, Scott ^[4]; Li, Sherry ^[5]; Loe, Jennifer ^[2]; Osei-Kuffuor, Daniel ^[6]; Tomov, Stan ^[3]; Tsai, Yaohung M. ^[3]; Yang, Ulrike Meier ^[6]

Karlsruhe Inst. of Technology (KIT) (Germany)
Sandia National Lab. (SNL-NM), Albuquerque, NM (United States)
Univ. of Tennessee, Knoxville, TN (United States)
Tech-X Corp., Boulder, CO (United States)
Lawrence Berkeley National Lab. (LBNL), Berkeley, CA (United States)
Lawrence Livermore National Lab. (LLNL), Livermore, CA (United States)

Publication Date:: 2021-01-21

Research Org.:: Lawrence Livermore National Lab. (LLNL), Livermore, CA (United States)

Sponsoring Org.:: USDOE National Nuclear Security Administration (NNSA)

OSTI Identifier:: 1762902

Report Number(s):: LLNL-TR-816649
1026283

DOE Contract Number:: AC52-07NA27344

Resource Type:: Technical Report

Country of Publication:: United States

Language:: English

Subject:: 97 MATHEMATICS AND COMPUTING

Citation Formats


                    Anzt, Hartwig, Boman, Eric G., Gates, Marj, Kruger, Scott, Li, Sherry, Loe, Jennifer, Osei-Kuffuor, Daniel, Tomov, Stan, Tsai, Yaohung M., and Yang, Ulrike Meier. Towards Use of Mixed Precision in ECP Math Libraries.  United States: N. p., 2021. 
        Web.  doi:10.2172/1762902.

Copy to clipboard


                    Anzt, Hartwig, Boman, Eric G., Gates, Marj, Kruger, Scott, Li, Sherry, Loe, Jennifer, Osei-Kuffuor, Daniel, Tomov, Stan, Tsai, Yaohung M., & Yang, Ulrike Meier. Towards Use of Mixed Precision in ECP Math Libraries.  United States.  https://doi.org/10.2172/1762902

Copy to clipboard


                    Anzt, Hartwig, Boman, Eric G., Gates, Marj, Kruger, Scott, Li, Sherry, Loe, Jennifer, Osei-Kuffuor, Daniel, Tomov, Stan, Tsai, Yaohung M., and Yang, Ulrike Meier. 2021.  
        "Towards Use of Mixed Precision in ECP Math Libraries".  United States.  https://doi.org/10.2172/1762902.  https://www.osti.gov/servlets/purl/1762902.

Copy to clipboard


                    
@article{osti_1762902,

  title        = {Towards Use of Mixed Precision in ECP Math Libraries},

  author       = {Anzt, Hartwig and Boman, Eric G. and Gates, Marj and Kruger, Scott and Li, Sherry and Loe, Jennifer and Osei-Kuffuor, Daniel and Tomov, Stan and Tsai, Yaohung M. and Yang, Ulrike Meier},

  abstractNote = {The use of multiple types of precision in mathematical software has the potential to increase its performance on new heterogeneous architectures. The xSDK project focuses both on the investigation and development of multiprecision algorithms as well as their inclusion into xSDK member libraries. This report summarizes current efforts on including and/or using mixed precision capabilities in the math libraries Ginkgo, heFFTe, hypre, MAGMA, PETSc/TAO, SLATE, SuperLU, and Trilinos, including KokkosKernels. It contains both numerical results from libraries that already provide mixed precision capabilities, as well as descriptions of the strategies to incorporate multiprecision into established libraries.},

  doi          = {10.2172/1762902},

  url          = {https://www.osti.gov/biblio/1762902},
  journal      = {},
number       = ,

  volume       = ,

  place        = {United States},

  year         = {2021},

  month        = {1}

}

Copy to clipboard

Technical Report:

View Technical Report (3.45 MB)

https://doi.org/10.2172/1762902

Save / Share:

Export Metadata

Save to My Library

Similar records in OSTI.GOV collections:

Towards Use of Mixed Precision in ECP Math Libraries [Exascale Computing Project]

Technical Report Antz, Hartwig; Boman, Erik; Gates, Mark; ...

The use of multiple types of precision in mathematical software has the potential to increase its performance on new heterogeneous architectures. The xSDK project focuses both on the investigation and development of multiprecision algorithms as well as their inclusion into xSDK member libraries. This report summarizes current efforts on including and/or using mixed precision capabilities in the math libraries Ginkgo, heFFTe, hypre, MAGMA, PETSc/TAO, SLATE, SuperLU, and Trilinos, including KokkosKernels. It contains both numerical results from libraries that already provide mixed precision capabilities, as well as descriptions of the strategies to incorporate multiprecision into established libraries.
https://doi.org/10.2172/1735694

Full Text Available
Advances in Mixed Precision Algorithms: 2021 Edition

Technical Report Abdelfattah, Ahmad; Anzt, Hartwig; Ayala, Alan; ...

Over the last year, the ECP xSDK-multiprecision effort has made tremendous progress in developing and deploying new mixed precision technology and customizing the algorithms for the hardware deployed in the ECP flagship supercomputers. The effort also has succeeded in creating a cross-laboratory community of scientists interested in mixed precision technology and now working together in deploying this technology for ECP applications. In this report, we highlight some of the most promising and impactful achievements of the last year. Among the highlights we present are: Mixed precision IR using a dense LU factorization and achieving a 1.8× speedup on Spock; resultsmore »« less
https://doi.org/10.2172/1814447

Full Text Available
Advances in Mixed Precision Algorithms: 2021 Edition

Technical Report Abdelfattah, A.; Anzt, H.; Ayala, A.; ...

Over the last year, the ECP xSDK-multiprecision effort has made tremendous progress in developing and deploying new mixed precision technology and customizing the algorithms for the hardware deployed in the ECP flagship supercomputers. The effort also has succeeded in creating a cross-laboratory community of scientists interested in mixed precision technology and now working together in deploying this technology for ECP applications. In this report, we highlight some of the most promising and impactful achievements of the last year. Among the highlights we present are • Mixed precision IR using a dense LU factorization and achieving a 1.8× speedup on Spock;more »« less
https://doi.org/10.2172/1814677

Full Text Available
xSDK Foundations: Toward an Extreme-scale Scientific Software Development Kit

Journal Article Heroux, Michael; Bartlett, Roscoe; Demeshko, Irina; ... - Supercomputing frontiers and innovations

Here, extreme-scale computational science increasingly demands multiscale and multiphysics formulations. Combining software developed by independent groups is imperative: no single team has resources for all predictive science and decision support capabilities. Scientific libraries provide high-quality, reusable software components for constructing applications with improved robustness and portability. However, without coordination, many libraries cannot be easily composed. Namespace collisions, inconsistent arguments, lack of third-party software versioning, and additional difficulties make composition costly. The Extreme-scale Scientific Software Development Kit (xSDK) defines community policies to improve code quality and compatibility across independently developed packages (hypre, PETSc, SuperLU, Trilinos, and Alquimia) and provides a foundationmore »« less
https://doi.org/10.14529/jsfi170104

Full Text Available
Final Report: Towards Optimal Petascale Simulations (TOPS), ER25785

Technical Report Reynolds, Daniel

Multiscale, multirate scientific and engineering applications in the SciDAC portfolio possess resolution requirements that are practically inexhaustible and demand execution on the highest-capability computers available, which will soon reach the petascale. While the variety of applications is enormous, their needs for mathematical software infrastructure are surprisingly coincident; moreover the chief bottleneck is often the solver. At their current scalability limits, many applications spend a vast majority of their operations in solvers, due to solver algorithmic complexity that is superlinear in the problem size, whereas other phases scale linearly. Furthermore, the solver may be the phase of the simulation with themore »« less
https://doi.org/10.2172/1011379

Full Text Available

Similar Records