High throughput multidimensional tridiagonal system solvers on FPGAs

Kamalavasan Kamalakkannan; Gihan R. Mudalige; Istvan Z. Reguly; Suhaib A. Fahmy

doi:10.1145/3524059.3532371

High throughput multidimensional tridiagonal system solvers on FPGAs

Kamalavasan Kamalakkannan, Gihan R. Mudalige, Istvan Z. Reguly, Suhaib A. Fahmy

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

Abstract

We present a high performance tridiagonal solver library for Xilinx FPGAs optimized for multiple multi-dimensional systems common in real-world applications. An analytical performance model is developed and used to explore the design space and obtain rapid performance estimates that are over 85% accurate. This library achieves an order of magnitude better performance when solving large batches of systems than previous FPGA work. A detailed comparison with a current state-of-the-art GPU library for multi-dimensional tridiagonal systems on an Nvidia V100 GPU shows the FPGA achieving competitive or better runtime and significant energy savings of over 30%. Through this design, we learn lessons about the types of applications where FPGAs can challenge the current dominance of GPUs.

Original language	English (US)
Title of host publication	Proceedings of the 36th ACM International Conference on Supercomputing
Publisher	ACM
DOIs	https://doi.org/10.1145/3524059.3532371
State	Published - Jun 2022

Access to Document

10.1145/3524059.3532371

Cite this

@inproceedings{6afd75e7a0434466b9cd3826e9bd7d94,

title = "High throughput multidimensional tridiagonal system solvers on FPGAs",

abstract = "We present a high performance tridiagonal solver library for Xilinx FPGAs optimized for multiple multi-dimensional systems common in real-world applications. An analytical performance model is developed and used to explore the design space and obtain rapid performance estimates that are over 85% accurate. This library achieves an order of magnitude better performance when solving large batches of systems than previous FPGA work. A detailed comparison with a current state-of-the-art GPU library for multi-dimensional tridiagonal systems on an Nvidia V100 GPU shows the FPGA achieving competitive or better runtime and significant energy savings of over 30%. Through this design, we learn lessons about the types of applications where FPGAs can challenge the current dominance of GPUs.",

author = "Kamalavasan Kamalakkannan and Mudalige, {Gihan R.} and Reguly, {Istvan Z.} and Fahmy, {Suhaib A.}",

note = "KAUST Repository Item: Exported on 2022-06-20 Acknowledgements: Gihan Mudalige was supported by the Royal Society Industry Fellowship Scheme (INF/R1/1800 12). Istvan Reguly ´ was supported by National Research, Development and Innovation Fund of Hungary (PD 124905), under the PD 17 funding scheme. We are grateful to Xilinx for their hardware and software donations and Jacques Du Toit and Tim Schmielau at NAG UK for their advice and making the SLV application avaialble for this work.",

year = "2022",

month = jun,

doi = "10.1145/3524059.3532371",

language = "English (US)",

booktitle = "Proceedings of the 36th ACM International Conference on Supercomputing",

publisher = "ACM",

}

TY - GEN

T1 - High throughput multidimensional tridiagonal system solvers on FPGAs

AU - Kamalakkannan, Kamalavasan

AU - Mudalige, Gihan R.

AU - Reguly, Istvan Z.

AU - Fahmy, Suhaib A.

N1 - KAUST Repository Item: Exported on 2022-06-20 Acknowledgements: Gihan Mudalige was supported by the Royal Society Industry Fellowship Scheme (INF/R1/1800 12). Istvan Reguly ´ was supported by National Research, Development and Innovation Fund of Hungary (PD 124905), under the PD 17 funding scheme. We are grateful to Xilinx for their hardware and software donations and Jacques Du Toit and Tim Schmielau at NAG UK for their advice and making the SLV application avaialble for this work.

PY - 2022/6

Y1 - 2022/6

N2 - We present a high performance tridiagonal solver library for Xilinx FPGAs optimized for multiple multi-dimensional systems common in real-world applications. An analytical performance model is developed and used to explore the design space and obtain rapid performance estimates that are over 85% accurate. This library achieves an order of magnitude better performance when solving large batches of systems than previous FPGA work. A detailed comparison with a current state-of-the-art GPU library for multi-dimensional tridiagonal systems on an Nvidia V100 GPU shows the FPGA achieving competitive or better runtime and significant energy savings of over 30%. Through this design, we learn lessons about the types of applications where FPGAs can challenge the current dominance of GPUs.

AB - We present a high performance tridiagonal solver library for Xilinx FPGAs optimized for multiple multi-dimensional systems common in real-world applications. An analytical performance model is developed and used to explore the design space and obtain rapid performance estimates that are over 85% accurate. This library achieves an order of magnitude better performance when solving large batches of systems than previous FPGA work. A detailed comparison with a current state-of-the-art GPU library for multi-dimensional tridiagonal systems on an Nvidia V100 GPU shows the FPGA achieving competitive or better runtime and significant energy savings of over 30%. Through this design, we learn lessons about the types of applications where FPGAs can challenge the current dominance of GPUs.

UR - http://hdl.handle.net/10754/678003

UR - https://dl.acm.org/doi/10.1145/3524059.3532371

U2 - 10.1145/3524059.3532371

DO - 10.1145/3524059.3532371

M3 - Conference contribution

BT - Proceedings of the 36th ACM International Conference on Supercomputing

PB - ACM

ER -

High throughput multidimensional tridiagonal system solvers on FPGAs

Abstract

Access to Document

Other files and links

Fingerprint

Cite this