28922

Optimization of Ported CFD Kernels on Intel Data Center GPU Max 1550 using oneAPI ESIMD

Mohammad Zubair, Aaron Walden, Gabriel Nastac, Eric Nielsen, Christoph Bauinger, Xiao Zhu
Old Dominion University, Norfolk, Virginia, USA
Proceedings of the SC ’23 Workshops of The International Conference on High Performance Computing, Network, Storage, and Analysis (SC-W’23), 2023

@inproceedings{zubair2023optimization,

   title={Optimization of Ported CFD Kernels on Intel Data Center GPU Max 1550 using oneAPI ESIMD},

   author={Zubair, Mohammad and Walden, Aaron and Nastac, Gabriel and Nielsen, Eric and Bauinger, Christoph and Zhu, Xiao},

   booktitle={Proceedings of the SC’23 Workshops of The International Conference on High Performance Computing, Network, Storage, and Analysis},

   pages={1705–1712},

   year={2023}

}

Download Download (PDF)   View View   Source Source   

639

views

We describe our experience porting FUN3D’s CUDA-optimized kernels to Intel oneAPI SYCL. We faced several challenges, including foremost the suboptimal performance of the oneAPI code on Intel’s new data center GPU. Suboptimal performance of the oneAPI code was due primarily to high register spills, memory latency, and poor vectorization. We addressed these issues by implementing the kernels using Intel oneAPI’s Explicit SIMD SYCL extension (ESIMD) API. The ESIMD API enables the writing of explicitly vectorized kernel code, gives more precise control over register usage and prefetching, and better handles thread divergence compared to SYCL. The ESIMD code outperforms the optimized SYCL code by up to a factor of 3.6, depending on the kernel. We also compared the performance of three ESIMD kernels on the Intel Data Center Max 1550 GPU with the CUDA-optimized versions on NVIDIA V100 and A100 GPUs. We found the performance of a single tile of the Intel GPU using ESIMD greater than NVIDIA V100 and similar to NVIDIA A100.
No votes yet.
Please wait...

Recent source codes

* * *

* * *

HGPU group © 2010-2024 hgpu.org

All rights belong to the respective authors

Contact us: