Optimization of mixed-precision iterative refinement using parallelized direct methods

Optimization of mixed-precision iterative refinement using parallelized direct methods
复制标题

使用并行直接方法优化混合精度迭代细化

DOI:
10.1109/iceet56468.2022.10007230
复制
发表时间:
2022
期刊:
2022 International Conference on Engineering and Emerging Technologies (ICEET)
影响因子:
--
通讯作者:
Kouya Tomonori
Kouya Tomonori
中科院分区:
--
文献类型:
--
作者:
H. Sone;Y. Tamura;and S. Yamada;伊藤綾音,照井章;Kouya Tomonori

文献摘要

参考文献

相似文献

解线性方程组是科学计算中最关键的任务之一,可以使用Linpack测试来评估TOP500超级计算机。我们已经用AVX2实现了SIMD化的基本线性计算,并通过在x86-计算环境中的基准测试证实了它的良好性能,证明了SIMD化可以用于加速LU分解。在这项研究中,进一步证明了用OpenMP并行化的SIMD化LU分解比串行化的速度更快,并且用于获得四重(QD,212位尾数)近似的混合精度迭代求精是可优化的。因此,用于迭代求精过程的Double-Double(DD,106位尾数)和QD算法的组合比DDMPFR 212位组合更有效。
Solving a linear system of equations is one of the most critical tasks in scientific computing, which can be performed using the LINPACK test to evaluate TOP500 supercomputers. We have already implemented SIMDized basic linear computation with AVX2 and confirmed that it performs well via benchmark tests in the x86-64 computing environment, demonstrating that SIMDized can be used to accelerate LU decomposition. In this study, it is further demonstrated that parallelized SIMDized LU decomposition with OpenMP is faster than the serial version, and that the mixed-precision iterative refinement used to obtain quad-double (QD, 212-bit mantissa) approximation is optimizable. As a result, the combination of double-double (DD, 106 bits mantissa) and QD arithmetic for the iterative refinement process is more efficient than the DDMPFR 212-bit combination.
使用 FMA、SIMD 和 OpenMP 加速交互式多精度算术工具箱 MuPAT
DOI: 10.3233/apc200069
发表时间: 2019
期刊: International Conference on Parallel Computing
影响因子: --
作者:
Hotaka Yagi;E. Ishiwata;H. Hasegawa
通讯作者: H. Hasegawa
DOI: 10.1109/arith51176.2021.00021
发表时间: 2021
期刊: 2021 IEEE 28th Symposium on Computer Arithmetic (ARITH)
影响因子: --
作者:
Kouya Tomonori
通讯作者: Kouya Tomonori
使用嵌入式公式控制步长的高阶多精度全隐式龙格库塔方法的实际实现
DOI: --
发表时间: 2013
期刊:
影响因子: --
作者:
T. Kouya
通讯作者: T. Kouya
DOI: 10.1109/tc.2019.2918451
发表时间: 2019
影响因子: 3.7
作者:
Nicolas Fabiano;J. Muller;Joris Picot
通讯作者: Joris Picot