Shared-memory parallelization of MTTKRP for dense tensors

Shared-memory parallelization of MTTKRP for dense tensors
复制标题

密集张量的 MTTKRP 共享内存并行化

DOI:
10.1145/3178487.3178522
复制
发表时间:
2018
期刊:
23rd ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming
影响因子:
--
通讯作者:
Tobia, Michael J.
Tobia, Michael J.
中科院分区:
--
文献类型:
--
作者:
Hayashi, Koby;Ballard, Grey;Jiang, Yujie;Tobia, Michael J.

文献摘要

参考文献

被引文献

相似文献

矩阵化张量乘Khatri-Rao积(MTTKRP)是计算张量CP分解算法的计算瓶颈。在这项工作中,我们开发了共享内存的并行算法MTTKRP涉及密集张量。该算法将几乎所有的计算都转换为矩阵运算,以便使用优化的BLAS子例程,并且它们避免了在内存中重新排序张量条目。我们使用我们的并行实现来计算一个神经成像数据集的CP分解,并实现了高达7.4倍的速度比现有的并行软件。
The matricized-tensor times Khatri-Rao product (MTTKRP) is the computational bottleneck for algorithms computing CP decompositions of tensors. In this work, we develop shared-memory parallel algorithms for MTTKRP involving dense tensors. The algorithms cast nearly all of the computation as matrix operations in order to use optimized BLAS subroutines, and they avoid reordering tensor entries in memory. We use our parallel implementation to compute a CP decomposition of a neuroimaging data set and achieve a speedup of up to 7.4X over existing parallel software.
DOI: 10.1145/2807591.2807671
发表时间: 2015-11
期刊: SC15: International Conference for High Performance Computing, Networking, Storage and Analysis
影响因子: --
作者:
Jiajia Li;Casey Battaglino;Ioakeim Perros;Jimeng Sun;R. Vuduc
通讯作者: Jiajia Li;Casey Battaglino;Ioakeim Perros;Jimeng Sun;R. Vuduc