A Parallel Tensor Infrastructure (ParTI!), is to support fast essential sparse tensor operations and tensor decompositions on multicore CPU and GPU architectures. These basic tensor operations are critical to the overall performance of tensor analysis algorithms (such as tensor decomposition). ParTI! is formerly known as SpTOL.

This is the ParTI!2 branch, which is a rewrite of the original ParTI!. Both ParTI! and ParTI!2 will be actively developed for a considerably long time.

• Optimizing Sparse Tensor Times Matrix on GPUs. Yuchen Ma, Jiajia Li, Xiaolong Wu, Chenggang Yan, Jimeng Sun, Richard Vuduc. Journal of Parallel and Distributed Computing (Special Issue on Systems for Learning, Inferencing, and Discovering). [pdf] [bib]

• Optimizing Sparse Tensor Times Matrix on multi-core and many-core architectures. Jiajia Li, Yuchen Ma, Chenggang Yan, Richard Vuduc. The sixth Workshop on Irregular Applications: Architectures and Algorithms (IA^3), co-located with SC’16. 2016. [pdf]

• Yuchen Ma (Contact: [email protected])
• Jiajia Li (Contact: [email protected])