Sparse MTTKRP Acceleration for Tensor Decomposition on GPU
Fuente:
arXiv
Salvato in:
| Autori principali: | , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866910446389297152 |
|---|---|
| author | Wijeratne, Sasindu Kannan, Rajgopal Prasanna, Viktor |
| author_facet | Wijeratne, Sasindu Kannan, Rajgopal Prasanna, Viktor |
| contents | Sparse Matricized Tensor Times Khatri-Rao Product (spMTTKRP) is the bottleneck kernel of sparse tensor decomposition. In this work, we propose a GPU-based algorithm design to address the key challenges in accelerating spMTTKRP computation, including (1) eliminating global atomic operations across GPU thread blocks, (2) avoiding the intermediate values being communicated between GPU thread blocks and GPU global memory, and (3) ensuring a balanced distribution of workloads across GPU thread blocks. Our approach also supports dynamic tensor remapping, enabling the above optimizations in all the modes of the input tensor. Our approach achieves a geometric mean speedup of 1.5x, 2.0x, and 21.7x in total execution time across widely used datasets compared with the state-of-the-art GPU implementations. Our work is the only GPU implementation that can support tensors with modes greater than 4 since the state-of-the-art works have implementation constraints for tensors with a large number of modes. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2405_08470 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Sparse MTTKRP Acceleration for Tensor Decomposition on GPU Wijeratne, Sasindu Kannan, Rajgopal Prasanna, Viktor Distributed, Parallel, and Cluster Computing Hardware Architecture Sparse Matricized Tensor Times Khatri-Rao Product (spMTTKRP) is the bottleneck kernel of sparse tensor decomposition. In this work, we propose a GPU-based algorithm design to address the key challenges in accelerating spMTTKRP computation, including (1) eliminating global atomic operations across GPU thread blocks, (2) avoiding the intermediate values being communicated between GPU thread blocks and GPU global memory, and (3) ensuring a balanced distribution of workloads across GPU thread blocks. Our approach also supports dynamic tensor remapping, enabling the above optimizations in all the modes of the input tensor. Our approach achieves a geometric mean speedup of 1.5x, 2.0x, and 21.7x in total execution time across widely used datasets compared with the state-of-the-art GPU implementations. Our work is the only GPU implementation that can support tensors with modes greater than 4 since the state-of-the-art works have implementation constraints for tensors with a large number of modes. |
| title | Sparse MTTKRP Acceleration for Tensor Decomposition on GPU |
| topic | Distributed, Parallel, and Cluster Computing Hardware Architecture |
| url | https://arxiv.org/abs/2405.08470 |