{"ID":2892692,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2507.15121","arxiv_id":"2507.15121","title":"AMPED: Accelerating MTTKRP for Billion-Scale Sparse Tensor Decomposition on Multiple GPUs","abstract":"Matricized Tensor Times Khatri-Rao Product (MTTKRP) is the computational bottleneck in sparse tensor decomposition. As real-world sparse tensors grow to billions of nonzeros, they increasingly demand higher memory capacity and compute throughput from hardware accelerators. In this work, we present AMPED, a multi-GPU parallel algorithm designed to accelerate MTTKRP on billion-scale sparse tensors. AMPED scales beyond the limits of a single GPU, meeting both the memory and performance requirements of large-scale workloads. We introduce a partitioning strategy combined with a dynamic load balancing scheme to distribute computation and minimize GPU idle time. On real-world billion-scale tensors, AMPED achieves a 5.1x geometric mean speedup in total execution time over state-of-the-art GPU baselines using 4 GPUs on a single CPU node.","short_abstract":"Matricized Tensor Times Khatri-Rao Product (MTTKRP) is the computational bottleneck in sparse tensor decomposition. As real-world sparse tensors grow to billions of nonzeros, they increasingly demand higher memory capacity and compute throughput from hardware accelerators. In this work, we present AMPED, a multi-GPU pa...","url_abs":"https://arxiv.org/abs/2507.15121","url_pdf":"https://arxiv.org/pdf/2507.15121v2","authors":"[\"Sasindu Wijeratne\",\"Rajgopal Kannan\",\"Viktor Prasanna\"]","published":"2025-07-20T21:01:13Z","proceeding":"cs.DC","tasks":"[\"cs.DC\"]","methods":"[]","has_code":false}