{"id":"arxiv/2602.03067","url":"https://arxiv.org/abs/2602.03067","tags":["cs.LG","cs.AI","math.NA"],"score":0,"title":"FlashSinkhorn: IO-Aware Entropic Optimal Transport on GPU","domain":"cs.LG","source":"arxiv","authors":["Felix X. -F. Ye","Xingjie Li","An Yu","Ming-Ching Chang","Linsong Chu","Davis Wertheimer"],"reviews":0,"subject":"cs.LG","summary":"Entropic optimal transport (EOT) via Sinkhorn iterations is widely used in modern machine learning, yet GPU solvers remain inefficient at scale. Tensorized implementations suffer quadratic HBM traffic from dense $n\\times m$ interactions, while existing online backends avoid storing dense matrices but still rely on generic tiled map-reduce reduction kernels with limited fusion. We present \\textbf{FlashSinkhorn}, an IO-aware EOT solver for squared Euclidean cost that rewrites stabilized log-domain Sinkhorn updates as row-wise LogSumExp reductions of biased dot-product scores, the same…","category":"research","comments":0,"disputed":false,"division":"cs","mentions":0,"discussed":0,"published":"2026-02-03","reactions":{}}