{"ID":6138920,"CreatedAt":"2026-07-09T01:07:32.349475501Z","UpdatedAt":"2026-07-11T00:42:18.236081736Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2607.06881","arxiv_id":"2607.06881","title":"Multiple Double Arithmetic on NVIDIA Tensor Cores","abstract":"A multiple double is an unevaluated sum of doubles. An NVIDIA tensor core is a specialized high performance compute core for matrix multiplication. The Ampere A100, released in 2020, introduced tensor cores capable of 64-bit floating-point arithmetic. Every multiple double arithmetical operation requires renormalization, which involves branching, for which tensor cores are unsuited. To solve this problem caused by renormalization, we apply a solution similar to the Ozaki scheme [Ozaki et al, Numerical Algorithms, 2012]. Our software is available under the GPU GPL license on github.","short_abstract":"A multiple double is an unevaluated sum of doubles. An NVIDIA tensor core is a specialized high performance compute core for matrix multiplication. The Ampere A100, released in 2020, introduced tensor cores capable of 64-bit floating-point arithmetic. Every multiple double arithmetical operation requires renormalizatio...","url_abs":"https://arxiv.org/abs/2607.06881","url_pdf":"https://arxiv.org/pdf/2607.06881v1","authors":"[\"Howard Chen\",\"Jan Verschelde\"]","published":"2026-07-08T00:44:25Z","proceeding":"cs.MS","tasks":"[\"cs.MS\",\"cs.DC\",\"math.NA\"]","methods":"[]","has_code":false}