{"id":"arxiv/2606.20195","url":"https://arxiv.org/abs/2606.20195","tags":["cs.PF","math.NA"],"score":0,"title":"On the Robustness of Sparse Randomized Sketching to FP16 Rounding on GPUs","domain":"cs.PF","source":"arxiv","authors":["Aryaman Jeendgar","Clément Flint","Hartwig Anzt"],"reviews":0,"subject":"cs.PF","summary":"Randomized sketching replaces large matrices with smaller surrogates that preserve geometry for downstream numerical linear algebra. Sparse sketches promise faster formation than dense projections, but irregular GPU atomic updates become costly when each input row contributes to several output rows, as in SparseStack. FP16 accumulation halves sketch storage and update traffic, relieving cache pressure. How much does this rounding affect sketch geometry and downstream accuracy? For SparseStack with four nonzeros per column, cuSPARSE's FP16 path is slower than its FP32 baseline. Our custom…","category":"research","comments":0,"disputed":false,"division":"cs","mentions":0,"discussed":0,"published":"2026-06-18","reactions":{}}