2025/09/28 by Piotr Luszczek, Piotr Łuszczek, Luszczek, Piotr +42 · 2 voices · 1 citation
Computer Science · Engineering · Mathematics · #Advanced Wireless Communication Techniques #Distributed and Parallel Computing Systems #FOS: Mathematics #Numerical Analysis (math.NA) #Parallel Computing and Optimization Techniques #math.NA
paper · pdf · doi:10.48550/arxiv.2509.23565
openalex publication_date 2025/09/28 · arxiv published 2025/09/28 · arxiv updated 2025/09/28 · openalex created_date 2025/10/19 · openalex updated_date 2026/07/28
Mixing precisions for performance has been an ongoing trend as the modern hardware accelerators started including new, and mostly lower-precision, data formats. The advantage of using them is a great potential of performance gain and energy savings. The disadvantage are the numerical issues not present in the standard-mandated floating-point formats. Split integer emulation of FP64 takes this to an extreme with the computation performed only by fixed-point tensor core units. We present the new issues the emulation faces for practical cases involving dense linear solver. We show extensive numerical tests indicating the effect of extended numerical range of matrix entries. We also scaled the input sizes to study the performance and numerical profiles on the NVIDIA Hopper GPUs.