NVIDIA cuBLAS adds Ozaki-II and 2M FP64 algorithms

Nvidia has introduced new FP64 algorithms, including Ozaki-II and the 2M method, in the cuBLAS library within CUDA Toolkit 13.4.2 to accelerate scientific simulations. By breaking double-precision numbers into integer pieces and utilizing faster integer Tensor Cores, these algorithms deliver full double-precision accuracy with significant performance gains for real-world applications like quantum chemistry and materials science. Open article →