NVIDIA Updates cuBLAS Library with New FP64 Algorithms
NVIDIA has introduced new FP64 algorithms in its cuBLAS library. The algorithms target high-precision dense general matrix multiplications that dominate execution time in many scientific simulations including molecular chemistry and materials science.
The update is described in an NVIDIA technical blog post as a means to improve performance of those core computations. No independent performance measurements or availability details are provided in the supplied source.
primary-source authority · single-source report · new story
Details and supporting sources
NVIDIA has introduced new FP64 algorithms in its cuBLAS library.
NVIDIA Developer ↗The algorithms target high-precision dense general matrix multiplications used in scientific simulations such as molecular chemistry and materials science.
NVIDIA Developer ↗The update is described in an NVIDIA technical blog post as a means to improve performance of those core computations.
NVIDIA Developer ↗