/kernel_float

CUDA header-only library for working with vector types (half2, float4, double2) and reduced precision math (half, e5m2) inside kernel code

Primary LanguageC++Apache License 2.0Apache-2.0

Issues