Kernel Float

CUDA natively offers several reduced precision floating-point types (__half, __nv_bfloat16, __nv_fp8_e4m3, __nv_fp8_e5m2) and vector types (e.g., __half2, __nv_fp8x4_e4m3, float3). However, working with these types is cumbersome: mathematical operations require intrinsics (e.g., __hadd2 performs addition for __half2), type conversion is awkward (e.g., __nv_cvt_halfraw2_to_fp8x2 converts float16 to float8),
and some functionality is missing (e.g., one cannot convert a __half to __nv_bfloat16).

Kernel Float resolves this by offering a single data type kernel_float::vec<T, N> that stores N elements of type T. Internally, the data is stored as a fixed-sized array of elements. Operator overloading (like +, *, &&) has been implemented such that the most optimal intrinsic for the available types is selected automatically. Many mathetical functions (like log, exp, sin) and common operations (such as sum, range, for_each) are also available.

By using this library, developers can avoid the complexity of working with reduced precision floating-point types in CUDA and focus on their applications.