[cfp-interest 3585] name of bfloat16 functions

[Paul Zimmermann]

       Hi,

now that we have a complete set of binary16 functions in CORE-MATH,
we start to work on bfloat16 functions.

For the binary16 functions we used the "f16" suffix, as for example
in the Intel Math Library: expf16, logf16, sinf16, ...

For the bfloat16 functions, I thought using the "bf16" suffix, but
this yields a conflict: logbf16 could be logb for f16 or log for bf16.

I see CUDA uses the "h" prefix:

https://docs.nvidia.com/cuda/cuda-math-api/cuda_math_api/group__CUDA__MATH____BFLOAT16__FUNCTIONS.html

Is there a standard? If not, what do you recommend?

Paul