iam currently working on an automatic OpenCL to CUDA kernel translator based on Clang.
In order to support translation of OpenCLs built-in functions i would like to know if the following function mapping is correct:As for the functions stated in https://www.khronos.org/registry/OpenCL/sdk/1.0/docs/man/xhtml/mathFunctions.html
OpenCL functions defined without a half_ or native_ prefix will be translated to CUDAs Single Precision Mathematical Functions or Double Precision Mathematical Functions based on the input arguments type.
Functions defined with a half_ prefix will be mapped to CUDAs Half Precision Intrinsics and _native prefix functions to CUDAs Single and Double Precision Intrinsics
I would appreciate your help