Is there any documentation aside from the Metal Shading Language Specification that explains the SIMD group functions such as
- simd_or
- simd_and
- simd_xor
etc.
CUDA has functions comparable to the vote/ballot simd functions, but I am not aware of anything similar to the above. The MSL specification is vague in stating - for example - simd_or returns the bitwise OR across all active threads.
Does that mean T data is applied bitwise to the value in each SIMD lane? What does that return exactly? Do you get only the highest or lowest or a sum of the bitwise operation?