CUDA documentation typo?

There appears to be a typo in the Toolkit documentation (see screenshot). float3 is aligned to 16 and not 4 bytes. This would also be consistent with vector_types.h.
Perhaps float w is missing?

in which document is this?

Thanks for the reply. Here it is

That is giving examples of possible usage of the __align__() directive. It has nothing to do with the float3 type.