The nop and xchg in the GCC output are padding to make the next function aligned. You can ignore everything after the ret.

The execution of movzx should be negligible. The big difference between the code generated by GCC and the code generated by clang is that the clang code does everything using 8-bit registers, and zero-extends at the end, whereas the GCC code does everything with full 32-bit registers.

I actually had never seen the dil register before this (it's the low 8-bits of the [er]?di register). It's pretty cool that clang uses this when it knows the value in the first argument is byte-sized. I think it's also neat seeing how the compilers use the commutativity of binary-and differently, and end up anding in a different order. Overall I like the clang code better, as it seems closer to the asm someone might write by hand.