If a Go compiler reorders struct fields all the padded structs will not offer good cache line protection.

[fmstephe]

Currently values in the padded package isolate an inner field by putting enough data before and after to separate that field from any other memory locations. However, this only works if the Go compiler places struct fields in the same order as found in the source code. This is definitely not guaranteed.

An alternative approach would be to make the entire struct a byte array and use unsafe to interpret some inner bytes as a particular value. We could have a function which returns a pointer to this value - for easy use with the stdlib atomic package.