threads vs memory layout

What do POSIX and other thread APIs have to say about memory layout
of data accessed by different treads?  E.g.:

    struct Foo {
        pthread_mutex_t ma, mb;
        int a, b; /* protected by ma and mb respectively */
        int c; /* not mutex-protected, only accessed by 1 thread */
    };

I expect that is wrong if I replace int with char or short, since the
machine might not have instructions to access smaller units than a word.
OTOH I hope I can trust it with int - or sig_atomic_t.  But then there
are pointers, floats etc, with who knows what widths...  Is there e.g.
any reason to worry that someone might make a machine with 4-byte float
but 8-byte word, and 1-word minimum memory access?

-- 
Hallvard