RE: unaligned accesses
>memcpy(d, s, 4) should also generate reasonable code, I would have
>thought, but actually, it doesn't... I'll check that...
I have no issue with the memcpy approach, it's just less
obvious than a simple __unaligned qualifier.
>But these are just ugly kludges, unaligned accesses should simply be
True, but the architecture awareness for such discipline is
not widespread. The most prevalent architecture for Linux doesn't
have an issue with unaligned accesses, so most developers
are unaware. (Yes I know the most prevalent Linux architecture
benefits greatly from aligned accesses, but unaligned ones are handled
transparently and hence ignored).
To UNSUBSCRIBE, email to email@example.com
with a subject of "unsubscribe". Trouble? Contact firstname.lastname@example.org