Re: Unaligned accesses in sha1dc

From: Junio C Hamano <gitster@pobox.com>
Date: Jun 3, 2017, 00:13 UTC
Message-ID: <xmqq4lvyx71c.fsf@gitster.mtv.corp.google.com>
In-Reply-To: <CA+55aFwL9LQfx8t7tixYgV+2w_=_dAABxk54_GLJoGod-w=mbw@mail.gmail.com>

Linus Torvalds <torvalds@linux-foundation.org> writes:

> Dereferencing an unaligned pointer may be "undefined" in some
> technical meaning, but it sure as hell isn't undefined in reality, and
> compilers that willfully do stupid things should not be catered to
> overly. Reality is a lot more important.

Thanks for succinctly putting it this way. I think you said the above number of times, and I was looking for one of them to include a reference to in the response I was preparing, but this message made it unnecessary ;-)

> And I think gcc can be tweaked to use "movbe" on x86 with the right
> magic incantation (ie not just __builtin_bswap32(), but also the
> switch to enable the new instructions).  So having code to detect gcc
> and using __builtin_bswap32() would indeed be a good thing.