Date: Sun, 12 Nov 2006 18:10:45 -0800 From: Tim Kientzle <kientzle@freebsd.org> Cc: arm@freebsd.org, current@freebsd.org, keramida@freebsd.org Subject: Re: [head tinderbox] failure on arm/arm Message-ID: <4557D425.40101@freebsd.org> In-Reply-To: <20061112.154733.-432839119.imp@bsdimp.com> References: <20061112180758.GD4237@kobe.laptop> <20061112132105.6bac38d6@kan.dnsalias.net> <20061112192810.GC1173@rambler-co.ru> <20061112.154733.-432839119.imp@bsdimp.com>
next in thread | previous in thread | raw e-mail | index | archive | help
In message: <20061112192810.GC1173@rambler-co.ru>
Ruslan Ermilov <ru@freebsd.org> writes:
: On the other hand, having GCC align "struct foo { char x[11]; }"
: on a four-byte boundary (other than for backward compatibility)
: doesn't make too much sense to me.
It makes a lot of sense, actually, even on i386, which is
why the standard allows the compiler a lot of leeway to
pad and align such structure. For example, consider this
structure assignment:
struct foo a,b;
a = b;
The structure copy here is a LOT faster if the structures
are always suitably aligned and padded. If the compiler
pads your sample struct to 12 bytes and 4-aligns, the
structure copy here can be done with 3 aligned 32-bit moves.
Even on i386, aligned operations are a lot faster than
unaligned ones.
The code mentioned in the original post is simply broken;
it does not comply with the ISO C Standard.
Tim Kientzle
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4557D425.40101>