From mboxrd@z Thu Jan  1 00:00:00 1970
From: erik quanstrom <quanstro@quanstro.net>
Date: Fri,  9 May 2014 11:29:53 -0400
To: 9fans@9fans.net
Message-ID: <6e34eb4ac48233d6cee9edb0ce2062c3@ladd.quanstro.net>
In-Reply-To: <CAOw7k5iCbM_UUBv0t8OrgZJdf_P7V2gZc1oj_Hka6EJ7mkTNzA@mail.gmail.com>
References: <51d4d8ec419f841a99baf30ede300c25@brasstown.quanstro.net>
	<CAOw7k5iCbM_UUBv0t8OrgZJdf_P7V2gZc1oj_Hka6EJ7mkTNzA@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Subject: Re: [9fans] ureg alignment
Topicbox-Message-UUID: e4d422f2-ead8-11e9-9d60-3106f5b1d025

> > on 64-bit machines, the unions in the ureg.h can lead to internal
> > padding.
>
> although one tactic not yet implemented, but soon ..., is to make
> struct and union alignment match the least alignment required, as is
> done for arrays, the approach on amd64 and arm64 is simply not to
> bother with the unions.  if it's sp, it's sp.  link is link.  i think
> the need for the alias has gone.  there aren't any references in port,
> which i think might have caused the aliasing in the first place, and
> xyz/trap.c should know what's what.

good point. i hadn't considered that it might simply be unnecessary.
perhaps because i was also fretting about this not being done consistently
across various codebases.  do you think that's perhaps worth leaving the
#define in as a temporary measure?

- erik