From mboxrd@z Thu Jan  1 00:00:00 1970
Message-ID: <41BF5500.4040606@chunder.com>
Date: Wed, 15 Dec 2004 08:02:56 +1100
From: Bruce Ellis <brucee@chunder.com>
User-Agent: Mozilla Thunderbird 1.0 (Windows/20041206)
MIME-Version: 1.0
To: Fans of the OS Plan 9 from Bell Labs <9fans@cse.psu.edu>
Subject: Re: [9fans] (no subject)
References: <9f1dbc022076092306218e169144847b@terzarima.net>
In-Reply-To: <9f1dbc022076092306218e169144847b@terzarima.net>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Topicbox-Message-UUID: 15c33d40-eace-11e9-9e20-41e7f4b1d025

it gets even fuunier the harder people try to get gcc to
align things "correctly".  two examples are the gcc for
the ps2 (where 128 bit alignment is often good/needed)
and the hoops in fftw to try and align the stack for
SIMD optimization.  C99 lacks GOK and maybe Egreg.

brucee

Charles Forsyth wrote:
> Re: [9fans] plan9ports and freebsd 4.x
> 
>>>you're just mad because they didn't put hjdicks in C99.
> 
> that reminds me that, prompted by an aside that brucee made
> last time hjdicks came up (so to speak),
> i had a quick look then at the __packed attribute (or whatever it was)
> that gcc implemented, on a few non-x86 platforms.
> it seemed to me that it didn't implement it
> properly either.  in particular, on architectures on which
> alignment traps might or will occur for certain unaligned
> accesses, the code generated made no
> attempt to avoid them.  neither does ?[acl] as it happens,
> so at least we're completely compatible with gcc in one thing.