[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <bc4d4f88727b359df5a8cf65bc857922@smtp.hushmail.com>
Date: Tue, 5 Feb 2013 00:54:35 +0100
From: magnum <john.magnum@...hmail.com>
To: john-dev@...ts.openwall.com
Subject: Re: OSX 32-bit assembler warnings.
On 4 Feb, 2013, at 23:49 , magnum <john.magnum@...hmail.com> wrote:
> I made some 32-bit builds on OSX just to test the legacy MMX/SSE builds out (and some bugs were indeed found and fixed).
>
> Due to my habit of always supplying -s to make, I saw these warnings:
> md4-mmx.S:280:Alignment too large: 15. assumed.
> md4-mmx.S:314:Alignment too large: 15. assumed.
> md5-mmx.S:48:Alignment too large: 15. assumed.
>
> That made me laugh, just like the guy in the first google hit:
> http://lists.cs.uiuc.edu/pipermail/llvmdev/2010-January/028881.html
>
> Do I understand right that it will it align to 32K? That's a lot but I presume it will work fine. And the build does work fine.
>
> If we said .balign instead, would some other assemblers not understand it? Or perhaps we could/should use some #ifdef stuff? I see now there is some DO_ALIGN stuff in x86.S but I can't say I understand it. Does it depend on defines in arch.h? OSX does not have it's own.
OK, using #ifdef __APPLE__ works fine - but can I assume that test will always do the right thing? It's better leaving it at 1<<15 than ending up insufficiently aligned on some other version of OSX toolchain that parses it the other way.
Also, I do not see any complaints about .align 16 on -m64... maybe it actually aligns to 1<<16 without complaining?
magnum
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
