[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Mon, 23 Jul 2012 12:11:07 +0200
From: Frank Dittrich <frank_dittrich@...mail.com>
To: john-dev@...ts.openwall.com
Subject: Re: Atom: bf P1 asm results
On 07/23/2012 11:13 AM, Solar Designer wrote:
> On Mon, Jul 23, 2012 at 09:57:22AM +0200, Frank Dittrich wrote:
>> On 07/09/2012 01:31 AM, Solar Designer wrote:
>>> Also, if you omit -march=native, does it still include any Atom-specific
>>> defines on your system?
>>
>> Without --march=native, atom doesn't appear among the definitions, just
>> #define __i386 1
>> #define __i386__ 1
>> #define __i686 1
>> #define __i686__ 1
>> #define __pentiumpro 1
>> #define __pentiumpro__ 1
>> #define i386 1
>>
>> With --march=native, the i36 definitions remain, but i686 and pentiumpro
>> disappear.
>
> i686 disappearing with -march=native on Atom sounds like a gcc bug to me.
> Maybe check if it's present on a clean gcc build (not a Linux distro's)
> and report it upstream if so. Also check for it on non-Atom (but also a
> 32-bit system).
To summarize, with -march=native:
$ gcc -march=native -dM -E - < /dev/null|grep "define _*[aip]"
#define __i386 1
#define __i386__ 1
#define __atom 1
#define __atom__ 1
#define i386 1
Without:
$ gcc -dM -E - < /dev/null|grep "define _*[aip]"
#define __pentiumpro__ 1
#define __i686 1
#define __i686__ 1
#define __i386 1
#define __i386__ 1
#define __pentiumpro 1
#define i386 1
Checking vanilla gcc and reporting upstream might take a while. During
the last weeks I neglected other duties, so I'll have to spend time to
make up for that.
BTW: gcc.gnu.org/bugzilla/show_bug.cgi?id=51017 doesn't show any
progress for half a year.
Status: UNCONFIRMED
This certainly encourages bug reports.
Frank
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.
