Re: multiplication does not work!!!

[Dmitri A. Sergatskov]

On Wed, May 11, 2011 at 2:45 PM, Thomas Weber <address@hidden> wrote:
> On Wed, May 11, 2011 at 08:01:36AM -0500, Dmitri A. Sergatskov wrote:
>> On Wed, May 11, 2011 at 3:20 AM, maclair <address@hidden> wrote:
>> > I did as you suggested and sent the bug report:
>> > "Bug #780966 in Octave: Octave crashes with libatlas"
>> >
>>
>>
>> This appears to be the same bug as
>> https://bugzilla.redhat.com/show_bug.cgi?id=573074
>
> They are probably related, but not the same. The problem in Fedora was
> that their generic amd64 (aka x86_64) package used SSE3 instructions;
> the relevant CPUs however are only guaranteed to have SSE2.
>
> Matyas however has an i386 CPU and the atlas package he has should
> indeed work on that architecture without SSE.
>

The ATLAS is a sneaky bastard and it tries to figure out optimal
compile flags on its own. So if the binary were compiled on the
hardware that had support for sseXX , ATLAS would use them.
That was the issue with Fedora bug and I *suspect* is the problem with
this Ubuntu bug.

>        Thomas
>


Sincerely,

Dmitri.
--