On Fri, Nov 13, 2009 at 10:01 AM, Georg S. Weber
<georgswe...@googlemail.com> wrote:
>
>
>
> On 13 Nov., 10:05, "David M. Monarres" <dmmonar...@gmail.com> wrote:
>> Looking into the 'abort trap' problem and found
>>
>> http://trac.macports.org/ticket/20491
>>
>> which relates the problem to numpy linking to a system blas/atlas.
>>
>> So I poked around the blas and atlas packages and noticed that they were 
>> disabled on the Darwin platform and the wisdom of this decision was 
>> confirmed when the numpy didn't build after I altered the packages to build. 
>> Since macports seem to be able to get numpy + blas + atlas all living in 
>> harmony on 10.6 I assume that sage can also.
>>
>> Has anybody so far looked down this road and reached a dead end? I just want 
>> to check to make sure that I am not barking up the wrong tree.
>>

I think you and I are perhaps (?) the only one who has done any work
on this Abort Trap issue with Sage.   I don't think the problem is
related to numpy or that thread... but I also wasn't able to solve the
problem (yet), so don't listen to me!
That thread you link to doesn't even mention OS X 10.6 anywhere, does
it, and it is before 10.6 was released.  I think the error "Abort
Trap" could arise for a billion different reasons...

>> David Monarres
>> dmmonar...@gmail.com

>
> Somehow,
>
> I've got the feeling that this leads in the right direction. Building
> and integrating BLAS and ATLAS (and LAPACK ?) into the numpy from Sage
> under Mac OS X should be doable. Does anybody know why this isn't the
> case already?

I see no point at all in building ATLAS on OS X, since OS X already
ships a highly tuned "Apple-certified" multi-core aware build of ATLAS
systemwide.  The ATLAS's we build/ship with Sage are never multicore
aware.  It's simply much better as far as I can tell for Sage to use
the system-wide ATLAS.

Georg, you seem to have the impression above that Sage's numpy isn't
using ATLAS.  It is -- it's just that it is using the Apple build of
ATLAS.

William

--~--~---------~--~----~------------~-------~--~----~
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to 
sage-devel-unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org
-~----------~----~----~----~------~----~------~--~---

Reply via email to