#9356: make SAGE_ATLAS_LIB work on Solaris
-----------------------------+----------------------------------------------
Reporter: jhpalmieri | Owner: drkirkby
Type: defect | Status: positive_review
Priority: minor | Milestone: sage-4.5
Component: solaris | Keywords:
Author: John Palmieri | Upstream: N/A
Reviewer: David Kirkby | Merged:
Work_issues: |
-----------------------------+----------------------------------------------
Comment(by drkirkby):
I can't see any logic for having both shared and static libraries, though
it would never suprise me if one bit of Sage will not work with one type,
so we need both.
If I understand correctly the static (.a) libraries are created and
converted to shared libraries (.so), on all platforms except Solaris,
since the conversion was not working in Solaris. I note that Mathematica
ships with the shared library only on !OpenSolaris on x64 - I don't have a
Solaris 10 SPARC switched on with Mathematica on it, but I can check
later. I suspect Wolfram Research only ship the shared library. The
Solaris operating system does not ship with a single static library -
their use is unpopular for their many disadvantages.
The main advantage for the shared library would be a reduced memory
footprint when multiple users make use of the same library. An ideal
solution would be to create shared libraries properly on all platforms,
and delete the static ones.
I did look at updating ATLAS, but I had problems. I spent several hours on
it, and got nowhere. Perhaps I was being too ambitions - trying to build
ATLAS in parallel. If one looks at the ATLAS package, one can see that
multiple threads are disabled, so the linear algebra will be solved more
slowly than necessary. ATLAS is designed to be built in parallel - there
are options to the Configure script for this. As are there options to
control the number of threads used at runtime.
I'd really like to resolve the issue with Sage building properly on
!OpenSolaris before devoting a lot of time to ATLAS. The !OpenSolaris is
annoying, as it builds, but does not run (segfault's at startup). I don't
have much of a clue on how to debug it. I was hoping William would, but he
has not looked at it despite I created him an account on my !OpenSolaris
machine with a built, but non-working version of Sage waiting to debugged.
I think this ticket is a useful addition to Sage, but long term it would
be better to have only the shared libraries, and so treat ATLAS the same
ways as other platforms.
I suspect, but have not confirmed, that the shared libraries will actually
work properly on Solaris - they are certainly created, then deleted. I
suspect that Micheal deleted them since he used the GNU linker (I know he
used it), despite even the GCC docs telling you to use the Sun linker. (Of
course, trying to build Sage with the GNU linker was easier, as one did
not have to rid Sage of so many GNUisms. But ultimately, getting rid of
the GNUisms and using the Sun linker was preferable.) Now Sage builds with
the Sun linker, it would be worth investigating if the shared library work
on Solaris, and we just delete the static libraries.
Dave
--
Ticket URL: <http://trac.sagemath.org/sage_trac/ticket/9356#comment:6>
Sage <http://www.sagemath.org>
Sage: Creating a Viable Open Source Alternative to Magma, Maple, Mathematica,
and MATLAB
--
You received this message because you are subscribed to the Google Groups
"sage-trac" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to
[email protected].
For more options, visit this group at
http://groups.google.com/group/sage-trac?hl=en.
