Okay, that's two in a row, from two different sources even.
New rules:
People who send help-providing messages to me alone when they should
be Cc:ing libmesh-users will get forwarded to libmesh-users, but may
get it pointed out in a teasing/patronizing preface like this one if
I'm in a mean mood.
People who send help-requesting messages to me alone may get forwarded
to libmesh-users if I'm in a good mood, may get a one-line reply to
whatever URL archives the message I'm typing now, or may just get
ignored if I'm in a bad mood.
Hitting "reply-to-all" is still fine; I don't mind getting extra
copies of things, I just want to make sure that the search engines all
get to see the Q&A too. In my fantasy world everybody hits Google
before asking technical questions of real humans, and I'd like to make
that as effective as possible.
This is just a little added encouragement for posting to libmesh-users
and libmesh-devel when appropriate. Using the mailing lists was
*already* a good idea, as Cody's quick helpful response demonstrates.
I know sometimes it *seems* like I'm a lone font of helpful
information who never goes on vacation or gets too busy or loses track
of email, but really none of those things are true, ;-)
---
Roy
---------- Forwarded message ----------
Date: Thu, 25 Aug 2011 02:59:46 +0000
From: [email protected]
To: Roy Stogner <[email protected]>
Subject: Re: Re: [Libmesh-users] linking problems installation (fwd)
Hi Robert,
My experience with the BIG machines is that many of them don't support dynamic
linking. This static build produces bigger
binaries but simplifies the runtime environment on the compute nodes. Most of
the Cray machines are designed this way including
Jaguar. I can't say for sure what IBM does since I don't have experience with
that architecture but before you dive very far
into this, I would suggest you try configuring libmesh with "--disable-shared"
and see if that resolves the problem.
Best Regards,
Cody
On , Roy Stogner <[email protected]> wrote:
---------- Forwarded message ----------
Date: Thu, 25 Aug 2011 00:23:35 +0200
From: robert [email protected]>
To: Roy Stogner [email protected]>
Subject: Re: [Libmesh-users] linking problems installation
Once again I have to bother you with a question:
I moved to a BlueGene P machine an am trying to get to run libmesh
there. MPI, PETSC and SLEPC should be fine (example tests worked).
However, I get the following error:
----------------------------------------------
----- Done Building Contributed Packages -----
----------------------------------------------
make[1]: Leaving directory
`/home/bodner/software/libmesh-0.7.1/libmesh/contrib'
Compiling C++ (in optimized mode) src/apps/amr.C...
Linking bin/amr-opt
/bgsys/drivers/V1R4M2_200_2010-100508P/ppc/gnu-linux/lib/gcc/powerpc-bgp-linux/4.1.2/../../../../powerpc-bgp-linux/bin/ld:
attempted static link of dynamic object
`/home/bodner/software/libmesh-0.7.1/libmesh/lib/powerpc64-unknown-linux-gnu_opt/libmesh.so'
collect2: ld gab 1 als Ende-Status zurück
make: *** [bin/amr-opt] Fehler 1
I an sorry to bother you all the time but I can't get it run with MPI.
Robert
Am Freitag, den 19.08.2011, 13:43 -0500 schrieb Roy Stogner:
On Fri, 19 Aug 2011, robert wrote:
Am Freitag, den 19.08.2011, 13:31 -0500 schrieb Roy Stogner:
Is it possible that your PETSc build is configured
with the "separate everything into libpetscsnes/ksp/vec/etcetera"
option but that libMesh somehow thinks you're using the "include every
object in libpetsc" option? What's
"ls -l /home/robert/include/petsc-3.1-p8/linux-gnu/lib" show?
robert@robert:~/include/petsc-3.1-p8$ ls
-l /home/robert/include/petsc-3.1-p8/linux-gnu/lib
insgesamt 66628
-rw-r--r-- 1 robert robert 988754 2011-08-19 19:17 libfblas.a
-rw-r--r-- 1 robert robert 10935510 2011-08-19 19:17 libflapack.a
-rw-r--r-- 1 robert robert 53811598 2011-08-19 20:31 libpetsc.a
-rw------- 1 robert robert 2404352 2011-08-19 20:31 stdXEqfD
robert@robert:~/include/petsc-3.1-p8$
I have no idea what's not working, then, but I have a suggestion for a
possible workaround - try building PETSc/SLEPc with shared libraries
and see if you see the same problem.
---
Roy
------------------------------------------------------------------------------
EMC VNX: the world's simplest storage, starting under $10K
The only unified storage solution that offers unified management
Up to 160% more powerful than alternatives and 25% more efficient.
Guaranteed. http://p.sf.net/sfu/emc-vnx-dev2dev
_______________________________________________
Libmesh-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libmesh-users
