Re: [OMPI users] mpicc link shouldn't add -ldl and -lhwloc
On May 15, 2012, at 10:37 AM, Orion Poplawski wrote: > $ mpicc -showme:link > -pthread -m64 -L/usr/lib64/openmpi/lib -lmpi -ldl -lhwloc > > -ldl and -lhwloc should not be listed. The user should only link against > libraries that they are using directly, namely -lmpi, and they should > explicitly add -ldl and -lhwloc if their code directly uses those libraries. > There does not appear to be any references to libdl or libhwloc symbols in > the openmpi headers which is the other place it could come in. I just read this a few times, and I admit that I'm a little confused. libmpi does use symbols from libdl; we use it to load OMPI's plugins. So I'm not sure why you're saying we shouldn't -ldl in the wrapper compiler...? libhwloc might be a little questionable here -- I'll have to check to see whether 1.6 uses hwloc only in a plugin or whether it's used in the base library (I don't remember offhand). -- Jeff Squyres jsquy...@cisco.com For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/
Re: [OMPI users] Need MPI algorithms, please help
You haven't really stated what kind of algorithms you want, other than the fact that they need to be parallel. :-) Your best bet is Google. There's lots of high-quality MPI-based middleware out there such that you might not need re-invent the wheel. Good luck. On May 21, 2012, at 3:02 PM, Mudassar Majeed wrote: > Dear MPI people, > I need a set of algorithms for > calculating the same thing using different distributed (MPI) algorithms. The > algorithms may need the different data distribution and their execution times > are sensitive to the problem size, number of processes and data distribution > used. We are working on composition of parallel programs over the clusters, > so we need such distributed algorithms written in OpenMPI that take different > execution time for different parameters (data distribution, problem size and > number of processes). Like a variant (algorithm) A may be good for a certain > data distribution and problem size and number of processes but at certain > other parameter values some other variant is better. We will do the online > selection of the best algorithm for the given parameters. But we need good > algorithms for that. Please provide me a link where I can find different > implementations of a certain computation (distributed algorithms in OpenMPI/C > or C++). > > best regards > > --Mudassar > ___ > users mailing list > us...@open-mpi.org > http://www.open-mpi.org/mailman/listinfo.cgi/users -- Jeff Squyres jsquy...@cisco.com For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/
Re: [OMPI users] mpicc link shouldn't add -ldl and -lhwloc
Thank you! This is on my list to investigate; I'm sorry I haven't had a chance to look at it in detail yet. I hope to look at this in detail later this week. On May 21, 2012, at 4:06 PM, Orion Poplawski wrote: > On 05/15/2012 10:37 AM, Orion Poplawski wrote: >> See https://bugzilla.redhat.com/show_bug.cgi?id=814798 >> >> $ mpicc -showme:link >> -pthread -m64 -L/usr/lib64/openmpi/lib -lmpi -ldl -lhwloc >> >> -ldl and -lhwloc should not be listed. The user should only link against >> libraries that they are using directly, namely -lmpi, and they should >> explicitly add -ldl and -lhwloc if their code directly uses those libraries. >> There does not appear to be any references to libdl or libhwloc symbols in >> the >> openmpi headers which is the other place it could come in. >> >> configure appears to add them to opal_WRAPPER_EXTRA_LIBS which then makes its >> way into this list. >> >> This report is with openmpi-1.5.5. Haven't tested with 1.6, but configure >> looks to do the same in it. > > Filed https://svn.open-mpi.org/trac/ompi/ticket/3103 > > -- > Orion Poplawski > Technical Manager 303-415-9701 x222 > NWRA, Boulder Office FAX: 303-415-9702 > 3380 Mitchell Lane or...@nwra.com > Boulder, CO 80301 http://www.nwra.com > ___ > users mailing list > us...@open-mpi.org > http://www.open-mpi.org/mailman/listinfo.cgi/users -- Jeff Squyres jsquy...@cisco.com For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/
Re: [OMPI users] mpirun error with " The MPI_Comm_f2c() function was called before MPI_INIT was invoked."
On May 21, 2012, at 2:40 PM, Reuti wrote: >> I encountered a question about openmpi running. I have compiled >> successfully a paralle VASP program with openmpi 1.4.2. But when I run > > Which version of VASP are you using? By default I can't spot any direct call > to MPI_Comm_f2c in the source. Do you want to compile the ScaLAPACK or plain > MPI version? FWIW, this is probably an internal call to MPI_Comm_f2c -- i.e., calling a Fortran MPI_ function, which, in turn, calls MPI_Comm_f2c to translate a comm parameter to its C equivalent. -- Jeff Squyres jsquy...@cisco.com For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/
Re: [OMPI users] mpicc link shouldn't add -ldl and -lhwloc
On 05/15/2012 10:37 AM, Orion Poplawski wrote: See https://bugzilla.redhat.com/show_bug.cgi?id=814798 $ mpicc -showme:link -pthread -m64 -L/usr/lib64/openmpi/lib -lmpi -ldl -lhwloc -ldl and -lhwloc should not be listed. The user should only link against libraries that they are using directly, namely -lmpi, and they should explicitly add -ldl and -lhwloc if their code directly uses those libraries. There does not appear to be any references to libdl or libhwloc symbols in the openmpi headers which is the other place it could come in. configure appears to add them to opal_WRAPPER_EXTRA_LIBS which then makes its way into this list. This report is with openmpi-1.5.5. Haven't tested with 1.6, but configure looks to do the same in it. Filed https://svn.open-mpi.org/trac/ompi/ticket/3103 -- Orion Poplawski Technical Manager 303-415-9701 x222 NWRA, Boulder Office FAX: 303-415-9702 3380 Mitchell Lane or...@nwra.com Boulder, CO 80301 http://www.nwra.com
[OMPI users] Need MPI algorithms, please help
Dear MPI people, I need a set of algorithms for calculating the same thing using different distributed (MPI) algorithms. The algorithms may need the different data distribution and their execution times are sensitive to the problem size, number of processes and data distribution used. We are working on composition of parallel programs over the clusters, so we need such distributed algorithms written in OpenMPI that take different execution time for different parameters (data distribution, problem size and number of processes). Like a variant (algorithm) A may be good for a certain data distribution and problem size and number of processes but at certain other parameter values some other variant is better. We will do the online selection of the best algorithm for the given parameters. But we need good algorithms for that. Please provide me a link where I can find different implementations of a certain computation (distributed algorithms in OpenMPI/C or C++). best regards --Mudassar
Re: [OMPI users] mpirun error with " The MPI_Comm_f2c() function was called before MPI_INIT was invoked."
Hi, Am 13.05.2012 um 14:37 schrieb Dongshan Wei: > Hi All, > > I encountered a question about openmpi running. I have compiled > successfully a paralle VASP program with openmpi 1.4.2. But when I run Which version of VASP are you using? By default I can't spot any direct call to MPI_Comm_f2c in the source. Do you want to compile the ScaLAPACK or plain MPI version? -- Reuti > the compiled program with mpirun, I got the following information at > very beginning: > > *** The MPI_Comm_f2c() function was called before MPI_INIT was invoked. > *** This is disallowed by the MPI standard. > *** Your MPI job will now abort. > > I believe this error will not related to the VASP program, but only to > openmpi. But I don't how to handle. Does anybody give some solution? > > Thanks in advance! > > Dongshan > > -- > - > Dongshan Wei, Dr. > Chongqing Institute of Green and Intelligent Technology, CAS > 85 Jinyu Ave, Yubei Area, Chongqing 401120, China > Email: ds...@cigit.ac.cn > Phone: (+86) 23-63063762 > - > ___ > users mailing list > us...@open-mpi.org > http://www.open-mpi.org/mailman/listinfo.cgi/users
Re: [OMPI users] MPI books and resources
Am 12.05.2012 um 12:18 schrieb Rohan Deshpande: > Hi, > > Can anyone point me to good resources and books to understand the detailed > architecture and working of MPI. > > I would like to know all the minute details. http://mitpress.mit.edu/catalog/item/default.asp?ttype=2=3945 http://mitpress.mit.edu/catalog/item/default.asp?ttype=2=3614 http://mitpress.mit.edu/catalog/item/default.asp?ttype=2=3490 http://www.mcgraw-hill.co.uk/html/0071232656.html
