Hey Ben, It looks like HCOLL v2.x is too old for OMPI v1.10. Could you download HPCx from http://mellanox.com/poducts/hpcx which contains latest hcoll and use it to compile OMPI?
(also HPCx contains OMPI sources and precompiled w/ HCOLL) On Thu, Dec 24, 2015 at 2:29 AM, Ben Menadue <ben.mena...@nci.org.au> wrote: > Hi Mike, > > > > 11:18 bjm900@raijin7 ~ > rpm -qi hcoll > > Name : hcoll Relocations: > /opt/mellanox/hcoll > > Version : 2.0.472 Vendor: mellanox > > Release : 1 Build Date: Sun Jan 5 > 23:40:18 2014 > > Install Date: Fri Jun 12 15:00:23 2015 Build Host: localhost > > Group : Applications Source RPM: > hcoll-2.0.472-1.src.rpm > > Size : 14074151 License: Proprietary > > Signature : DSA/SHA1, Sun Feb 23 23:20:58 2014, Key ID c5ed83e26224c050 > > Packager : mellanox > > URL : http://www.mellanox.com > > Summary : Hierarchical collectives (HCOLL) > > Description : > > HCOLL... > > > > I also just realised that it doesn’t work for any OpenMPI 1.8.x series or > 1.10.0 either – while it is able to build (unlike for 1.10.1) it fails at > runtime with a symbol lookup error: > > > > symbol lookup error: /apps/openmpi/1.8.8/lib/openmpi/mca_coll_hcoll.so: > undefined symbol: hcoll_check_mem_release_cb_needed > > > > That symbol doesn’t appear to be in any of the .so files in > /opt/mellanox/hcoll, so I’m guessing it’s from a newer version of hcoll. > This would also explain why it warned about an implicit declaration of this > symbol during the builds: > > > > ../../../../../../../../ompi/mca/coll/hcoll/coll_hcoll_module.c:263: > warning: implicit declaration of function > 'hcoll_check_mem_release_cb_needed' > > > > Cheers, > > Ben > > > > *From:* users [mailto:users-boun...@open-mpi.org] *On Behalf Of *Mike > Dubman > *Sent:* Thursday, 24 December 2015 7:14 AM > *To:* Open MPI Users <us...@open-mpi.org> > *Subject:* Re: [OMPI users] hcoll API in 1.10.1 > > > > Hi, > > hcoll is part of MOFED or comes from HPCx. > > what version of hcoll do you have on your system? > > > > Thx > > > > On Wed, Dec 23, 2015 at 4:58 AM, Ben Menadue <ben.mena...@nci.org.au> > wrote: > > Hi, > > It's probably in plain sight somewhere and I missed it, but is there a > minimum version of hcoll needed to build 1.10.1? > > We have 2.0.0, which allows us to build 1.10.0, but 1.10.1 fails with > missing entries in the hcoll_collectives_t structure: > > CC coll_hcoll_module.lo > ../../../../../../../../ompi/mca/coll/hcoll/coll_hcoll_module.c: In > function > 'mca_coll_hcoll_mem_release_cb': > ../../../../../../../../ompi/mca/coll/hcoll/coll_hcoll_module.c:60: > warning: > implicit declaration of function 'hcoll_mem_unmap' > ../../../../../../../../ompi/mca/coll/hcoll/coll_hcoll_module.c: In > function > 'hcoll_comm_attr_del_fn': > ../../../../../../../../ompi/mca/coll/hcoll/coll_hcoll_module.c:166: > warning: implicit declaration of function 'hcoll_group_destroy_notify' > ../../../../../../../../ompi/mca/coll/hcoll/coll_hcoll_module.c: In > function > 'mca_coll_hcoll_comm_query': > ../../../../../../../../ompi/mca/coll/hcoll/coll_hcoll_module.c:263: > warning: implicit declaration of function > 'hcoll_check_mem_release_cb_needed' > ../../../../../../../../ompi/mca/coll/hcoll/coll_hcoll_module.c:317: error: > 'hcoll_collectives_t' has no member named 'coll_alltoallv' > ../../../../../../../../ompi/mca/coll/hcoll/coll_hcoll_module.c:318: error: > 'hcoll_collectives_t' has no member named 'coll_gatherv' > ../../../../../../../../ompi/mca/coll/hcoll/coll_hcoll_module.c:324: error: > 'hcoll_collectives_t' has no member named 'coll_igatherv' > > It looks like PR567 in the 1.10 branch that added the new references. > > Cheers, > Ben > > > _______________________________________________ > users mailing list > us...@open-mpi.org > Subscription: http://www.open-mpi.org/mailman/listinfo.cgi/users > Link to this post: > http://www.open-mpi.org/community/lists/users/2015/12/28200.php > > > > > > -- > > > > Kind Regards, > > > > M. > > _______________________________________________ > users mailing list > us...@open-mpi.org > Subscription: http://www.open-mpi.org/mailman/listinfo.cgi/users > Link to this post: > http://www.open-mpi.org/community/lists/users/2015/12/28202.php > -- Kind Regards, M.
