Programs and libraries for graph, mesh and hypergraph partitioning

Edit Package scotch

Programs and libraries for graph, mesh and hypergraph partitioning Its purpose is to apply graph theory, with a divide and conquer approach, to scientific computing problems such as graph and mesh partitioning, static mapping, and sparse matrix ordering, in application domains ranging from structural mechanics to operating systems or bio-chemistry.

The SCOTCH distribution is a set of programs and libraries which implement the static mapping and sparse matrix reordering algorithms developed within the SCOTCH project.

Refresh
Refresh
Source Files
Filename Size Changed
_multibuild 0000000357 357 Bytes
scotch-Makefile.inc.in 0000000455 455 Bytes
scotch.changes 0000005315 5.19 KB
scotch.spec 0000014300 14 KB
scotch_6.0.7.tar.gz 0006136154 5.85 MB
Revision 21 (latest revision is 52)
Atri Bhattacharya's avatar Atri Bhattacharya (badshah400) accepted request 726716 from Ana Guerrero's avatar Ana Guerrero (anag) (revision 21)
- Update to version 6.0.7:
  * bugfixes and some new minor features, such as an extension of 
    the target architecture API and MeTiS v5 compatibility.
Comments 2

Mark Olesen's avatar

It doesn't seem possible to link in a serial scotch library with any of the SLE_15 variants. Is this an oversight, or are there symbol conflicts with the respective MPI variants?

Would expect something like this around line 360:

# Don't build non-HPC on SLE, but serial is still ok
%if !0%{?is_opensuse} && !0%{?with_hpc:1} && "%{flavor}" != "serial"
ExclusiveArch:  do_not_build
%endif

Mark Olesen's avatar

General note for scotch maintainer(s):

with OpenFOAM we noticed some regressions with scotch-7.0.1 (perhaps other versions too) and are thus currently sticking locally with scotch-6.1.0 - so some caution may be needed if/when update versions here.

In homebrew (for example), they have aggressively enabled -DSCOTCH_PTHREAD_MPI, which means that any program using ptscotch will fail if MPI is not initialized with MPI_THREAD_MULTIPLE (we normally do not use MPI threading for performance reasons). It would be nice to avoid that here.

openSUSE Build Service is sponsored by