Search Criteria
Package Details: metis 5.2.1-2
Package Actions
Git Clone URL: | https://aur.archlinux.org/metis.git (read-only, click to copy) |
---|---|
Package Base: | metis |
Description: | Serial Graph Partitioning and Fill-reducing Matrix Ordering |
Upstream URL: | https://github.com/KarypisLab/METIS |
Licenses: | Apache |
Submitter: | arojas |
Maintainer: | carlosal1015 |
Last Packager: | carlosal1015 |
Votes: | 28 |
Popularity: | 1.03 |
First Submitted: | 2022-11-21 21:22 (UTC) |
Last Updated: | 2025-04-07 16:53 (UTC) |
Dependencies (3)
Required by (39)
- apbs
- coin-or-coinmumps
- coin-or-couenne
- coin-or-hsl
- colmap
- colmap-git
- deal-ii (optional)
- elemental-gmp-git
- freefem (optional)
- gmsh
- gmsh (make)
- gmsh-docs (make)
- gmsh-git
- gmsh-git (make)
- gmsh-git-docs (make)
- kahip
- lib32-metis
- libmesh-petsc
- medcoupling
- mfem
- mumps
- mumps-cmake
- mumps-par
- netgen-nogui
- openfoam-com-git
- parmetis
- parmetis-git
- parmetis-mpich
- pastix
- petsc (optional)
- petsc-complex (optional)
- petsc-git (optional)
- python-pyfr (optional)
- python-pyfr-git (optional)
- siesta
- sparselizard
- spral
- tnl-git
- xyce-shylu
Latest Comments
1 2 Next › Last »
medaminezghal commented on 2025-04-17 20:18 (UTC)
@lahwaacz scotch does support 64bits integer (you can check by reading
INSTALL.txt
from source. It’s not necessary to for the amount of data to be twice the size and I don’t think it’s a big deal.If you don’t mind, could you build metis, scotch and pastix with 64bits support and test if it will cause any problems?
I’ve done the same work in my machine to use petsc and it works well.
lahwaacz commented on 2025-04-17 19:58 (UTC)
@medaminezghal I think you underestimate the amount of work needed here. E.g.
-DPASTIX_INT64=ON
will not work for pastix because it also depends on scotch which has 32-bit ints at the moment.Also note that switching the types will make pretty much all data structures take about twice the amount of memory. The performance effects are really case-by-case depending on many factors.
medaminezghal commented on 2025-04-17 19:07 (UTC) (edited on 2025-04-17 19:19 (UTC) by medaminezghal)
@lahwaacz I think it's better and faster to use 64bits support rather than 32bits.
I have checked that only few packages need to be rebuilt to use the 64bits version: petsc, kahip, pastix, spral. The others have if else case to support both.
carlosal1015 commented on 2025-04-17 18:42 (UTC)
Many thanks, therefore will keep the same configuration.
lahwaacz commented on 2025-04-17 18:40 (UTC)
@carlosal1015 metis can be built with either 32-bit types or 64-bit types, but not both at the same time. This change would affect all packages that depend on metis. Some packages have multiple dependencies like metis that force either 32-bit or 64-bit types and will not work if one but not all packages change the size of types.
carlosal1015 commented on 2025-04-17 18:09 (UTC)
Thanks @medaminezghal, if nobody objects, I'll add it in two weeks.
medaminezghal commented on 2025-04-17 17:25 (UTC) (edited on 2025-04-17 17:34 (UTC) by medaminezghal)
@carlosal1015 I have some suggestion:
1- Add 64bits support by Adding
to the prepare function and
i64=1 r64=1
to themake config
command.2- Add OpenMP support (it won't affect the usage if you don't use
ENABLE_OPENMP
variable) by addingopenmp=1
to themake config
command and OpenMP as dependency.xantares commented on 2025-04-07 16:22 (UTC)
its the same problem than gklib, the first line of cmakelists.txt should be modified:
cmake_minimum_required(VERSION 3.10)
Libellisth commented on 2025-03-31 23:29 (UTC) (edited on 2025-03-31 23:37 (UTC) by Libellisth)
I ran into a problem building with CMake 4.0
So I ran
makepkg -o
, changed the value to 3.5 insrc/METIS-5.2.1/CMakeLists.txt
, then continued with withmakepkg -ei
lahwaacz commented on 2023-08-02 10:31 (UTC)
@SvenMeyer The problem is that it runs
/home/sum/.local/bin/cmake
instead of/usr/bin/cmake
. You need to solve that on your own.1 2 Next › Last »