slackbuilds/libraries/trilinos
fourtysixandtwo 982d6ceb86
libraries/trilinos: Fix redundant deps.
Signed-off-by: Willy Sudiarto Raharjo <willysr@slackbuilds.org>
2023-12-22 07:59:53 +07:00
..
README libraries/trilinos: Added (trilinos project) 2022-10-01 09:01:14 +07:00
slack-desc libraries/trilinos: Added (trilinos project) 2022-10-01 09:01:14 +07:00
trilinos.SlackBuild libraries/trilinos: Align to 15.0 template. 2023-02-18 10:06:57 +07:00
trilinos.info libraries/trilinos: Fix redundant deps. 2023-12-22 07:59:53 +07:00

README

 The Trilinos Project is an effort to develop algorithms and enabling 
technologies within an object-oriented software framework for the 
solution of large-scale, complex multi-physics engineering and 
scientific problems. A unique design feature of Trilinos is its focus 
on packages.

The package has two building methods, the default is for a single 
system (without parallelism).

* The trilinos can also be compiled against individual packages 
(i.e. amd, umfpack and metis) outside of suitesparse.SlackBuild. 
** The suitesparse.SlackBuild, doesn't build metis by default. 
For that you can use the parMETIS.SlackBuild.
For more details about dependeces the Xyce webpage have a 
list of required depences.

This package is mantained for building Xyce at 32 and 64 bit
architectures with serial and parallel mode. Nonetheless, some 
required packages for building Xyce are not compiling at 32-bit
architectures, this is Kokkos. This SlackBuild includes a workaround 
for compiling Trilinos at 32-bit architectures, disabling some packages 
that depends on Kokkos, i.e. Amesos2 and Stokhos. These packages 
enable optional features in Xyce. However, according to the Xyce 
building instructions the ShyLU package is a requirement for 
Parallel Xyce but also depends on Kokkos.
* The workaround solution may not allow total funtionality of 
Serial and Parallel Xyce on 32-bit architecture.

Optional dependences:
For enable parallelism set PARALLEL=yes. This requires the 
ParMETIS library and a MPI implementation (openmpi or mpich).
To enable CUDA set CUDA=yes (Additional requirements: CUDA
SDK and CUSPARSE).