https://github.com/sandialabs/seacas
The Sandia Engineering Analysis Code Access System (SEACAS) is a suite of preprocessing, postprocessing, translation, and utility applications supporting finite element analysis software using the Exodus database file format.
https://github.com/sandialabs/seacas
scr-10 scr-126 scr-127 scr-532 scr-59 scr-638 scr-69 scr-70 scr-75 scr-76 scr-864 scr-869 scr-870 scr-90 scr-91 scr-92 scr-93 snl-applications snl-comp-science-libs snl-science-libs
Last synced: about 1 month ago
JSON representation
The Sandia Engineering Analysis Code Access System (SEACAS) is a suite of preprocessing, postprocessing, translation, and utility applications supporting finite element analysis software using the Exodus database file format.
- Host: GitHub
- URL: https://github.com/sandialabs/seacas
- Owner: sandialabs
- License: other
- Created: 2015-06-05T16:44:12.000Z (almost 10 years ago)
- Default Branch: master
- Last Pushed: 2024-06-11T15:55:31.000Z (11 months ago)
- Last Synced: 2024-06-11T17:01:08.332Z (11 months ago)
- Topics: scr-10, scr-126, scr-127, scr-532, scr-59, scr-638, scr-69, scr-70, scr-75, scr-76, scr-864, scr-869, scr-870, scr-90, scr-91, scr-92, scr-93, snl-applications, snl-comp-science-libs, snl-science-libs
- Language: C
- Homepage:
- Size: 356 MB
- Stars: 127
- Watchers: 9
- Forks: 74
- Open Issues: 15
-
Metadata Files:
- Readme: README-KOKKOS.md
- Contributing: CONTRIBUTING.md
- License: LICENSE
- Code of conduct: CODE-OF-CONDUCT.md
Awesome Lists containing this project
README
A capability in testing is to build an "on-node parallel" version of
SEACAS using the Kokkos package. This includes a "Kokkos-aware" version of
the Ioss library. The modifications to the build process described in the
README file are shown below:
## Additional Libraries
You will need the following library if you want to use Kokkos with CUDA as the backend:
* [CUDA](#cuda)
### CUDA
* CUDA is already installed on many platforms. In many cases, typing something like `module load cuda` should be sufficient. Loading the module would typically set environment variables, such as CUDA_ROOT.
* If installing yourself, see www.nvidia.com/object/cuda_home_new.html
## Configure, Build, and Install SEACAS
Build as described in README-PARALLEL.md if using MPI or as described in README.md
otherwise. If you are using the `cmake-config-kokkos` script, change `MPI` to
`ON` or `OFF` as appropriate, change `CUDA` to `ON` or `OFF`
as appropriate, and then source the configure script.
```bash
source cmake-config-kokkos
```
If using your own cmake script or directly calling cmake, be sure to do the following.
* Enable the Kokkos package.
* If using OpenMP as the Kokkos backend, enable OpenMP and disable Pthread.
If using CUDA as the backend, OpenMP can still be enabled.
```bash
-D Seacas_ENABLE_OpenMP:Bool=ON
-D TPL_ENABLE_Pthread:Bool=OFF
```
* If using CUDA as the Kokkos backend, enable CUDA, set the CUDA root directory, and disable Pthread.
```bash
-D TPL_ENABLE_CUDA:Bool=ON
-D CUDA_TOOLKIT_ROOT_DIR:Path=${CUDA_PATH}
-D TPL_ENABLE_Pthread:Bool=OFF
```
* If using CUDA as the Kokkos backend, some environment variables need to be set. These can be set in your configure script, but then be sure to source the script rather than just running it so that your environment is correct when running `make`.
```bash
export OMPI_CXX=/packages/kokkos/config/nvcc_wrapper
export CUDA_MANAGED_FORCE_DEVICE_ALLOC=1
```
Finally, build and install
```bash
make
make install
```