https://github.com/sandialabs/affect
Read/write and post-process results of the finite element method in Python and C++.
https://github.com/sandialabs/affect
scr-2244 snl-data-analysis snl-science-libs
Last synced: 3 months ago
JSON representation
Read/write and post-process results of the finite element method in Python and C++.
- Host: GitHub
- URL: https://github.com/sandialabs/affect
- Owner: sandialabs
- License: other
- Created: 2014-03-14T15:11:51.000Z (about 11 years ago)
- Default Branch: master
- Last Pushed: 2018-12-18T16:13:30.000Z (over 6 years ago)
- Last Synced: 2025-01-11T15:25:20.502Z (4 months ago)
- Topics: scr-2244, snl-data-analysis, snl-science-libs
- Language: JavaScript
- Homepage:
- Size: 8.27 MB
- Stars: 1
- Watchers: 3
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
affect
======
The affect Python package provides a API/library for processing meshes and field results of a finite element
analysis.
License
-------
BSD. Specifically the "new/modified BSD" license, also called the "BSD three clause license".
See [license file](https://github.com/kevincopps/affect/blob/master/LICENSE).
Documentation
-------------
See the [affect documentation here](https://kevincopps.github.io/affect).
Basic Goals and Capabilities
----------------------------
* Designed for use on many-core shared memory workstations or servers using
threaded parallelism.
* Easy to use within a Python (_Jupyter_) Notebook and targets modern Python
(>3.5).
* Support for faster operations on very large analysis results on
unstructured grids (billions of elements/nodes) using a compact storage scheme and CPU threaded parallelism.
* Fast skinning of mesh/block boundaries.
* A clean Python Database class for reading mesh and fields from ExodusII files directly into Numpy arrays.
* Some optimized built in postprocessors, like the FRF (frequency response function).
Using the Library
-----------------
### Dependencies
`affect` requires Python 3.6+ with additional package dependencies. The easiest way to get going
is by installing and using the [Anaconda](https://www.continuum.io/downloads) python environment.
* [numpy](htttps://www.numpy.org)
* [numexpr](https://github.com/pydata/numexpr)
* [netcdf4](https://pypi.python.org/pypi/netCDF4)
The netcdf4 package has additional dependencies including the hdf4 and hdf5 libraries.
For a complete list of dependencies see the `requirements.txt` file.
### OpenMP and thread affinity
Setting OpenMP environment values for binding threads may improve performance. The matter of thread affinity becomes
important on multi-socket nodes. Optimal settings of these environment variables will vary by platform.
Putting threads far apart, i.e. on different sockets
* May improve the aggregated memory bandwidth available to your application
* May improve the combined cache size available to your application
* May decrease performance of synchronization constructs
Putting threads close together, i.e. on two adjacent cores which possibly shared some caches
* May improve performance of synchronization constructs
* May decrease the available memory bandwidth and cache size
See the OpenMP documentation for documentation on setting the two variables `OMP_PROC_BIND` and `OMP_PLACES`.
These can take on the following values:
One of `OMP_PROC_BIND={false|true|master|close|spread}`
And one of `OMP_PLACES={sockets|cores|threads}` or an explicit list of places that are described by non-negative
numbers.
For example,
```
export OMP_PLACES=threads
export OMP_PROC_BIND=close
```
Developing and Improving affect
-------------------------------
### Using the makefile
For building, testing, and documentation, the easiest way to work is using the `makefile`
in the root directory of the project.
```
make clean; make build
make test
make docs
[...etc]
```
### Build warnings
Some warnings are hidden during the `make build` process, notably the warning about the
deprecated numpy API that always occur with `cython` builds. These warnings are filtered using
the `remove_warning.py` script.
The following type of warning occurs when building on Linux with gcc:
```
cc1plus: warning: command line option "-Wstrict-prototypes" is valid for Ada/C/ObjC but not for C++
```
This is normal, and a side effect of using setuptools and distutils with the
`Extension(language='c++')` parameter. These warnings may also be hidden during `make build`.
### Setting up your environment
There are a few other dependencies required only for development purposes for
building the code and documentation. `affect` is developed with Cython and the
documentation is built using sphinx. For building a conda recipe using
`conda build` you may also need jinja2.
* cython >= 0.24.1
* jinja2
* pytest >= 3.0.5
* pytest-cov >= 2.3.1
* pytest-runner >= 2.11.1
* sphinx >= 1.5.1
* sphinx_rtd_theme >= 0.1.9
Other command line tools required or that you may want to use on Mac or Linux
are:
* ccache
* gcc6
* make
Using the Conda distribution is the easiest way to develop and manage the
Python package dependencies. Each of these may be installed similar to:
```
conda install pytest-cov
```
If you find any of the dependencies are not in the default conda channel,
first try adding the conda-forge channel (you may also wish to disable SSL
verification behind firewalls or web proxies):
```
conda config --set ssl_verify false
conda config --add channels conda-forge
```
If the package is still not found, fall back to using pip:
```
pip install sphinx-autodoc-typehints
```
### Conda Recipe
To build the conda recipe, make sure you install the prerequisites:
`conda install conda-build`
The file `meta.yaml` in the root directory controls the building of the conda package.
Build the recipe using
`conda build `
Immediate Plans
---------------
* Write meshes in the _ExodusII_ format (unstructured meshes)
* Read ExodusII meshes split across files, thus avoiding the need for
executing the _Seacas_ `epu` tool.
* Read and write _CGNS_ mesh format (structured meshes)
* Provide a local least squares recovery for mesh fields. This is a smoothing
algorithm that can be used to, for example, either promote a element field
(piecewise constant) to a piecewise linear field, or provide a local error
indicator for spatial interpolation error on each element.
Future Plans
------------
* Fast visualization of mesh/block boundary.
* Divide mesh boundary (block skin) into separate non-intersecting
surface regions at sharp edges. (This can be thought of as automatic
creation of sidesets ExodusII terms).
* Fast search algorithm for nearest neighbor searches for element sides
and nodes on the boundary surface.