Getting Started#
Hardware Requirements#
Warning
CyRSoXS requires an NVIDIA GPU.
Installation on Linux#
Note
This installation guide uses the Anaconda Python distribution. Any Python
installation version >=3.6 should work, you will just need to point CMake to your
specific installation. git, conda, and pip are also used, and it is assumed you have
working installations of all three.
CyRSoXS v1.1.5.0 is now Conda installable and no longer requires building from source. If you need to compile from source (to enable double-precision, for example) the instructions are provided below.
Conda Installation#
Clone NRSS from the github repository:
git clone https://github.com/usnistgov/NRSS.git
Use the environment.yml file to create a new virtual environment, and activate it:
conda env create -f environment.yml
conda activate nrss
CyRSoXS is listed as a dependency in environment.yml, and will automatically be installed.
Now we can pip install NRSS, which will also install PyHyperScattering as a dependency:
pip install .
The conda-forge distribution of CyRSoXS includes the executable and Python Shared Library File. You can use the CyRSoXS executable from the shell:
CyRSoXS <Morphology HDF5>
or import CyRSoXS to a python script or jupyter notebook:
import CyRSoXS
After importing, you should see the following output:
CyRSoXS
============================================================================
Size of Real : 4
Maximum Number Of Material : 32
__________________________________________________________________________________________________
| Thanks for using Cy-RSoXS |
|--------------------------------------------------------------------------------------------------|
| Copyright : Iowa State University |
| License : NIST |
| Acknowledgement : ONR MURI |
| |
| Developed at Iowa State University in collaboration with NIST |
| |
| Please cite the following publication : |
| Comments/Questions : |
| 1. Dr. Baskar GanapathySubramanian ([email protected]) |
| 2. Dr. Adarsh Krishnamurthy ([email protected]) |
| 3. Dr. Dean DeLongchamp ([email protected]) |
--------------------------------------------------------------------------------------------------
Version : <version_number>
Git patch : <git_patch_number>
Building CyRSoXS from source#
NOTE: These installation instructions use Conda to install the required dependencies. If you prefer to manually install and manage these dependencies, please see the installation instructions at usnistgov/cyrsoxs
Dependencies
Required Dependencies
A C++ compiler with C++14 support is required.
gcc >= 7 (CUDA specific versions might have GCC requirements )
Cuda Toolkit (>=9)
HDF5
OpenMP
libconfig
Python >= 3.6 (only for building with Pybind)
Optional Dependencies
Doxygen
Docker
Clone CyRSoXS from the github repository:
git clone https://github.com/usnistgov/cyrsoxs.git
Use the environment-build.yml file to create a new virtual environment, and activate it:
conda env create -f environment-build.yml
conda activate cyrsoxs-build
Building CyRSoXS without Pybind
cd $CyRSoXS_DIR
mkdir build;
cd build;
cmake .. -DCMAKE_BUILD_TYPE=Release
make
Create a bin directory and move the CyRSoXS executable inside:
mkdir bin
mv CyRSoXS bin/
Add CyRSoXS to your PATH:
cd bin
echo "export PATH=$PATH:`pwd`" >> ~/.bashrc
source ~/.bashrc
At this point you should have a working CyRSoXS installation. If you also want to import CyRSoXS as a Python library, you need to compile with Pybind.
Building CyRSoXS with Pybind
cd $CyRSoXS_DIR
mkdir build_pybind;
cd build_pybind;
cmake .. -DCMAKE_BUILD_TYPE=Release -DPYBIND=Yes -DUSE_SUBMODULE_PYBIND=No
Depending on where your python installation is, you may need to point CMake to it by including the following compile flags:
-DPYTHON_EXECUTABLE=[path_to_anaconda]/anaconda/bin/python
-DPYTHON_LIBRARY=[path_to_anaconda]/anaconda/lib/libpython3.9.so
-DPYTHON_INCLUDE=[path_to_anaconda]/anaconda/include/python3.9/
If this still doesn’t work, you can edit the CMakeLists.txt file on line 82 to include the three set commands:
if (PYBIND)
set(Python_EXECUTABLE [path_to_anaconda]/anaconda/bin/python)
set(Python_INCLUDE_DIR [path_to_anaconda]/anaconda/include/python3.9)
set(Python_LIBRARIES [path_to_anaconda]/anaconda/lib/libpython3.9.so)
find_package(Python COMPONENTS Interpreter Development REQUIRED)
Once the CMake files have been generated run the following command:
make
This will generate a shared library CyRSoXS.so file. Create a lib directory and move CyRSoXS.so inside:
mkdir lib
mv CyRSoXS.so lib/
Add to your PYTHONPATH:
cd lib
echo "export PYTHONPATH=$PYTHONPATH:`pwd`" >> ~/.bashrc
source ~/.bashrc
Now you can import CyRSoXS in a python script or jupyter notebook:
import CyRSoXS
Again, you should see the following output:
CyRSoXS
============================================================================
Size of Real : 4
Maximum Number Of Material : 32
__________________________________________________________________________________________________
| Thanks for using Cy-RSoXS |
|--------------------------------------------------------------------------------------------------|
| Copyright : Iowa State University |
| License : NIST |
| Acknowledgement : ONR MURI |
| |
| Developed at Iowa State University in collaboration with NIST |
| |
| Please cite the following publication : |
| Comments/Questions : |
| 1. Dr. Baskar GanapathySubramanian ([email protected]) |
| 2. Dr. Adarsh Krishnamurthy ([email protected]) |
| 3. Dr. Dean DeLongchamp ([email protected]) |
--------------------------------------------------------------------------------------------------
Version : <version_number>
Git patch : <git_patch_number>
Optional CMake Flags
-DPYBIND=Yes # Compiling with Pybind:
-DMAX_NUM_MATERIAL=64 # To change the maximum number of materials (default is 32)
-DDOUBLE_PRECISION=Yes # Double precision mode
-DPROFILING=Yes # Profiling
-DBUILD_DOCS=Yes # To build documentation
-DCMAKE_CXX_COMPILER=icpc -DCMAKE_C_COMPILER=icc # Compiling with the Intel compiler (does not work with Pybind)
NRSS Tutorials#
Several tutorials are available in the tutorials folder. Most of these tutorials include Jupyter notebooks with explanatory prose, example code, and code use strategies.
For all tutorials, it is recommended to copy the entire tutorial folder out of the installed repository location before running, because some tutorials will generate file changes and new serialized objects within repository directories.
kkcalc for NRSS#
This Jupyter notebook series describes how to convert Near Edge X-ray Absorption Fine Structure (NEXAFS) data into a complex index of refraction for use in NRSS computations. Conditioning of the NEXAFS and the use of the kkcalc library to develop this complex index are demonstrated. Both notebooks additionally include calculation code to predict binary contrast from pairs of complex indices in real materials.
The tutorial series has two notebooks: 1. Scalar complex index of refraction. determination from NEXAFS of poly(styrene) (PS) and poly(methyl methacrylate) (PMMA)
2. Tensor (uniaxial) complex index of refraction. determination from NEXAFS of PM6 and Y6, common materials used in organic photovoltaic devices. It is recommended to master the scalar notebook before attempting the tensor notebook.
Core-Shell Radial Disk in the NRSS#
This Jupyter notebook is a basic introduction to the NRSS. It has two different versions. The pybind version is recommended, as it demonstrates the preferred NRSS pybind workflow.
Both tutorials describe a radial disk scattering object, a simple 2D structure with a radial orientation of the extraordinary component of a tensor uniaxial index of refraction.
This tutorial describes how to create the radial disk morphology and serialize the model to hard drive, submit a CyRSoXS job via slurm, and examine the simulation result.
This tutorial describes how to create the radial disk morphology, submit the simulation job within Python, and examine the simulation result. This tutorial relies on index of refraction information stored in the commandline / slurm queue version subdirectory, so it is recommended to copy both at the same time into the same parent directory.
Polymer Adsorption on Sphere in the NRSS#
An in-depth Jupyter notebook series describing the adsorption of a polymer on single sphere. Six notebooks guide new users through basic NRSS model construction, creating libraries for models, automating model parameter sweeps, and ultimately rendering animated visualizations of parameter sweep results. These tutorials feature poly(styrene) (PS) and poly(methyl methacrylate) (PMMA) adsorbed to an SiO2 particle. All tutorials within this series use the recommended NRSS pybind workflow.
From a scientific perspective, the single-sphere nature of this tutorial series makes it more of a fundamental study into tensor index of refraction effects and their manifestation in pure core-shell form factor scattering. More realistic simulations might include multiple spheres, potential structure factor effects between them, dispersity in sphere and shell dimensions, etc.
Tutorial notebooks in this series include: 1. Basic model creation: describes how to create composition fields and Euler angle fields for radial and tangential chain orientation, run simulation within Python, and examine the simulation result.
2. Models with functions: Move model construction from Jupyter cells into functions with arguments for model creation and functions for running models.
3. Model variable sweeps: Sweep model variables using Python loop structures, and save the results using pickle.
4. Model variable sweeps - streamlined: Advanced sweep infrastructure in which most of the code to run variable sweeps are functions imported from a library.
5. Visualizing simulation results: Examining and plotting NRSS simulation results using the matplotlib library.
6. Visualizing sweeps as animations: Collate result visualization collections of variable sweep campaigns into mp4 video using ffmpeg.