escatter/e-scatter

Fast simulator for physical scattering processes in materials

e-scatter

Fast CUDA-enabled simulator for electron scattering processes in materials.

Building

Dependencies of the core simulator:

• C++14 compiler (tested with GCC 5)
• CUDA 8.0 or newer
• CUB
• CMake 3.8 or newer

Build the program:

cd e-scatter
mkdir build
cd build
cmake ..
make

To also build all of the utilities, append -DWITH_UTILS=ON to the cmake command. Dependencies required to build the utilities:

• Boost
• muParser
• ImageMagick
• FFTW
• SDL2
• OpenGL
• GLEW

Usage

The main program is located at bin/csdem. It requires a geometry file, which can be generated with bin/line-gen, and an exposure file, which can be generated with scripts/pri-gen.py.

Testing

To run the unit tests, the googletest library should be available in the 3rdparty directory; clone it from github and run cmake with -DWITH_TEST=ON.

cd 3rdparty
git clone https://github.com/google/googletest.git
mkdir -p ../build
cd ../build
cmake .. -DWITH_TEST=ON
make
make test

Creating materials

The material files contain most of the physics involved.

• ELSEPA can be downloaded from the Computer Physics communications Program library. It has an attribute-only license for non-commercial use. We use ELSEPA to compute Mott cross-sections.

• Livermore database EPDL97 or ENDF/B-VII.1

e-scatter's Issues

Doesn't compile with latest GSL

Some things have changed:

• the include has moved from <gsl.h> to <gsl/gsl>
• the gsl::as_span function has been renamed

Unfortunately I could not get it to work myself.

Remove CSTool

We should provide integration with eScatter/cstool and remove the cstool in this repository.

• Remove GSL dependency
• Switch back C++ standard to C++11 from C++14

Rework scripts for handling material data (crosssections)

Rework means:

• translate C++ parts into Python
• move to HDF5 as intermediate storage, in stead of XML
• wrap Elsepa into Python interface
• Access ENDF data directly from Python
• Connect elements using Workflow techniques

C++14 code but compiled with -std=c++11 flag

It is impossible to build the project since some C++14 code was introduced but GCC is set to C++11 mode. We cannot switch the entire project to C++14 since CUDA doesn't support that. Is it a good idea to mix C++11 and C++14? (different compiler flags for different subdirectories/libraries)

Look at Fast-BVH for an example of a BVH implementation.

Implement a tool to create inelastic cross-sections from optical data

Make scripts/inelastic-gen.py able to generate inelastic cross-sections from optical data using Ashley.
Also provide (reference to) optical data so all cross-sections can be created from scratch.

Add python interface to cdsem

After #18 has been resolved we should provide a python library to interface with cdsem so there will be an easy-to-use all-python interface available.

Data table serialisation and storage

options: google protobuf, hdf5, netcdf
include python scripts for text dump of binary data.

update preliminary documentation attempt.

investigate github large file storage

Use Eigen for SSE accelerated vector operations