This repo includes XMDS2 source code and simulation instances for numerical exploration of optical Gradient Echo Memory (GEM) in a room-temperature gas cloud. These make use of a custom helper tool for managing multiple XMDS2 simulations and different combinations of parameters.
Different schemes and phenomena are visualised in the included Jupyter notebook (python).
XMDS2 source files are based on the original 1DGEM2L.xmds file. You can see the group's repository here.
See Convenient Flags for how to autocompile and/or run all simulations at once.
For a single instance:
root-dir/bin: xmds2 ../src/my-sim.xmds. This compiles the XMDS file to thebin/directory.root-dir: ./run.sh path/to/my-instance.toml. You may need to first makerun.shexecutable usingchmod.
An instance .toml file specifies the name of the compiled binary to be run (e.g. my-sim-name), and any arguments to be passed. Output(s) will appear in a sibling folder to my-instance.toml, named my-instance. Existing instance files have been placed in ./instances.
If no arguments are specified as lists, this will contain the output (.h5 and .xsil file pair) of a single simulation, with filename specified in the .xmds source. Otherwise, a separate simulation will be run for every combination of list-specified arguments, and the output files of each will be renamed to indicate this combination. Unspecified arguments will be set to their default, and neither unspecified nor single-valued arguments will be included in the
output file name (which is convenient when a simulation takes many
parameters).
simulate="double-gaussian"
[arguments]
pulsewidth=[0.5, 1.5]
pulsesep=[2,4]
bandwidth=4
tswitch=8
tin=4
This runs four simulations, one for each combination of pulsewidth and pulsesep.
To automatically (re)compile the appropriate .xmds file to the bin directory before simulating, add the -c flag:
root-dir: ./run.sh ./instances/my-instance.toml -c
For autocompiling to work, the name of the source file and executable binary must match, except for the .xmds extension.
To run all .toml files in a directory, specify the directory name and then the -a flag:
root-dir: ./run.sh ./instances/ -a
Using -c with -a will run all instances, (re)compiling all necessary simulations.
The XMDS Homebrew formula (described here) does not work on my machine as of 2022. Instead, I successfully followed the manual install instructions. Below is a brief overview, which may need to be adjusted for different machines and is described more thoroughly in the instructions.
Ensure you have Python 3 and a C++ compiler. The compiler clang can be obtained (amongst other things) by installing the XCode developer tools:
xcode-select --install
Download the XMDS2 source and move the unzipped directory to ~/.xmds-3.1.0. In the next step, Homebrew is convenient for installing required libraries:
brew install hdf5 fftw open-mpi
pip3 install h5py lxml numpy (optional)
brew install gsl (optional)
cd ~/.xmds-3.1.0
sudo ./setup.py develop
You will likely need to add ~/.xmds-3.1.0/bin/ to your PATH and give the contained files execute permission using chmod. On an Intel Mac, I am then able to run xmds2 --help in terminal. On Apple Silicon however, Homebrew uses a different default install directory which (as of 2022) will not be automatically found by XMDS2. The manual install instructions contain steps for explicitly specifying the locations of missing libraries (e.g. FFTW).
The two advantages of storing the XMDS-generated C and binary files in a separate bin/ directory are:
- Allow them to be
.gitignored. - Avoid cluttering the
src/directory.
- add a
-rflag to remove old files from the output directory.