The GLobal airglOW Model 0.981 -- aurora and dayglow, IR-VIS-UV-EUV
This command automatically compiles the Fortran code to access from Python on any platform:
pip install -e .
Run self-test with:
Much more data is readily available, just let me know.
While many users use the Python interface to Glow, users on HPCC may
want to use MPI directly in Fortran using the Makefiles or CMake.
Here's how to compile the
basic example with CMake
cd bin cmake ../src cmake --build . ctest -V cd ..
testdrvPython intfc test
With regard to precision, at first try I occasionally see the least digit of precision in the text output files differ. This can be related to Stan using Intel Fortran compiler and I'm using Gfortran. E.g. I get 4.34e-9 and Stan got 4.33e-9.
I also get the message:
Note: The following floating-point exceptions are signalling: IEEE_UNDERFLOW_FLAG IEEE_DENORMAL
Which is a warning sign that different compilers and platforms may yield different results of unpredictable variance.
Aurora example (night)
Compare your results vs. Stan's results:: with
./basic < in.basic.aur > aur.basic.out meld reference/aur981.basic.out <(./basic < in.basic.aur)
Dayglow Example (day)
./basic < in.basic.day > day.basic.out
meld reference/out.basic.day <(./basic < in.basic.day)
Allows parallel execution of GLOW Fortran code for HPC:
apt install libopenmpi-dev
allows reading/writing data in NetCDF (optional):
apt install libnetcdf-dev libnetcdff-dev
Select Fortran compiler
Simply use the variable
FC=ifort cmake ../src cmake --build .
NOTE: Using the Intel compiler requires that you have built NetCDF using Intel Fortran. This is an issue ANYTIME you use the Intel Compiler and NetCDF.