In Situ Python Analysis: Plummer

This test problem Plummer demonstrates the in situ Python analysis feature in GAMER.

---

1. Install the required packages for this demo.

• libyt and yt_libyt: the in situ analysis library and its yt frontend.
  • This example assumes it is using libyt interactive mode. Please compile libyt in interactive mode.
• yt: the core analysis tool.
• FFTW: needed in this test problem.
• HDF5: this is optional since we are not going to output any data.

2. Edit the Makefile to validate the following settings and set the paths FFTW_PATH, MPI_PATH, and LIBYT_PATH. You can also use the Python script configure.py to tailor the Makefile (an example can be found at example/test_problem/Hydro/Plummer/generate_make.sh).

SIMU_OPTION += -DGRAVITY

SIMU_OPTION += -DPARTICLE

# number of user-defined passively advected scalars
# --> set it to 0 or comment it out if none is required
# --> useless for RTVD
SIMU_OPTION += -DNCOMP_PASSIVE_USER=2

#SIMU_OPTION += -DSERIAL

SIMU_OPTION += -DLOAD_BALANCE=HILBERT

# support FFTW library
# --> SUPPORT_FFTW must be defined when GRAVITY is enabled
# --> use FFTW3 for fftw3 support
#     use FFTW2 for fftw2 support
SIMU_OPTION += -DSUPPORT_FFTW=FFTW2

# support yt inline analysis
SIMU_OPTION += -DSUPPORT_LIBYT

# support libyt interactive mode
# --> this activates python prompt and does not shut down a simulation when there are
#     errors in an inline python script
# --> must compile libyt with INTERACTIVE_MODE
# --> must enable SUPPORT_LIBYT
SIMU_OPTION += -DLIBYT_INTERACTIVE

# switch to MPI compiler
#CXX        = g++                             # serial compiler
CXX         = $(MPI_PATH)/bin/mpicxx          # MPI compiler

# set library paths
CUDA_PATH    :=
FFTW_PATH    := ${YOUR_FFTW_PATH}
MPI_PATH     := ${YOUR_MPI_PATH}
HDF5_PATH    :=
GRACKLE_PATH :=
GSL_PATH     :=
LIBYT_PATH   := ${YOUR_LIBYT_PATH}

3. Compile the code in the src folder.

> make clean
> make -j 4
   ...
   ...
Compiling GAMER --> Successful!

4. Create a working directory in the bin folder and copy the GAMER executable and Plummer test problem files.

mkdir Plummer                                       # create a folder for this test
cd Plummer
cp -r ../../example/test_problem/Hydro/Plummer/* .  # copy test problem settings
cp ../../src/gamer .                                # copy GAMER executable

5. In Input__Parameter, set OPT__OUTPUT_TOTAL to 0 since we don't need to dump data to disk in this demo. (If you want to make GAMER dump data snapshots, enable SUPPORT_HDF5 and set HDF5_PATH in Makefile. Keep OPT__OUTPUT_TOTAL to 1.)

# data dump
OPT__OUTPUT_TOTAL       0     # output the simulation snapshot: (0=off, 1=HDF5, 2=C-binary) [1]

6. In Input__TestProb, set Plummer_Collision to 1 for simulating colliding Plummer clouds.

Plummer_Collision       1     # (0/1)--> single Plummer cloud/two colliding Plummer clouds
Plummer_AddColor        1     # assign different colors to different clouds (must turn on Plummer_Collision
                              # and set NCOMP_PASSIVE_USER to 2 in the Makefile) [0]

7. Create a LIBYT_STOP file. This demo assumes libyt is in interactive mode. It will only enter interactive Python prompt if errors occur while running Python codes or it detects the file LIBYT_STOP.

touch LIBYT_STOP

8. Run GAMER.

OMPI_MCA_osc=sm,pt2pt mpirun -np 4 ./gamer

9. Results from the inline script.

Inline script inline_script.py:

import yt_libyt
import yt

yt.enable_parallelism()

def yt_inline():
    # Get data
    ds = yt_libyt.libytDataset()

    # Do ProjectionPlot to field Cloud0.
    sz = yt.ProjectionPlot(ds, 'z', ('gamer', 'Cloud0'), center='c')

    # Do ParticlePlot
    par = yt.ParticlePlot(ds, 'particle_position_x', 'particle_position_y', 'particle_mass', center='c')

    if yt.is_root():
        sz.save()
        par.save()

def yt_inline_inputArg( fields ):
    pass

Projection of field Cloud01 along the z-axis:

Particle plot using particle position x and y as axes with particle mass mapped to a colorbar:

10. Since we have run libyt in interactive mode, the program should now pause and wait for user input.

=====================================================================
  Inline Function                              Status         Run
---------------------------------------------------------------------
  * yt_inline                                  success         V
  * yt_inline_inputArg                         success         V
=====================================================================
>>>

11. You can explore interactive prompt or simply press Crtl + C to shut down the process.