Code for:

A parallel & automatically tuned algorithm for multispectral image deconvolution.
R. Ammanouil, A. Ferrari, D. Mary, C. Ferrari and F. Loi.
Submitted to MNRAS, 2019.

Verification demos for centralized and distributed versions of MUFFIN

• example_class.py: executes centralized version of muffin takes as input the path to the data and the prefix of the data. The data is supposed to be named data/prefix_dirty prefix_sky prefix_psf.

python3 example_class.py -fol data -nam m31_3d_conv_10db

• example_class_mpi_sure.p: executes & compares the centralized & distributed version of muffin + SURE. Similarly to example_class.py it takes as input the path to the data folder and the prefix of the data.

mpirun --np 6 python3 example_class_mpi_sure.py -fol data -nam m31_3d_conv_10db

Running tests of MUFFIN on a single node

• run_tst_mpi.py: executes & saves results of the distributed muffin. You have to set the algorithm parameters from the terminal.

mpirun --np 3  python3 run_tst_mpi.py -L 2 -N 4 -mu_s 0.2 -mu_l 7 -mu_w 10 -stp_s 0.3 -stp_l 10000 -data m31_3d_conv_10db -fol data -pxl_w 1 -bnd_w 1

• plot_figures_Run_tst_mpi.py: Loads results saved in run_test_mpi.py and plots Figures (SNR, PSNR, MSE, cost, SUGAR, ... )

python3 plot_figures_Run_tst_mpi.py -res_fol output/1881888

Testing MUFFIN on a cluster using SLURM

• run_sigamm.py: adapted version of run_tst_mpi.py to run on a cluster using SLURM
• run_sigamm.slurm: sets the number of nodes, and all algorithm parameters, and wall time for running run_sigamm.py on a cluster using SLURM

arguments=" -s 1 -L 256 -N 5000 -mu_s 0.005 -mu_l 3 -mu_w 10 -stp_s 0 -stp_l 0 -pxl_w 1 -bnd_w 1 -data M31_skyline220db -fol data/data_david -sav 1 -init 0 -fol_init output_sigamm/7844989"; 
 
export arguments
sbatch run_sigamm.slurm
squeue -u rammanouil

• example with dependency

arguments=" -s 1 -L 256 -N 5000 -mu_s 0.005 -mu_l 3 -mu_w 10 -stp_s 0 -stp_l 0 -pxl_w 1 -bnd_w 1 -data M31_skyline220db -fol data/data_david -sav 1 -init 1 -fol_init output_sigamm/7844989"; 

export arguments
sbatch --dependency=afterok:7645246 run_sigamm.slurm

MUFFIN algorithm & related functions

• deconv3d.py : code for the muffin iterative algorithm written using a class framework
• deconv3d_mpi.py : distributed version of deconv3d.py using MPI
• deconv3d_tools.py : module with fonctions called by deconv3d.py and deconv3d_mpi.py