First, I'm sorry for my poor english.
I compiled rosetta applications via mpi options
(/home/ael/Desktop/rosetta_bin_linux_2020.28.61328_bundle/main/source/bin/relax.mpi.linuxgccrelease)
and I tried to use relax application to prepare structure
Then, when i start relax application, my cpus (16cores, 32threads) rus fully and it calculates same calculations 32times
(one output pdbs and 32 scores in .sc file)
how can i solve this problems?
*I read this post but I can't understand what they say :(
https://www.rosettacommons.org/node/10669
---my script----
mkdir Starting_PDB
mpiexec -use-hwthread-cpus relax.mpi.linuxgccrelease -s ./D1Frag_PDB/3qwq_D1.pdb \
-out:suffix _prep -nstruct 1 \
-relax:constrain_relax_to_start_coords -relax:coord_constrain_sidechains -relax:ramp_constraints false \
-out:path:pdb ./Starting_PDB -out:path:score ./Starting_PDB
----------------------------------------------------------
----score file
SEQUENCE:
SCORE: total_score coordinate_constraint dslf_fa13 fa_atr fa_dun fa_elec fa_intra_rep fa_intra_sol_xover4 fa_rep fa_sol hbond_bb_sc hbond_lr_bb hbond_sc hbond_sr_bb lk_ball_wtd omega p_aa_pp pro_close rama_prepro ref yhh_planarity description
SCORE: -405.633 123.102 -2.780 -1458.619 456.389 -398.157 2.662 51.867 159.163 797.808 -41.637 -103.976 -20.028 -40.516 -26.510 19.702 -60.587 1.293 57.131 77.682 0.378 3qwq_D1_prep_0001
SCORE: -405.580 123.080 -2.782 -1458.113 456.556 -397.907 2.662 51.870 158.796 797.500 -41.636 -103.955 -20.011 -40.542 -26.568 19.730 -60.597 1.292 57.075 77.682 0.289 3qwq_D1_prep_0001
SCORE: -405.637 123.121 -2.787 -1458.404 456.443 -398.064 2.662 51.868 159.112 797.660 -41.628 -103.956 -20.014 -40.514 -26.579 19.707 -60.621 1.294 57.027 77.682 0.354 3qwq_D1_prep_0001
SCORE: -405.496 122.701 -2.525 -1458.187 457.178 -398.226 2.639 51.830 158.226 797.851 -41.625 -103.890 -20.018 -40.705 -26.417 19.984 -60.709 1.277 57.207 77.682 0.232 3qwq_D1_prep_0001
.....
-------------------------------------------------------
---mpi compilation script
sudo apt install openmpi-bin (maybe?)
sudo apt install mpich
sudo apt install zlib1g-dev
#Compiling Rosetta
cd rosetta*/main/source
./scons.py -j30 bin mode=release extras=mpi
--------------------------------------------------------
Increase nstruct to the number of MPI cores you are using.
I increased nstruct number but, I have same symptom
Hmm. So, what is this option? -use-hwthread-cpus
It allows use hyperthreading
If I don't use that option, my cpu use only 16 threads (same symptom but do same job 16times)
I use ryzen 9 3950x cpu (16cores, 32threads)
so i think do same job 32times
some lines when i run my script
Looks like you are only running on on the head node. Try giving mpiexec the number of processors to use with the -np option. I think that may work.