ERROR from LAMMPS compiler (Intel Compiler 2017)

[Leon_W_1]

Dear There,

I got an error shown below when I compiled LAMMPS with optional packages, reax, poems, and meam. I do not understand the error origin and do not know how to fix it. Any suggestion would be highly appreciated.

OS: CentOS 7

Compiler: Intel® Parallel Studio XE 2017 (cluster version)

Makefile: default file (Makefile.mpi) and optional file (Makefile.intel_cpu_mpi)

LAMMPS version: stable one released by Nov 17, 2016 and beta version released by Feb. 13, 2017

Appreciatively,

Leon

mpiicpc -g -qopenmp -xHost -O2 -fp-model fast=2 -no-prec-div -qoverride-limits -L../../lib/reax -L../../lib/meam -L../../lib/poems  -L/opt/intel/compilers_and_libraries_2017.1.132/linux/mpi/intel64/lib    -L/opt/intel/fce/10.0.023/lib  fix_deform.o rerun.o compute.o atom.o fix_press_berendsen.o pair_reax.o minimize.o fix_reax_bonds.o min.o fix_print.o math_extra.o change_box.o angle.o pair_buck.o fix_drag.o compute_chunk_atom.o compute_rdf.o pair_dpd.o pair_lj_cut_coul_dsf.o fix_gravity.o compute_msd_chunk.o fix_shear_history.o fix_nvt_sllod.o fix_tmd.o compute_property_local.o compute_dipole_chunk.o compute_angle_local.o neigh_list.o random_mars.o fix_lineforce.o min_fire.o imbalance.o atom_vec_body.o compute_com.o pair_soft.o delete_bonds.o fix_wall.o library.o update.o pair_lj_gromacs_coul_gromacs.o image.o fix_spring_self.o fix_nve_noforce.o pair_hybrid.o dihedral.o angle_hybrid.o compute_angle.o dump_movie.o atom_vec_line.o input.o pair_lj96_cut.o atom_vec_charge.o compute_pe_atom.o pair_buck_coul_cut.o compute_reduce.o min_hftn.o compute_pair_local.o compute_temp_sphere.o pair_zero.o compute_heat_flux.o compute_temp_region.o angle_zero.o imbalance_var.o fix_ave_time.o fix_npt_sphere.o fix_spring_rg.o compute_angmom_chunk.o timer.o fix_ave_atom.o neigh_bond.o fix_wall_region.o dihedral_zero.o compute_slice.o fix_halt.o compute_temp_com.o fix_wall_lj126.o fix_move.o fix_nve_sphere.o region_cone.o neigh_respa.o compute_gyration_chunk.o compute_hexorder_atom.o pair_lj_smooth.o atom_vec_hybrid.o irregular.o compute_vcm_chunk.o memory.o fix_setforce.o atom_map.o fix_nh_sphere.o replicate.o fix_nve.o fix_recenter.o fix_wall_lj93.o compute_temp_deform.o fix_langevin.o reader.o pair_lj_cut.o fix_box_relax.o read_restart.o fix_store.o compute_group_group.o compute_msd.o pair_beck.o verlet.o fix_spring.o bond_zero.o min_quickmin.o pair_born_coul_wolf.o fix_wall_lj1043.o pair_coul_wolf.o pair_coul_dsf.o compute_reduce_region.o compute_pe.o thermo.o improper_zero.o compute_contact_atom.o comm_tiled.o fix_nh.o region_sphere.o fix_wall_harmonic.o pair.o region_intersect.o fix_restrain.o compute_omega_chunk.o output.o comm.o variable.o body.o neigh_half_bin.o pair_meam.o fix_planeforce.o fix_ave_correlate.o bond.o compute_cna_atom.o neigh_request.o domain.o fix_controller.o universe.o compute_improper.o write_dump.o pair_table.o fix_store_state.o region.o fix_enforce2d.o respa.o compute_temp_partial.o info.o dump_custom.o fix_temp_csld.o pair_lj_cut_coul_cut.o fix_poems.o neigh_half_multi.o fix_deprecated.o create_bonds.o region_union.o pair_dpd_tstat.o neighbor.o imbalance_group.o compute_erotate_sphere.o fix_vector.o dump_cfg.o molecule.o dihedral_hybrid.o fix_temp_csvr.o write_coeff.o velocity.o fix_nph.o min_cg.o procmap.o compute_cluster_atom.o pair_zbl.o neigh_half_nsq.o pair_born_coul_dsf.o imbalance_neigh.o integrate.o fix_ave_histo_weight.o pair_lj_smooth_linear.o math_special.o comm_brick.o force.o fix_balance.o finish.o read_data.o compute_dihedral_local.o fix_ave_histo.o displace_atoms.o improper.o reader_xyz.o compute_com_chunk.o neigh_shardlow.o min_linesearch.o imbalance_time.o pair_lj_cut_coul_debye.o pair_mie_cut.o imbalance_store.o atom_vec.o fix_wall_reflect.o neigh_gran.o bond_hybrid.o compute_stress_atom.o fix_respa.o modify.o dump_local.o fix_spring_chunk.o fix_viscous.o dump_image.o compute_erotate_sphere_atom.o compute_pair.o rcb.o balance.o compute_temp.o error.o compute_pressure.o set.o fix_property_atom.o delete_atoms.o fix.o neigh_stencil.o pair_lj_cubic.o compute_displace_atom.o compute_dihedral.o dump.o main.o create_box.o atom_vec_sphere.o compute_bond_local.o compute_improper_local.o fix_dt_reset.o compute_inertia_chunk.o fix_addforce.o citeme.o lammps.o random_park.o compute_vacf.o region_plane.o region_cylinder.o compute_centro_atom.o atom_vec_tri.o write_data.o pair_lj_gromacs.o compute_bond.o fix_store_force.o create_atoms.o compute_torque_chunk.o pair_coul_debye.o fix_external.o lattice.o compute_property_chunk.o compute_coord_atom.o fix_group.o fix_npt.o special.o pair_hybrid_overlay.o compute_ke_atom.o fix_read_restart.o compute_property_atom.o pair_lj_expand.o neigh_derive.o fix_minimize.o improper_hybrid.o pair_gauss.o pair_morse.o write_restart.o dump_dcd.o min_sd.o region_prism.o pair_coul_streitz.o fix_aveforce.o fix_nph_sphere.o fix_nvt_sphere.o dump_atom.o fix_temp_rescale.o kspace.o compute_temp_chunk.o atom_vec_ellipsoid.o fix_indent.o compute_ke.o pair_coul_cut.o fix_momentum.o read_dump.o run.o group.o dump_xyz.o pair_yukawa.o compute_temp_ramp.o fix_nvt.o fix_heat.o fix_adapt.o compute_gyration.o fix_nve_limit.o reader_native.o atom_vec_atomic.o fix_temp_berendsen.o compute_temp_profile.o neigh_full.o pair_born.o compute_orientorder_atom.o fix_ave_chunk.o region_block.o -lreax -lmeam -lpoems   -LKLROOT/lib/intel64/ -lmkl_intel_ilp64 -lmkl_sequential -lmkl_core      -lifcore  -lifcore -lsvml -lompstub -limf  -ltbbmalloc -o ../lmp_mpi
ld: cannot find -lompstub
make[1]: *** [../lmp_mpi] Error 1
make[1]: Leaving directory `/home/group/Downloads/lammps-17Nov16/src/Obj_mpi'
make: *** [mpi] Error 2

[TimP]

It looks like a bad idea to attempt linking against two widely different versions of icpc libraries (17.1 and 10.0), or against a library older than the one supplied with your compiler.  If you mean to link against the OpenMP stubs library (all omp parallel support stubbed out so that program runs single thread), you would need -liompstubs5, assuming that libiompstubs5 reference (but not the long ago libompstubs) is present in your 17.1 compiler installation along with libiomp5.

I don't know whether it is normal to link LAMMPS MPI version with OpenMP stubs library (so that each MPI rank runs single thread even though OpenMP calls are compiled in).

The most common reason for setting -fp-model fast=2 is for the fast complex double effect you would obtain by -complex-limited-range (in-line shortcuts for complex absolute value, divide, etc).  A second reason might be to permit emulation of division without support for non-finite operands.  So those are options you use only after determining that your cases can run faster (and correctly) than without them.

[Leon_W_1]

Dear Tim,

Thank you for your suggestion. I used the default settings from LAMMPS. Any suggestion for revise Makefile? 

Best,
Leon