NWChem 7.0.2 XC B3LYP-Grimme2 PSPW segfault crash

[Alchem334]

Description

Plane-wave DFT calculations with exchange-corellation potential "B3LYP-Grimme2" crash with segfault on NWChem 7.0.2.

Settings

OS xUbuntu 20.04

NWChem compiled from source https://github.com/nwchemgit/nwchem/releases/download/v7.0.2-release/nwchem-7.0.2-release.revision-b9985dfa-srconly.2020-10-12.tar.bz2

Compiled with with commands

export NWCHEM_TOP=/home/alchem/Sources/nwchem-7.0.2/
export NWCHEM_TARGET=LINUX64
export USE_MPI=y
export NWCHEM_MODULES="all python"
export MRCC_METHODS=FALSE
export CCSDTQ=FALSE
export USE_PYTHONCONFIG=Y
export BLASOPT="-lopenblas64 -lpthread -lrt"
export LAPACK_LIB="-lopenblas64 -lpthread -lrt"
export BLAS_SIZE=8

make nwchem_config

make

Log file

log.txt

To reproduce

Input file

nwchem.txt

Steps to reproduce:

1. Execute nwchem with the input file attached

Additional context

Running through gdb generates the following backtrace

Thread 1 "nwchem" received signal SIGSEGV, Segmentation fault.
0x000055555604c8d2 in becke_smalln_correction_ ()
(gdb) bt
#0  0x000055555604c8d2 in becke_smalln_correction_ ()
#1  0x00005555560508fb in gen_b3lyp_bw_restricted_ ()
#2  0x0000555556044fa8 in v_bwexc_ ()
#3  0x000055555603e95f in v_bwexc_all_ ()
#4  0x0000555556034105 in electron_gen_scf_potentials_ ()
#5  0x0000555556035651 in electron_run_ ()
#6  0x0000555556017bcf in psi_1get_tgradient_ ()
#7  0x0000555556115695 in cgminimize_ ()
#8  0x000055555610c8c5 in cgsd_energy_ ()
#9  0x0000555556103d70 in cgsdv5_ ()
#10 0x0000555555577194 in task_gradient_doit (rtdb=0, theory=..., energy=6.0134700169990685e-154, gradient=..., _theory=_theory@entry=32)
    at task_gradient.F:511
#11 0x0000555555577cfb in task_gradient (rtdb=0) at task_gradient.F:120
#12 0x00005555555683cb in task (rtdb=0) at task.F:373
#13 0x0000555555560f19 in nwchem () at nwchem.F:320
#14 0x00005555555614ea in main (argc=argc@entry=2, argv=0x7fffffffd57e) at nwchem.F:404
#15 0x00007ffff51060b3 in __libc_start_main (main=0x5555555614cb <main>, argc=2, argv=0x7fffffffd038, init=<optimized out>, fini=<optimized out>, 
    rtld_fini=<optimized out>, stack_end=0x7fffffffd028) at ../csu/libc-start.c:308
#16 0x000055555555f52e in _start ()

Error is in src/nwpw/pspw/lib/exchange-correlation/b3lyp.F:647. This is where the Becke_smalln_correction subroutine is called.

              call Becke_smalln_correction(n,n_thrd,agr,beta,lda_c,
     >                                   chi,chi2,chiSQ,K,F1,F2,
     >                                   xe,fdnx,fdagrx)

One argument is missing in the call to Becke_smalln_correction. The function is defined in the file src/nwpw/pspw/lib/exchange-correlation/blyp.f:13 and it takes 15 arguments, but in its calls in the b3lyp.F file only 14 arguments are passed to it.

      subroutine Becke_smalln_correction(n,n_thrd,agr,fac,beta,lda_c,
     >                                   chi,chi2,chiSQ,K,F1,F2,
     >                                   xe,fdnx,fdagrx)

I honestly don't understand how gfortran compiles this.

[ebylaska]

This might be fixed in the development tree.  Have you tried this?

…

On Nov 27, 2020, 11:15 PM -0800, Alchem334 ***@***.***>, wrote: Description Plane-wave DFT calculations with exchange-corellation potential "B3LYP-Grimme2" crash with segfault on NWChem 7.0.2. Settings OS xUbuntu 20.04 NWChem compiled from source https://github.com/nwchemgit/nwchem/releases/download/v7.0.2-release/nwchem-7.0.2-release.revision-b9985dfa-srconly.2020-10-12.tar.bz2 Compiled with with commands export NWCHEM_TOP=/home/alchem/Sources/nwchem-7.0.2/ export NWCHEM_TARGET=LINUX64 export USE_MPI=y export NWCHEM_MODULES="all python" export MRCC_METHODS=FALSE export CCSDTQ=FALSE export USE_PYTHONCONFIG=Y export BLASOPT="-lopenblas64 -lpthread -lrt" export LAPACK_LIB="-lopenblas64 -lpthread -lrt" export BLAS_SIZE=8 make nwchem_config make Log file log.txt To reproduce Input file nwchem.txt Steps to reproduce: 1. Execute nwchem with the input file attached Additional context Running through gdb generates the following backtrace Thread 1 "nwchem" received signal SIGSEGV, Segmentation fault. 0x000055555604c8d2 in becke_smalln_correction_ () (gdb) bt #0 0x000055555604c8d2 in becke_smalln_correction_ () #1 0x00005555560508fb in gen_b3lyp_bw_restricted_ () #2 0x0000555556044fa8 in v_bwexc_ () #3 0x000055555603e95f in v_bwexc_all_ () #4 0x0000555556034105 in electron_gen_scf_potentials_ () #5 0x0000555556035651 in electron_run_ () #6 0x0000555556017bcf in psi_1get_tgradient_ () #7 0x0000555556115695 in cgminimize_ () #8 0x000055555610c8c5 in cgsd_energy_ () #9 0x0000555556103d70 in cgsdv5_ () #10 0x0000555555577194 in task_gradient_doit (rtdb=0, theory=..., energy=6.0134700169990685e-154, gradient=..., ***@***.***=32) at task_gradient.F:511 #11 0x0000555555577cfb in task_gradient (rtdb=0) at task_gradient.F:120 #12 0x00005555555683cb in task (rtdb=0) at task.F:373 #13 0x0000555555560f19 in nwchem () at nwchem.F:320 #14 0x00005555555614ea in main ***@***.***=2, argv=0x7fffffffd57e) at nwchem.F:404 #15 0x00007ffff51060b3 in __libc_start_main (main=0x5555555614cb <main>, argc=2, argv=0x7fffffffd038, init=<optimized out>, fini=<optimized out>, rtld_fini=<optimized out>, stack_end=0x7fffffffd028) at ../csu/libc-start.c:308 #16 0x000055555555f52e in _start () Error is in src/nwpw/pspw/lib/exchange-correlation/b3lyp.F:647. This is where the Becke_smalln_correction subroutine is called. call Becke_smalln_correction(n,n_thrd,agr,beta,lda_c, > chi,chi2,chiSQ,K,F1,F2, > xe,fdnx,fdagrx) One argument is missing in the call to Becke_smalln_correction. The function is defined in the file src/nwpw/pspw/lib/exchange-correlation/blyp.f:13 and it takes 15 arguments, but in its calls in the b3lyp.F file only 14 arguments are passed to it. subroutine Becke_smalln_correction(n,n_thrd,agr,fac,beta,lda_c, > chi,chi2,chiSQ,K,F1,F2, > xe,fdnx,fdagrx) I honestly don't understand how gfortran compiles this. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or unsubscribe.

[Alchem334]

Tried commit d9c7972 now. Attached log
log2.txt

The calculation passed without error.

[edoapra]

Confirmed that commit 0ec0cfc fixes this issue