I was trying to build the VASP GPU version with VTSTCODE v190 and was using the Nvidia HPC compiler. I am getting the following messages/errors in the pyamff_fortran section and was wondering if this version is supported with the Nvidia HPC compiler.
make[2]: Entering directory '/share/apps/vasp/6.3.2-build-gpu/src-vtst-beef/build/std/pyamff_fortran'
mpif90 -acc -gpu=cc70,cuda11.0 -c fptype.f90 atoms.f90 neighborlist.f90 behlerParrinello.f90 nntype.f90 trainType.f90 normalization.f90 fingerprints.f90 neuralnetwork.f90 nnmodule.f90 pyamffType.f90 lossgrad.f90 adam.f90 lbfgs_ml.f90 rprop.f90 opts.f90 training.f90 model.f90
fptype.f90:
atoms.f90:
neighborlist.f90:
NVFORTRAN-W-0093-Type conversion of expression performed (neighborlist.f90: 188)
0 inform, 1 warnings, 0 severes, 0 fatal for calcnlist
behlerParrinello.f90:
nntype.f90:
trainType.f90:
normalization.f90:
fingerprints.f90:
NVFORTRAN-S-0027-Illegal integer constant: 4294967296 (fingerprints.f90: 1817)
NVFORTRAN-S-0027-Illegal integer constant: 4294967291 (fingerprints.f90: 1819)
NVFORTRAN-S-0034-Syntax error at or near identifier result (fingerprints.f90: 1823)
NVFORTRAN-S-0310-Illegal statement in the specification part of a MODULE (fingerprints.f90: 1828)
NVFORTRAN-S-0310-Illegal statement in the specification part of a MODULE (fingerprints.f90: 1830)
NVFORTRAN-S-0302-Unmatched ENDFUNCTION statement (fingerprints.f90: 1832)
NVFORTRAN-S-0302-Unmatched ENDSUBROUTINE statement (fingerprints.f90: 1834)
neuralnetwork.f90:
nnmodule.f90:
pyamffType.f90:
lossgrad.f90:
adam.f90:
lbfgs_ml.f90:
NVFORTRAN-W-0093-Type conversion of expression performed (lbfgs_ml.f90: 175)
0 inform, 1 warnings, 0 severes, 0 fatal for lbfgs_step
rprop.f90:
opts.f90:
training.f90:
model.f90:
NVFORTRAN-S-0188-Argument number 4 to train_init: type mismatch (model.f90: 127)
NVFORTRAN-S-0188-Argument number 4 to train_init: type mismatch (model.f90: 129)
0 inform, 0 warnings, 2 severes, 0 fatal for pyamff_init
NVFORTRAN-S-0188-Argument number 6 to trainexec: type mismatch (model.f90: 269)
0 inform, 0 warnings, 1 severes, 0 fatal for pyamff_step
NVFORTRAN-S-0188-Argument number 3 to normalizefps2: type mismatch (model.f90: 508)
0 inform, 0 warnings, 1 severes, 0 fatal for pyamff_calc
make[2]: *** [makefile:8: all] Error 2
make[2]: Leaving directory '/share/apps/vasp/6.3.2-build-gpu/src-vtst-beef/build/std/pyamff_fortran'
make[1]: *** [makefile:145: pyamff_fortran] Error 2
make[1]: Leaving directory '/share/apps/vasp/6.3.2-build-gpu/src-vtst-beef/build/std'
make: *** [makefile:17: std] Error 2
Jeff
VTSTCODE v190 Nvidia comipiler issue
Moderator: moderators
Re: VTSTCODE v190 Nvidia comipiler issue
Hi Jeff,
Thank you for the report. VTSTCODE v190 is pretty new so we haven't tested the version with the Nvidia HPC compiler yet.
I would recommend to use v184 in the meantime. I will attach v184 just in case you need. For installation, you can check this forum's reply: viewtopic.php?p=25129#p25129.
Let me know if you still have an issue.
Thanks,
Jiyoung
Thank you for the report. VTSTCODE v190 is pretty new so we haven't tested the version with the Nvidia HPC compiler yet.
I would recommend to use v184 in the meantime. I will attach v184 just in case you need. For installation, you can check this forum's reply: viewtopic.php?p=25129#p25129.
Let me know if you still have an issue.
Thanks,
Jiyoung
- Attachments
-
- vtstcode-184.tgz
- (155.46 KiB) Downloaded 15442 times
Re: VTSTCODE v190 Nvidia comipiler issue
Hi Jeff,
You can try v191 if you want. The type mismatch error issue should be fixed.
Thanks,
Jiyoung
You can try v191 if you want. The type mismatch error issue should be fixed.
Thanks,
Jiyoung