Page 1 of 1

Possible Bug of vtst6.3 compiled with GPU

Posted: Wed May 10, 2023 2:49 pm
by zongtang
Hello,

Recently I tried to run VTST jobs with the version of vtst6.3 compiled with VASP 6.3.2 GPU.
I was able to run VTST and the gpu performance was good.

(1) However, it stopped in the middle:

==================================================
5 F= -.12687222E+04 E0= -.12687027E+04 d E =-.163681E-03 mag= 82.0002
quench: g(F)= 0.000E+00 g(S)= 0.000E+00 dE (1.order)= 0.000E+00
bond charge predicted
N E dE d eps ncg rms rms(c)
DAV: 1 -0.122522856198E+04 -0.21798E-03 -0.31654E-03 3744 0.497E-02 0.165E-02
RMM: 2 -0.122522856832E+04 -0.63420E-05 -0.63933E-05 3636 0.840E-03 0.367E-02
RMM: 3 -0.122522856752E+04 0.79512E-06 -0.97767E-06 2654 0.372E-03 0.182E-02
RMM: 4 -0.122522856921E+04 -0.16823E-05 -0.15687E-06 2069 0.210E-03 0.911E-03
RMM: 5 -0.122522856943E+04 -0.22334E-06 -0.40661E-07 1978 0.116E-03
6 F= -.12687224E+04 E0= -.12687028E+04 d E =-.171151E-03 mag= 82.0001
quench: g(F)= 0.000E+00 g(S)= 0.000E+00 dE (1.order)= 0.000E+00
reached required accuracy - stopping structural energy minimisation
-----------------------------------------------------------------------------
| _ ____ _ _ _____ _ |
| | | | _ \ | | | | / ____| | | |
| | | | |_) | | | | | | | __ | | |
| |_| | _ < | | | | | | |_ | |_| |
| _ | |_) | | |__| | | |__| | _ |
| (_) |____/ \____/ \_____| (_) |
| |
| internal error in: vhdf5.F at line: 110 |
| |
| HDF5 call in vhdf5.F:110 produced error: 209 |
| |
| If you are not a developer, you should not encounter this problem. |
| Please submit a bug report. |
| |
-----------------------------------------------------------------------------
==========================

also there was a warning:
========================================
-----------------------------------------------------------------------------

Warning: ieee_invalid is signaling
Warning: ieee_divide_by_zero is signaling
Warning: ieee_underflow is signaling
Warning: ieee_inexact is signaling
1
Warning: ieee_invalid is signaling
Warning: ieee_divide_by_zero is signaling
Warning: ieee_underflow is signaling
Warning: ieee_inexact is signaling
1
Warning: ieee_invalid is signaling
Warning: ieee_divide_by_zero is signaling
Warning: ieee_underflow is signaling
Warning: ieee_inexact is signaling
1
Warning: ieee_invalid is signaling
Warning: ieee_divide_by_zero is signaling
Warning: ieee_underflow is signaling
Warning: ieee_inexact is signaling
1
Warning: ieee_invalid is signaling
Warning: ieee_divide_by_zero is signaling
Warning: ieee_underflow is signaling
Warning: ieee_inexact is signaling
1
Warning: ieee_invalid is signaling
Warning: ieee_divide_by_zero is signaling
Warning: ieee_underflow is signaling
Warning: ieee_inexact is signaling


(2) Another issue was the calculation generated lots of *.t1 files with one file for one core. please check out the attached.
Not sure if it is from VTST itself or compile. There was no this issue for the version without VTST.

Thanks a lot for your help.

Best,

Zongtang

Re: Possible Bug of vtst6.3 compiled with GPU

Posted: Wed May 10, 2023 3:26 pm
by zongtang
It looks like some one brought the same problem on VASP :

https://vasp.at/forum/viewtopic.php?p=23408