the output discrepancy of VASPsol for CPU and GPU results

questions related to VASP with GPU support (vasp.5.4.1, version released in Feb 2016)

Moderators: Global Moderator, Moderator

Post Reply
Message
Author
User avatar
cenwanglai
Newbie
Newbie
Posts: 16
Joined: Mon Jan 18, 2010 12:35 pm
License Nr.: 1148
Location: Hangzhou City, Zhejiang Province, P.R.China

the output discrepancy of VASPsol for CPU and GPU results

#1 Post by cenwanglai » Thu Mar 29, 2018 11:08 pm

The OSZICAR for GPU VASPsol+VASP.5.4.4
N E dE d eps ncg rms rms(c)
DAV: 1 -0.147820402104E+02 -0.14782E+02 -0.50630E-02 40 0.744E-01 0.165E-01
DAV: 2 -0.147820811285E+02 -0.40918E-04 -0.88552E-03 40 0.481E-01 0.853E-02
DAV: 3 -0.147818774719E+02 0.20366E-03 -0.37629E-04 40 0.812E-02 0.534E-02
DAV: 4 -0.147816775664E+02 0.19991E-03 -0.27196E-04 32 0.708E-02 0.136E-02
DAV: 5 -0.147817050491E+02 -0.27483E-04 -0.20922E-05 48 0.199E-02 0.818E-03
DAV: 6 -0.147817523064E+02 -0.47257E-04 -0.65675E-06 48 0.105E-02 0.283E-03
DAV: 7 -0.147817896297E+02 -0.37323E-04 -0.19369E-06 48 0.541E-03 0.251E-03
DAV: 9 -0.147818696444E+02 -0.80015E-04 -0.58268E-06 32 0.761E-03 0.138E-03
DAV: 9 -0.147818824096E+02 -0.12765E-04 -0.33145E-07 48 0.209E-03 0.116E-03
DAV: 10 -0.147819030599E+02 -0.20650E-04 -0.84742E-07 40 0.308E-03 0.416E-04
DAV: 11 -0.147819058458E+02 -0.27859E-05 -0.55338E-08 40 0.887E-04 0.260E-04
DAV: 12 -0.147819082609E+02 -0.24151E-05 -0.68597E-08 40 0.823E-04 0.110E-04
DAV: 13 -0.147819087980E+02 -0.53709E-06 -0.92515E-09 48 0.283E-04
SOL: 1 -0.40112E-01 0.36096E-01 -0.40153E-02 105
1 F= -.14781909E+02 E0= -.14781909E+02 d E =-.379556E-53

The SOL step only presents at the end.

The OSZICAR for CPU VASPsol+VASP.5.4.4:

N E dE d eps ncg rms rms(c)
SOL: 1 -0.40285E-01 0.36096E-01 -0.41887E-02 106
DAV: 1 -0.147876960632E+02 -0.14788E+02 -0.44498E-02 32 0.886E-01 0.122E-01
SOL: 2 -0.42594E-01 0.36037E-01 -0.65569E-02 90
DAV: 2 -0.147871362365E+02 0.55983E-03 -0.21731E-03 24 0.244E-01 0.586E-02
SOL: 3 -0.44404E-01 0.35995E-01 -0.84084E-02 90
DAV: 3 -0.147869546788E+02 0.18156E-03 -0.32428E-04 24 0.112E-01 0.342E-02
SOL: 4 -0.47482E-01 0.35937E-01 -0.11546E-01 92
DAV: 4 -0.147865012941E+02 0.45338E-03 -0.16873E-04 24 0.783E-02 0.123E-02
SOL: 5 -0.48283E-01 0.35945E-01 -0.12338E-01 86
DAV: 5 -0.147864661719E+02 0.35122E-04 -0.30578E-05 24 0.304E-02 0.711E-03
SOL: 6 -0.49117E-01 0.35950E-01 -0.13167E-01 86
DAV: 6 -0.147864265215E+02 0.39650E-04 -0.16257E-05 32 0.235E-02 0.286E-03
SOL: 7 -0.49202E-01 0.35953E-01 -0.13249E-01 74
DAV: 7 -0.147864742531E+02 -0.47732E-04 -0.23956E-06 32 0.108E-02 0.185E-03
SOL: 8 -0.49411E-01 0.35969E-01 -0.13442E-01 79
DAV: 8 -0.147866038528E+02 -0.12960E-03 -0.14538E-05 32 0.280E-02 0.103E-03
SOL: 9 -0.49384E-01 0.35968E-01 -0.13416E-01 74
DAV: 9 -0.147866282159E+02 -0.24363E-04 -0.70285E-07 24 0.575E-03 0.544E-04
SOL: 10 -0.49370E-01 0.35966E-01 -0.13404E-01 73
DAV: 10 -0.147866458513E+02 -0.17635E-04 -0.13877E-06 24 0.692E-03 0.337E-04
SOL: 11 -0.49361E-01 0.35964E-01 -0.13397E-01 71
DAV: 11 -0.147866366463E+02 0.92050E-05 -0.53655E-07 24 0.536E-03 0.375E-04
SOL: 12 -0.49380E-01 0.35964E-01 -0.13416E-01 69
DAV: 12 -0.147866341141E+02 0.25322E-05 -0.19092E-08 32 0.848E-04 0.149E-04
SOL: 13 -0.49378E-01 0.35964E-01 -0.13414E-01 63
DAV: 13 -0.147866340056E+02 0.10858E-06 -0.36499E-08 32 0.115E-03
SOL: 14 -0.49376E-01 0.35964E-01 -0.13413E-01 57
1 F= -.14786634E+02 E0= -.14786634E+02 d E =-.134999E-47

The SOL steps present alternatively with DAV steps.

james.almeida
Newbie
Newbie
Posts: 2
Joined: Sun Sep 29, 2019 11:43 pm
License Nr.: 5-426

Re: the output discrepancy of VASPsol for CPU and GPU results

#2 Post by james.almeida » Wed Nov 06, 2019 5:41 pm

I confirm this bug.

I have experienced the same problem, the solvent is only applied after the SCF cycle converges.
Thus, the VASPSol calculations on the GPU are useless.

Is there any plan to address this GPU/VASPSol bug?

james.almeida
Newbie
Newbie
Posts: 2
Joined: Sun Sep 29, 2019 11:43 pm
License Nr.: 5-426

Re: the output discrepancy of VASPsol for CPU and GPU results

#3 Post by james.almeida » Mon Jan 20, 2020 4:02 pm

Dear VASP users and developers,

For anyone looking at this bug, as it is not clear if it is on the VASP side or the VASPSol plugin side, I have also reported at the VASPSol github project: https://github.com/henniggroup/VASPsol/issues/40

I hope someone looks at this bug, using the CPU version feels sluggish after experiencing the VASP on GPU.

Post Reply