[QE-users] One bug of pw2wannier90 in parallel execution

[박민규]

Dear all,

I attach the result of example11 of wannier90 for saving time to reproduce the problem.

This is what I hope to get and is obtained without any option,
                   Extraction of optimally-connected subspace
                   ------------------------------------------
 +---------------------------------------------------------------------+<-- DIS
 |  Iter     Omega_I(i-1)      Omega_I(i)      Delta (frac.)    Time   |<-- DIS
 +---------------------------------------------------------------------+<-- DIS
       1      20.71708776      20.71708776       0.000E+00      0.00    <-- DIS
       2      20.71708776      20.71708776       0.000E+00      0.00    <-- DIS
       3      20.71708776      20.71708776       0.000E+00      0.00    <-- DIS

             <<<      Delta < 1.000E-10  over  3 iterations     >>>
             <<< Disentanglement convergence criteria satisfied >>>

        Final Omega_I    20.71708776 (Bohr^2)

If I use -nb 2 option, I instead get
                   Extraction of optimally-connected subspace
                   ------------------------------------------
 +---------------------------------------------------------------------+<-- DIS
 |  Iter     Omega_I(i-1)      Omega_I(i)      Delta (frac.)    Time   |<-- DIS
 +---------------------------------------------------------------------+<-- DIS
       1    -425.57456453    -425.57456453      -1.110E-16      0.00    <-- DIS
       2    -425.57456453    -425.57456453      -1.110E-16      0.00    <-- DIS
       3    -425.57456453    -425.57456453      -1.110E-16      0.00    <-- DIS

             <<<      Delta < 1.000E-10  over  3 iterations     >>>
             <<< Disentanglement convergence criteria satisfied >>>

        Final Omega_I  -425.57456453 (Bohr^2)

Furthermore, If I use -nb 12 I get
                   Extraction of optimally-connected subspace
                   ------------------------------------------
 +---------------------------------------------------------------------+<-- DIS
 |  Iter     Omega_I(i-1)      Omega_I(i)      Delta (frac.)    Time   |<-- DIS
 +---------------------------------------------------------------------+<-- DIS
       1   **************   **************       0.000E+00      0.00    <-- DIS
       2   **************   **************       0.000E+00      0.00    <-- DIS
       3   **************   **************       0.000E+00      0.00    <-- DIS

             <<<      Delta < 1.000E-10  over  3 iterations     >>>
             <<< Disentanglement convergence criteria satisfied >>>

        Final Omega_I ************** (Bohr^2)

============================================
Minkyu Park
Research Institute of Basic Sciences, University of Ulsan,
Republic of Korea
============================================

On 9 Mar 2020, at 11:53 PM, 박민규 <minkyupark at ulsan.ac.kr<mailto:minkyupark at ulsan.ac.kr>> wrote:

Dear all users,

Although this might be already known, I want to report my observation in case someone has the same problem but cannot find any hint.

I found that if nband (or nb) option is used in pw2wannier90 step, wannier90 incorrectly gives negative values for Omega_I which is positive quantity by definition. The values also depend on how many cores I used.
I used following command after scf and nscf calculations (you can try to reproduce simply using e.g. example11 of wannier90)
mpirun -n 36 ~/wannier90-3.1.0/wannier90.x -pp wannier
mpirun -n 36 ~/qe-6.5/bin/pw2wannier90.x -nb 3 -nt 3 -in pw2wan.in > pw2wan.out
mpirun -n 36 ~/wannier90-3.1.0/wannier90.x wannier

Someone already reported this negative Omega_I problem long time ago in a different context:
https://lists.quantum-espresso.org/pipermail/wannier/2008-September/000196.html

If I change second one (pw2wannier90.x) to this
mpirun -n 36 ~/qe-6.5/bin/pw2wannier90.x -nt 3 -in pw2wan.in > pw2wan.out
Omega_I has positive value and does not depend on how many cores I used.
Actually nt option does not change the computation time much.

I guess it is good to add an error or warning message like the error message when nk option is used:
Error in routine pw2wannier90 (9):
     pools not implemented

Best regards,
============================================
Minkyu Park
Research Institute of Basic Sciences, University of Ulsan,
Republic of Korea
============================================

_______________________________________________
Quantum ESPRESSO is supported by MaX (www.max-centre.eu/quantum-espresso<http://www.max-centre.eu/quantum-espresso>)
users mailing list users at lists.quantum-espresso.org<mailto:users at lists.quantum-espresso.org>
https://lists.quantum-espresso.org/mailman/listinfo/users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20200309/8e6cf269/attachment.html>