Hi all, When I performed phonon calculation in QE 5.3.0 on BGQ, I got set_irr_sym_new %%%%%%%%%%%%%%%%%%%%%%%%%%%%% Error in routine set_irr_sym_new (211): wrong representation %%%%%%%%%%%%%%%%%%%%%%%%%%%%% This error doesn't appear all the time but on some specific q points in a grid.
By comparing with the same calculation performed on x86, I noticed that some of the eigenvalues (autoval) computed for the q points emitting errors are quite different. So I went to check how those values are computed. I found that the cdiagh (in PW/src/cdiagh.f90) are branched using ZHPEV in ESSL or ZHEEV in LAPACK. So I changed the route to force the LAPACK one and my issue disappears. Though this solution seems solve my issue but I was not sure it address the key problem. My library linking order is SCALAPACK ESSL LAPACK ESSL (for FFT and BLAS) I think this is the right order for BGQ. Since, the cdiagh is not reported at the end of pw or ph output, I guess it takes little time during the run and I won't get any performance penalty. Any comments or suggestions? particularly from developers. PS: I send an email a few weeks ago and got no reply. I hope someone can also answer these questions. 1) Any update for releasing DFT+U phonon? I saw the paper was out years ago. 2) Any ways to reduce the phonon disk I/O to 'none' similar to pw.x? Reduce_io seems help very little during the calculation. Best regards, Ye =================== Ye Luo, Ph.D. Leadership Computing Facility Argonne National Laboratory
_______________________________________________ Pw_forum mailing list Pw_forum@pwscf.org http://pwscf.org/mailman/listinfo/pw_forum