I find that in the version svn rev. 12863 the corrections suggested by Simone is required.
In PW/src/electrons.f90 comment line numbers 1235 - 1236 In PW/src/exx.f90 comment line numbers 2825 - 2826 Best wishes, Biswajit Biswajit Santra Mobile: +1-609-227-9202 http://www.princeton.edu/~bsantra/ On Fri, Sep 16, 2016 at 5:58 PM, Biswajit Santra <[email protected]> wrote: > Dear Paolo, > > Thanks you very much. > > Does the svn rev. 12863 already contain the correction suggested by > Simmone? I found that those lines are uncommented. > > I used the flag -D__EXX_ACE in DFLAGS, so all files were compiled with > this flag. I did not see any error in the compilation. Inside the output of > the job I ran I do see the ACE is called. E-ACE is printed and the > cpu timings in the related subroutines are also printed. > aceupdate : 1.04s CPU 1.31s WALL ( 72 calls) > vexxace : 32.13s CPU 34.44s WALL ( 1062 calls) > aceinit : 431.72s CPU 587.74s WALL ( 72 calls) > > Best wishes, > Biswajit > > Biswajit Santra > Mobile: +1-609-227-9202 > http://www.princeton.edu/~bsantra/ > > On Fri, Sep 16, 2016 at 4:26 PM, Paolo Giannozzi <[email protected]> > wrote: > >> >> On Fri, Sep 16, 2016 at 4:58 PM, Biswajit Santra <[email protected]> >> wrote: >> >>> >>> So far, I just did one test calculation with ACE. I did not see any >>> significant change in the computational time. Could you tell me what is >>> your experience in this regard? >>> >> >> it is definitely faster. Please verify that ACE is really compiled: >> contrary to what I said, you need to compile with option __ACE_EXX >> electrons.f90 and h_psi.f90 in addition to exx.f90. The svn version already >> contains fixes for the parallel case. >> >> Paolo >> >>> >>> Best wishes, >>> Biswajit >>> >>> >>> Biswajit Santra >>> Mobile: +1-609-227-9202 >>> http://www.princeton.edu/~bsantra/ >>> >>> On Fri, Sep 16, 2016 at 3:23 AM, Simone Piccinin <[email protected]> >>> wrote: >>> >>>> Dear Biswajit, >>>> >>>> I'm testing ACE in QE6.0-beta too, and I noticed the total energy >>>> changes with the number of pools. >>>> I contacted Ivan Carnimeo here at SISSA, the guy who coded ACE, and it >>>> turns out there's a bug. >>>> >>>> You need to comment these 2 lines: >>>> >>>> CALL mp_sum( exxenergyace, intra_bgrp_comm) >>>> CALL mp_sum( exxenergyace, inter_pool_comm ) >>>> >>>> inside exxenergyace, as well as: >>>> >>>> CALL mp_sum( eexx, intra_bgrp_comm) >>>> CALL mp_sum( eexx, inter_pool_comm ) >>>> >>>> inside aceinit. >>>> >>>> I did not yet test whether these changes give a npoool-independent >>>> total energy. >>>> >>>> Best regards, >>>> Simone >>>> >>>> >>>> On Sep 14, 2016, at 22:21 PM, Biswajit Santra wrote: >>>> >>>> Hello, >>>> >>>> I find that Lin Lin's ACE (Adaptively Compressed Exchange) for hybrid >>>> functionals is implemented in the espresso-6.0-beta version. >>>> >>>> I am interested in using/testing this functionality. Is it possible to >>>> provide the necessary keywords related to ACE or the place to find it? >>>> >>>> Thank you very much. >>>> >>>> Best wishes, >>>> Biswajit >>>> >>>> Biswajit Santra >>>> Mobile: +1-609-227-9202 >>>> http://www.princeton.edu/~bsantra/ >>>> _______________________________________________ >>>> Pw_forum mailing list >>>> [email protected] >>>> http://pwscf.org/mailman/listinfo/pw_forum >>>> >>>> >>>> --- >>>> Simone Piccinin >>>> CNR-IOM c/o SISSA >>>> Via Bonomea 265 >>>> 34136 Trieste >>>> Tel: +39-040-3787317 >>>> email: [email protected] >>>> >>>> >>> >>> _______________________________________________ >>> Pw_forum mailing list >>> [email protected] >>> http://pwscf.org/mailman/listinfo/pw_forum >>> >> >> >> >> -- >> Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche, >> Univ. Udine, via delle Scienze 208, 33100 Udine, Italy >> Phone +39-0432-558216, fax +39-0432-558222 >> >> >> _______________________________________________ >> Pw_forum mailing list >> [email protected] >> http://pwscf.org/mailman/listinfo/pw_forum >> > >
_______________________________________________ Pw_forum mailing list [email protected] http://pwscf.org/mailman/listinfo/pw_forum
