[Pw_forum] Band parallelization of EXX ACE in QE 6.1

[WF]

Thanks, this is a small case (16 Silicon) with ONCV pseudo potential. 
Run with 16 processors per node and 4 nodes (64 procs in total), with
nb=1,2,4,8 the results are all the same, and the ACE energies in the output
file are 

     ACE energy    -9.83550603
     ACE energy    -9.85593641
     ACE energy    -9.86091525
     ACE energy    -9.86195787
     ACE energy    -9.86217893
     ACE energy    -9.86222150

but with nb=16, the result is different:
     ACE energy    -9.57847448
     ACE energy    -7.73229301

==========================
&control
 calculation='scf',
 outdir='./',
 prefix="Si8",
 verbosity='high',
 disk_io = "none"
/

&system
 ibrav=2,
 celldm(1)=20.526,
 ntyp = 1,
 nat=16,
 ecutwfc=90,
 input_dft = "pbe0",
 exx_fraction = 0.1449
 exxdiv_treatment = 'gygi-baldereschi'
 ecutfock = 180
/
&electrons
  electron_maxstep=600,
  mixing_beta=0.7,
  conv_thr=1.0D-8,
/
ATOMIC_SPECIES
Si 28.0855 Si_ONCV_PBE-1.1.upf
ATOMIC_POSITIONS alat
Si 0 0 0
Si 0.125 0.125 0.125
Si 0.25 0.25 0
Si 0.375 0.375 0.125
Si 0.25 0 0.25
Si 0.375 0.125 0.375
Si 0 0.25 0.25
Si 0.125 0.375 0.375
Si 0.25 0.25 0.5
Si 0.375 0.375 0.625
Si 0.25 0.5 0.25
Si 0.375 0.625 0.375
Si 0.5 0.25 0.25
Si 0.625 0.375 0.375
Si 0.5 0.5 0.5
Si 0.625 0.625 0.625
K_POINTS automatic
 4 4 4 0 0 0
====================

Feng

>Message: 6
>Date: Tue, 18 Jul 2017 16:59:52 +0200
>From: Paolo Giannozzi <p.giannozzi at gmail.com>
>Subject: Re: [Pw_forum] Band parallelization of EXX ACE in QE 6.1
>To: PWSCF Forum <pw_forum at pwscf.org>
>Message-ID:
>	<CAPMgbCvE1LXnbzGp=2GFZQjRPJNNzfmAVno9e8GXySyy9W+J1A at mail.gmail.com>
Content-Type: text/plain; charset="utf-8"
>
>There are no special restrictions on band parallellization of exact
>exchange, but since this has undergone a major reorganization recently,
>there might be corner cases that do not work as expected. If you have a
>reasonably-sized case that does not behave as expected, please submit it
>
>Paolo
>
>On Mon, Jul 17, 2017 at 11:49 PM, WF <kalamaillist at gmail.com> wrote:
>
>> Dear everyone,
>>
>>          I am trying the new EXX-ACE in QE6.1 on a system consists of 215
>> atoms, and 933 electrons of spin up and 932 down, gamma point only. When
no
>> parallelization level is specified in the command line, it runs well.
With
>> ?nb 2, it runs much faster with the same result. But with ?nb 4, it gives
a
>> wrong result in the first iteration of EXX and everything followed is
>> wrong. Is there any restriction of the band parallelization here? Thanks.
>
>>
>>
>> Best regards,
>>
>> Feng
>>
>> --------------------------------------------------
>>
>> Dr. Feng Wu
>>
>> Postdoctoral researcher
>>
>> Department of Chemistry and Biochemistry
>>
>> University of California, Santa Cruz
>>
>> Phone: 831-459-2874 <(831)%20459-2874>
>>
>>
>>
>> _______________________________________________
>> Pw_forum mailing list
>> Pw_forum at pwscf.org
>> 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
>-------------- next part --------------
>An HTML attachment was scrubbed...
>URL:
http://pwscf.org/pipermail/pw_forum/attachments/20170718/2b5b5754/attachment
-0001.html