odd "quiet death" when running GW on 5.0.4

You can find here problems arising when using old releases of Yambo (< 5.0). Issues as parallelization strategy, performance issues and other technical aspects.

Moderators: Davide Sangalli, andrea.ferretti, myrta gruning, andrea marini, Daniele Varsano, Conor Hogan

Locked
chwolf
Posts: 44
Joined: Mon Jul 11, 2016 2:11 pm
Location: South Korea
Contact:

odd "quiet death" when running GW on 5.0.4

Post by chwolf » Wed Oct 20, 2021 1:54 pm

Dear all,

I am facing a weird issue; I have run a calculation for G0W0 on this system (same cell, positions, pseudos, cutoffs etc) with yambo 4.3.2 without any issue, but with the new version of yambo it crashes without any error every time at the following step:

Code: Select all

 [06] Local Exchange-Correlation + Non-Local Fock
 ================================================

 [VXC] Plane waves :   7075
 [EXS] Plane waves :   61093
  
 QP @ state[ 1 ] K range:  1  1
 QP @ state[ 1 ] b range:   45   55
  
 [FFT-HF/Rho] Mesh size:   93   93   69

the log in this case will say

Code: Select all

 <02m-48s> P1: [X-CG] R(p) Tot o/o(of R):   523   539   100
 <02m-48s> P1: Xo@q[1] |                                        | [000%] --(E) --(X)
 <02m-53s> P1: Xo@q[1] |########                                | [020%] 05s(E) 24s(X)
 <02m-58s> P1: Xo@q[1] |################                        | [041%] 10s(E) 24s(X)
 <03m-03s> P1: Xo@q[1] |########################                | [061%] 15s(E) 24s(X)
 <03m-08s> P1: Xo@q[1] |#################################       | [082%] 20s(E) 24s(X)
 <03m-12s> P1: Xo@q[1] |########################################| [100%] 24s(E) 24s(X)
 <03m-13s> P1: [PARALLEL distribution for X Frequencies on 15 CPU] Loaded/Total (Percentual):1/2(50%)
 <03m-13s> P1: X@q[1] |                                        | [000%] --(E) --(X)
 <03m-29s> P1: X@q[1] |########################################| [100%] 16s(E) 16s(X)
 <03m-29s> P1: [PARALLEL distribution for RL vectors(X) on 3 CPU] Loaded/Total (Percentual):1287075/3861225(33%)
 <03m-30s> P1: [06] Local Exchange-Correlation + Non-Local Fock
 <03m-30s> P1: Self_Energy parallel ENVIRONMENT is incomplete. Switching to defaults
 <03m-30s> P1: [PARALLEL Self_Energy for QPs on 5 CPU] Loaded/Total (Percentual):2/11(18%)
 <03m-30s> P1: [PARALLEL Self_Energy for Q(ibz) on 1 CPU] Loaded/Total (Percentual):1/1(100%)
 <03m-30s> P1: [PARALLEL Self_Energy for G bands on 3 CPU] Loaded/Total (Percentual):19/55(35%)
 <03m-30s> P1: [PARALLEL distribution for Wave-Function states] Loaded/Total(Percentual):21/55(38%)
 <03m-30s> P1: [FFT-HF/Rho] Mesh size:   93   93   69
 <03m-30s> P1: EXS |                                        | [000%] --(E) --(X)
there is no other error, the RAM consumption is rather marginal (~10% of the available RAM) - any ideas?

compiled using intel-19.

Thanks in advance for any help :)

Chris
Christoph Wolf

Postech university, South Korea
chwolf@postech.ac.kr

User avatar
Daniele Varsano
Posts: 3816
Joined: Tue Mar 17, 2009 2:23 pm
Contact:

Re: odd "quiet death" when running GW on 5.0.4

Post by Daniele Varsano » Wed Oct 20, 2021 2:02 pm

Dear Christoph,
actually not easy to spot the problem.
Any error message from the log files of the other CPUs?

If not, can you try to recompile the code using internal linear algebra (--enable-int-linalg in the configure) and see if this solve the problem?

Best,
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/

Locked