[Wien] Gd5Si4 calculations freeze

Peter Blaha pblaha at theochem.tuwien.ac.at
Mon Nov 1 11:06:02 CET 2010


case.output1_1  or similar (depending on spin,...)

Am 01.11.2010 09:40, schrieb Maxim Rakitin:
> Dear Prof. Blaha,
>
> Where can I find timings for parallel running programs? In my
> calculations I usually get such strings in case.dayfile:
>
>     Summary of lapw1para:
>     node-09-07 k=0 user=0 wallclock=0
>     0.692u 1.028s 14:50.98 0.1% 0+0k 0+0io 0pf+0w
>
> But this 'time' command output is for lapw1para script, not for actual
> lapw1c_mpi programs. The same situation is for lapw0/2.
>
> Thank you.
>
> Best regards,
>     Maxim Rakitin
>     Email:rms85 at physics.susu.ac.ru
>     Web:http://www.susu.ac.ru
>
>
> 01.11.2010 11:48, Peter Blaha пишет:
>> When the EF WARNING occurs during the scf cycle, but not at the end,
>> this is no problem.
>>
>> However, your lines:
>> >> lapw1 (14:40:19) 800.850u 14.192s 54:24.88 24.9% 0+0k 0+0io 0pf+0w
>> >> lapw0 (14:37:49) 34.906u 4.344s 2:28.51 26.4% 0+0k 0+0io 0pf+0w
>>
>> are suspicious. You get only 25% of the CPU. lapw1 should finish after
>> 14 min, but took almost one hour.
>> Either 3 other jobs are running on this single cpu machine, or you run
>> out of memory (you have very large system times !) and the machine
>> pages. (Check with the top command during execution.
>>
>> If memory is all used, either reduce RKMAX or go to a different machine.
>>
>>> I have tried again to run a regular LDA calculations. Again as in the
>>> LDA+U case, lapw2 takes very long time (calculating like 2 days
>>> already), and there is an interesting error:
>>>
>>> dayfile
>>> ***********************************************************************
>>>
>>> 0.461365296892454 211.999990982969 212.000016483888
>>>> lapw2 (15:34:45) WARNING: EF not accurate, new emin,emax,NE-min,NE-max
>>>> 0.461365291265349
>>>> lapw1 (14:40:19) 800.850u 14.192s 54:24.88 24.9% 0+0k 0+0io 0pf+0w
>>>> lapw0 (14:37:49) 34.906u 4.344s 2:28.51 26.4% 0+0k 0+0io 0pf+0w
>>>
>>> cycle 1 (Thu Oct 28 14:37:49 EDT 2010) (1/99 to go)
>>>
>>> start (Thu Oct 28 14:37:49 EDT 2010) with lapw0 (1/99 to go)
>>>
>>> ***********************************************************************
>>> I have seen it before for this compound during earlier trials. In
>>> general, somehow this compound Gd5Si4 takes maybe 10 times longer to go
>>> through the initial lapw0, lapw1 cycles than other transition metal
>>> containing compounds I was calculating with about 10 times finer k-mesh!
>>> And sizes of the unit cells are roughly comparable.
>>>
>>> I have attached a scf2 file from this cycle.
>>>
>>> Gd5Si4_np_2nd.scf2
>>> ***********************************************************************
>>> :GMA : POTENTIAL AND CHARGE CUT-OFF 12.00 Ry**.5
>>> Bandranges (emin - emax) and occupancy:
>>> :BAN00204: 204 0.457469 0.458922 2.00000000
>>> :BAN00205: 205 0.458380 0.459246 2.00000000
>>> :BAN00206: 206 0.458579 0.459997 2.00000000
>>> :BAN00207: 207 0.459122 0.460319 2.00000000
>>> :BAN00208: 208 0.459122 0.460319 2.00000000
>>> :BAN00209: 209 0.459929 0.460891 2.00000000
>>> :BAN00210: 210 0.459944 0.460891 2.00000000
>>> :BAN00211: 211 0.460551 0.461480 1.89644048
>>> :BAN00212: 212 0.460552 0.461646 1.73748489
>>> :BAN00213: 213 0.460552 0.462067 0.30738641
>>> :BAN00214: 214 0.461259 0.462270 0.05869553
>>> :BAN00215: 215 0.461611 0.462536 0.00000000
>>> :BAN00216: 216 0.461611 0.462545 0.00000000
>>> :BAN00217: 217 0.461923 0.462744 0.00000000
>>> :BAN00218: 218 0.461923 0.462895 0.00000000
>>> :BAN00219: 219 0.462510 0.463225 0.00000000
>>> Energy to separate low and high energystates: 0.02312
>>>
>>>
>>> :NOE : NUMBER OF ELECTRONS = 424.000
>>>
>>> :FER : F E R M I - ENERGY(TETRAH.M.)= 0.46137
>>> ***********************************************************************
>>>
>>> Looking forward to your advises and ideas,
>>> Thank you,
>>> Volodymyr
>>>
>>>
>>>
>>> _______________________________________________
>>> Wien mailing list
>>> Wien at zeus.theochem.tuwien.ac.at
>>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>>
>
>
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien

-- 
Peter Blaha
Inst.Materials Chemistry
TU Vienna
Getreidemarkt 9
A-1060 Vienna
Austria
+43-1-5880115671


More information about the Wien mailing list