[Wien] delay and sleepy parameter in lapw1/2para

[Khuong P. Ong]

Dear Wien2k users,

  Could someone give us an advice on setting up parameters delay and sleepy 
in lapw1/2para? What is the best value for these parameters?

  Our calculation crashed at LAPW1 with errors

cycle 4         (Tue Apr 19 15:08:43 SGT 2005)  (37/37 to go)

 >   lapw0 -p    (15:08:43) starting parallel lapw0 at Tue Apr 19 15:08:43 
SGT 2005
--------
running lapw0 in single mode
8.560u 0.120s 0:08.73 99.4%     0+0k 0+0io 2515pf+0w
 >   lapw1  -up -p       (15:08:52) starting parallel lapw1 at Tue Apr 19 
15:08:52 SGT 2005
->  starting parallel LAPW1 jobs at Tue Apr 19 15:08:52 SGT 2005
running LAPW1 in parallel mode (using .machines)
4 number_of_parallel_jobs
      opto033(45)      opto033(45)      opto033(45)      opto033(46) 
Summary of lapw1para:
    opto033       k=45    user=0  wallclock=45
0.090u 0.200s 7:18.22 0.0%      0+0k 0+0io 29908pf+0w
 >   lapw1  -dn -p       (15:16:11) starting parallel lapw1 at Tue Apr 19 
15:16:11 SGT 2005
->  starting parallel LAPW1 jobs at Tue Apr 19 15:16:11 SGT 2005
running LAPW1 in parallel mode (using .machines.help)
4 number_of_parallel_jobs
**  LAPW1 crashed!
0.110u 0.240s 7:16.55 0.0%      0+0k 0+0io 26223pf+0w

 >   stop error

the error appeared only in file:

  <http://darwin:6500/util/edit.pl?SID=501065&file=/home1/mic/cc0501/khuong/CdCrO2-AFM/dnlapw1.error>dnlapw1.error 

lapw1.error:**  Error in Parallel LAPW1
lapw1.error:**  check ERROR FILES!
:parallel:**  check ERROR FILES!

When we increase the delay =5, sleepy=10  in lapw1/2para and number of 
k-points is smaller then SCF calculation can survive till cycle 22.

We notice here that everything is Ok with TiC and other simple materials. 
But with materials that each SCF cycle taking more than 5 minutes then 
Lapw1 will crash soon.

  Thanks for reading this.

Regards,
Khuong 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20050419/b7c1533a/attachment.html