[Pw_forum] calculation crashes after reading charge density

[Sergey Lisenkov]

 dear all,

  I do have the following problem (it occurs only on ibm power4 machine, code is pw 3.2.0). I am performing the atomic relaxation and calculations stop (due to timelimit of queue) after several successful bfgs steps. I resume calculation using new atomic coordinates and put " startingpot='file' " in my input file. However, in this case the code crashed immediately :

     G cutoff =   31.6598  ( 697757 G-vectors)     FFT grid: ( 64, 64,360)
     G cutoff =    9.0456  ( 106713 G-vectors)  smooth grid: ( 36, 36,192)

     nbndx  =  1532  nbnd   =   383  natomwfc =   456  npwx   =     829
     nelec  = 638.00  nkb   =   888  ngl    =   18256

     The initial density is read from file :
     intf.save/charge-density.xml


     starting charge  640.36115, renormalised to  638.00000

     negative rho (up, down):  0.900E-05 0.000E+00
     Parameters of the lda+U calculation:
     Number of iteration with fixed ns =  0
     Starting ns and Hubbard U :
 enter write_ns
....
     total cpu time spent up to now is    177.60 secs

     WARNING: integrated charge=   636.11804878, expected=   638.00000000

 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
     from electrons : error #         1
     charge is wrong
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

     stopping ...

It does not happen on other machine (Xeon cluster) using the same input file. Any hints?

 thanks,
   sergey