[CPMD-list] increased CPU times for wavefunction iterations

[Axel Kohlmeyer]

On Wed, 20 Jul 2005, Martin Konopka wrote:

MK> Dear CPMD developers,

hi martin,

MK> May I have few questions on wavefunction optimization, in particular CPU
MK> times for individual iterations? The usual pattern is that first part of
MK> them takes one value of CPU time (about 82 s in the example below) but the
MK> other part spends significantly more time (here 121 s). I see that this
MK> larger value occurs when wavefunctions are almost converged; often it is
MK> only in last one or two iterations but sometime in more of them.

yes. i've noticed something similar as well. also during MD runs
where it seems to come and go. are you by any chance running on
a dual-opteron machine? is this reproducable across platforms
or algorithms?

MK> I can notice this behaviour at least when using Vanderbilt
MK> pseudopotentials.
MK> 
MK> What is a more detailed reason for this?

i don't know the reason. but one possible explanation would be
a not so favorable memory allocation/access pattern for some
part of the data. at least on multiprocessor opteron machines
without pinning an mpi task to a cpu and its memory, this is
quite plausible. depending on the MPI library you are using,
you should also compare the TCPU (which is only user-cpu time)
value with the total time. its value may be misleading, i.e.
each iteration takes about the same time, but only for some
steps it is fully accounted to the master-cpu, which in turn
may depend on the memory allocation scheme.

sorry for not being able to offer a real solution.

best regards,
	axel.

MK> Is it possible to avoid the incresed CPU times somehow (except cutting
MK> down the number of iterations)?
MK> 
MK> 
MK> RESETTING WAVEFUNCTION TOLERANCE FACTOR TO:  1.0
MK>    1  1.274E-05   2.093E-06   -2003.766203    6.517E-06     82.41
MK>    2  1.418E-05   7.662E-07   -2003.766206   -3.192E-06     81.77
MK>    3  1.423E-05   6.226E-07   -2003.766206   -3.597E-07     81.00
MK>    4  1.120E-05   3.037E-07   -2003.766207   -3.911E-07     82.17
MK>    5  9.869E-06   2.593E-07   -2003.766207   -1.469E-07     81.38
MK>    6  9.743E-06   2.190E-07   -2003.766207   -2.256E-07     81.62
MK>    7  8.522E-06   1.914E-07   -2003.766208   -3.139E-07     81.72
MK>    8  7.540E-06   1.659E-07   -2003.766208   -2.474E-07     83.64
MK>    9  6.717E-06   1.455E-07   -2003.766208   -1.815E-07     82.54
MK>   10  5.885E-06   1.346E-07   -2003.766208   -1.956E-07     82.62
MK>  RELAXING WAVEFUNCTION TOLERANCE, NEW FACTOR:  2.0
MK>   11  5.428E-06   1.274E-07   -2003.766208   -1.691E-07     82.66
MK>   12  4.997E-06   1.185E-07   -2003.766208   -1.303E-07    121.48
MK>   13  4.565E-06   1.093E-07   -2003.766209   -1.725E-07    121.36
MK>   14  4.126E-06   9.982E-08   -2003.766209   -1.418E-07    120.97
MK>   15  3.850E-06   8.996E-08   -2003.766209   -1.009E-07    120.63
MK>   16  3.854E-06   8.264E-08   -2003.766209   -1.117E-07    120.15
MK>   17  3.879E-06   7.954E-08   -2003.766209   -9.304E-08    121.96
MK>   18  3.662E-06   7.727E-08   -2003.766209   -7.082E-08    121.41
MK>   19  3.680E-06   7.516E-08   -2003.766209   -8.924E-08    119.64
MK>   20  3.410E-06   7.326E-08   -2003.766209   -8.909E-08    120.40
MK> 
MK> Thanks for your help.
MK> Martin Konopka.
MK> 
MK> 
MK> 
MK> ------------------------------------------------------------------------
MK> Dr. Martin Konopka             http://www.ccms.elf.stuba.sk/konopka.html
MK> Department of Physics, CCMS                      tel: +421 (0)2 60291714
MK> Slovak University of Technology (FEI STU)        fax: +421 (0)2 65411483
MK> Ilkovicova 3, 812 19 Bratislava, Slovakia
MK> ------------------------------------------------------------------------
MK> _______________________________________________
MK> CPMD-list mailing list
MK> CPMD-list at cpmd.org
MK> http://cpmd.org/mailman/listinfo/cpmd-list
MK> 
MK> 

-- 

=======================================================================
Dr. Axel Kohlmeyer   e-mail: axel.kohlmeyer at theochem.ruhr-uni-bochum.de
Lehrstuhl fuer Theoretische Chemie          Phone: ++49 (0)234/32-26673
Ruhr-Universitaet Bochum - NC 03/53         Fax:   ++49 (0)234/32-14045
D-44780 Bochum  http://www.theochem.ruhr-uni-bochum.de/~axel.kohlmeyer/
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.