wierd output from parallel CP2K

hawk2012 hawk2... at gmail.com
Thu Dec 27 21:21:33 UTC 2007


Hi, everyone:
We have successfully compiled a parallel CP2K executable cp2k.popt
with MPICH2+Intel Fortran compiler on our 64 Itanium CPUs  SMP
machine. We ran several tests with different number of CPUs using the
built-in H2O.inp as the input file under ~/cp2k/tests/QS. However, we
found that the content of output log files is very strange after we
checked them carefully. In general, we believe that the current
executable is still an serial executable although we can see several
cp2k.popt processes running at the same time. We did 3 tests with 1, 2
and 4 CPUs, respectively.
First of all, the total running time is always about 2
minutes and 10 seconds. The increase of number of CPUs
does not help reduce the total running time.
Second, the line ' GLOBAL| Total number of message
passing processes                             1' is
always the same no matter how many CPUs we are
actually using.
Third, almost every output part is just simply repeated
the same number of times as the number of CPUs.
For example, we can see the following in the log file with 2 CPUs:
... ...
  ----------------------------------- OT
--------------------------------------
     1  OT DIIS        0.15E+00    2.93
0.3610914591      -15.3342911053
     1  OT DIIS        0.15E+00    2.92
0.3610914591      -15.3342911053
     2  OT DIIS        0.15E+00    4.80
0.2696279745      -15.9344405484
     2  OT DIIS        0.15E+00    4.80
0.2696279745      -15.9344405484
     3  OT DIIS        0.15E+00    4.81
0.1188136007      -16.9597299341
     3  OT DIIS        0.15E+00    4.81
0.1188136007      -16.9597299341
     4  OT DIIS        0.15E+00    4.79
0.0728931898      -17.0674533793
     4  OT DIIS        0.15E+00    4.80
0.0728931898      -17.0674533793
     5  OT DIIS        0.15E+00    4.80
0.0652124778      -17.1410844505
     5  OT DIIS        0.15E+00    4.80
0.0652124778      -17.1410844505
... ...
, while we can see the below in the log file with 4 CPUs:
  ----------------------------------- OT
--------------------------------------
     1  OT DIIS        0.15E+00    2.91
0.3610914591      -15.3342911053
     1  OT DIIS        0.15E+00    2.91
0.3610914591      -15.3342911053
     1  OT DIIS        0.15E+00    2.90
0.3610914591      -15.3342911053
     1  OT DIIS        0.15E+00    2.90
0.3610914591      -15.3342911053
     2  OT DIIS        0.15E+00    4.77
0.2696279745      -15.9344405484
     2  OT DIIS        0.15E+00    4.78
0.2696279745      -15.9344405484
     2  OT DIIS        0.15E+00    4.78
0.2696279745      -15.9344405484
     2  OT DIIS        0.15E+00    4.78
0.2696279745      -15.9344405484
     3  OT DIIS        0.15E+00    4.78
0.1188136007      -16.9597299341
     3  OT DIIS        0.15E+00    4.78
0.1188136007      -16.9597299341
     3  OT DIIS        0.15E+00    4.78
0.1188136007      -16.9597299341
     3  OT DIIS        0.15E+00    4.79
0.1188136007      -16.9597299341
     4  OT DIIS        0.15E+00    4.76
0.0728931898      -17.0674533793
     4  OT DIIS        0.15E+00    4.77
0.0728931898      -17.0674533793
     4  OT DIIS        0.15E+00    4.77
0.0728931898      -17.0674533793
     4  OT DIIS        0.15E+00    4.77
0.0728931898      -17.0674533793
     5  OT DIIS        0.15E+00    4.77
0.0652124778      -17.1410844505
     5  OT DIIS        0.15E+00    4.77
0.0652124778      -17.1410844505
     5  OT DIIS        0.15E+00    4.77
0.0652124778      -17.1410844505
     5  OT DIIS        0.15E+00    4.78
0.0652124778      -17.1410844505
... ...

If we set np = 20 when we started the mpirun, we can see the above
output 20 times.

Does any one know what the problem might be or we should use another
input file to test parallel CP2K?


More information about the CP2K-user mailing list