[CPMD-list] test calculation problems.

[Axel Kohlmeyer]

On Tue, 27 Apr 2004, Sombat Ketrat wrote:

SK> Dear CPMD-Community,
SK> 
SK> I'm a beginner on CPMD code. I've installed CPMD 3.7.2 on SuSE 9.0 linux 
SK> PC with pgf77 and also with ifc complier (in the case of ifc I used the 
SK> LAPACK/BLAS/ATLAS library from Dr. Axel Kohlmeyer). The complation step 
SK> is seem to be well done. I submitted the test calculations with the 
SK> input provided by cpmd download page. The output which I found are a bid 
SK> differ from the output from provided (the provided outputs were 
SK> calculated by CPMD 3.7.1). Onething I'm very suprising is the time for 
SK> calculation. The jobs which calculated by my binary took much more time 
SK> (about 5 min from the provided output and 8 hours 11 min. from my 
SK> calculation.
SK> 
SK> the questions is,
SK> 1) anyone know the type of mechines which calculated on the provided 
SK> outputs.

hi,

the reference machine was a probably a 32 processor ibm power4 
processor based SMP-machine ('regatta'). check out the PARAPARAPARAPARA
section. your job on the other hand was obviously running with a serial 
executable, and since it seems to be a 32-bit pc machine, with much less 
memory bandwidth, which is rather crucial to good CPMD performance for 
such large jobs. 

SK> 2) Is it the resonable output for my calculation?

with different architectures, you'll probably have different fft 
grids, slightly different (precision impeding) optimizations, so
this will affect the absolute energies and the convergence a little 
bit. also with parallel vs. serial jobs there are likely to be small
differences. your result seems to be within reasonable agreement.

regards,
	axel kohlmeyer.

SK> 
SK> if one have any comments or suggestions please share your idea. Thanks 
SK> in advanced.
SK> 
SK> regard,
SK> Sombat
SK> 
SK> ps. I attached the input and output from me and provider in this email. 
SK> It's the 2 molecule of C60 in the std directory of untar cpmd-test.tar.Z 
SK> file.
SK> 
SK> 

-- 


=======================================================================
Dr. Axel Kohlmeyer                        e-mail: axel.kohlmeyer at rub.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/
=======================================================================