[Pw_forum] something wrong with vc-relax & relax on my clusters
Axel Kohlmeyer
akohlmey at cmm.chem.upenn.edu
Tue Jun 24 14:54:50 CEST 2008
On Tue, 24 Jun 2008, lewvega wrote:
VL> Dear Users
dear vega lew,
thanks for your detailed e-mail, but before making
any claims that something is broken, you should fix
the obviously broken stuff. in this case it is your
compilers. the intel 10.1.008 compilers are _horribly_
broken and miscompile almost any scientific code that
i came across. you have to update to at least 10.1.015
and try again.
btw: don't worry about MKL's FFT not being detected.
there is no directly supported anyways and both the fftw3
and the fftw2 wrappers don't work reliably.
in all my tests it turned out to be slower than the
bundled FFTW, except for power of two grids (which confirms
results with BenchFFT). more importantly, you have to make
certain that the OMP_NUM_THREADS environment variable is set
to 1 or else it will multi-thread across all local cpu cores,
regardless of whether you parallelize with MPI (see the MKL
documentation for details).
[...]
VL> celldm(1) = 19.293941717,
VL> celldm(2) = 1.109509628630600744472407458828,
VL> celldm(3) = 1.7484889932972017886534804648227,
another remark: please read up double precision floating point
arithmetic (e.g. on wikipedia) and realize that specifying more
than 14 significant digits is a completely pointless execise...
(not to mention that the errors from DFT, using grids and an
incomplete basis set by far exceed the possible numerical accuracy.
cheers,
axel.
--
=======================================================================
Axel Kohlmeyer akohlmey at cmm.chem.upenn.edu http://www.cmm.upenn.edu
Center for Molecular Modeling -- University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582, fax: 1-215-573-6233, office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.
More information about the Pw_forum
mailing list