FFT3D Error

Message

Sebast2600 · #1 Post by **Sebast2600** » Sat Nov 19, 2005 1:52 am

Hello,

I'm trying to compile the serial version of Vasp 4.6.12 with Ifort 9.0 and Mkl 8.0. The compilation is fine, but after sending a job the program crashes with the following error:

LDA part: xc-table for Ceperly-Alder, standard interpolation
POSCAR, INCAR and KPOINTS ok, starting setup
WARNING: random wavefunctions but no delay for mixing, default for NELMDL
entering main loop
N E dE d eps ncg rms rms(c)
INIT FFT3D: IERR = 2

I'm running Vasp on a P3. This error happens with both FFTW 3.0.1 and FFT Furth.
I have tryed many compilation options, but none solved my problem. I'm really desesperate. Anyone can help?

Thank you in advance,

Seb

wistey · #2 Post by **wistey** » Sun Nov 20, 2005 11:45 am

I just ran into a similar crash (INIT FFT3D IERR=4) doing ion relaxation. I found if I commented out the NGX, NGY, & NGZ lines in INCAR, to use the defaults, it ran fine. Ironically, I only defined NGX etc. because of the wrap warning in OUTCAR, but apparently I was using values which were too small. (35 was suggested in OUTCAR, but 36 worked.) I'm running the precompiled Linux serial version. ENMAX and PREC didn't seem to matter.

Sebast2600 · #3 Post by **Sebast2600** » Mon Nov 21, 2005 5:27 pm

Wistey, thank you for your reply.

I changed back NGX, NGY, and NGZ to the default values and it worked. (I previously used the values given in OUTCAR, because of the warp warning).

However, in the prarallel code everything works fine with my values, the problem only happens in the serial code. I don't understand why.

Seb

#4 Post by **admin** » Wed Nov 23, 2005 3:18 pm

NGX NGY and NGZ have to be factorizable into integer multiples of
2,3,4,5,6, or 7 in the fft-routines which are called in the serial code
version of Juergen Furthmueller's FFT, but NOT in the parallel version,
IERR=2 is the error core if this factorisation fails.