It's hard to guess. We can only see that for every second case the
k-indices are nonsense.
As far as I see, you do k-parallel and mpi-parallel simultaneously.
While this is fine in principle, I do not see how this will work with
MPIRUN:mpiexec _EXEC_
Where/how do you specify on how much cores do you parallelize ?
Usually, mpiexec is called automatically by mpirun, but one needs to
specify -np XXX somewhere ?
Does it work when you do NOT use k-parallel, but only mpi-parallel ?
Am 28.09.2012 13:31, schrieb Lionel Calmels:
Dear Wien2k developers and users,We have tried to compile Wien2k (version
12.1) on a supercomputer (with Intel ifort 12.0.013 +mkl) The compilation
didn?t display any errors.Here are the options we choosed :current:FOPT:-FR
-mp1 -w -prec_div -pc80 -pad -ip -DINTEL_VML ?tracebackcurrent:FPOPT:-FR -mp1
-w -prec_div -pc80 -pad -ip -DINTEL_VML -DFFTW3
-tracebackcurrent:LDFLAGS:$(FOPT) -L$(MKLROOT)/lib/$(MKL_TARGET_ARCH)
-pthreadcurrent:DPARALLEL:'-DParallel'current:R_LIBS:-lmkl_lapack95_lp64
-lmkl_intel_lp64 -lmkl_intel_thread -lmkl_core -openmp
-lpthreadcurrent:RP_LIBS:-lmkl_scalapack_lp64 -lmkl_solver_lp64
-lmkl_blacs_lp64 -L/opt/software/SGI/fftw/3.3a/lib -lfftw3_mpi -lfftw3
$(R_LIBS) -lmpicurrent:MPIRUN:mpiexec _EXEC_
When we tried to test a calculation that works well on other computers, we
got the following error :severe (41): Insufficient virtual memorysevere(179):
Cannot allocate array - overflow on array size calculationWe suppose it
happens due to unusually high (iff1,iff2,iff3) and max k
indices.Fe3O4.output2up_1: max k indices: 12 12 12Fe3O4.output2up_2: max k
indices: 196607 196607 196607Fe3O4.output2up_3: max k indices: 12 12
12Fe3O4.output2up_4: max k indices: 262143 262143 262143Fe3O4.output2up_5:
max k indices: 196608 196608 196608Fe3O4.output2up_6: max k indices: 196607
196607 196607Fe3O4.output2up_7: max k indices: 13 13 13Fe3O4.output2up_8: max
k indices: 196608 196608 196608Fe3O4.output2up_1: n,iff1,iff2,iff3 2085 54 54
54Fe3O4.output2up_2: n,iff1,iff2,iff3 2095 787320 787320
787320Fe3O4.output2up_3: n,iff1,iff2,iff3 2097 54 54 54Fe3O4.output2up_4:
n,iff1,iff2,iff3 2085 1049760 1049760 1049760Fe3O4.output2up_5:
n,iff1,iff2,iff3 2126 787320 787320 787320Fe3O4.output2up_6!
: n,iff1
,iff2,iff3 2108 787320 787320 787320Fe3O4.output2up_7: n,iff1,iff2,iff3 2102 60
60 60Fe3O4.output2up_8: n,iff1,iff2,iff3 2114 787320 787320 787320This only
happens when running a k-points parallelization (and not for a sequential
calculation on 1 core).
If somebody has any suggestions to solve this problem, it would be really
appreciate.
Best wishes
Lionel CalmelsCEMES-Toulouse
___Wien mailing listWien at
zeus.theochem.tuwien.ac.athttp://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
--
Peter Blaha
Inst.Materials Chemistry
TU Vienna
Getreidemarkt 9
A-1060 Vienna
Austria
+43-1-5880115671