Firefly and PC GAMESS-related discussion club

 
Learn how to ask questions correctly  
 
 

 

Alex Granovsky
gran@classic.chem.msu.su

this problem has been traced down to a bug in MKL matrix diagonalization
routine when working in multi-threaded mode and running on your particular
hardware (Core 2 Quad). This problem can be avoided by adding the following line to your input:

 $smp dianp=1 $end

Kind regards,
Alex



On Thu Mar 30 '17 9:46pm, Bernhard Dick wrote
---------------------------------------------
>Dear Alex,
>here is a ZIP with the requested files.
>best regards,
>Bernhard
>
>
>On Tue Mar 28 '17 3:33pm, Alex Granovsky wrote
>----------------------------------------------
>>Dear Bernhard,

>>I'm sorry for delay with my reply. Could you please send me the entire
>>input file including $vec group, and also your basis set definition file?

>>All the best,
>>Alex
>>
>>
>>On Wed Mar 22 '17 0:37am, Bernhard Dick wrote
>>---------------------------------------------
>>>Dear Alex,
>>>I get a crash of an MCSCF run with the error message

>>> Fatal error in DOVESHIFTX, error code =        -1

>>>immediately after the printout of the " DENSITY MATRIX OVER CANONICALIZED ACTIVE MO-S". The next part of the output should be
>>>          -----------------------
>>>          -MCHF- NATURAL ORBITALS
>>>          -----------------------
>>>...

>>>The error is reproducible with this particular input. However, I have successfully run two other rather similar jobs (only singlets instead of triplets, and different ISTSYM). Here is the command line part of the input:

>>> $SYSTEM TIMLIM=60000 MWORDS=200 MKLNP=2 NP=4 $END
>>> $CONTRL SCFTYP=MCSCF RUNTYP=ENERGY EXETYP=RUN MULT=3 ICHARG=0
>>>  MAXIT=200 COORD=UNIQUE fstint=.t. gencon=.t. inttyp=hondo icut=11
>>>  NZVAR=0 MPLEVL=2 NOSYM=0 D5=.T. $END
>>> $BASIS GBASIS=cc-pvdz extfil=.t. $END
>>> $STATPT METHOD=GDIIS NSTEP=30 NOREG=5 HSSEND=.F. $END
>>> $MOORTH nostf=1 nozero=1 syms=1 symden=1 symvec=1 $end
>>> $SCF DIRSCF=.TRUE. DIIS=.TRUE. $END
>>> $MCSCF CISTEP=GUGA fullnr=.F.  MAXIT=50 MINMEM=.T.
>>>   fors=.t. acurcy=1d-7 engtol=1d-12 $END
>>> $DRT GROUP=C2 ISTSYM=2 NMCC=63 NDOC=5 NALP=2 NVAL=5 FORS=.T. $END
>>> $trans dirtrf=.t. $end
>>> $GUGDM2 wstate(1)=1  cutoff=1d-12 $end
>>> $GUGDIA ITERMX=2000 NSTATE=1 cvgtol=1d-7 $END
>>> $gugem pack2=.t. cutoff=1d-12 $end
>>> $XMCQDPT EDSHFT=0.02 ISTSYM=2 KSTATE(1)=1 $END
>>> $GUESS GUESS=MOREAD NORB=75  $END

>>>This is what Firefly says about the Computer:

>>> Intel Core2/ Win32  Firefly version running under Windows NT
>>> Running on Intel CPU:  Brand ID  0, Family  6, Model  23, Stepping 10
>>> CPU Brand String    :  Intel(R) Core(TM)2 Quad CPU    Q9400  @ 2.66GHz
>>> CPU Features        :  x87 FPU, CMOV, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, MWAIT, EM64T
>>> Data cache size     :  L1 32 KB, L2 3072 KB, L3     0 KB
>>> max    # of   cores/package :   4
>>> max    # of threads/package :   4
>>> max     cache sharing level :   2
>>> Operating System successfully passed SSE support test.

>>>and somewhat later:

>>> This job is executing on     1 unique host(s)
>>> Minimum number of processes per host is:    1
>>> Maximum number of processes per host is:    1

>>> On master's host, detected    4 CPU core(s) in aggregate

>>> DGEMM will use                                   2 threads.

>>> Matrix diagonalization and inversion will use    2 threads.

>>> SMP/multicore aware parts of program will use    4 threads.

>>> Creating thread pool to serve up to            128 threads.

>>> Activating Call64 option.

>>>The two successful runs used the same computer. I am not so familiar with the MKLNP and NP parameters, I usually use parallel run with P2P, but I wanted to try XMCQDPT, and that does apparently not run parallel.

>>>best regards,
>>>Bernhard
>>>