G: Internal consistency error detected in FileIO
Dear CCL,
I trying to compute interactions in a small complex with MP5 method - 4 atoms,
160 basis functions and 14 electrons. I did MP4 without
any problems.
However during MP5 calcs I get an error:
Saving the triples amplitudes on disk, using 189613920 words of disk space.
Time for triples= 942.87 seconds.
Disk space used for TT scratch files : 447697125 words
E5TTaaa = 0.31800067D-04
Internal consistency error detected in FileIO for unit 1 I= 9 J= 0.
dumping /fiocom/, unit = 1 nfiles =108 sizext = 524288 winblk = 512
defal = T lstwrd =**********
number 0 0 0 0 0 0 0
base 305152 352256 395264 298496 4865536 27059712 592384
end 324096 353792 398336 300544 4922368 40986624 2508800
end1 324096 353792 398336 300544 4922368 40986624 2508800
wr pntr 305152 352256 395264 298496 4884661 27644937 826474
rd pntr 305152 352256 395264 298496 4884661 27641344 823808
number 0 0 1 2 3 4 5
base ****************** 4034560 2685952135401472 394752 3448832
end ****************** 4150840 3271177135517752 394807 4034057
end1 ****************** 4151296 3271680135518208 395264 4034560
wr pntr319396440********* 4034560 2685952135401472 394752 3448832
rd pntr319396440********* 4150840 3271177135517752 394807 4034057
(... more error analysis ... )
I tried different options: full direct, semi-direct, Trans(conventional),
Int(NoRaff) etc. And the result is such that Gaussian 94 sometimes
crashes in link 804 and sometimes in link 915 ...
Does anybody know a cure for this error ? Is this rather hardware problem
or software (gaussian) problem? Or maybe it is a platform dependent problem
?
I use NEC SX-4 and 1Gb of memory and 30GB of disk.
Thanks in advance.
G.Bakalarski
Warsaw University
Poland