Ceilidh ... Re^2: $LIBE Roundoff problems

Firefly and PC GAMESS-related discussion club

Re^2: $LIBE Roundoff problems

Big thanks Alex! Coordinates in input you have provided worked fine. I guess i will completely avoid $LIBE in future and use planar bands, as you have suggested. By the way, output you provided shows that you used version 8.0.0, i'm using 7.1.G and it states, that my version is outdated, when the latest version will be released? I have Nvidia GTX 460 GPU in my system but i can't use it because Firefly detects my GPU with CUDA support but states:

DP capable Async Timeout Emulated Useable
GeForce GTX 460 Yes Yes Yes No No

I've seen somewhere below in forum that somebody was able to use CUDA on GTX 460 using latest Firefly 8. Where i can obtain this recent version of program?

On Sun Jul 10 '11 1:04pm, Alex Granovsky wrote
----------------------------------------------
>Hi,

>Linear bends in Firefly (as well as in GAMESS (US)) are strange
>and odd beasts. Generally, I would not recommend to use them.

>Constructing good z-matrix for such degenerate and
>pathological cases as this molecule is could be a bit tricky.
>Attached please find the example of good coordinates.

>Regards,
>Alex Granovsky
>
>
>
>On Sun Jul 10 '11 7:54am, Alexander V. Malutin wrote
>----------------------------------------------------
>>Hi. I'm constantly getting bored by strange problems with $LIBE group. Every time i try to optimize molecule with linear bended region in internal coordinates, geometry search is almost constantly failing to locate stationary point then i'm using OPTTOL=1d-5 or below. Visualization of search steps shows oscillatory behavior near linear bend geometry, so i'm pretty much convinced that using linear bend $LIBE group is essential. However then i do use it, it results in error on the first step:

>> TRANSFORMING DISPLACEMENT FROM INTERNALS TO CARTESIANS
>> THE ROOT MEAN SQUARE ERROR IN ITERATION 1 IS 0.00004979
>> THE ROOT MEAN SQUARE ERROR IN ITERATION 2 IS 0.00006993
>> THE ROOT MEAN SQUARE ERROR IN ITERATION 3 IS 0.00013865
>> *** ROUNDOFF ERROR IN LIBE - TEST,DOTP= -0.2563072398E-06 -0.7003238419E-03

>>How can i solve this problem? May be i am missing something and should supply input files with some special keys to avoid such errors? As far as i can see such errors have unpredictable character, sometimes $LIBE supported input works just fine, but very often run fails due this nasty roundoff problem, witch is extremely annoying. Will grateful for any help regarding this matter.

>>Best regards
>>Alexander

>>Input file:

>> $SYSTEM MEMORY=128000000 $END
>> $CONTRL SCFTYP=MCSCF RUNTYP=OPTIMIZE $END
>> $CONTRL UNITS=ANGS MULT=1 NZVAR=8 $END
>> $BASIS GBASIS=N31 NGAUSS=6 NPFUNC=1 NFFUNC=1 $END
>> $MCSCF CISTEP=GUGA FULLNR=.TRUE. FORS=.FALSE. $END
>> $DRT GROUP=C1 SOCI=.TRUE. NMCC=5 NDOC=6 NVAL=3 NEXT=2 $END
>> $STATPT NSTEP=100 OPTTOL=1d-6 HSSEND=.TRUE. $END
>> $FORCE NVIB=2 VIBANL=.TRUE. PURIFY=.TRUE. TEMP(1)=0.001, 298.15, 398.15 $END
>> $GUESS GUESS=MOREAD NORB=67 $END
>> $DATA
>> Cyanamide NH2-CN MCSCF.
>> C1
>> N 7.0 0.0000000000 0.0000000001 1.2870668159
>> C 6.0 0.0000000001 0.0000000033 0.1424955716
>> N 7.0 0.0000000000 -0.0000000042 -1.1681804833
>> H 1.0 -0.0000000001 0.8498848363 -1.6742591359
>> H 1.0 -0.0000000001 -0.8498848188 -1.6742591918
>> $END
>> $ZMAT IZMAT(1)=1,1,2, 1,2,3, 1,3,5, 1,3,4,
>>2,2,3,5, 2,5,3,4, 3,2,3,4,5, 5,1,2,3 $END
>> $LIBE APTS(1)=1.0,0.0,0.0 $END
>> $VEC
>> ....
>> ....
>> ....
>> $END

Sun Jul 10 '11 11:19pm

This message read 918 times