Re:Re: [SIESTA-L] CG optimizing problem. geometriesin.STRUCT_NEXT_ITER file

krishna Mohan Grangadhran Prasanna Thu, 28 Jul 2011 22:57:06 -0700

Dear Hongyi,

I ran your files with Siesta. But, this does not seem to solve the problem, yet. Because It run the SCF calculations but it doesn’t do CG steps, as I mentioned in my previous mail.

Here, I reduced the time by setting:

SpinPolarized F

PAO.BasisSize SZP

DM.Tolerance 1.d-3

in the Frelax.fdf, and by running, it give the same error as of my first mail. Ie. In 10xGraphene.Final_.STRUCT_NEXT_ITER it print the NaN

21.450000000 12.384163052 0.000000000

21.450000000 -12.384163052 0.000000000

0.000000000 0.000000000 20.000000000

210

1 6 NaN NaN NaN

…

and hence creating the error log:

forrtl: severe (174): SIGSEGV, segmentation fault occurred

Image PC Routine Line Source

siesta 00000000006B0542 Unknown Unknown Unknown

siesta 00000000006AC92F Unknown Unknown Unknown

siesta 000000000058C8D3 Unknown Unknown Unknown

...

Additionally, you can simply change CG to FIRE (using my first post's fdf file), then it will work (but in this case it didn’t correctly catch the constraint flags!).

If you have any idea please reply,

With best regards

------------ Original Message ------------
From : "Hongyi Zhao"<[email protected]>
To : [email protected]
Mail Subject : Re: [SIESTA-L] CG optimizing problem. geometriesin.STRUCT_NEXT_ITER file
Sent : Fri, 29 Jul 2011 12:31:35 +0800

On 07/29/2011 11:29 AM, krishna Mohan Grangadhran Prasanna wrote:
> Dear Marcos, and Hongyi
>
>
> Thank you very much for your helps. Here are some additional infos:
>
> *
> *
>
> *1]* I used O2 flags for compiling Siesta (siesta-3.0-rc2).
>
> *2] *I removed the white spaces in myFDFpart.fdf, but it couldn’t help
> me!

The advice I given to you is based the fact that the GDIS ( svn version
) cann't read your fdf files correctly unless I revise the fdf according
to the things I told you. In fact, this advice just my guess as I've
told in the last post.

> I also think that the file was successfully read by Siesta to start
> SCF, that’s why I got the CG=0 step.
>
> *3] *I replace CG to FIRE, and it worked! (which used my attached
> versions of files)

Where is the file, I cann't find it.

> eventhough it didnt give the desired constraint
> optimization!
>
> *4] *When I removed the /constraint block/, the CG worked for my
> original files. There were NO error messages...

You can check all my files posted to you in order to compare with the
matter you encountered. Wish you can find some hints and if so, please
let me know ;-)

Regards
>
>
> Anyway, Hongyi, can you sent me your edited *myFDFpart.fdf* file to
> gpkmohanpostech.ac.kr ? If so, I can make a close examination...
>
>
> Thanks in advance
>

--
Hongyi Zhao
Institute of Semiconductors, Chinese Academy of Sciences
GnuPG DSA: 0xD108493

Re:Re: [SIESTA-L] CG optimizing problem. geometriesin.STRUCT_NEXT_ITER file

Responder a