[gmx-users] No convergence with g_nmeig

[KARAKAS Esra]

Dear gromacs users,

I still have the same problem with the vector convergence. There is a link with the size of my protein (18 081 atoms) ?
I tried with a small molecule composed with 826 atoms and it seems to work.

Any idea, please.

Best

Esra



________________________________________

Hi,

Thank you for your response. I used a supercomputer and all tools are compiled with mpi, I used 1 cpu and I have
also the same result.

Best,

Esra

Message: 2
Date: Thu, 5 Nov 2015 14:45:27 +0000
From: Joaquim Rui de Castro Rodrigues <joaquim.rodrigues at ...<http://gmane.org/get-address.php?address=joaquim.rodrigues%2d7rSwZE0QHb3QUPosyRRdSA%40public.gmane.org>>
To: "gmx-users at ...<http://gmane.org/get-address.php?address=gmx%2dusers%2d0nhw1vs2%2b5JAfugRpC6u6w%40public.gmane.org>" <gmx-users at ...<http://gmane.org/get-address.php?address=gmx%2dusers%2d0nhw1vs2%2b5JAfugRpC6u6w%40public.gmane.org>>
Subject: Re: [gmx-users] No convergence with g_nmeig
Message-ID:
        <99DCD30D03D1124CBDFCE15D7659B63183F0AA9A75 at ...<http://gmane.org/get-address.php?address=99DCD30D03D1124CBDFCE15D7659B63183F0AA9A75%2dTBb6hYIbNRuPJ8G44WGNSF6hYfS7NtTn%40public.gmane.org>>
Content-Type: text/plain; charset="us-ascii"

Hi,

You seem to be running the calculation in parallel (g_nmeig_mpi_d). I don't think this is possible. Have
you tried the nmeig serial version?

HTH,
Rui Rodrigues

________________________________________
De: gromacs.org_gmx-users-bounces at ...<http://gmane.org/get-address.php?address=gromacs.org%5fgmx%2dusers%2dbounces%2dmrEvjEli4ZDCMOUXqa3p%2bTAV6s6igYVG%40public.gmane.org>
[gromacs.org_gmx-users-bounces at ...<http://gmane.org/get-address.php?address=gromacs.org%5fgmx%2dusers%2dbounces%2dmrEvjEli4ZDCMOUXqa3p%2bTAV6s6igYVG%40public.gmane.org>] Em Nome De
KARAKAS Esra [Esra.Karakas at ...<http://gmane.org/get-address.php?address=Esra.Karakas%2dKCE40YydGKI%40public.gmane.org>]
Enviado: quinta-feira, 5 de Novembro de 2015 11:08
Para: gromacs.org_gmx-users at ...<http://gmane.org/get-address.php?address=gromacs.org%5fgmx%2dusers%2dmrEvjEli4ZDCMOUXqa3p%2bTAV6s6igYVG%40public.gmane.org>
Assunto: Re: [gmx-users] No convergence with g_nmeig

Any idea, please.

Best

Esra KARAKAS

Dear gromacs users,

I am trying to diagonalize my Hessian matrix obtained from mdrun (normal mode
calulation) but I have this message and my vector 1 doesn't converge :

GROMACS:      gmx nmeig, VERSION 5.0.4 (double precision)
Executable:   /usr/local/gromacs-5.0.4/bin/gmx_mpi_d
Library dir:  /usr/local/gromacs-5.0.4/share/gromacs/top
Command line:
  g_nmeig_mpi_d -s nm.tpr -f nm.mtx -v eigenvec_TEST -of eigenfreq_TEST -ol
eigenval_TEST

Reading file nm.tpr, VERSION 5.0.4 (double precision)
Reading file nm.tpr, VERSION 5.0.4 (double precision)
Reading double precision matrix generated by Gromacs VERSION 5.0.4

Diagonalizing to find eigenvectors 1 through 50...
Calculation Ritz values and Lanczos vectors, max 10000000 iterations...
^MIteration    1:   0 out of  50 Ritz values converged.^MIteration    2:   0
out of  50 Ritz values converged.^MIteration    3:   0 out of  50 Ritz values
converged.....

I found a similar problem on the gmx-users but any response
https://mailman-1.sys.kth.se/pipermail/gromacs.org_gmx-users/2006-February/020179.html

Do you know why my vector 1 doesn't converge ?

Best regards,

Esra KARAKAS