envmc matrix inversion not converged

General discussion of the Cambridge quantum Monte Carlo code CASINO; how to install and setup; how to use it; what it does; applications.

envmc matrix inversion not converged

Postby Katharina Doblhoff » Tue Jul 11, 2017 2:59 pm

Dear Casino users!
I have a system for which I constantly get the message
Warning: [EMIN_MATRIX_ALGEBRA] Could not invert S (stage 1).

when performing an emin optimization of the Jastrow.

Looking at the code, this error is thrown if the divide and conquer algorithm in the singular value decomposition of the S matrix (I think it is eqn 226 in the manual) does not converge or if the input values are forbidden.

Does anybody have a good idea of what to DO against this (except for exchanging gesdd by gesdv in the code, which would - I guess - be slower but maybe more stable)?

Thanks,
Katharina
Katharina Doblhoff
 
Posts: 83
Joined: Tue Jun 17, 2014 6:50 am

Re: envmc matrix inversion not converged

Postby Katharina Doblhoff » Fri Aug 11, 2017 8:31 am

I solved the issue: It seems to be a problem with early lapack versions...
So in case anybody else stumbles into this issue: Try using newer libraries!
Katharina Doblhoff
 
Posts: 83
Joined: Tue Jun 17, 2014 6:50 am


Return to The CASINO program

Who is online

Users browsing this forum: No registered users and 2 guests

cron