The ABINIT Users Mailing List ( CLOSED )

[verstraete@pcpm.ucl.ac.be]

Hello,

unfortunately, we've also had the hermitian bug, but only on one specific
platform: a cluster of 2-processor intel boxes linked with Myrinet (cox on
the UCLouvain HPcomputing web site), and the calculations had to be
parallel.

My theory is either 1) a bug in GM-MPI from Myrinet, and a resulting
overwriting of the hamiltonian matrix, or 2) a bug in our MPI coding. 2)
would be surprising, since it should appear on other platforms, and an
identical run does not crash on other clusters. We haven't seen the bug
for a few months, but it definitely depends on your system.

So: what's your system and platform? Usually the breaking of hermiticity 
is much smaller, but if you increase the tolerance in hermit.f it just 
gets bigger.

I know, it doesn't help...

Matthieu


On Fri, 4 Apr 2003 priya-g@umail.ucsb.edu wrote:

> Hi All,
>   I was trying to do some band structure calculations. I did the SCF  total 
> energy calculations and then ran a non-SCf calculation for the band 
> structure. I got the following error in the output..
> 
>    cgwf: converged with tan2th=????????????????
>  cgwf: converged with tan2th=????????????????
>  cgwf: converged with tan2th=????????????????
>  cgwf: converged with tan2th=????????????????
>  cgwf: converged with tan2th=????????????????
>  cgwf: converged with tan2th=????????????????
>  cgwf: converged with tan2th=????????????????
>  cgwf: converged with tan2th=????????????????
>  cgwf: converged with tan2th=????????????????
>  cgwf: converged with tan2th=????????????????
>  cgwf: converged with tan2th=????????????????
>  cgwf: converged with tan2th=????????????????
>  cgwf: converged with tan2th=????????????????
> 
>  hermit: BUG -
>   Input Hermitian matrix has nonzero Im part on diagonal:
>   for component     1 Im part is  7.254468056006+257.
>   Action : contact ABINIT group.
> 
>  Delivered    9 WARNINGs and   0 COMMENTs to log file.
> 
>  leave_new : decision taken to exit ...
> 
>    --Priy
> 

-- 
===================================================================
Matthieu Verstraete               mailto:verstraete@pcpm.ucl.ac.be 
PCPM, Boltzmann, pl. Croix du Sud, 1            tel: 010/ 47 86 81
B-1348 Louvain-la-Neuve  Belgium                fax: 010/ 47 34 52