[Pw_forum] difficulties with 'ortho'

[Silviu Zilberman]

Hi Paolo,

In the example I posted, it fails in 'ortho_gamma' at the initial part, 
right after setting rho, before starting any iteration. It seems that 
subroutine 'rhoset' does not work very well and there must be NaN's in 
it's outcome (rho). As a result it fails in 'ortho_gamma' at the first 
NaN trap.

Thanks, Silviu.

Paolo Giannozzi wrote:
> On Tuesday 21 March 2006 14:06, Silviu Zilberman wrote:
>
>   
>> I have been testing the cvs version on an IBM Blue Gene machine.
>> In some cases everything works fine, but in many other I keep 
>> having difficulties with the 'ortho' procedures in CP [...] Any ideas?
>>     
>
> support for IBM Blue Gene is very preliminary. Please try to find out 
> when exactly ortho doesn't work and why. It is an iterative procedure 
> and it is quite easy to see if it is converging or "going bananas".
>
>   
>> In addition, did anyone get PW (cvs version) to work on AIX?
>>     
>
> everybody who tried, I guess. I for sure did. Parallel subspace
> diagonalization in the Gamma-only case doesn't work on SP5 
> and has been disabled.
>
> Paolo
>