The ABINIT Users Mailing List ( CLOSED )

[Matteo Giantomassi <gmatteo@pcpm.ucl.ac.be>]

On Fri, 2007-06-22 at 17:58 +0200, snd2mit@yahoo.co.in wrote:

Dear Madan, 

which kind of parallelism are you using, over k-points i.e gwpara=1 
or over bands i.e gwpara=2?

> I come across two problems while running / testing the GW calculation
>  on Silicon.
> 
> System : IBM AIX 5.2 PowerPC Regetta
> Abinit Version : 5.3.4
> 
> 1. In parallel code, using 1 processor, it gave correct resutls
> 
> ---Output---
>  Perturbative Calculation
>  
>  k =   -0.125   0.000   0.000
>  Band     E0 <VxcLDA>   SigX SigC(E0)      Z dSigC/dE  Sig(E)    E-E0  
>      E
>     4   5.616 -11.116 -12.334   1.257   0.775  -0.290 -11.085   0.030  
>  5.646
>     5   8.357 -10.140  -5.951  -3.336   0.779  -0.284  -9.476   0.664  
>  9.021
> 
>  E^0_gap          2.741
>  E^GW_gap         3.375
>  DeltaE^GW_gap    0.634
> ----------------------------------------------------
> 
> 2. In parallel code using more than 1 processor, the it gave totally
>  wrong values.
> 
> -----Output-----
>  Perturbative Calculation
> 
>  k =   -0.125   0.000   0.000
>  Band     E0 <VxcLDA>   SigX SigC(E0)      Z dSigC/dE  Sig(E)    E-E0  
>      E
>     4   5.616 -11.116  -3.307   0.308   0.926  -0.080  -3.599   7.516
>   13.132
>     5   8.357 -10.140  -1.651  -0.857   0.927  -0.079  -3.068   7.072
>   15.429
> 
>  E^0_gap          2.741
>  E^GW_gap         2.297
>  DeltaE^GW_gap   -0.444
> 
> ---------------
> 
> 3. As per the suggestions given, I modified only the 'csigme.F90' in 
> 'src/15GW' and the error persists, but the values are different : 
> 
Sorry I don't understand: what did you modify in csigme?
By the way could you attach your input file in the next mail, so we 
can have a better understanding of the particular GW calculation you 
are performing.


Regards
Matteo Giantomassi
> ----------------------------
>  Perturbative Calculation
> 
>  k =   -0.125   0.000   0.000
>  Band     E0 <VxcLDA>   SigX SigC(E0)      Z dSigC/dE  Sig(E)    E-E0       
> E
>     4   5.616 -11.116  -5.675   0.311   0.884  -0.131  -6.030   5.086  
> 10.702
>     5   8.357 -10.140  -3.094  -1.389   0.886  -0.129  -5.130   5.010  
> 13.367
> 
>  E^0_gap          2.741
>  E^GW_gap         2.665
>  DeltaE^GW_gap   -0.076
> 
> ----------------------
> 
> What could have gone wrong ? The input file remains the same in both
> cases and I used 'abinip' to run them. 
> 
> Please help me.
> 
> Thanks
> 
> - Madan
> 
> This is how I compiled the code. If I do not use the flags "-O0", the
>  job crashes after few steps giving segmentation fault.
> 
>