28/11/2010
Running tests of SGW that do full Sternheimer calculation of Screened Coulomb interaction:
some notes on scaling:
with 11 imaginary frequencies:
    GWHS         :     8h 4m CPU time
with 7 imaginary frequencies:
    GWHS         :     5h28m CPU time
basically scales linearly with frequency as expected.
There is no Plasmon peak showing up in the spectral functions when the q-grid is only 4x4x4. could be the coarse q-mesh? Switching to 5x5x5 sampling which might push this up to a 16-20 hour job. We will see...

30/11/2010 
For a 5x5x5 grid on 16 processors  the total time was:
    GWHS         :    17h47m CPU time
5x5x5 q mesh did indeed resurrect the plasmon peak partially. This could be more of that long wavelength excitation business. Also the ImSigma and ReSigma energies exhibit much more structure and appear closer to what is in Feliciano's paper. It shouldn't be too surprising that a more accurate calculation produces more accurate results but it's nice that things are doing what I expected them to. 

03/12/2010 
Cleaned up the examples directory for fulltests in branches. Now the only directory corresponds to the 555 qmesh run on silicon. I also am doing a run with the Cohen Bergstresser parameters for Ge so that I have more things to compare. Should look at sorting out the parallel implementation of SGW. 

Also I've opened a GWTC branch to start recoding the whole DFPT-GW approach from scratch making explicit use of all (or as much as possible) of Quantum Espresso's functions: symmetry reduction, pseudopotentials, optimized parallelisation, etc. First step is to run some model jobs with PW.x and PH.x to get a hang of those. Also the output file in quantum espresso provides a list of all the functions called during the calculation which will speed things up considerably I think.