CHANGES OF VERSION 5.0

 

1.  Ports to new machines or machine specific changes

2.  BUGS

3. INPUT RELATED CHANGES

4. DEFAULTS in input and throughout the code

5. OUTPUT related changes

6. CLEANING up

7. MINOR or relatively insignificant changes

8. ALGORITHMS: additions or changes  to existing functional ability

 

1. PORTS to new machines, re-ports, or machine specific changes

 

·        added new timing for rs6k. Works on sp3nersc. Others to be tested. – David Raczkowski

·        Re-ported to the SP3 at NPACI. – David Roundy

·        Fixed sysvars for origin2k. – David Raczkowski

·        updated port to origin200 and origin 2000 – David Roundy

·        ported to sp3 at NERSC – David Raczkowski

·        extended the allowed FFT grids for Alpha – Peter Haynes

 

2. BUGS

 

·        Fixed a bug in the relax_recycle_H feature.  Now it should work properly whether or not there is a CHECKPOINT file present.

·        Fixed a bug that made the pressure always relax to zero even when it should have been relaxing to a nonzero pressure.

·        fixed old bug in mixer to make vxc0 the same on all machines

·        fixed transpose problem in matinvert.

·        Fixed a bug in angular.f that seemed to only affect modi4, which was found by Soung Hoon.  Yay! This bug caused a crash on modi4 when calculating the angularly resolved density of states, but annoyingly didn’t affect other machines.

·        Fixed a memory leak in the calculation of the density of states. This probably was not a big problem, since it only leaked 8kB per processor per DOS calculation.

 

 

3. INPUT related changes

 

·        made it not crash running relax_recycle_H if there is no CHECKPOINT

file available.

·        Added a mini-change to the preprocessing of the adjustfft function so that it may be easier to support multiple fft options on the same platform.

·        added diagonlization_method  as a keyword in esdf_mod.f90

·        Added a feature to not include fractional translations, if you set number_symmetry_ops -2.

·        made it so if the lattice vectors are left-handed than the 2 and 3rf LV are switched. The coordinates of the atoms and the grid and shift the kpoints are also switched.

·        Inputread converted to F90 (Peter Haynes) with help of esdf of Chris Pickard

 

4. DEFAULTS in input and throughout the code

·        changed accuracy of fermilevel to give # electrons accurate to 10^-12.

more consistent with the uccupation cutoff of 10^-12  - David Raczkowski

·        made default polarization 0.2 from 0.0 – David Raczkowski

·        changed to allow up to 99 k-points and up to 999 bands (previously restricted to 9 and 99) in the wavefunction visualization – Peter Haynes

 

5. OUTPUT related changes

·        added html to makefile

·        Added output to check whether the Broyden mixing is actually changing

the potential.

·        for making dist now  prompts for a new version number, changes the version number and then tars up a distribution labelled by the new version number and gzips the tar file.

6. CLEANING up

·        removed most of the unused files in scilib.f. Only sscal was used and it now resides in the routine that calls it(symm_ident).

·        made all calls to a 3x3 inverse to be to matinert. Removed 2 instances of invers(). Matinvert is coded by hand to give the inverse and return the value of the  determinant.

·        Relax converted from C to F90 – Peter Haynes

 

7. MINOR or relatively insignificant changes

·        Got rid of oldfiles.

·        Fixed the estimated memory usage, although I’m still not sure it’s correct.Changes by DJR:

·        Major update to documentation.

 

8. ALGORITHMS: additions or changes  to existing functional ability

·        A new flag set to do more than one band fft at the time. – Michel Cote

·        modified the band structure calculation so the code allocates only one wavefunction since we do not need to keep the information.

·        enables a new option to 'occupy_levels' called 'excited' which just sets HOMO and LUMO occupations to 0.5 – Peter Haynes

·