Merge new functionality in Nick's branch
Merge of new objects, netcdf, openmp, mixing, etc
Blueprint information
- Status:
- Complete
- Approver:
- Siesta Maintainers
- Priority:
- High
- Drafter:
- Nick Papior
- Direction:
- Approved
- Assignee:
- Nick Papior
- Definition:
- Approved
- Series goal:
- Accepted for trunk
- Implementation:
- Implemented
- Milestone target:
- None
- Started by
- Nick Papior
- Completed by
- Nick Papior
Whiteboard
- Fully re-implemented transiesta
* N-terminal support in transiesta
* Added different mechanisms for integration control of the
contours in transiesta
* Made the convergence of transiesta much better.
This is accommodated by better handling of the Hartree potential.
* Different ways of handling charge-reductions in SCF is enabled
* All electrodes have settings _per electrode_ for full customization
* Greatly reduced memory usage
* Does not use xij as phases, enables skewed axis
* Implemented MUMPS method for inversion
* Implemented LAPACK for inversion
* Implemented BTD method for inversion (extremely fast)
* Implemented two different ways of calculation Gf.G.Gf MM
* 12 different weighting schemes are available
* Many different pivoting routines available
* OpenMP threaded
* Start directly from transiesta enabled
* Temperature gradients as non-equilibrium a possibility
- tbtrans fully reimplemented
* EXTREME SCALE version (BTD-only)
- Memory consumption _only_ dependent on "device" region
* N-electrode support
* region projections for transmission through "eigenstates"
* Custom change of the self-energies
* Enabled tight-binding using MPI-parallelised tbtrans (easily)
* k -> k' transmissions
* Interpolation of bias Hamiltonians
* Bond-currents
* DOS and bulk transmission for electrodes
* Gf-DOS/spectral-DOS for device region
- Special considerations for siesta
* New and more robust mixing schemes (Pulay/
* Constraints are verbose and many new ways of using constraints exists
* Added complete NetCDF4 file format for siesta (parallel IO)
* LUA scripting for MD simulations
- can also be used for regular constraints
* Threading for a few routines in siesta
* Phonon calculations now produces TSHS files more conveniently
* Enabled gate-calculations (also for transiesta)
- Charge gate
- Hartree gate
* Extended class objects for reference counting objects
* Allowed fdf-flags on the command line:
-fdf <flag>:
only possible for non-blocks.
* Changed default DM.MixSCF1 to .true.
In many cases this is preferred, especially for restarts.
* Will die if there are more processors than orbitals
- Utilities
* All make-files are now made to enable parallel builds
- this makes compilation *MUCH* faster:
make -j4
will compile using 4 cores.
* Grimme utility
- easy creation of FDF block for Grimme parameters
* Sp2METIS, convert a sparsity pattern to METIS pivoting
* SpPivot, pivot sparsity pattern using several different routines
* TS/** different utilities related to transiesta
[Alberto]
We have to review the issue of the new supercell definition.
Yes, this is important.
[Alberto]
I guess it is not possible to partition the merge into smaller pieces...
One could, but that would mean a lot of work, I think we should discuss this in case you want to wait with certain features.
Work Items
Dependency tree
* Blueprints in grey have been implemented.