scnc
/
abinit
mirror of https://github.com/abinit/abinit.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
abinit/doc/release_notes_obsolete/release_notes_v6.12.html

284 lines
14 KiB
HTML
Raw Permalink Blame History

<html>
<head>
<title>Release notes</title>
</head>
<body>
<h5>Copyright (C) 2010-2025 ABINIT group (XG)
<br> This file is distributed under the terms of the GNU General Public License, see
~abinit/COPYING or <a href="http://www.gnu.org/copyleft/gpl.txt">
http://www.gnu.org/copyleft/gpl.txt </a>.
<br> For the initials of contributors, see ~abinit/doc/developers/contributors.txt .
</h5>
<pre>
Many thanks to the contributors to the ABINIT project between
September 2011 and January 2012. These release notes
are relative to improvements of ABINITv6.12 with respect to v6.10.
The list of contributors includes :
F. Abreu, B. Amadon, A. Berger, J.-M. Beuken, F. Bottin, J. Bouchet, E. Bousquet,
F. Bruneval, D. Caliste, F. Dahm, H. Dammak, M. Delaveau, T. Deutsch,
C. Espejo, G.A. Franco, G. Geneste, L. Genovese, M. Giantomassi, Y. Gillet,
X. Gonze, A. Jacques, F. Jollet, G. Jomard, M. Marques,
S. Ponce, Y. Pouillon, A. Romero,
M. Stankovski, M. Torrent, M. Verstraete, D. Waroquiers, Bin Xu, J. Zwanziger .
It is worth to read carefully
all the modifications that are mentioned in the present file,
and examine the links to help files or test cases ...
This might take some time ...
Xavier
__________________________________________________________________
Version 6.12, released on January 30, 2012.
Changes with respect to version 6.10 :
A. WARNINGS.
B. Most noticeable achievements (for users)
C. Changes in the package, for developers
D. Other changes (or on-going developments, not finalized)
__________________________________________________________________
A. WARNINGS AND IMPORTANT REMARKS
A.1 As mentioned in B.2, the standard version of LibXC is now 1.1.0 instead of 1.0.0 .
There has been a few numerical changes, esp. in the PBE functional and the mGGAs ...
Also, functional number 104 is no more the Becke 86 Xalfa,beta,gamma (reoptimized)
but Herman Xalphabeta GGA.
__________________________________________________________________
B. Most noticeable achievements
B.1 The Ground-state calculations can benefit from GPUs, with CUDA library.
See the input variable use_gpu_cuda, as well as the tests contained in the tests/gpu
subdirectory.
This functionnality is still in beta stage and has to be used carefully.
Note that to take benefit of the GPU efficiency the computation load has to be large enough.
Abinit has to be compiled with the NVidia Cuda package (v3.2+) as well as the Magma library (http://icl.cs.utk.edu/magma/).
Relevant configure options: --enable-gpu, --with-gpu-flavor, --with-gpu-prefix, NVCC, NVCC_FLAGS
and --with-linalg-flavor (for Magma activation).
When Abinit has been compiled with Cuda and an available Graphics card is detected, ground-states calculations
are automatically done on the GPU. This can de deactivated by setting "use_gpu_cuda 0" in the input file.
Note that, when doing a GS calculation one GPU, some defaults are automatically changed: LOBPCG algorithm use
is forced (wfoptalg=14), Ylm version of non-local operator is selected, FFT algorithm is forced to be 401.
All these changes do not modify the final result of ground-state calculations.
(F. Dahm and M. Torrent, with some testing by JM Beuken)
B.2 The LibXC version 1.1.0 is now interfaced with ABINIT, instead of LibXC version 1.0.0 .
There are numerous new LDA/GGA functionals of which the portability has been tested.
New functionals that can be used with ABINIT (numbers refer to the index given in the description of the ixc input variable):
- 3 LDA (number 22,23,24)
- 9 X-GGA (number 139,140,141,142,144,145,149,183,184)
- 4 C-GGA (number 143, 147, 148, 186)
- 3 XC-GGA (number 146, 170, 171)
The documentation is up-to-date, and also include the documentation
of different functionals that were previously already available (125,126, 172->181)
See the automatic tests libxc#17,18,19 .
See also the warning A.1
(work by XG + YPouillon and JMBeuken, and of course, the LibXC team, esp. M. Marques)
B.3 Tutorials for parallelism (also Bethe-Salpether)
For the needs of the ABINIT hands-on tutorial for High-performance computing at the TGCC in Bruyeres-le-chatel,
November 2011, http://sites.google.com/site/abinit2011tutorial, a set of six new lessons for the ABINIT tutorial has been created :
ground state with plane waves, ground state with wavelets, molecular dynamics, parallelism on "images",
response functions, Many-Body Perturbation Theory .
They are available on the Web, as well as in the package, entry point doc/tutorial/welcome.html.
A new directory for testing the has been created : tests/tutoparal.
The old lesson on parallelism, obsolete, has been cleaned and is now up-to-date, and coherent with the other lessons.
The tutorials all demonstrate, for different capabilities of ABINIT, that a speed-up of more than one hundred
is achievable.
By the way, in the process of setting up the tutorials, tests and improvements of the scaling were made, for the
different functionalities.
There is also another new tutorial (sequential) on the use of the Bethe-Salpether equation to compute optical properties,
created for v6.12 : see doc/tutorial/lesson_bse.html
Also, the tutorial on GW has been considerably improved.
Work by M. Torrent, F. Jollet, B. Amadon, G. Geneste, J. Bouchet, G. Jomard, M. Stankovski, M. Giantomassi, F. Bruneval,
F. Bottin, D. Caliste, L. Genovese.
B.4 The latest version of the BigDFT library has been interfaced with ABINIT : switch from BigDFT 1.2 to BigDFT 1.6 .
This allowed to develop the tutorial on BigDFT in parallel.
Work by D. Caliste (and of course the whole BigDFT tesm, esp. L. Genovese and T. Deutsch).
__________________________________________________________________
C. Changes for the developers
C.1 All the "allocate/deallocate" instructions have been replaced by "ABI_ALLOCATE/ABI_DEALLOCATE"
(ABI_DATATYPE_(DE)ALLOCATE for pointers).
This project, lead by David Waroquiers, aims to allow memory control and monitoring.
C.2 Previously a set of blanks might have been introduced by abiauty, changing
!$OMP
to something like
! $OMP
depending on the alignment with other neighboring lines.
The abiauty script has been modified, so that the lines starting with
!$OMP
will now be unchanged by abiauty. This is the expected behaviour of abiauty for future developments inside
ABINIT, in order not to have abiauty destroying the OMP directives.
(By the way, comments can also be leaved unaffected by abiauty, if they start with !!OMP )
C.3 There have been many improvements for the test farm.
Among other developments :
- new "on_push" server to test quickly the public branch freshly committed
(already active but not for production
- new nightly builders :
- tikal with pathscale 4.0
- max2 with Open64
- ktulu with native deb packages ( gcc44, fftw3, atlas, openmpi,...)
- new "on_demand only" builders :
- buda_gcc45_checks to check quickly buildsys
- buda_gcc45_cuda to test GPU/CUDA
- shiva_gcc45_abiauty
- tikal_intel11_openmp to test OpenMP
- removed nightly builder : bigmac_gcc43
(JM Beuken)
C.4 Improvement/modifications of the build system :
- Switched from plugins to connectors / fallbacks.
- Added connector for Levenberg-Marquardt nonlinear least squares
algorithms (levmar).
- Added build-system support for flexible MPI compiler wrappers.
- Automated the triggering of defines and conditionals.
- Added build-system support for multiple shared-object file extensions.
- Added build-system support for the Magma library.
- Fully modularized the build system into 5 blocks: core, doc, fallbacks,
gui, and tests, to allow for asynchronous improvements on each block
while preserving the overall robustness.
- Rewrote makemake as a configurable Python script. Obsoleted the
clean-source-tree script.
- Fixed various consistency issues.
- Enhanced and refactored build-system outputs.
(Y. Pouillon)
C.5 The name of the routine is now defined automatically, thanks to the ABI_FUNC macro. The abilint script
automatically creates it.
(M. Giantomassi)
__________________________________________________________________
D. Other changes
(or on-going developments, not yet finalized).
D.1 Switch to atompaw 3.1.0.3
Switch to ETSF-IO 1.0.4
(YPouillon, XG)
D.2 Fox 4.1.0 has been replaced by Fox4.1.2 . It is now tested satisfactorily on more builders than previously : testf_gcc44,
testf_gcc44_serial, buda_gcc43-mpiio, green_g95, ibm_xlf12, inca_gcc44_sdebug, ktulu_gcc44,
littlebuda_gcc45_gpu, max2_gcc46, shiva_gcc45_distchk .
Seem NOT to work with Intel, Open64, Pathscale compilers, though.
(XG, YPouillon, JMBeuken).
D.3 The input variable bs_loband has been introduced.
See the tests Tv6#91,94 and tutorials Bethe-Salpether, cases 2 to 5, as well as
tutorial Many-Body Theory, test case 5.
(M Giantomassi)
D.4 More work on the Contour deformation approach to GW calculations (see section B.3 of the
release notes for version 6.10), with the introduction of the following input variables :
cd_use_pole_src, cd_reconstr_scr, cd_npoles (still to be documented and tests to be provided).
The new test Tv6#101 has been defined, to test the use of cd_customimfrqs, cd_imfrqs,
cd_full_grid, nfreqim, nfreqre, cd_halfway_freq, cd_max_freq, gw_use_freqre_tangrid
Enabling pole-fit in sigma calculations, using the Levenberg-Marquardt algorithm
(introduction of the levmar library).
(M. Stankovski)
D.5 Work on the electron-phonon coupling part of ABINIT, especially the tools to analyse its effects.
New input variables for ABINIT : fermie_nest.
New input variable for ANADDB : ep_extrael.
New tests : Tv6#103 and 104.
(Bin Xu, M. Verstraete)
D.6 The preconditioners defined by non-zero iprcel values (extrapolar-type) can now be used used safely
with PAW. See the input variable pawsushat, and the test paral#W .
(M. Torrent)
D.7 Computation of the chemical shielding, on-going development. Definition of the input variable
prtcs, not operational yet, though.
Orbital magnetization and Chern number can be computed, but at present, only in the nspinor=nsppol=1 case.
More possibilities (esp. related to spin/spinor) under development.
(J. Zwanziger).
D.8 To avoid numerical problems with some XC functionals (see point B.2), a new input varaible, xc_denpos,
has been introduced, to allow avoiding the density to vanish, and avoid spurious divergences.
See test Libxc#13 .
(XG)
D.9 Development of a new timing technique, based on the writing of _TIMER files.
(M. Giantomassi)
D.10 Work on the geometry optimisation preconditioner.
By G.A. Franco
D.11 Work on the GW part of ABINIT, to allow to use OpenMP to speed up the code and also reduce the memory needs.
(M. Giantomassi)
D.12 Decomposition of the square of phonon frequencies at each q-point in short-range and long-range contributions,
inside ANADDB. New input variable prtsrlr. Test Tv6#105.
(E. Bousquet)
D.13 Some rudimentary automatic test of the ABINIT GUI is currently being developed.
By F. Abreu, Y. Pouillon, JM Beuken, X Gonze.
D.14 Implementation inside of ANADDB of Gaal Nagy algorithm to refine the set of q points in the Brillouin Zone, in development.
Input variable : qrefine. See test v6#102. Not working yet ...
(M. Verstraete)
D.15 Work on automation of the distribution of the work load for the parallelism
(no need to explicitly specify the splitting of processors in npband / npfft / npspin / npkpt ...)
An automatic test should be available ...
By F. Jollet and M. Torrent.
D.16 Work on Shirley interpolation, to speed up the BSE code.
(Should provide documentation and test cases ...)
(M. Giantomassi)
D.17 On-going effort to provide vdW functionals for ABINIT.
(see D.7 of the 6.4 release notes)
Included Silvestrelli method to vdW (version 2), following Ambrosetti & Silvestrelli, http://arxiv.org/abs/1111.6737
Work by C. Espejo, Y. Pouillon, AH. Romero and X. Gonze.
D.18 A genetic algorithm to find new structures is under development.
(see D.13 of the 6.4 release notes)
By AH. Romero and X. Gonze
D.19 Phonon calculations using DFPT can be done reliably with the PAW methodology,
for the insulating case.
The metal case is operational (under the same conditions), except for q=0 (not fully implemented).
The electric field response was already available previously.
Still, for the time being, mixing of electric field and atomic displacement is not allowed.
Not yet possible to do calculations with spinors.
Not yet all possibilities that are present in PAW (e.g. pawxcdev is not allowed).
Work done by M. Torrent.
D.20 Electric field gradient computation with spin-orbit has been coded. See test v6#49, for the ICl molecule.
(J. Zwanziger)
D.21 The use of LDA+Dynamical Mean Field Theory is in progress.
(see D.6 of the 6.4 release notes)
By B. Amadon
D.22 Work on parallelisation/distribution on the atoms in the PAW case
By M. Delaveau, M. Torrent and F. Bottin.
D.23 Work on the computation of polarisation (Berry phase) and finite electric-field computations,
with spinorial wavefunctions.
(J. Zwanziger)
D.24 Numerous miscellaneous additional bug fixes (to the sources, as well as to the build system, including the plug-ins),
and improvements of documentation by
A. Berger, J.M. Beuken, F. Bottin, E. Bousquet, D. Caliste, H. Dammak, M. Delaveau,
G. Franco, Y. Gillet, A. Jacques, F. Jollet,
S. Poncé, A.H. Romero, M. Torrent, M. Verstraete, Bin Xu, J. Zwanziger, D. Waroquiers.
</pre>
Reference in New Issue View Git Blame Copy Permalink