11
|
|
Siesta-1.0.62 -- almost 1.1 (after PDOS, .ion files)
(CHANGES from 1.0.50 to 1.0.62 follow)
Version: 1.0.62 Date: 2001/09/24 12:00 GMT Title: Projected Density of States. Author: Javier Junquera <javier.junquera@uam.es>
Description:
The Total Density Of States (Total DOS), and its projection on the basis orbitals (Projected Density Of States, PDOS), are computed and dumped to files ('.TOT.DOS' and '.PDOS' respectively).
This only works in a serial run. An error message will be issued when we try to use it in a parallel execution.
It is only implemented when the hamiltonian is solved by diagonalization with k-points.
Files: siesta.F Call to pdos subroutine is added pdos.f New subroutine, interface between siesta and the one that computes the PDOS pdosk.f New subroutine. Computes the PDOS Makefile pdos.o and pdosk.o added to object list Docs/siesta.tex, Docs/siesta.ind Documentation of new block usage. ======================================================================== Version: 1.0.61 Date: 2001/09/16 18:00 GMT Title: New input/output facilities for ps and basis information Author: Alberto Garcia <wdpgaara@lg.ehu.es>
Description:
Pseudopotential information can now be read by SIESTA in formatted form from '.psf' files (which are now also generated by ATOM). This will make it easier to transport ps files among computers. The old *.psatom.data files are no longer valid.
Also, the PAO, KB projector, and Vlocal and charges data are dumped to '.ion' (and '.ion_nc' if NetCDF is available) files, which can later be used to re-generate the atomic information if the 'user-basis' (or 'user-basis-netcdf') option is specified in SIESTA. A header section containing information about the specifications for basis construction and about the pseudopotential used is also included.
Files: chemical.f New module to centralize the chemical species info. pseudopotential.f New routine to read formatted '.psf' files. New routine to print a header containing basic ps info. basis_io.F Modifications to the routines which read and write the '.ion' files. atm_types.f basis_specs.f basis_types.f atm_types.f initatom.f transfer.f Minor changes to support the above changes.
Pseudo/atom/wrapup.f New file VPSFMT output Pseudo/atom/Samples/pg.sh File VPSFMT copied to <name>.psf Pseudo/atom/Docs/User.Guide Updated
Docs/siesta.tex, Docs/siesta.ind Documentation of new file usage. ======================================================================== Version: 1.0.60 Date: 2001/09/11 17:00 GMT Title: Corrections and modification to limit runaway memory Author: Julian D. Gale <j.gale@ic.ac.uk>
Description:
This is a collection of corrections for a few bugs, typos, and problems noticed on certain platforms. It appears that on Linux boxes, under Absoft f90, and on DEC platforms the f90 implementation fails to properly free memory when deallocated. Because of the size of phi/listp2/lstpht, this can rapidly lead to the memory being exhausted on some machines. Hence a workaround has been added to PhiOnMesh so that these arrays are only redimensioned when absolutely necessary - i.e. they are too small. A tolerance has also been added to the initial dimension (of 1%) to hopefully minimise reallocations.
Files: atom.f Spelling of "analysis" corrected matel.f Dimensions of arrays adjusted to prevent crash when running in parallel (contraction prevented). Debug statements commented out. meshsubs.F Reallocation of phi/lstpht/listp2 avoided one possible to avoid memory increasing on platforms that fail to properly deallocate arrays from memory. mulliken.F Spelling of "analysis" corrected Loops over 1,Nodes corrected to 0,Nodes-1 redata.F Spelling of "analysis" corrected / alignment of write tidied siesta.F Call to ioeig corrected to pass no_u instead of no_l ======================================================================== Version: 1.0.59 Date: 2001/07/10 10:00 GMT Title: Changes to support pgf90 and non-MPI version Author: Alberto Garcia <wdpgaara@lg.ehu.es>
Description:
A. The pgf90 compiler had trouble compiling certain sections of the code. These had been rewritten slightly to help it.
Files: Src/fdf/parse.f: Use token_arr instead of tokens to avoid compiler bug. Src/fdf/fdf_mod.f, fdf.f : Use fdf_block_old wrapper routine to avoid compiler bug that led to an infinite loop. Src/alloc.F90: Fix declaration of character variable. Sys/pgf90*.make: Name NetCDF library explicitly, instead of using -L. B. The process to generate a de-MPI'ized version suitable for general users has been streamlined somewhat.
Files:
build_serial.sh: Use the same Makefile. Copy also .f90 files Src/Makefile: Test for existence of MPI directory before cleaning. ======================================================================= Version: 1.0.58 Date: 2001/06/29 12:00 GMT Title: removed local variables NSM and NMESHG in cellxc Author: Jose M. Soler <jose.soler@uam.es>
Description:
Variables NSM and NMESHG are now imported from module mesh and should not be redeclared locally.
Files: cellxc.F: removed local variables NSM and NMESHG ======================================================================= Version: 1.0.57 Date: 2001/05/18 20:15 GMT Title: serial version of cellxc.F and xc.f returned to f77 Author: Jose M. Soler <jose.soler@uam.es>
Description:
The serial part of cellxc and atomxc was returned to f77, to make it more transportable as a separated package. Required node and mesh arguments for parallel esecution are now obtained directly within cellxc, rather than passed as arguments.
Files: cellxc.F: serial part returned to f77 Node and Nodes obtained by calling MPI NMESHG and NSM obtained from module mesh dhscf.F: call to cellxc forhar.f: call to cellxc meshsubs.F: added static valiables NMESHG and NSM siesta.F: removed save declarations xc.f: returned to f77 ======================================================================= Version: 1.0.56 Date: 2001/05/18 20:00 GMT Title: Corrected bug in MATEL which produced numerical errors Author: Jose M. Soler <jose.soler@uam.es>
Description:
A change in the call to RADFFT was missing in basis_io.F and read_user_basis.F. Instead, RADFFT has been returned to its previous state, to avoid reproducing this error. This corrects the small numeric differences in the output introduced in version 1.0.50. Additionally, calls to realloc were simplified in MATEL.
Files: matel.f: simplified realloc procedures radfft.f: commented out return of spillage in last element ======================================================================= Version: 1.0.55 Date: 2001/05/11 Title: Grid cell sampling bug fix Author: Emilio Artacho <Emilio.Artacho@uam.es>
Description:
The energy was not being averaged by the grid-cell-sampling option, since the calculation of Etot (and Eharris and FreeEner) was done only in the before-last iteration, before grid-cell sampling. Fixed.
Files: siesta.F (customary version.F, CHANGES and siesta.tex) ======================================================================= Version: 1.0.54 Date: 2001/05/01 Title: Mesh memory correction Author: Julian Gale <j.gale@ic.ac.uk>
Description:
Out of bounds memory writes were occuring for a small number of arrays in routines relating to the mesh for some cases. This was related to the local variable "maxloc" being used in two different contexts with a common value. A new separate value has now been introduced for the case where a value which was too small was being assigned.
Files: dfscf.f : Sizing of ibc, C, gC and xgC corrected. rhoofd.f : Sizing of ilc and Clocal corrected. vmat.f : Sizing of ilc and Clocal corrected. ======================================================================= Version: 1.0.53 Date: 2001/04/26 Title: New inver Author: William Mattson <wmattson@uiuc.edu> (committed by JDG)
Description:
It turns out that the subroutine inver fails for certain matrices and returns "NaN"s. This was observed by William Mattson from Richard Martin's group and he has kindly provided a routine that fixes the problem.
Files: inver.f : Removed in favour of inver.f90 inver.f90 : Replaces inver.f ======================================================================= Version: 1.0.52 Date: 2001/03/27 08:45 GMT Title: Matel fixes Author: Julian D. Gale <j.gale@ic.ac.uk>
Description:
A number of corrections have been made, primarily to matel. The main change to matel is to explicitly zero INDF and INDFF when reallocated in case this is not the default action and also to trap when IOPER increases so as to increase the size of the appropriate arrays. A few other bugs have been cleared up for parallel execution.
Files: ksv.F : Dimension of ek changed to nuotot. matel.f : INDF / INDFF zeroed after realloc calls. : Test for INDF being associated added to avoid : call to SIZE when INDF is a null pointer. : Change in third dimension (IOPER) now trapped. parallel.f : New subroutine to set the default blocksize : added - checks that if the number of orbitals : is small then the blocksize is lowered to ensure : some work on each processor. siesta.F : Setting of level for alloc routines corrected : for parallel execution. : Call to new routine for setting default blocksize : added and calls for parallel setup moved until : later in the routine so that no_u is defined. ======================================================================= Version: 1.0.51 Date: 2001/02/28 05:00 GMT Title: Improved memory for parallel order N Author: Julian D. Gale <j.gale@ic.ac.uk>
Description:
The key routines for order N calculation have been modified to lower the peak memory usage by reordering certain operations to reduce the amount of local workspace at the expense of a communication overhead in some cases for multiprocessor systems. A few other changes have been made as a result of compilation and testing of the latest version on the Cray and under the pgf90 compiler.
Files: ener3.F : Memory usage reduced for parallel execution denmat.F : Memory usage reduced for parallel execution gradient.F : Memory usage reduced for parallel execution dhscf.f : Typo in message corrected fdf/fdf_mod.f : Local variable default2 added to fdf_sp to : handle type conversion on calling fdf_single periodic_table.f: lmax lowered to 3 in cnfig to avoid write beyond : bounds error when called by routines where lmax is : already 3 Makefile : atmparams.o added to object list for make of base ======================================================================= Version: 1.0.50 Date: 2001/01/31 17:30 GMT Title: Miscellaneous corrections Author: Jose M. Soler <jose.soler@uam.es>
Description:
More modular reading of .RHO files in plrho. Calls memory from realloc to keep the correct total allocated memory. Authorship correction of some periodic_table routines.
Files: alloc.F90 : Call memory. Corrected CRAY sizes. arw.F : Document moving of some routines iorho.F : Returns correct sizes instead of stopping matel.f : Last function value made zero after radfft call parallel.f : Default initialization of Blocksize and ProcessorY periodic_table.f: Authorship correction of some routines Plrho/iorho.f : Serial version of iorho Plrho/plrho.f : Use iorho to read .RHO files Plrho/plrho.f77 : Fortran77 version Plrho/plrho_guide.txt : Document previous changes =======================================================================
|
Alberto Garcia |
19 years ago
|
|
|