if the number of processors are less or equal the number of atoms,
nothing is changed... atoms are distributed to processors
if the number of processors is greater than the number of atoms
the same atom can be assigned to more than one processor so that
the work for each atom can be shared.
All this is required for scalability of PAW simulation
- OMP parallelization
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@5750 c92efa57-630b-4861-b058-cf58834340f0
causing misterious segmentation faults when called under reproducible but
apparently irrelevant conditions. The bug may be related to some misbehaviour
of derived types.
Changing the subroutine to "contained" (inside paw_xc_potential) seems to
work around the problem and have no collateral effects.
LP
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@5167 c92efa57-630b-4861-b058-cf58834340f0
strange problems with more recent versions of openmpi libraries.
Removed unnecessary debug prints from gener_pseudo
LP
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@5061 c92efa57-630b-4861-b058-cf58834340f0
LP
UPF file format updated completely, UPFv2 introduced:
* ld1.x can still produce old format, with the switch upf_v1_format=.true. in inputp
this is disabled by default, but we can discuss if it should be the opposite.
* pw.x cp.x and all utilities should notice no difference
* some utilities in upftools still need to be updated, anyway conversion UPFv1 to UPFv2
is very easy, so this should be no big issue
* starting from now to produce an UPF file you need to fill the pseudo_upf derivedd type
and feed it to write_upf woutine in upf_module (Modules/upf.f90)
* extensive use of iotk
I have tried to make the new format as self contained as possible, e.g. there should be
minimal need for post-processing after the data is read, no more reconstruction of known
quantities, and no more odd syntax to save negligible quantity of space. Also the human
readable section is a bit richer, all the rest is more machine readable.
I hope this will not cause any throuble, and tried really hard to, all examples and all
tests works as fine as before and gives (what really looks like) the same results.
Other changes that I needed to make:
* radial grids are now allocatable, they management is a bit less of a hack too
* paw and uspp augmentation are stored in the same place
* paw print total all-electron energy if all atoms are paw, not very useful, but nice
* most of the pseudopotential-writing reading files have been renamed to some more
logical name, I spare you the list. E.g. read_oldpseudo -> read_pseudo_rrkj3
* paw_t derived type was only used in atomic, so I have put it there (as the pseudo_type
module take ages to recompile it was awkward to leave it there).
PAW tests inserted in test/ there are 6 of them, as a consequence I have also put 5 paw
pseudopotentials in the pseudo/ directory.
I will update the PAW scf examples soon, by deleting them (as running a pw with a PAW
pseudopotential requires no option at all). PAW generation examples should be updated.
A lot of small bugfixes here & there mostly uninitialized variables or unallocated
pointers used as subrotuine arguments.
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@4769 c92efa57-630b-4861-b058-cf58834340f0
becsum is now saved on file, to allow restart with PAW; it still needs some testing.
As a consequence read/write of lda+U ns coeffs have been unified with becsum and rho.
Subroutines read_rho and write_rho from "io_rho_xml.f90" are now managed by an interface:
if param #1 is rho(nrxx,nspin) it will call the old subrt; if param #1 is of scf_type
it will save becsum and ns too in separate files (xml is not used, but can be added
easily now).
Fixed uninitialized variable in dynamics_module caused crashes in parallel MD.
Added error handling in paw part of read_upf.
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@4683 c92efa57-630b-4861-b058-cf58834340f0
Bugfix: PAW stress should work in parallel, even with pools
Bugfix: vc-bfgs should have a consistent convergence threshold (by SdG)
(its existence should be reported in INPUT_PW too!)
Note: I had to move some initialization from allocate_nlpot to a new function
pre_init that I've put in PW/setup.f90; it's called by setup and read_file
Note: the scf_mod part of PAW is now much better, becsum and rho%bec are NOT
the same any more (rho%bec is mixed and recollected among pools,
becsum is not).
LP
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@4657 c92efa57-630b-4861-b058-cf58834340f0
Symmetrization of becsum simplified a bit introducing mapping to "triangular" composite index.
Read upf now uses default (*) format for paw part.
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@4529 c92efa57-630b-4861-b058-cf58834340f0
global auxiliary variable saved%v removed and replace by a smaller local array.
ddd_PAW has now the same indexing structure as becsum.
miscellaneous:
- defined undefined variable ze2=0 in ascheqps
- ibnd_up/dw initialized in print_ks_energies.f90
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@4517 c92efa57-630b-4861-b058-cf58834340f0
Precision of dylmr2 increased a bit, maybe it's not necessary
mix_rho messed up a bit, included the possibility to normalize mixing matrix (disabled by default)
Added a currently unused function in paw_init to increase integration precision on the fly
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@4498 c92efa57-630b-4861-b058-cf58834340f0
for a bug, before I realized that there were two copies of "gamma_only",
I wasted two more hours to iuse a single value of "gamma_only" (the one
oin control_flags).
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@4468 c92efa57-630b-4861-b058-cf58834340f0
Mods: removed read_paw module and .PAW temporary format support. Makefiles and
make.depend updated accordingly.
Fixes in ld1: test of upf types was not working anymore (same as reason as CP),
added deallocation of pawsetup. As testing of PAW without generation may not
work yet I have put a big error for it.
Other: simpson_cp90 was pretending to be simpson in error messages
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@4461 c92efa57-630b-4861-b058-cf58834340f0
routines, actually results should be the same, and execution time should be
a lot shorter.
* each atom type now integrated to a different max(l)
* max(l) tuned in order to exclude the non local channel of pseudo-
potential, if it's empty
* several implementation optimization
* all occurences of ALLOCATABLE changed to POINTER inside derived types
* added original Bloechl shape for augmentation charge, at the moment it
is not useful but if someone wants to implement analylitical correction
to overlaps...
* PAW now uses UPF format, the format is still being tested: it will
probably change a bit in the near future
* as a consequence radial grids are now pointed from upf type and can be
updated directly in read_upf, upf_to_internal don't have to set the grid
anymore (at the moment only for pseudos read from upf files)
* added a non-destructive check for abs(zeta)>1. in vxc_t
LP
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@4449 c92efa57-630b-4861-b058-cf58834340f0
cessary for PAW, the latter has been moved to GIPAW/gipaw_module.f90, as it is only used there.
Added a little trick to free some memory after the initialization is completed (only for parallel .and. PAW, but similar method could be
used for US too; is it worth the hassle?)
Minor changes in paw_symmetrize (added parallel) and paw_newd.
LP
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@4431 c92efa57-630b-4861-b058-cf58834340f0
This is the first step of code cleanup. Actually nothing should change.
rad_paw_routines -> paw_onecenter (2 functions went to paw_init)
grid_paw_variables -> paw_variables (with some additions from rad_paw_routines)
grid_paw_routines -> paw_init (merged with 2 init functions from rad_paw)
Probably there are a few random fixes around.
Next step: remove ndmx and ntpx.
LP
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@4422 c92efa57-630b-4861-b058-cf58834340f0
ccavazzoni