Recent SeqQuest code version history
- Version 2.61k (Mar08-Jun10)
- Automated optimization of net spin polarization
- Molecular dynamics NVT thermostats matured, more user-controls
- Some improved grid-routine parallelism
- Bugfixes/patches:(a) 18Apr08 – bugfix: void at origin with floating orbitals (born: old)- with sizable void at cell origin with floaters, grid errors(b) 2May08 – parallel version consistency bug (born: 2.61)- parallel multi-k metallic population code would crash(c) 17May08 – parallel spin memory allocation error(born: 2.61)- parallel spin calculation would give (obviously) nonsense results(d) 24May08 – tweak memory allocations to enable large spin calcs- adjusted pointers to enable larger spin calcs (esp. parallel)(e) 2Jun08 – fix to patch (a) (born 2.61a)- if triggered, caused code to stop with improperly set flag(f) 28Aug08 – another patch to (a), and ASD tune- if triggered, real solver failed (leading to failed SCF);
and ASD initial atomic masses forced to be one (not atomic mass)(g) 4Sep08 – error fixed: GGA with left-handed cell vectors (born 2.55a)- with left-handed cell vectors, GGA calculation would fail
(previously fixed in 2.53h, bug reinserted 2.55a)(h) 23Sep08 – error fixed: grid memory assembly (born 2.61c)- if triggered, code asked for too much memory; else, no effect(i) 30Sep08 – error fixed: pop uninitialized variable (born 2.61)- if triggered, code would stop with ER/pop error; else, no effect(i) 26Jan09 – error fixed: “do post” crash on select platforms- compiler/platform-dependent, if triggered crashed with seg-fault(j) 26Jan09-PAS: bugfix: spsteps input bug; rhosave-as-subroutine- minor cosmetics(k) 29May10-PAS: bugfix: dosdump, lcao.post file format for prop1es- various repairs to harmonize with updated prop1e code
- Version 2.60f (Sep07-Mar08)
- Expanded task-parallelism
- AM05 functional (spinless only) – AEM
- Molecular Dynamics matured – RPM
- parallel-parallel NEB (multi-processor per image) – APT
- test suites – RPM
- Bugfixes/patches:(a) 2Oct07-PAS – Full Sandia copyright notice inserted- no change in code behavior(b) 16Nov07-PAS – parallel NEB error fixed (born 2.60)- crashed input of parallel-NEB with seg-fault(c) 10Dec07-PAS – portability issues- some compiler/OS portability problems fixed(d) 4Jan08-PAS – parallel-parallel NEB problem- parallel eigensolves within parallel NEB would crash(e) 28Jan08-PAS – parallel file access error- if triggered, would cause crash of parallel GGA(f) 12Feb08-PAS – index error in nearby function- if triggered, very small error in energy on grid
- Version 2.59b (Jul06)
- NEB updates: features, analysis, documentation – RMV
- Linear dependence handling: singular value decomposition – RPM
- Finite Defect Supercell Model – LMCC, sampling, bulk polarization
- Task-parallel version – faster, but not bigger
- Bugfixes/patches:(a) 17Jul06: tp/pgrid0xc – uninitialized variable (born 2.59/tp)Unexpected seg-fault; downstream impact possible but not seen(b) 9Apr07: dosdump – update for consistency with post-procNo effect in main code; installing Art Edwards’ post-proc updates
- Version 2.58b (Feb05)
- Molecular dynamics – RPM: NVE, Berendsen, scaled temp, Hoover, …
MD capability is “beta” code:
input/output subject to change; Hoover code not fully verified - Some improved internal exception/error handling
- Bugfixes/patches:(a) 2Dec04: phantom GGA stress term added for better grid convergence of stress calculationcomputed GGA stress will change, giving better convergence with respect to grid dimensions(a) 2Dec04: atom proximity diagnostic check fixed- thanks to A.C. Pineda for fixing this error(b) 10Feb05: completion status for mpi-NEB fixed- uninitialized variable could cause crash
- Molecular dynamics – RPM: NVE, Berendsen, scaled temp, Hoover, …
- Version 2.57d (Mar04)
- Spin-polarized GGA stress implemented (AEM)
- External applied electric field (with help of JTK)
- Jost model bulk polarization added to bulk charged defect
- Step limits (both SCF and geometry) removed
- Default Broyden history length set to 10 (SCF and geometry).
– trajectories longer than 10 steps will change - Bugfixes/patches:(a) 5Feb04: Broyden file+limit history bug fixed (from 2.55)- for larger number of steps or larger problem, code would unexpectedly crash in blend, typically with end-of-file error(b) 16Feb04: Reduced-symmetry for grid fixed (original)- code exited rather than properly reducing symmetry(c) 6Mar04: Gap diagnostic in complex calc fixed (from 2.49)- if triggered, code crashed OR reported (incorrect) zero gap(d) 18Mar04: History file reported incorrect cell-opt energy- no downstream impact
- Version 2.56 (Aug03)
- Internal handling of k-point sampling modified
(very) small changes in numerics result - Spin-polarization added to BLYP (AEM)
- Timing and big-I/O modules polished (with new output diagnostics)
default I/O record lengths were changed - Cosmetic changes to NEB output
- Internal handling of k-point sampling modified
- Version 2.55a (Aug03)
- Input format change:
Setup data section now begins with “setup data”, changed from “input data” - Limited-history, limited-file-size built into Broyden.
- Limited-history BFGS minimizer installed.
- Geometry history output for post-processing (e.g. MOLDEN).
- Altered makefile (simplified machine dependence in “utl”)
- BLYP functional installed (Ann E. Mattsson)
- Input format change:
- Version 2.54e (Jan03) – NEB transition state finder
- NEB transition state finder implemented
- Trivial MPI by NEB image implemented, Cplant-compatible
- Small optimization of setup-phase non-local integrals
- Version 2.53h (Oct02) – cell optimization
- (g) **** error fixed: GGA with left-handed cell vectors
- cell optimization
- GGA stress completed
- cutoff input moved to run phase (from setup phase)