Last Modified 18 Dec. 1998
DoD Tight-Binding
Software Trouble Reports
| Title: | Ambiguity in Documentation of static
code
|
| STR# |
CCM3-97-0001
|
| Date Detected: |
16 Dec 1997
|
| Date Closed: |
5 March 1998
|
| Priority: |
3
|
| Category: |
e
|
| When Discovered (Phase): |
Pre-Alpha
|
| Notes: |
The documentation for Example
1 was unclear about what was meant by the phrase "total number
of electrons". The documentation was modified to indicate that this
means the total number of electrons in the unit cell.
|
| Title: | Exit error in search2 routine of static
|
| STR# |
CCM3-98-0001
|
| Date Detected: |
5 March 1998
|
| Date Closed: |
6 March 1998
|
| Priority: |
2
|
| Category: |
e
|
| When Discovered (Phase): |
Prior to Alpha test
|
| Notes: |
The routine search2.f in static is to cause the program to
shut down when the number of pairs found is larger than the
dimension of the pair array. When ported to MPI, however, this
shutdown did not properly close the MPI part of the run. This has
now been changed so that search2 indicates to the calling program
that it has detected an error and the program should halt execution
properly.
|
| Title: |
Misplaced close statement in scalable static
routine.
|
| STR# |
CCM3-98-004
|
| Date Detected: |
11 March 1998
|
| Date Closed: |
11 March 1998
|
| Priority: |
2
|
| Category: |
e
|
| When Discovered (Phase): |
Prior to Alpha test.
|
| Notes: |
When collecting eigenvalues for use in band structure or Density of
states calculations, static writes results onto a temporary
binary file, one per processor. The code had been written in such a
way that these files were not guaranteed to be closed when the
master processor started reading them for final output. The code
was changed so the files are closed just before the last collective
MPI operation, guaranteeing that the files are closed before the
master processor can operate on them.
|
| Title: | Failure of
angular momentum decomposition in static's
QLMT generation.
|
| STR# |
CCM3-98-0004
|
| Date Detected: |
20 May 1998
|
| Date Closed: |
20 May 1998
|
| Priority: |
1
|
| Category: |
e
|
| When Discovered (Phase): |
After Alpha test, in preparation for Beta test
|
| Notes: |
On the IBM RS6000 the angular momentum decomposition of the
eigenstates occasionally gave NANQ values. On a Pentium Pro running
Linux/Absoft Fortran the program would halt at that point with a
Floating Point Exception (FPE). The problem was traced to the
"diag.f" subroutine. In that routine the overlap matrix is copied,
then the original array is destroyed in the diagonalization. The
copy is then used in the angular momentum decomposition. The error
occured when the copy was used in the diagonalization rather than
the original. The copy was then filled with nonsense numbers,
causing incorrect results for the angular momentum decomposition.
The NANQ and FPEs occured when this junk copy of the overlap matrix
produced a negative eigenvalue. This condition was not checked
because negative overlap matrix eigenvalues are caught in the
diagonalization phase. The problem has now been corrected in both
the 0.91 and 1.05 versions of static.
|
| Title: | Badly Defined Fraction in static
Version 1.05 setlat.f routine
|
| STR# |
CCM3-98-0005
|
| Date Detected: |
27 May 1998
|
| Date Closed: |
27 May 1998
|
| Priority: |
1
|
| Category: |
e
|
| When Discovered (Phase): |
After Alpha test, in preparation for Beta test
|
| Notes: | The
parameter variable "two3rd" was found to be defined as 1/3 rather
than 2/3. This only affected results in the mode 4 strain of a
lattice, leading to a non-volume conserving strain and an incorrect
c/a ratio. The problem was fixed by setting two3rd to 2/3.
|
| Title: |
Error in Birch fitting routines
|
| STR# |
CCM3-98-006
|
| Date Detected: |
9 June 1998
|
| Date Closed: |
9 June 1998
|
| Priority: |
1
|
| Category: |
Documentation
|
| When Discovered
(Phase): |
After Alpha test, during CCM workshop
|
| Notes: |
Prefix of the cubic term in the Birch fit routine used in examples
1, 3, and 10 was missing a factor of 1/2. As a result, evaluated
values of Bo'-4 were too small by a factor of 2. Fitting scripts
and examples in manual were fixed.
|
| Title: | Documentation Error in static
manual |
| STR# |
CCM3-98-0007
|
| Date Detected: |
10 Nov 1998
|
| Date Closed: |
10 Nov 1998
|
| Priority: |
1
|
| Category: |
d
|
| When Discovered (Phase): |
Prior to beta test
|
| Notes: |
Found that strain type 6 was misdefined on the strain
types page, in the hcp
C44 upper bound example, and in the hcp
relaxed C44 example, with e3 having the
wrong sign. The code in setlat.f
has it right. Fixed the manual to match the code.
|
| Title: | Documentation Error in static
manual |
| STR# |
CCM3-98-0008
|
| Date Detected: |
12 Nov 1998
|
| Date Closed: |
12 Nov 1998
|
| Priority: |
1
|
| Category: |
d
|
| When Discovered (Phase): |
Prior to beta test
|
| Notes: |
Found that the primitive lattice vectors a2 and
a3 defined in equation (3) of both
the unrelaxed and
relaxed C44 calculations had extraneous "-" signs.
Documentation has been fixed.
|
| Title: |
Lack of Dimension Checking in the rdspgp.f routine of static.
|
| STR# |
CCM3-98-0008
|
| Date Detected: |
18 Dec 1998
|
| Date Closed: |
18 Dec 1998
|
| Priority: |
1
|
| Category: |
e
|
| When Discovered (Phase): |
Prior to beta test
|
| Notes: | The routine
rdspgp.f lacked a check to insure that the number of spacegroup
operations, NOP, was no larger than the allowed dimension, NOPD.
The check was added to version 1.06 of
static. |
Send trouble reports to Michael Mehl
(mehl@dave.nrl.navy.mil) (Privacy Advisory)
This page is maintained by
Michael Mehl,
e-mail mehl@dave.nrl.navy.mil (Privacy Advisory)
Link to DoD Tight-Binding Home Page.