***** TEST INPUT DECKS / OUTPUT FILES *****

  Several test input and output decks are included and these are summarized
below.  For test5.* it is necessary to use a VSTAR potential routine supplied
in file test5.f; this must be "linked" before the compiled code in order
to replace the dummy VSTAR in v12.f. Note that test6 uses the VRTP
potential mechanism and is designed to work with the VRTP routine supplied
in v12.f.  A summary of the test input decks is listed below; these are files 
testN.input; where N=1,2,3,5,6,8 (except test3.v12.input).
Most of these run several 'subcases'. Resulting 
output can be found in files testN.v12.out.
  
                potential              ITYPE          INTFLG
             ------------------------  ------       -----------
test1
  a          std. powers/exponentials    1(CC)         4
  b                   "                 21(CS)         2
  c                   "                101(IOS)        6

test2
  a          std. powers/exponentials   22(CS)         8 
  b                   "                102(IOS)        8

test3
  a          std. powers/exponentials   23(IDENT=1)    3
  b          std. powers/exponentials   23(IDENT=0)    8
(Note: in developing version 14 it was discovered that the 
 potential used in test3.v12.input is ill-conditioned at 
 short range and a newer test case is now implemented.)

test5
             VSTAR must be supplied     25(CS)         6

test6
  a          std. powers/exponentials   26(CS)         6
  b          projected "internal" VRTP  26(CS)         6
  c          std. powers/exponentials    6(CC)         7
  d          std. powers/exponentials  106(IOS)       -1
  e          projected "internal: VRTP 106(IOS)       -1

test8        std. powers/exponenetials   8             2
          
                     ***** DOCUMENTATION *****

   The documentation is in file version_12.doc.  Documentation
for version 13 is in version_13x.doc; this version has more efficient code
for itype=2 cases, some extended capabilites in POTENL to allow for
automatic projection of the angular components of a potential supplied
via the VRTP mechanism (see Section 4.2 of the documentation), and various
other upgrades.  Those interested in this version should contact S. Green
at agxsg@nasagiss.giss.nasa.gov.

                      ***** SOURCE CODE *****

  The codes in this directory are running on an IBM RS/6000 workstation.  
Since MOLSCAT is very close to standard FORTRAN 77, the same code should
give little trouble on other machines.  For the CRAY, however, a 'single
precision' version should be used.

  MOLSCAT, version 12 FORTRAN source is in v12.f. (NOTE:  The code 
in v12.f differs from the "standard" distribution in the OPEN
statements.  The code here does NOT have FILE='fname' parameters so that
AIX will read/write to files fort.XX in the current directory, where XX
is the unit number.  The "standard" distribution code is retained as 
comment cards and could be readily revived.  Cf. Section 10.8 of the 
documentation.)

  Routines for the IBM RS/6000 workstations which should give cpu and wall 
clock time/date are in timers.f and timers_c.c; the latter is a C routine.  
The timers must be "linked" before the v12 compiled deck in order to replace 
"dummies" in v12.f.  

   The LAPACK and BLAS routines called by MOLSCAT are NOT included in v12.f.
As discussed in Section 10.7 of the documentation, it is best to use code 
optimized for each platform.  The BLAS routines are available on most IBM 
RS/6000 machines by using "xlf -lblas". If needed, FORTRAN versions of the 
BLAS routines used by MOLSCAT are in dblas.f.  FORTRAN versions of the LAPACK 
routines used by MOSLCAT are in lapack.f.  Unix "compressed" versions of
these are available as dblas.f.Z and lapack.f.Z.

  File syminv.f has a simple FORTRAN routine that is often found to be faster 
than the LAPACK routines called by SYMINV supplied in v12.f; it can be 
compiled and "linked" before v12 object deck in order to replace the
v12.f version (see Section 10.7 of the documentaion).  Similarly 
diag_eispack.f has FORTRAN source for matrix diagonalization based on the
EISPACK library which might be faster than the LAPACK routines.