Molpro version 2002.3 release notes
===================================

This version contains mostly bug fixes for version 2002.1.

29 April 2002


Detailed patch history of preceding version:

11 Jan 2002 bench.2002.2
Tweak procedures for benchmarks

13 Jan 2002 configure_tmpdir.2002.2
Trap for invalid TMPDIR corrected.

13 Jan 2002 newconfig2.2002.2
Initial implementation of generating different executables in the same Molpro tree.
Usage:
1.) run configure for each mpp case you want (presently tcgmsg and mpi are possible). Of
course only if you make an mpp version.
2.) run configure for each processor type you want (presently p3, p4, athlon are possible)
The last configure will hold the default values, used by make. But all libraries from
previous runs of configure are remembered (unless CONFIG is deleted).
3.) run make. Without any arguments it produces molpro.exe and molpro.rc as before,
using the last configure.
Optionally, one can use
make ARCH=arch MPPLIB=mpplib, where arch can be p3, p4, or athlon and mpplib can be tcgmsg or mpi.
Of course, configure must have done with the corresponding options, otherwise wrong libs might
be used. This will then produce molpro_arch_mpplib.exe and molpro_arch_mpplib.rc. The latter
is copied to molpro.rc as well. Eventually, molpro should allow to use different molpro.rc files.
Bug: executable is always remade when make is done with the same options. My knowledge of
make is limited, and I don't know how to fix (I tried, but does not work).

13 Jan 2002 make_veryclean.2002.2
make veryclean destroyed testjobs/*.out.gz. This patch prevents this.

13 Jan 2002 runbench_shell.2002.2
Make runbench try to run itself using an advanced shell, eg bash or ksh, if
these can be found

14 Jan 2002 linux_licence.2002.2
Fix (?) bugs in linux licencser.
Id: getinfo.c,v 2001.16 2001/11/15 10:24:00 peterk Exp 
Id: licence_check.c,v 2002.2 2002/01/14 08:26:00 peterk Exp 
Id: strcmpp.c,v 2001.1 2001/05/01 07:43:32 peterk Exp 

14 Jan 2002 molver_sun.2002.2
Work around funny behaviour of Sun tr command in non-default LOCALE

14 Jan 2002 normal_dft.2002.2
Get rid of double call to int in bench/normal_dft.com

14 Jan 2002 patcher.2002.2
Fix malloc error.

14 Jan 2002 poisson.2002.2
Fixes illegal jump in poisson

14 Jan 2002 sunconf3.2002.2
Changes -O to -fast in sun.conf
Removes ucb libs from sun.conf

14 Jan 2002 eom.2002.2
fixes incorrect jumps in cceom.f and dolresi.f

14 Jan 2002 sunc2.2002.2
Disables temporary sun sections in patcher.c and timing_molpro.c

15 Jan 2002 patcher2.2002.2
Bug(s) fixed: 203
scandir on Sun now returns 0 instead of -1 for an empty directory.

15 Jan 2002 dftfun.2002.2
Bug(s) fixed: 200
Fix syntax errors unmasked by Fujitsu compiler

15 Jan 2002 mrcibench.2002.2
Only single memory statement in normal_mrci

15 Jan 2002 bench_views.2002.2
Remove view blas from bench/views.

15 Jan 2002 bib2.2002.2
Another go at the Molpro citation

15 Jan 2002 accw.2002.2
Fixed buffer size in accw to awoid crash on sun.

16 Jan 2002 newconfig3.2002.2
Some fixes and cleaning-up of configure.
Batch not yet tested!

18 Jan 2002 cfit_norm.2002.2
Adds the option to constrain densities in density fitting to have the correct total charge.

18 Jan 2002 cfit_jab.2002.2
Adds a missing 'd0' to the end of a constant in the old cfit code; all previous Coulomb fitting results will be slightly wrong, since this constant was used in the single-centre (s|s) special case. This code is still used in the Coulomb fitting code, though not for much longer.

20 Jan 2002 eom_all4.2002.2
All recent (end 2001) changes to eom directory

22 Jan 2002 ccsd_mpp4.2002.2
For shared file implementation, do ccmp2 only on processor 0.
Additional timing info for kext and hfma.
Variable CCSD_IMPL for ccsd scratch implementation.
Activate mpptune,ccsd_eaf (if 1, force df in all cases,
if 0, force df unless default is sf, if -1 use default)
mpptune,sfreadp=1 enables parallel reading of sf files
sfreadp=0 (default) reads on proc. 0 and brdcst

23 Jan 2002 blastest5.2002.2
Corrects blastest.f for mpp. Implements separate tests for blas1, blas2, blas3
Modifies blasbench to perform tests for blas1, blas2, blas3, and mxma_mpp.

25 Jan 2002 for.2002.2
Reflect the demise of tcpc3 in the destination for depositpatch

27 Jan 2002 geminal2.2002.2
Elininates some huge arrays from common cgem

30 Jan 2002 newconfig4.2002.2
Some hack to GNUmakefile

31 Jan 2002 nightmare.2002.2
1) fix in SR rspdoms (opdom.f).
rsp domain of a strong pair domain [kl] relative to 3ext domain
[i], where [kl] is only partly contained within [i] (and which will never
be used) caused overwriting of previous rsp domain specification.
Severity: very high, all lccsd and l(t) calculations using i3ext=1
carried out so far, are potentially wrong (probably only large molecules
and/or molecules with high domain fragmentation were affected)
Solution: Explicitely drop strong pair domains that are only partly
contained in a 3ext domain

1 Feb 2002 locfiles2.2002.2
Removes unnecessary reserves in lmp2 case and reduces record
lengths for nkcp/nkcpi in lccsd case (only strong pairs).
ccdiis in lmp2 only on processor 0.

2 Feb 2002 luqcisd3.2002.2
Implments luqcisd, lucisd
option iopdom=0,1 for luqcisd, lucisd, luccsd

2 Feb 2002 kext.2002.2
Fixes fujitsu case in kext for mpp

3 Feb 2002 hess_cpmcscf.2002.2
Fixes incorrect (huge) reserve in hess_cpmcscf.f and reserves record 8000 in hess_pople.f

4 Feb 2002 bccd_mpp.2002.2
Fixes direct bccsd mpp for impl=ga

5 Feb 2002 Slapaf.2002.2
Gives Fortran filenames in Slapaf Process specific names and deletes them
after geometry optimization

8 Feb 2002 undefs.2002.2
Initializes some undefined variables

8 Feb 2002 fujitsu2.2002.2
Adds support for Fujitsu PRIMEPOWER SPARC servers and Fujitsu vector (VPP5000 and VPP300) platforms.

12 Feb 2002 lmp2force.2002.2
Implements force integrals needed for the gradient in fitted (L)MP2.
The two-index force routines are fairly sensible, but the 3-index
force routine basis_3idx_force_3idx is horribly inefficient and is
intended only to test the LMP2 gradients. In the end this routine
will be replaced by a version that deals with suitable batches of
integrals/coefficients, rather than the whole lot.

14 Feb 2002 molden.2002.2
Bug(s) fixed: 216
Fix bug in Molden dump associated with incorrect MO coefficients
for d and higher basis functions.

15 Feb 2002 bug214.2002.2
Bug(s) fixed: 214
replaces DUM by dimensioned ORIG in "call read_op", the use of just DUM
caused an error in co_cordip.test with Intel compiler (dipnuc(3) was wrongly
put to 0)

15 Feb 2002 cfitshift.2002.2
Changes default setting for the linear algebra shift in CFIT/poisson code.

17 Feb 2002 cidensav.2002.2
Bug(s) fixed: 218

18 Feb 2002 hybrid_dft_ci4.2002.2
Robin Strange's stuff on DFT CI hybrid

18 Feb 2002 matrop2.2002.2
Bug(s) fixed: 219
Fixes mult command in matrop with factors

18 Feb 2002 ifci8b.2002.2
64-bit support for Intel Fortran Compiler under linux.
blas2.f:
The "blas4" mode is only correct if system blas routines (in 32 bit) do not call each other.
This happens in the "ghenry" blas routines. Routine DTRSV is called from within
DTRSM.
Since DTRSV is not used by MOLPRO, it is simply deactivated by cstart .not. blas4 // cend.
This may not be necessary for other blas libs, e.g. the INTEL mkl package.
machines.h:
I included support for 64-bit under LINUX (if I64 is defined,
FORTINT should become "long long" and the architecture string
should include "unix-i8"
system.c:
Ugly bug which kills test job system.test: "int SHELL_COMMAND" must of course
be corrected to FORTINT SHELL_COMMAND.
getmachine:
Allow 8-byte-integers on LINUX if the INTEL compiler is available
linux.conf:
Compiler options for 64-bit mode.
Precautions which might or might not be necessary:
Do not use GMALLOC since I am not sure about the FORTINT's.
Do not use openmp since I am not sure about the library interfaces allowing 64 bit.

14 Feb 2002 lrccsd.2002.2
implements lrccsd for weak pairs
fixes bugs in lrccsd for local=1,4 in nosingle option
fixes bugs for distant pairs
files
ccdtucc_loc.f src/ccsd/
ccmp2_ucc_loc.f src/local/
luccsd_util.f src/local/
rmp2.f src/ccsd/
rmp2_loc.f src/local/
spinpro_loc.f src/local/
uccsd.f src/ccsd/
uccsd_loc.f src/lccsd/

19 Feb 2002 for.2002.2
Make buildpatch properly look for header lines in a configurable number
of lines (default 20)

19 Feb 2002 myrinet3.2002.2
the third attempt to produce a conflict free patch...
.
The idea is to have a patch which contains the multiple executable
implementation of newconfig[2-4] and the myrinet library stuff.
...
So this is what I wrote along with patch myrinet:
This patch is another one in the series of patches newconfigX, that
have the aim to allow having different molpro_TARGET.exe in one molpro
directory. Another point is the incorporation of myrinet.
...
Essentially I added some new categories to configure. There are new
library search paths for the myrinet stuff and special link time
options. The latter also affected utilities/linux.conf. Also
new are parameters that allow having different parallel library
executables in one molpro environment. I added the opportunity to
specify runtime wrapper paths.
...
Of course bin/makegalib had to become aware of the myrinet library.
...
molpro.c underwent major changes. First of all I had to convert compile
time switches into runtime parameters because molpro now has to handle
different parallel versions of molpro.exe with different runtime
wrappers and different procgroup formats. Thus I added new command
line parameters to molpro.c: --mpi, --myrinet, --tcgmsg change the
procgroup format of molpro. These parameters don't change the runtime
wrapper used.
...
The resulting molpro executable still has a preference for the
main parallel library as specified in CONFIG. But it's also capable
of driving other parallel library versions provided with the
respective procgroup format switch and wrapper path.
...
I also added mechanisms to specify a non-standard rc-file; namely
the -rc <filename> command line option and the environment variable
MOLPRO_RCFILE. The command line takes precedence over MOLPRO_RCFILE.
I changed bin/makerc in order to incorporate the procgroup format
specifier and the -l option according to the wrappers specified
in CONFIG.
...
I had a pretty hard time to learn that the myrinet variant of
mpirun (gm version 1.5) doesn't tunnel environment variables. So
I had to convert environment variables into command line parameters.
This feature might change with future versions of the myrinet
library. So one should keep in mind that this environment
variable fudging might become obsolete one day.
...
Another point to remember is, that I hard-wired the choice of
myrinet ports into molpro.c. This works fine on our clusters because
there are no concurrent jobs that might hog the chosen port. At
other sites that might pose a problem. In principle we would need
a robust mechanism to determine available ports at runtime. Okay,
just wanted to let you know that trouble could arise here.
...
Some of our new clusters have various network capabilities
that lead to multiple node names for each box. I had to take care
for the case that hostname and network interface name differ. We
solved this problem by adding a distinct hostname category to the
-N option. For example I call molpro now with
molpro -rc molpro_tcgmsg.rc -N:aresm5:2:::ares5,:aresm8:2:::ares8 h2o_vdz.test
ares5 and ares8 being the hostnames of the nodes; aresm5 and aresm8
specify the network interface to use.

19 Feb 2002 mpptimes.2002.2
Additional timings to scf for mpp tuning

19 Feb 2002 newconfig5.2002.2
Restores GNUmakefile and linkprog that might be lost on some
systems due to my superceding of newconfig[2-4]. No changes,
just the files as taken from newconfig4.

20 Feb 2002 child.2002.2
child

20 Feb 2002 loc_eom2.2002.2
Add missing common in ini_ipair1d and ini_ipair1e. Without this common Tij was
used instead of Cij for selected weak T pairs in local EOM. Add option CFORWEAK
allowing to use Cij or Tij for weak T pairs in local EOM.

21 Feb 2002 tru64_5.2002.2
Minor changes, including a recompiled parse file, to enable parallel
on Tru64 version 5

21 Feb 2002 ccdomain_pointer.2002.2
Bug(s) fixed: 221
Fix pointer coding error in ccdomain.f

21 Feb 2002 cilop.2002.2
Bug(s) fixed: 222
Fix proper zeroing of spacers in matrices produced by cilop.

21 Feb 2002 k2dfock2.2002.2
Removes multiple computation of k2 quantities in dscf

22 Feb 2002 libmolc.2002.2
Repair the damage done by the Fujitsu patches to the linking of
utilities/libmol.c and src/util/libmol.c

22 Feb 2002 darwin.2002.2
This patch contains all changes required for building Molpro on
the Macintosh under OS X (Darwin). Currently the Absoft Fortran90
compiler is required. The files utilities/ftc.c and utilities/libmol.c
must be manually edited to supply the correct definition for
MOLPROBASE. An optimized Atlas blas library can be downloaded from
the molpro web site. Before building, the user should issue the
command unlimit from the terminal app. At this date the performance
is about a factor of 2 slower than expected based on the Atlas
dgemm timings. The reasons for this are not known at this time.

22 Feb 2002 diag2.2002.2
Make dsyev default diagonalization
Implement orbital rotation method in scf

25 Feb 2002 aosort.2002.2
Bug(s) fixed: 223
Removes compilation problem with different integer types in max/min functions

25 Feb 2002 patcher5.2002.2
Bug(s) fixed: 210
patcher now reapplies any child patches that had to be reverted because a
parent had to be reverted.
Fix problem in patch_order if patch[12].conflict is NULL.
Support for zlib scanning of patch files for increased speed.
Reverse date ordering of patches to what it should be (I hope).
Attempted fix-up round bug 210.

25 Feb 2002 sew_molpro.2002.2
Add a missing parameter to check_op in sew_molpro.f

26 Feb 2002 lmp2_mpp4.2002.2
Reduces I/O and communication in lmp2
Parallelizes multipole treatment for non-shared case
Fixes default file_implementation to be df for files 5 and 6

26 Feb 2002 compaq_sc.2002.2
Configuration for Compaq SC (prun support)

27 Feb 2002 optg.2002.2
Bug(s) fixed: 226
Fixes overwriting bug in optg, subr. cons1

27 Feb 2002 basis_check_dimension.2002.2
Bug(s) fixed: 220
Fix bad (for 32-bit) dimension in src/util/aoint.f:basis_check.

28 Feb 2002 uhfnat.2002.2
Bug(s) fixed: 226
New implementation of natural charge orbitals in uhf.
Fix some initailizations of noc, in order to recover previous occupations in optimization.

28 Feb 2002 newconfig6.2002.2
Fixes a bug in molpro.c that was introduced with the multi library
configure stuff. The effect only occured under AIX.

1 Mar 2002 config_ftcflags.2002.2
Bug(s) fixed: 227
Get rid of trailing blanks in FTCFLAGS in CONFIG produced byu configure,
otherwise ftc goes wrong

1 Mar 2002 lmp2_mpp6.2002.2
Avoids gasum in lmp2 iterations using global_acc
Option mppacc=1: use global_sum as before.
option mppacc=2: use global_acc, ga evenly distributed
option mppacc=3: use global_acc, ga on node 0.
New mpptune option minbatch_lmp2 [10]
Reduce numbatch (mpptune(7)) from 10 to 6.

2 Mar 2002 optrgf.2002.2

4 Mar 2002 dkroll.2002.2
Gets douglas-Kroll integrals working for molpro
This is activated by setting variable dkroll=1

5 Mar 2002 newconfig7.2002.2
This patch merges the AlphaServer SC changes with the
multilibrary and myrinet stuff. It also fixes a minor
bug in the environment variable fudging for myrinet.

5 Mar 2002 Molcas_updates2.2002.2
Updates to Molcas 5.3, patch level 4.3.2002

6 Mar 2002 dkroll2.2002.2
Documentation, test, and example for Douglas-Kroll
Make sure that DK integrals are computed after setting dkroll=1
and in dummy calculations. Fix bug in dkrelint.f

6 Mar 2002 rimp2grad.2002.2
Preparation for ri-mp2 gradients

7 Mar 2002 hybrid_dft_ci5.2002.2
Delete debugging write in ciexit

7 Mar 2002 Molcas_missing.2002.2
Adds missing files to Molcas

8 Mar 2002 signal.2002.2
Bug(s) fixed: 232
Avoid infinite recursion from signal handler.

8 Mar 2002 musave3.2002.2
Gives error message and exit if gradient calculation attempted for
sa-mcscf without cpmcscf. Fixes print of state number in kraft/krafta
for single-state calculations for excited states. Fixes print of
wave function type in kraft.

8 Mar 2002 Molcas_fix.2002.2
Fixes too small parameter in Molcas/input.f

8 Mar 2002 linuxconf.2002.2
Removes special Stuttgart specific library paths from linux.conf (except /software/linux)

11 Mar 2002 make_executable.2002.2
Bug(s) fixed: 233
Fix mistake in PNAME= definition in ./GNUmakefile which caused a link on
every make.

11 Mar 2002 dft_grid_reset.2002.2
Bug(s) fixed: 199
Harder resetting of DFT grid cache when the geometry changes. This will
lead to discontinuous, but globally OK, potential energy surfaces!
If variable GRID_FREEZE is set, then the numbers of grid points will be
frozen as before.

12 Mar 2002 configure_libz.2002.2
Automatic detection of libz for use by patcher and anything else that
wants it.

12 Mar 2002 multierr.2002.2
Calls error if multi does not converge in geometry optimization of
frequency calculation

12 Mar 2002 mpptune.2002.2
Reduced numbatch to 4 and increases maxbatch to 5000 in order to reduce
overhead for shared counter

12 Mar 2002 dftopt.2002.2
Freeze dft grid in frequency calculation and in geometry optimization
once optconv is smaller than 0.01 (default, or what is set by thr grid option).

12 Mar 2002 zmatrix_indexed_variable2.2002.2
Bug(s) fixed: 198
Improve (perfect?) handling of indexed variables, index.ne.1, in geometry
description

13 Mar 2002 grid_radial_put_get3.2002.2
Implement grid;radial,put and grid;radial,get to allow messing with
radial grids

14 Mar 2002 doc_integral3.2002.2
Fix latex syntax in integral.tex after patch dkroll2

15 Mar 2002 Slapaf2.2002.2
Replaces .molcas_info by molcas_info

16 Mar 2002 multipspace3.2002.2
Some further attempts to fix problems with pspace
initialize irefx, icga, icgb in muinp, print p-space
configs in case of too many p-space configs.
This does not fix p-space problem in mudet!

16 Mar 2002 lmp2bench2.2002.2
modifies lmp2 benchmarks
Use orbital rotation method in dscf
eliminate lmp2 gradients
modify thresholds

18 Mar 2002 newconfigure3.2002.2
Includes mppname and archname by default into executable name.
Makes INSTLIB-name machine and integer-type dependent

18 Mar 2002 sewoneel2.2002.2
Fixes bug causing calculation of 2-el. integrals in direct case
Adds variable sew_oneel. If sew_oneel=, the property integrals are
computed with seward.

18 Mar 2002 pspace_input_det.2002.2
Bug(s) fixed: 237
mxact instead of 31 was needed for dimension of icga,icgb in muinp2.f:pspace!!

19 Mar 2002 myrinet4.2002.2
This patch adds the myrinet mpirun options to molpro.c.
--gm-kill 10 is added as default option. It can be
switched off with '--mpirun--gm-kill-no'. Another
interesting option might be '--mpirun--gm-use-shmem'.
...
A glitch with multiple network interfaces was
resolved. The 'local node missing in node list'
message now reports the local node name.

20 Mar 2002 muaughes.2002.2
Attempt to fix problems with negative eigenvalues of hessian in aughes.
Select eigenvector giving smallest step.

20 Mar 2002 optnum8.2002.2
Adds variable numstp to driver.f for use in opt.
Sets variable NUMSTEP=numstp in numerical gradient or hessian calculation
Implements storage and reuse of numerical hessian in optg.
Not finished: reset maxupd if hessian is recomputed

20 Mar 2002 iow_error_stdout.2002.2
Bug(s) fixed: 242
Send iow_direct error messages to stdout not stderr

22 Mar 2002 gopenmol2.2002.2
Interface to gOpenMol. Some tidying of basic utilities to allow long file names
for fortran files.

25 Mar 2002 gopenmol_driver6.2002.2
(a) GOPENMOL command in driver.f; (b) 8 bytes instead of 6 for vers in /cvers/.

25 Mar 2002 f77version.2002.2
Fill in the F77VERSION stanza in CONFIG for more machine types

25 Mar 2002 testjobs_hostname.2002.2
Allow forcing of a host name for testjobs report.

25 Mar 2002 intface.2002.2
Fixes undef variable iout in subr intface

26 Mar 2002 addinfo3.2002.2
Disables add_info, avoiding generation of file add_info
Updates common.gam.inc avoiding undefined variables in gammaf.f, gammat.f

26 Mar 2002 configp4.2002.2
Adds -p4 option to configure

26 Mar 2002 instguide2.2002.2
Bug(s) fixed: 215
Adds description for multiple executables

26 Mar 2002 aoint_matrix3.2002.2
Fix for large integral files with 64 bit integers. inter*4 ibuck now
only stores relative adress in segement.
Truncate integral record once integrals have been written to avoid
exessive integral file length
Moves call to ao_integral_put_finish outside the loop over segments.
Fixes bug in ao_integral_matrix_get.
Reserve of buffer space for ao_integral_matrix_get done by
AO_Integral_Matrix_Get_initialize

26 Mar 2002 molproc8.2002.2
Fixes problem with "Local machine not in node list".
setenv MOLPRO_EXECUTABLE to basename of molpro.exe file.

26 Mar 2002 tuningrc3.2002.2
Bug(s) fixed: 248
Adds command saverc to blastest.f. If set (without arguments),
get program name from MOLPRO_EXECUTABLE and save to bin/name.rc
Corrects GNumakefile to print correct rc file in make tuning

26 Mar 2002 compressacc.2002.2
Fixes problem with compression accuracy. If two subsequent calculations with
different basis sets were performed, the first compression threshold was kept.
Now it is always resets the threshold before computing new integrals.
Fix call to cmpr_init in drv2el_compress (missing dummy argument).
Remove call to cmpr_default from argos.
Remove obsolye ao_integral_get and ao_integral_put from aoint.f

28 Mar 2002 bench2.2002.2
Bug(s) fixed: 252 253
Corrects (modifies) input small2.com (weight zero for second
state in multi to avoid cpmcscf for gradients).
Sets etarget in big_direct_lmp2.bigcom

28 Mar 2002 buildpatch_tar6.2002.2
Bug(s) fixed: 247
buildpatch -tar (default) and buildpatch -shar (the old thing).
buildpatch_tar3: restore possibility to unpack to a directory; sneaky
stuff to check that patcher is up to date, and to crash out otherwise;
patcher itself again now notices when patch application failed, and
also spits out patcher.log whenever anything goes wrong.
buildpatch_tar4: attempt at more robust removal of patch directory
on reversion.

28 Mar 2002 bigrec4.2002.2
Implements big (>16GB) records for integer*4.
New routines reserve_big, writem_big, readm_big
Modifies ao-integral interfaces and aosort to handle
big integral records (1300).

28 Mar 2002 cipt2states.2002.2
Error message if cipt2 is attempted with nstate.gt.1

29 Mar 2002 hpconf2.2002.2
Updates hp.conf. integer*8 now default for hp and sun.
Enables lapack for hp 64 bit, and adds new support for hp
blas- and lapack libs downloaded from molpro homepage
Updates INSTLIB defaults for Stuttgart in configure.

3 Apr 2002 maketab.2002.2
Sorts data before spline fit as required

3 Apr 2002 signal_honour2.2002.2
Bug(s) fixed: 238
Transmit 'nohup' to the molpro.exe processes.

3 Apr 2002 muacc.2002.2
Tighter convergence thresholds in gradient calculations

3 Apr 2002 unused.2002.2
Bug(s) fixed: 254
Get rid of unused/ (kept in master) and OLD/ (deleted for ever) files

3 Apr 2002 molproc9.2002.2
This patch corrects the order of reading diverse
.rc files.

3 Apr 2002 optnum13.2002.2
Adds variable numstp to driver.f for use in opt.
Sets variable NUMSTEP=numstp in numerical gradient or hessian calculation
Implements storage and reuse of numerical hessian in optg.
Improvements of transition state and reaction path optimization
Bofill hessian update now default for this case.

4 Apr 2002 systemc.2002.2
Fix bad syntax in system.c that caused compilation failure on HP

4 Apr 2002 ifc_compile_parallel.2002.2
Bug(s) fixed: 204
Skirt round the bug in Intel fortran 5.0.1 that means that in a given
directory only one compilation should be happening at once

4 Apr 2002 compression2.2002.2
Bug(s) fixed: 16
Modest (10%) speed-up in uncompress_double

4 Apr 2002 hstart.2002.2
Fix for hstart with dummy atoms in optg.f and numforces.f
Fix for save of hessian to file 1 in frequencies.f
Fix for error exit if no convergence in optgeo.f

5 Apr 2002 mpptimes2.2002.2
Additional mpp timings for lmp2 and (t)
modified defaults for lmp2 parallelization

6 Apr 2002 hess_perm2.2002.2
Bug(s) fixed: 255
Fixes problems in analytical hessian (scf, mcscf)
with center permutations and dummy atoms. Still,
numerical frequency calculations cannot be done with dummies.

6 Apr 2002 ccdiis_mpp3.2002.2
parallelizes ccdiis, use this in lmp2
maxtsk=1 for mpptype=1 in drv2el_llmp2

6 Apr 2002 mpptim_save3.2002.2
Save mpp timings into a file given by variable mpptim_save.

7 Apr 2002 restartoldfiles.2002.2
Fixes restart of files generated before patch bigrec4.
In most cases, the file type should be recognized
automatically. In the unlikely event of problems,
variable oldstyle=1 can be used to force attempt of
old style restart, oldstyle=-1 to force attempt of
new-style restart. This variable must be set before the
file cards,

8 Apr 2002 basis_shells.2002.2
New Coulomb integral shell routines, new <r_{12}> integral shells, and various others to extend the class of problems treatable with auxiliary basis sets.

8 Apr 2002 bench3.2002.2
mpptype=2 for big_direct_lmp2

8 Apr 2002 orbdom.2002.2
removes diag2 and smh, reads smh instead.
Add some timing prints

9 Apr 2002 diabrefs.2002.2
References for diabatization

9 Apr 2002 doc_ccsd.2002.2
Adds note about saving density matrix.
Corrects the subsection structure for density matrix save and natural orbitals

9 Apr 2002 make_flavours8.2002.2
Bug(s) fixed: 249 256
Cleaning up of make procedures for multiple flavours (architecture, mpplib etc)
and the corresponding installation procedures.
make_flavours4: incorporate makepath; restore -l in installed version where
PARALLEL not defined.
make_flavours5: now INSTBIN/molpro* is a script and everything else is in INSTLIB.
make_flavours6: add support for MOLPRO_CONFIG and MOLPRO_OPTIONS.
make_flavours7: fix bug 256.
make_flavours8: replace '.' by '_' in version number from bin/pname; more robust linking of bin/molpro and bin/molpro.rc.

9 Apr 2002 basis_variable_doc.2002.2
Bug(s) fixed: 118
Add documentation of how basis set changes are handled, in particular
where variables affecting the basis set are changed.

10 Apr 2002 shells.2002.2
Bug(s) fixed: 259
Replace erroneous type
DOUBLE PRECISION, PARAMETER, SAVE :: pi =3.141592653589793238d0
with
DOUBLE PRECISION, PARAMETER :: pi =3.141592653589793238d0
and similar cases.

10 Apr 2002 mxma_mpp6.2002.2
Modifies mxma_mpp.f. Limit number of processors to a maximum, determined
by mpptune(11) (maxproc_mxma), mpptune(12) (min row dimension), cpu-speed,
latency, and transfer speed.
CPU speed taken from MOLRPO_FLOPDGM, Latency MOLPRO_MPPLAT (or mpptune(25)),
Speed MOLPRO_MPPSPEED (or mpptune(26)).
blastest changed to produce mpplat and mppspeed.

9 Apr 2002 r12.2002.2
Correction to code for <r12> integral shells for density fitting.

11 Apr 2002 del2n.2002.2
Adds a routine to evaluate integral shells of any operartor of the form del^{2n}.

11 Apr 2002 lrcckext2.2002.2
implements calculation of 4ext-int. with K(C(ij,1)) for lrccsd
fixes bug in ucckext_loc
fixes bugs in ccmp2_ucc_loc for RCCD and for LRCCSD with weak pairs
files
ccmp2_ucc_loc.f src/local/
ccv_ucc_loc4.f src/lccsd/
luccsd_util.f src/local/
spinpro_loc.f src/local/
ucckext_loc.f src/lccsd/
uccsd.f src/ccsd/
uccsd_loc.f src/lccsd/
ucckext2.f src/lccsd/
rcc3ext_loc1.f src/lccsd/

11 Apr 2002 cfit_r12.2002.2
Implements density fitting using Gill's fitting criterion r_{12} which
amounts to fitting the density to least-squares minimize the error in
the Coulomb potential. This kind of fitting can be switched on using
the input cfit,fit='R'.

12 Apr 2002 tuning4.2002.2
Fixes crash of blastest.f in mppcase.
Modifies tuning.com to save mpp tuning data

12 Apr 2002 dderf.2002.2
Bug(s) fixed: 260
Fixes bug 260. basis_integral_shells.f called derf instead of dderf.

14 Apr 2002 c2h2.2002.2
Bug(s) fixed: 250
Fix up order of symmetry generators in C2h so that both binput and Seward
can be happy simultaneously. Reflect this choice of generator in spin-orbit
integral symmetry assignment.
Addresses also some of the issues raised in http://www.tc.bham.ac.uk/molpro/molpro-user/archive/msg00024.html
c2h2 eliminate test print in zmatrix.f

14 Apr 2002 ccmp2_loc.2002.2
Fixes an mpp problem in ccmp2_loc

15 Apr 2002 esp.2002.2
First implementation of electrostatic potential on a grid:
subroutine esp_grid_trace and interface to it from gopenmol;potential
and cube;potential.

15 Apr 2002 mpplat.2002.2
This patch moves the output of MPPLAT and MPPSPEED settings from
mpp_open to init_molpro.

15 Apr 2002 min_crash.2002.2
Bug(s) fixed: 258
Fix crash in min2d.f because unit argument to getvar wasn't declared
character in calling routine.

16 Apr 2002 dual_lmp2.2002.2
implements dual basis lmp2

16 Apr 2002 mxma_mpp7.2002.2
Bug(s) fixed: 264
Fixes syntax error in mxma_mpp.f

16 Apr 2002 ga_debug.2002.2
ga_debug logical flag only set if MOLPRO_GA_DEBUG.gt.1. If 1, only
separate output for each process is opened and memory info is printed.
Fixes bugs in cicore for direct/mpp (global_sum twice).

16 Apr 2002 reserve_ccsd.2002.2
Bug(s) fixed: 262
Fixes problem with irregular gas in reserve_ccsd.
Still all pairs are distributed, even in lccsd case when only strong
pairs are needed. This should be fixed once lccsd is parallelized.

16 Apr 2002 unini.2002.2
Fixes undefined variables

16 Apr 2002 patcher_rmdir.2002.2
Bug(s) fixed: 266
Remove all empty patch/* directories at start and end of patcher in case
they didn't get properly removed before.

16 Apr 2002 make_parallel.2002.2
Bug(s) fixed: 263
Less ridiculous set of dependencies so that make in parallel doesn't try
to simultaneously and multiply build utilities/molpro and bin/molpro.rc.

16 Apr 2002 progman.2002.2
Latex syntax errors fixed in iom.f; additional files added to programmer
manual, and makefile made a bit more robust towards additions.

17 Apr 2002 ftn_linkfc3.2002.2
Restore the behaviour that ftn uses LINKFC not FC to link mopro executable
even with the new generalisations of molpro.exe.

17 Apr 2002 task5a.2002.2
Offers nextask() mechanism as option dynamic_meth=5. Makes this
the default method if ccsd_impl != ga.

17 Apr 2002 ddr_doc.2002.2
Note in ddr documentation that STATE command necessary when sym.ne.1

17 Apr 2002 configure_lapi.2002.2
Extend the flavours stuff to the case of IBM SP, where one can now have
ether tcgmsg or lapi.

17 Apr 2002 task5b.2002.2
Reorganization of initask and endtask calls for dynamic_meth=5.
This moves the initask synchronization in front of vijm1 while
avoiding interference with the cc3ext nextask environment.

17 Apr 2002 report_moredetail.2002.2
MPPNAME and ARCHNAME bundled into report file name

17 Apr 2002 recmax.2002.2
Modifies parameter for max addressing using 32-bit integers

17 Apr 2002 configure_lapi2.2002.2
Extend the flavours stuff to the case of IBM SP, where one can now have
ether tcgmsg or lapi.

17 Apr 2002 ccsd_mpp5.2002.2
Static load balancing in vijm1p

18 Apr 2002 geminal6.2002.2

18 Apr 2002 maketidy.2002.2
Reduce unnecessary verbosity in src/GNUmakefile.subdir

18 Apr 2002 aix6.2002.2
Some stuff for AIX5/Power4.
makes config.guess execuatble (aix6)

18 Apr 2002 basis_efg.2002.2
Extend basis_shell_efg to electric field gradients in the general basis set code.

19 Apr 2002 make_squash_another_race.2002.2
Get rid of the double making of bin/molpro and bin/molpro.rc

19 Apr 2002 ccsd_loc_cleanup.2002.2
Separates 3ext, 4ext, and triples sections from ccsd_loc.

19 Apr 2002 geminal8.2002.2
Remove USE GEMINAL from ccsd_loc.f and ccdomain.f

19 Apr 2002 geminal10.2002.2
Remove USE GEMINAL from ccsd_loc.f and ccdomain.f

20 Apr 2002 configure_linux.2002.2
Fixes bug in configure (missing quotes in tests for instlib)
Removes prefetch compiler options for linux

22 Apr 2002 cckint_bank_zero.2002.2
Bug(s) fixed: 267
Fix floating point errors coming from mxm in cckint because
vkl not zeroed in serial mode .. bank padders not zero.

22 Apr 2002 patcher_force.2002.2
Implement -f option for patcher which is transmitted as -f option to patch
scripts, in order to force version number.

22 Apr 2002 buildpatch_tar7.2002.2
buildpatch -tar (default) and buildpatch -shar (the old thing).
buildpatch_tar3: restore possibility to unpack to a directory; sneaky
stuff to check that patcher is up to date, and to crash out otherwise;
patcher itself again now notices when patch application failed, and
also spits out patcher.log whenever anything goes wrong.
buildpatch_tar4: attempt at more robust removal of patch directory
on reversion.
buildpatch_tar7: 'B' option required for piped tar on SGI

22 Apr 2002 rsjc.2002.2
Limit threshold not to be smaller then 1.d-16 in rsjc. Avoids
infinite loop on sun 64bit

22 Apr 2002 sunconf.2002.2
Lower optimization for Molcas/darwin.f on sun

22 Apr 2002 sgiconf2.2002.2
ccdiis.f with lower optimization on sgi

22 Apr 2002 ar_new.2002.2
Avoid, for a little longer, shell argument overflow on first building
lib/libmolpro.a by doing it in 3 chunks instead of 2.

23 Apr 2002 hybrid_dft_ci6.2002.2
Robin Strange's stuff on DFT CI hybrid

24 Apr 2002 lmp2_print2.2002.2
Fixes print statement in lmp2

24 Apr 2002 maxrec_hls.2002.2
Bug(s) fixed: 269
Checks for max number of records on hls card

24 Apr 2002 sun_make_race.2002.2
Bug(s) fixed: 271
Cure race conflicts in parallel make that especially showed up on Sun.

24 Apr 2002 aosort_mem4.2002.2
Fixes integer overflow in aosortv.f and dynamic increase of memory if too many segments.
Replaces functions length_of_file and length_of_record by double precision functions
dlength_of_file and dlength_of_record, changes calls accordingly.

25 Apr 2002 molpro_debug.2002.2
Get rid of copy in of base/src/molpro.o in develop directory; instead,
let it be compiled with debug option. Essential for Intel ldb 6.0.
Also make use of shared libraries the default for Portland and for ifc 6.0
or higher.

25 Apr 2002 fci.2002.2
Fixes aoint_matrix_get_init in fci.f

25 Apr 2002 ddr_doc2.2002.2
Updates documentation for ddr.
Eliminates multiple labels.

25 Apr 2002 examples.2002.2
Fixes some example inputs
Replaces all example outputs

25 Apr 2002 testjob_outputs.2002.2
New outputs for testjobs. Some Cleanup of testjob directory.

26 Apr 2002 parse_aix_431.2002.2
Parse object for AIX 4.3.1 (made by Kirk Peterson).

26 Apr 2002 linuxconf_ifc6.2002.2
Detect Intel fortran compiler if ifc is in $PATH

26 Apr 2002 hpconf4.2002.2
Support for hp-ux 11.22.
hpconf4: put hp.conf in utilities/ not ./

26 Apr 2002 parconf2.2002.2
This patch implements a minimal node set for myrinet in the
absence of a ~/.gmpi/conf file. It fixes a problem with the
-l directive when calling 'make install'.

27 Apr 2002 massprint.2002.2
Adds option mass,print to print masses of all atoms

27 Apr 2002 ftcflush2.2002.2
Adds blastest.f to ftcflushlist

27 Apr 2002 sew_molpro2.2002.2
Fixes mpp_state problem in sew_molpro.f (end_pack)

27 Apr 2002 dkext_mpp.2002.2
Fixes floating point error in case that a procesor has no work

28 Apr 2002 sun_make_race2.2002.2
Cure race conflicts in parallel make that especially showed up on Sun and SGI.