Molpro version 2002.5 release notes from patch history of preceding version 1 May 2002 molproc_cray.2002.4 Bug(s) fixed: 278 Fix compilation problem in molpro.c on Cray 3 May 2002 test_cfit.2002.4 Test job to test cfit code and many aspects of the src/basis code. I have included all varieties of density fitting in kohn-sham theory, and basis sets of standard spherical gaussians, cartesians, contracted, uncontracted and poisson type functions. Also tested is the mechanism for reading basis sets from the libarary and explicitly. 3 May 2002 parse_ibm.2002.4 Replace corrupted licence file for RS6000 with AIX 4.3.1 6 May 2002 linuxconf.2002.4 Adds support for mkl library (default for p4 if available) Modifies configure to recognize p3, p4, athlon automatically Modifiles bench/configure to save CPU name in CONFIG 6 May 2002 linuxtune.2002.4 new optimized mxma for linux (mxma2). Support for mkl_p4 lib. Optimized tuning default values in syscon. Modified tuning.com including check of mxma2 makerc punch MOLPRO_BLASLIB 7 May 2002 myrinetports.2002.4 This patch adds flexibility to the port handling in myrinet networks. With the --myrinet_ports option one can specify the desired ports for each node at runtime. A basic consists of a colon separated list of . pairs (e.g. 7.0:7.1:6.0:6.1). For ports on card 0 the .0 part can be omitted. To specify different ports for individual nodes you can build comma separated lists of basic portstrings. 8 May 2002 opt_linear_triatomic.2002.4 Bug(s) fixed: 281 Fix up transformation matrix for gradient from cartesian to internal coordinates for the special case of the bond angle of a linear triatomic (indeterminate direction) so that the model hessian code can have a chance of working. 8 May 2002 fitmos.2002.4 Reorganizes ccdomain.f: First make orbital domains, then possibly modify orbitals, then project New fitmos, cutmos options Fix shell permutation in dual basis calculations 14 May 2002 molproc_ibm.2002.4 Incorporate recommendations for environment variables GT_RTQ and AIXTHREAD_SCOPE for IBM POE. 16 May 2002 p3_mkl.2002.4 Changes to linux.conf: a) tidy library detection, and include mkl for p3 (the default if available). Note that this means ghenry is not default, but was difficult to dectect this properly anyway. b) for ifc, -rpath the shared library directory in case LD_LIBRARY_PATH not set at run time. 21 May 2002 ecp_parse.2002.4 Bug(s) fixed: 284 Change so that for i=8 basis={ecp,i,ecp46mdf} you get iodine instead of centre 8 22 May 2002 chgfracfix.2002.4 fixes chgfrac.ge.1 case. full domains do not work with integral distributions based on center quads, so we go back from i4ext=1 to i4ext=0, and from i3ext=2,3 to i3ext=1 23 May 2002 cilssym.2002.4 Fixes transformation of property matrices to symmetry adapted SO basis 29 May 2002 rs2grad.2002.4 gets rs2 gradients working (but not yet for sa-mcscf reference and only for core,0 so far) 1 Jun 2002 drv2el_llmp2.2002.4 Bug(s) fixed: 288 Fix compilation error in drv2el_llmp2.f: < > izero_shell(kso).ne.0) goto 100 --- > > izero_shell(kso)) goto 100 3 Jun 2002 rccsd.2002.4 implements faster method for rccsd files ucckext2.f src/lccsd/ uccsd.f src/ccsd/ uccv.f src/ccsd/ rcc3ext_ucc.f src/ccsd/ rcckextconv.f src/ccsd/ isynpp.f src/ccsd/ common/ntnpp src/common/ 3 Jun 2002 rilmp2g_5.2002.4 stuff for RI-LMP2 Gradients 6 Jun 2002 ibm64_gmainv.2002.4 Bug(s) fixed: 158 291 Revert to using malloc in ibm i8 because shared memory stuff doesn't seem to function properly. 2000.9.ibm64_gmainv2 patch was corrupted, and we now properly bring in the change to machines.h. Additionally, make gmainv.c 64-bit compliant. 6 Jun 2002 t1diag_uccsd2.2002.4 Modifies t1 diagnostic for open shell ccsd 7 Jun 2002 cc_ifc.2002.4 Bug(s) fixed: 295 Fix up conflict of -Difc cc compilation option and system header files. 14 Jun 2002 linux_gmalloc.2002.4 Disables gmalloc replacement for linux (ia32) This makes possible to use large memory! 17 Jun 2002 fitmos6.2002.4 Fixes amd improves fitmos for lmp2. Makes lm2algo=0 default for non-orthogonal calcs (faster) 17 Jun 2002 linuxconf2.2002.4 Disable shared library stuff since infinitite loop with ifc 5.0.1 17 Jun 2002 linux_ia64.2002.4 Supplies a parse object for linux/ia64 and modifies linux.conf to anable Intel efc compiler for ia64. But it does not work due to unknown compiler errors. I found that getmem 'Free' does not release the memory in drvk2, and this problems does not disappear without optimization. hf and multi work, but mrci, ccsd are wrong. Compiling all of ccsd and blas without opt does not help. I therefore give up. getmachine modified to check for xunix-linux-ia64 in 4th field rather in 3rd. This patch might screw up support for ia64/linux/sgif90. To link with ifc, disable special code for unix-linux-ia64 in bild.f (call of ishell). This is not included in this patch. 18 Jun 2002 ga_avqz.2002.4 Bug(s) fixed: 296 Fix bug in Ga aug-cc-pvqz diffuse g function (was also present in EMSL library) 21 Jun 2002 fci_dm.2002.4 Implement density matrix for FCI. Fix benign bug in print of correlation energy in FCI when ncore.gt.0. Implement scaling of 2-electron integrals in FCI. THR and MAXIT commands in FCI. Testjob fci.test. 25 Jun 2002 cfit_test.2002.4 Updated density fitting testjob that avoids numerical problems with Gill's r_{12}^{+1} criterion. 27 Jun 2002 fitmos7.2002.4 Fix unwanted call of fitmo in case of squeeze 27 Jun 2002 austin_j4.2002.4 Fixes bug due to wrong roots and weights in the presence of g-functions 1 Jul 2002 basis_aoinfo4.2002.4 Provide a temporary interface to get ao group information. To be replaced later by proper stuff of pjk supercedes all cftrans patches 2 Jul 2002 basis_4_dflmp2.2002.4 New interface for obtaining all information about a basis set. Routines to precompute obtain all gaussian product data. Extensions of some information routines in basis.f to treat composite basis sets -- Peter, please note my modified basis_ngroup, basis_nprim and basis_nprim_sh. 2 Jul 2002 df-hf3.2002.4 Density fitting for coulomb problem in HF theory, and correction to do hybrid DFT. Reserves dump file earlier in scfpr2.f. Add timings for cfit 3 Jul 2002 batchs.2002.4 3-index integral batcher with new interface. 3 Jul 2002 basis4.2002.4 Write AO basis info in Argos to record 970 (subroutine basis_save) Read this in basis_aoinfo. Proper storage and normalization of contraction coefficients. Basis info printed using gprint,basis=1 5 Jul 2002 batchs2.2002.4 A 3-index integral batcher for the df-lmp2 project with several restrictions: (1) No contracted functions (2) s and p only in AO basis (but any in auxiliary basis) (3) No generalized shells (ie no s and p with same exponent) (4) Test densities can be passed but are ignored These limitations will be lifted as I write the contraction section of the routines. 7 Jul 2002 force_dflmp2b.2002.4 Next stage in 3-index force integrals for DF-LMP2. 8 Jul 2002 optdflmp2.2002.4 Adds df-lmp2 to the list of known methods for optimization with analytical gradients. Make df-lmp2 command (alias to ri-lmp2). Fixes some related problems. 9 Jul 2002 batchs4.2002.4 New version of 3-index batcher. Several bugs fixed. Case with >1 AO groups in batches fixed. 10 Jul 2002 basiscen.2002.4 Fixes bug in allocating centers to groups 10 Jul 2002 dftrans.2002.4 New df-transformation for lmp2. Can be activated using cfit=1 10 Jul 2002 batchs5.2002.4 Further enhancements to the 3-index integral engine. Mostly concenring contractions. Contraction and sphericals for auxiliary basis is working. Spherical transformation for AOs in written but not working because we have to wait for the AO information routines from PJK. AO contractions work for s-functions only owing to a quirk of HJW's temporary interface; when PJK's routines arrive this too should be corrected. 10 Jul 2002 rilmp2g_7.2002.4 Declares is_triang (only in debug section) logical to avoid problem on sun 10 Jul 2002 mxm_blas4.2002.4 Replaces 1 in calls to mxm[ab][34] by i1 to avoid problem with the blas4 version 12 Jul 2002 dflmp2_dual2.2002.4 Fixes problem with dual basis sets in df-lmp2 (wrong fock matrix read) Fixes further bug (wrong density matrix used for ref. energy) 15 Jul 2002 batchs7.2002.4 Bug in basis.f regarding AO contractions removed. Further improvements of the batcher: more screening, generalized contraction routines. 18 Jul 2002 basis_sphtr3.2002.4 Sets up transformation matrices for cartesian -> spherical harmonics See common/csphtr for further details. 19 Jul 2002 sphtr.2002.4 Renames commons in csphtr to avoid clash with commons used by the standard basis interface. This is necessary since (for now) the definitions of spherical harmonics (normalization and order) are *different* for AOs and for auxiliary basis sets. 23 Jul 2002 info.2002.4 Update info to reflect altered ordering procedure and new web address 25 Jul 2002 freqdump.2002.4 Implement option DUMPALL in frequencies to dump all modes to the frequenciy record. Implement THRESH=LOW command in frequencies to vary the threshold for low frequencies (default 150 cm-1). 25 Jul 2002 batchs8.2002.4 Bug in basis.f regarding AO contractions removed. Tested contraction routines. Current status: sphericals work for AO/auxiliary basis. Contraction works for AO/auxiliary basis *except* for AO contractions where there is >1 contraction in the group and the angular momentum is nonzero. Groups with ranges of ang mom still not allowed. These issues will be resolved in due course. 25 Jul 2002 batchs9.2002.4 Bug in basis.f regarding AO contractions removed. Tested contraction routines. Current status: sphericals work for AO/auxiliary basis. Contraction works for AO/auxiliary basis *except* for AO contractions where there is >1 contraction in the group and the angular momentum is nonzero. Groups with ranges of ang mom still not allowed. These issues will be resolved in due course. 26 Jul 2002 getscl.2002.4 Introduces a simple wrapper for getvar to deliver single real, integer or logical molpro variables without all this nonsense with unit. 27 Jul 2002 rs2grad2.2002.4 Correction for gradient with level shift. Gradient computed exactly only without back scaling correction. Store Davidson corrected cipt2 energy in ENERGD Decouple S1/R and I1/R spaces by default in cipt2 29 Jul 2002 rs2grad3.2002.4 Fix for excited state calculations (only one weigth non zero Proper implementation of cpmcscf,rs2 options in muinp1. MCSCF state information stored in variables. Set variable ENERGU to uncorrected energy in case of level shift calculations Fix for setting variable ENERGD in case of cipt2 calculations 30 Jul 2002 cirestrict.2002.4 Ignore restrict input for closed-shell orbitals 30 Jul 2002 fixlmp2g.2002.4 fixes problems for domain restart in lmp2g geometry optimizations when using 3ext integrals for construction of K(T)L 1 Aug 2002 basis_ao6.2002.4 bases(0) contains orbital basis info 1 Aug 2002 fixoptdomrest2.2002.4 fixes problem in the context of geometry optimizations with local methods and domain restart. Additional card NOPERM/LOCAL: prevents permutations of paos for (initial) local MP2 calc (for domain specification), to be compatible with subsequent lmp2 grad calcs in geo optimization... 1 Aug 2002 rs2tests.2002.4 test for caspt2 gradients added to h2o_optrs2.test cipt2 testjobs modified since core-reference couplings are now removed by default 1 Aug 2002 rs2grad4.2002.4 some fixes to remember correct symmetry in rs2 optimizations 2 Aug 2002 numfreq.2002.4 Zero zhessvar in the beginning of frequencies to avoid wrong results 2 Aug 2002 directluccsd.2002.4 implements direct option for luccsd 5 Aug 2002 shelldoc.2002.4 Correct a cstart/cend bug that broke codedoc 5 Aug 2002 getvardoc.2002.4 Documents the getvar, setvar and related routines for handling variables. 6 Aug 2002 basis_efg.2002.4 Corrected efg shell routine 6 Aug 2002 basisgen3.2002.4 Bug(s) fixed: 307 Generalization of some of the basis interface routines to accept lists of basis sets and explicit documentation for those which do not accept lists of basis sets. 7 Aug 2002 for.2002.4 Simple script to translate a patch from one version to another 8 Aug 2002 web4.2002.4 Migration to new web server 8 Aug 2002 doc2.2002.4 Changes to documentation reflecting changed web server address 12 Aug 2002 undefs.2002.4 Fixes undefined variables in some places 13 Aug 2002 matrop.2002.4 Corrects matrop operation oprd 14 Aug 2002 dftrans2.2002.4 Updates for new basis interface 15 Aug 2002 rs2grad5.2002.4 set correct norm for rs2 gradients in cidima 15 Aug 2002 batchs12.2002.4 At last a working version of the general 3-index batcher for LMP2 and related projects. This works with contracted, spherical basis sets (as far as I know) with the one caveat that general shells (eg s and p functions with the same exponents) will fail, possibly silently. I will correct defect this shortly. 15 Aug 2002 numfreq3.2002.4 Strcter convergence thresholds in scf and mrci/ccsd for numerical gradient and frequency calcs. Eenbale freq,symm=auto for open-shell ccsd Fix problems with wrong occupations during frequency/restart calculations in scf Error exit if no convergence in ccsd during optimization/frequency calc. 15 Aug 2002 cigptr_mpp.2002.4 Adds missing reservs to cigptr.f 21 Aug 2002 batchs13.2002.4 Some special cases inserted and other tricks to increase efficiency. Also bugs removed related to screening. 22 Aug 2002 basisr12.2002.4 Some transparent changes to the general integral routines, preparing the ground for the extra integrals needed for density fitted MP2-R12. 22 Aug 2002 test_dfmp2_2.2002.4 Test for density fitting mp2 Corrected to avoid use of variable EMP2 which is already used by the MP2 programs. 23 Aug 2002 schwarz.2002.4 First part of the Schwarz inequality code for the general basis interface. Upgraded integral selection through routine basis_inttype. Some integral routines extended to a wider class of integrals needed for explicit correlation. 27 Aug 2002 dummy.2002.4 Check symmetry after dummy atom input. Error exit if symmetry is not consisitent with dummies 28 Aug 2002 explicit.2002.4 Directory and makefile tweak for new explicit correlation program. Currently there are components relating to MP2-R12 and DF-MP2-R12. Documentation and interface will follow in the distant future. 28 Aug 2002 cckext_mpp.2002.4 fixes bug in cckext affecting mpp only 29 Aug 2002 explicit3.2002.4 Include missing routines for 4-index integrals needed in the explicit project. 30 Aug 2002 batchs15.2002.4 Fixes screening problem in basis_3idx_batchs Memory allocation and zeroing for basis_3idx_batchs in basis_3idx_batchs_init Memory release for basis_3idx_batchs in basis_3idx_batchs_term Remove thr from call to basis_3idx_batchs, now in basis_3idx_batchs_init Remove index searching in basis_3idx_batchs Count computed and significant integrals Improved blocking Implement density screening in basis_3idx_batchs and basis_GPT_info Activate special code in basis_integral_batchs. 1 Sep 2002 cnvrt.2002.4 Bug(s) fixed: 323 Avoid problem of utilities/cnvrt.exe crashing with ifc 6.0. Also fix up lib/Makefile to depend properly on RYSRW as well as ABDATA. 3 Sep 2002 trap_conflict.2002.4 Bug(s) fixed: 324 IBM xlf8 compiler has an intrinsic function called trap which conflicts with trap() defined by molpro 3 Sep 2002 explicit5.2002.4 Fully fitted df-mp2-r12 code. Extension of basis_3idx to more general integral/basis combinations. Correction of small bug in a previously un-checked case of basis_contract_matrix. 4 Sep 2002 patch_delobj2.2002.4 Bug(s) fixed: 283 298 Force deletion of object library members on reversion of a patch. Reluctantly done: it slows patch reversion down; however, it's necessary because we must remove from the library any new files introduced by the patch. Also now properly honour the delobj stanza: the effect now is to attempt to delete both the object and a file of the same name in the library. 4 Sep 2002 blas6.2002.4 Bug(s) fixed: 288 325 Proper typing of variables in intrinsic functions for blas4 4 Sep 2002 basis_cartesian.2002.4 Bug(s) fixed: 308 310 Implement attribute,-attrib and retention by default of the default attribute, eg poisson, for general basis input. Implement default cartesian attribute whenever global cartesian flag is on. files basis.f src/basis/ 4 Sep 2002 explicit6.2002.4 Implements 3-index integrals over [r_{12},hat t_1] and implements fully fitted mp2-r12 theory (ie no 4-index integrals anywhere). 4 Sep 2002 compress.2002.4 Bug(s) fixed: 306 Fixes problem with incomplete last record in compress 4 Sep 2002 rccsd4.2002.4 Enables denominator shift in first iteration of rccsd/uccsd 4 Sep 2002 cicore_aoint_init.2002.4 Bug(s) fixed: 273 286 Put in missing AO_Integral_Get_Init in cicore. 4 Sep 2002 rccsd5.2002.4 improvements for rcc3ext_ucc.f Save all files that will be overwritten by patch rccsd6 for later use 4 Sep 2002 rccsd6.2002.4 Reverts effect of patch rccsd, since this caused the following problems: (i) poor efficiency, cckext and ucc3ext_rcc called more than once per iteration (ii) rcisd does not work (iii) rccsd does not work with >1 processors (mpp) (iv) cckext1 is needed by conventional rccsd, but is in lccsd/ucckext2.f. This is not available for people who don't have module lccsd. 4 Sep 2002 fci_mpp.2002.4 Bug(s) fixed: 329 Fix bad blocking calculation in small fci calculations in parallel 4 Sep 2002 dftrans4.2002.4 Implements sparse storage of half transformed integrals in dftrans_loc. Also implements sparse blocked algorithm for second-half transformation. Makes dftrans_loc to be the default. Disables density screening by default Releases memory allocated in basis_shell_S 5 Sep 2002 ccmp2_loc.2002.4 Fixes wrong norm in case local=1 5 Sep 2002 gzgets3.2002.4 Bug(s) fixed: 287 321 some libz libraries on SGI machines do not have gzgets() function which causes utilities/patcher to fail on linking This patch changes configure to test for gzgets() and if it fails then it does not set the -DZLIB flag 5 Sep 2002 buildpatch_sgi.2002.4 Bug(s) fixed: 293 Using tar with pipes on SGI machines can cause problems with mismatched block sizes. The tar B flag is added on SGI machines to correct for this. 5 Sep 2002 kext4ext.2002.4 Bug(s) fixed: 301 Moves kext4ext.f from lccsd to local to avoid problems without module lccsd 6 Sep 2002 numfreq4.2002.4 Modify test threshold in o3_ccsdfreq.test. 7 Sep 2002 dfmp2b.2002.4 Implements new df-mp2 using batchs routines. 7 Sep 2002 ccsd_mpptim.2002.4 Fixes wrong setting of iprint(39) in ccsd, leading to wrong print of total times in mp2 7 Sep 2002 transform2.2002.4 set variables CPUINT_DTRAF,CPUTRA_DTRAF,CPUTRANS_DTRAF in transform 7 Sep 2002 basis_check2.2002.4 Fixes a bug in bascop, causing blanks in basis strings. This could lead to restart problems (orbitals belong to different basis set) Also, modify basis_check to ignore all blanks in basis comparisons, in order to cope with the previous bug 7 Sep 2002 orbital_attributes4.2002.4 Bug(s) fixed: 309 331 Parse and store attributes for the ORBITAL basis set orbital_attributes2: nullify bases(:)%attributes to get round HP compiler problem orbital_attributes4: basis_contract.f must be -O1 on HP 8 Sep 2002 dfmp2c.2002.4 Implements new df-mp2 using batchs routines. 9 Sep 2002 loc_eom.2002.4 correct loc_eom.test file to make it compliant with the ccmp2_loc patch 10 Sep 2002 dfmp2r12.2002.4 Improved 3-index integrals for the DF-MP2-R12 program that make it a viable method. All integral types (ie 1/r12, r12, [r12,t1]) computed simultaneously taking full advantage of all the shared precomputation. A bug in basis_contract is corrected in a hither-to unused case. A routine basis_GPT_PR is added to basis_GPT; this computes the coefficients that define the modified AO products that appear in commutator integrals. A new integral shell routine is added to compute 2-index r_12 and 1/r_12 integrals simultaneously. 11 Sep 2002 cfit.2002.4 Uses standard rather than shifted linear algebra in RKS program as the latter appears to be introducing convergence errors. 11 Sep 2002 input_mods3.2002.4 1.) enable the use of input blocks like command{...} or command={...}. If any input ket within {...} is not a valid subcommand of command, an error will result. 2.) Changes input for cfit: cfit parameters can be given on a cfit card either inside (locally valid) or outside (globally valid) of an command{...} block. Global cfit input does not imply use of density fitting in hf/ks any more. Density fitting is now enabled by one of the following possibilities: (i) use a prefix df-, ri-, or cf- to the program under consideration (mp2 or hf/ks). (ii) add key cfit after command, e.g., hf,cfit (iii) give a cfit input within the command{...} block In all the above, dfit is an alias for cfit. 3.) Modify and somewhat cleanup ciinp.f: cfit parameters cannot be given any more on command line. They can be modified only using cfit cards as described above. Local parameters can now be given on the ri-lmp2 command line in the same way as usual for lmp2. 4.) Fitting basis parameter in mp2 is now BASIS_MP2. Also BASIS_CCSD is defined. 5.) The CFIT parameters for enableing cfit are CFIT_SCF, CFIT_MP2, and CFIT_CCSD, respectively for the cporresponding methods. 6.) Group command in basis input 12 Sep 2002 dft_direct2.2002.4 Fixes wrong B3LYP in direct case. For the time being, dfock1 is used 12 Sep 2002 mp2r12.2002.4 Preparation for mp2-r12 14 Sep 2002 contraction4.2002.4 Bug(s) fixed: 319 Get the definition of contraction coefficients consistent in basis.f and the things that depend on it. The definition is now consistent for orbitals and other sets, and has been documented. contraction3: a) case-independent matching of basis sets b) basis set names now 16 characters. contraction4: basis set names now 24 for good measure, and make sure this is done in calling routines. 15 Sep 2002 dfexch7.2002.4 Implements df-exchange for scf. For cfit=3 also coulomb is computed. 15 Sep 2002 contraction_cfit.2002.4 Follow up the mess of contraction4 putting cfit_fock in the wrong directory 15 Sep 2002 df_forces.2002.4 use variables to detect density fitting possibilities in kraft and krafta set variables !DFSCF, !DFMP2, !DFBASIS_COUL, !DFBASIS_EXCH, !DFBASIS_MP2 as used in the corresponding programs. Gradient programs cfit_force and ri_lmp2g_accG pick these up. Also, replace thrdscf in cfit_fock by cfit paramter thrao_scf 15 Sep 2002 dfmp2_basisname.2002.4 Change character length of basis_sets to 24 16 Sep 2002 cfitmp2a.2002.4 changes CFIT to CFIT_MP2 in ccdomain.f and asmbl_k.f 17 Sep 2002 newmyrinetformat.2002.4 mpich-1.2.4..8 (the myrinet adapted mpich library) introduced a completely new set of input parameters; the procgroup file now is identical to the underlying mpich library. This patch includes the respective changes to molpro.c. When running ./configure one can specify the version of the mpich library. 17 Sep 2002 getinp.2002.4 Modifications of getinp.f: 1.) additional last parameter in default_inp* for "program" 2.) Alias names for input keys can be given separated by : New input set THRESH in preparation of registry stuff, and use of this in multi and mrci ini_inp now called within programs for MRCI, MULTI, SCFPRO and FORCES. 17 Sep 2002 getinp2.2002.4 Fixes bug in getinp.f:check_inp 17 Sep 2002 cfitmp2b.2002.4 Corrects bug in ciinp setting cfit_mp2 to wrong value 18 Sep 2002 registry7.2002.4 Implement registry. 18 Sep 2002 variable_registry.2002.4 Reads variable.registry and initializes system variables. Sets and handles types of variables. 18 Sep 2002 checkinp.2002.4 Checks variable settings in input file to avoid later errors due to attempts of overwriting read-only variables 19 Sep 2002 charge2.2002.4 Variable CHARGE can be used to define the total charge of the system. WF card input now processed via util.f:get_wf. Either old input can be used, or keys NELEC:ELEC=..., SPIN:MS2=, SYM[METRY=..., CH[ARGE]=... 19 Sep 2002 registry_speed5.2002.4 Speeds up load_registry_file: 1.) remove adjustl, len_trim etc by explicit pointer to the characters in buffer. This reduces elapsed time for load_registry_file by a factor of 4-5. 2.) Reduce length of result%values from 10240 to 1024 (temporary). This has a dramatic effect on the time needed for add_registry. Also, the timings for query_registry appear to be proportional to the length of result%values. For further improvement, the lengths of result%values should be stored and used in further processing. These lengths are already passed to add_registry but not yet used any further. Or maybe the long descritions should never be processed by molpro but only be a separate program for generating the documentation? Then all strings could be much shorter. Adds common/cmpp to load_registry_file (iprocs undefined). Initializes blaslib in syscon (undefined for hp) Fix load_registry_file: do not use lbound, ubound if variable is not allocated otherise crash on hp-itanium call express before syscon to avoid crash on hp (unknown reason) 20 Sep 2002 getvar2.2002.4 Fixes bug in getvar, overwriting constants increases character length of bef in aosortv.f 20 Sep 2002 registry_highspeed2.2002.4 Provide a sinmple set of routines to parse a registry file. Use this for the variables. 20 Sep 2002 getinp5.2002.4 Moves keyword lists from include files into gthresh_init 20 Sep 2002 getvar4.2002.4 fixes bug in setvar 20 Sep 2002 check_lic.2002.4 Adds a little subroutine to call express before syscon. This avois crashes on hp and sun for unknown reasons! 21 Sep 2002 read_defaults4.2002.4 Reads registered defaults Fixes modified sortint values for cpp Adds thresh.registry 21 Sep 2002 parse_registry_mpp.2002.4 Fixes errors for mpp 21 Sep 2002 aoint_mpp.2002.4 Fixes problems with compression setting for mpp 21 Sep 2002 timing.2002.4 Modifies timing to avoid setting of system variable. 22 Sep 2002 locinp5.2002.4 Cleans up input for local. Now all parameters in registry file and treated by getinp. 23 Sep 2002 hfma_dft_mpp.2002.4 Fixes a bug in hfma arising in dft with exchange and mpp 23 Sep 2002 lib_makefile_wildcards.2002.4 Adding wildcards to lib/Makefile so registry files are copied into installed lib directory 23 Sep 2002 getinp6.2002.4 check for correct parameter type in input parsing 24 Sep 2002 xlf_zerosize.2002.4 adds -qzerosize to compiler flags in utilities/ibm.conf 24 Sep 2002 ddv.2002.4 Circumvents compiler problem with IBM xlf 8.1 24 Sep 2002 ibmconf_power4a.2002.4 Adds pget2.f to default O2 list (fails with 8.1) Treats power as power3 Makes blas4 default if essl is present Removes all NO_OPT and F_OPT entries except pget2 for version 8.1 25 Sep 2002 basis_alloc.2002.4 Fixes properly the problem on the IBM identified by HJW with allocated space not being deallocated in some integrals routines. 25 Sep 2002 bench3.2002.4 Adjust benchmark inputs to new variable handling: set command used for setting system variable program, other old-style set replaced by normal variable settings. 25 Sep 2002 dtraf_flush.2002.4 Separate routine for index flushing before flushing buffer to avoid compilation optimization error. 26 Sep 2002 ibmconf.2002.4 Bug(s) fixed: 292 Ensure C symbol AIX42 is defined for AIX5 as well as AIX4. Without this, 64-bit I/O is impossible. 26 Sep 2002 abdata_parallel_make.2002.4 Bug(s) fixed: 351 when doing a parallel make in the lib directory both ABDATA and RYSRW tried to compile utilities/cnvrt.exe at the same time causing an error. 26 Sep 2002 block.2002.4 Routine to provide a matrix in a specified block pattern for the blocked df-lmp2 program and related projects. 27 Sep 2002 registryf5.2002.4 More updates to registry.f 27 Sep 2002 aosort_nread.2002.4 Fix erroneous global_sum_int of nread in aosortv: it is now real*8 28 Sep 2002 prepin.2002.4 Keeps blnak lines in xyz input so that comment line may be blank 29 Sep 2002 mccoreorb4.2002.4 Fixes bug in muinp1: core card was ignored if no record was given. Now uses most recent orbital record of the current geometry. scf reads wf card also in geometry optimizations, to allow different charges in scf and in multi. Replace iout by 6 for error messages in readdump, writedump, scfpr0. Modify putorb/getorb so that ityp is 1 for closed-shell and 2 for open shell, independent of usage of hf or rhf 29 Sep 2002 cpp_enuc4.2002.4 Fixes wong core energy after restart (polarisation contribution was forgotten). Data information was not written or written with wrong offset in semi_int and cpps. Sets variable ENUC whereever ekern is remade. Consisent use of ekern (current nuclear energy) and ekernt (initial nuclear energy without fields). 30 Sep 2002 dfmp2r12_2.2002.4 Updates to MP2-R12 program, particularly just cleaning things up, correcting bug in input, and bringing everything in line with the variable registry. 30 Sep 2002 tasks.2002.4 Bug(s) fixed: 352 Fixes rounding error in floating->integer conversion in tasks.f. This fixes error with 4 processors on ibm 1 Oct 2002 mp2r12_variable.2002.4 Bug(s) fixed: 355 Closed apostrophe in variable registry. 1 Oct 2002 registryf6.2002.4 Bug(s) fixed: 354 Fix syntax problem in registry.f. 1 Oct 2002 dftdebug.2002.4 Remove debugging output from grid restart 2 Oct 2002 basis_save.2002.4 Generalized version of basis_save 3 Oct 2002 patcher3.2002.4 Bug(s) fixed: 312 313 353 343 patcher: Fix dependency iteration problems in patcher. patcher2: Fix failure for new users if ~/.molpro doesn't already exist patcher3: Fix failing test for prerequisites 3 Oct 2002 thresh.2002.4 Adds input for thresholds to some benchmarks, since defaults in the presence of gradients changed. Modifies handling of thresholds in ciinp/cidflt. 3 Oct 2002 ibm_strip2.2002.4 Bug(s) fixed: 349 giving -X 32_64 flag to strip on ibm machines 3 Oct 2002 semi_int.2002.4 fixed undef intrel in semi_int.f 3 Oct 2002 cadpac_mpp.2002.4 parallelizes 2-electron part in cadpac! 3 Oct 2002 doc_getinp.2002.4 Bug(s) fixed: 360 fixed an uncommented _ in documentation of src/util/getinp.f 3 Oct 2002 tuning_format.2002.4 Bug(s) fixed: 357 malformed fortran write statements caused **'s to be printed in molpro.rc. Fixed by adding more digits. 3 Oct 2002 tuning2.2002.4 Fixes some errors in tuning.com Now I have to fix another one that I put in .. Next try! 4 Oct 2002 registry_doc.2002.4 Some automatically grepped short_documentation for registry files 4 Oct 2002 integral_interface.2002.4 Provides an interface for external integral programs. Geometry and basis info is passed to to this program 5 Oct 2002 use_registry.2002.4 Bug(s) fixed: 358 Start using the real registry routines 7 Oct 2002 ecpgrad.2002.4 8 Oct 2002 registry_static.2002.4 Bug(s) fixed: 358 Use some static, but checked, dimensions in registry routines 08 Oct 2002 ibmconf64.2002.4 Sidestep compiler optimization problem 64-bit, IBM, xlf 7.1, blas/mxm4.f 8 Oct 2002 dflccsd.2002.4 implementation of local CCSD with density fitting for 4-ext integrals (kext) 8 Oct 2002 brueck_save.2002.4 Saves scf occupation numbers (2.0) to brueckner record to get molden working 9 Oct 2002 prope.2002.4 Fixes wrong gradient in case that so pseudopotential is given in input 9 Oct 2002 transform3.2002.4 Reverts some temp changes related to dtraf_mpp4 (now superceded) 9 Oct 2002 mp2dip.2002.4 Changes memory handling in cphf to avoid error in mpp case 9 Oct 2002 ciinp.2002.4 Changes 6 to iout in print of convergence thresholds 9 Oct 2002 bench_direct.2002.4 Removes mpp dependent if in direct.com. Now dmp4 is parallel 9 Oct 2002 dtraf_mpp8.2002.4 parallelizes drv2el_pg.f, drv2el_pgk.f, and drv2el_half.f Removes parallelisation of sortopk in dkexts.f since not ok with df files. 11 Oct 2002 brdcst_length.2002.4 Restricts the length in calls to ga_brdcst to a maximum (adjustable parameter in global_brdcst) to avoid crashes in ga_brdcst for very long vectors. 11 Oct 2002 iom_formats2.2002.4 Inreases format length in iom.f record name (now i8) and length (i12) 11 Oct 2002 dtraf_bug366.2002.4 Bug(s) fixed: 366 adjusts the memory precomputations to take into account the buckets now allocated before the k4 loop 11 Oct 2002 df4ext_cache2.2002.4 implements caching of 2-ext integrals when constructing the 4-ext integrals 11 Oct 2002 hpfixes2.2002.4 some fixes for hp 13 Oct 2002 inicor_ibm.2002.4 Replaces unix-ibm by unix-ibm_old in inicor (util.f) and cclist (ciutil.f) to avoid different memory allocaltion on different processors. Modify prical (util.f): removes old interactive stuff, and add checks for consistent base addresses and free memory on different processors (mpp only). These checks are done with print,test. Add gprint,test option. 13 Oct 2002 molpro_help5.2002.4 Fixes parse_registry to properly handle strings. Adds documentation to thresh.registry. Adds command.registry. Implements molpro help command. See example/help.com or documentation. Disables registry.f since this caused various problems. Also, inivar is very slow in the mpp case if registry is used (12 sec. on ibm 4 proc). 14 Oct 2002 multi_canorbspin.2002.4 Allows to select spin in addition for state for natural/canonical orbitals 14 Oct 2002 reavar2.2002.4 Fixes reavar to treat things like fockdone, cppdone etc correctly 15 Oct 2002 lmp2bench2.2002.4 Sets mpptype=1 to avoid slightly different results for different number of processors (prescreening effect). Also sets the default mpptype=1, since also better performance. 15 Oct 2002 dflmp2gfix.2002.4 fixes fro DF-LMP2 gradient 16 Oct 2002 ibmconf_dtfint2.2002.4 Reduces optimization of dtfint2.f for xlf 8.1 since test isobutene_mp2 gives wrong results with -O3 17 Oct 2002 comment_registry.2002.4 Comments registry.f to avoid internal compiler error on ia64 linux 17 Oct 2002 gradnoprint.2002.4 printout of ecp and basis-set parameters is suppressed in ecp gradient calculations with alaska 17 Oct 2002 makepatchdef_findfiles3.2002.4 Bug(s) fixed: 370 looking for .registry and .test files in develop directory and adding to default file list 17 Oct 2002 iom_latex.2002.4 Bug(s) fixed: 368 src/util/iom.f failed with latex preprocessor codedoc. I moved cend flag to after subroutine set_record_mppstate to fix this. 17 Oct 2002 soft_links2.2002.4 soft links created to VERSION in develop and develop/progman directories were absolute rather than relative. Changed codedoc in develop to rectify the problem. 17 Oct 2002 inttypes2.2002.4 Bug(s) fixed: 359 inttypes.h not available on some dec platforms. Check using a test program in configure and if it is then add -DHAVE_INTTYPES_H to cdef and then changed machine.h with a #ifdef HAVE_INTTYPES_H statement. 18 Oct 2002 basis_list.2002.4 Bug(s) fixed: 369 corrects 3-index force code to work for lists of basis sets. 18 Oct 2002 opt_bccd.2002.4 Forces numerical gradient for bccd 19 Oct 2002 thrkex2.2002.4 Bug(s) fixed: 342 Fixes use of sparsity in kext.f if thrkex.ne.0. The length passed to kext was loo large, causing access of undefined numbers. 19 Oct 2002 cfit_opt2.2002.4 Bug(s) fixed: 341 Modify gradient program to take into account exact exchange for df-scf and df-ks with, e.g, b3lyp. Gradients should now work for all df-hf, df-rhf, df-uhf and the corresponding dft cases. 21 Oct 2002 dfexch8.2002.4 Sparse fitting algorithm in dfexch.f Adds missing goto 100 after cfit input in scfpr0.f 21 Oct 2002 matrop_multi.2002.4 Fixes recognition of multi command following matrop 23 Oct 2002 lccsd_discon+mpp.2002.4 Remove disconnected terms in lccsd for idiscon=0 Make idiscon=0 default for local, change testjob results accordingly. parallelize ccv_loc4.f 23 Oct 2002 necsx2.2002.4 Bug(s) fixed: 375 Port to NEC SX6 23 Oct 2002 lattice.2002.4 Gradient of energy with respect to movement of lattice points. Invoked by input 'force,lattice' for a quiet appending of the 3*n lattice gradients to record 800 (the standard place for lattice data) or 'force,lattice=ipr' with ipr=1 for low-level output and ipr=2 for all lattice forces. 23 Oct 2002 dec_build.2002.4 Fix build error with Compaq Fortran compiler on alpha Linux. 24 Oct 2002 pspace.2002.4 Increases max number of p-space configurations to 200 in mrci and to 100 in multi. Increases max number of spin couplings to 1500. Increases max number of state symmetries in multi to 16. New threshold thrpsp:pspace for checking p-space gradient in multi 24 Oct 2002 thresh_thrpsp.2002.4 Bug(s) fixed: 377 Make the threshold thrpsp valid for all programs 25 Oct 2002 fixdf4ext.2002.4 fixes bug in DF-4ext for DFLCC 25 Oct 2002 displace2.2002.4 Bug(s) fixed: 318 Provide a routine (displace) for generating symmetric vibrational coordinates for numerical geometry optimization 25 Oct 2002 dec_build2.2002.4 Fix build error with Compaq Fortran compiler on alpha Linux. 26 Oct 2002 dgmat.2002.4 Fixes symmetry bug in dgmat 27 Oct 2002 numopt3.2002.4 Use symmetrical displacements in numerical optimization unless z-matrix coordinates are used Replace write(6 by write(naus in scfpr0 Fix icphf in ciinp for mp3/mp4 cases New Testjobs for numerical optimization in displacement coordinates 27 Oct 2002 nocoul_gmat.2002.4 Implements test option NOCOUL_DGMAT to skip coulomb contributions in g(e) and 1-ext. contributions to beta in local=4 case. 28 Oct 2002 dgmat1.2002.4 Fixes symmetry bug in dgmat1 28 Oct 2002 sun_gmalloc.2002.4 makes gmalloc=0 default for all sun systems 28 Oct 2002 zmat_symcheck.2002.4 Bug(s) fixed: 318 zmatrix now always checks symmetry on restart 28 Oct 2002 fixopdom4.2002.4 fixes opdom==4 case in lccsd. i3ext=2,3 not implemented for opdom==4 case, so we go with i3ext=1 28 Oct 2002 cc3ext_loc_fix.2002.4 corrects calls to cc3ext_loc made wrong in all_eom 28 Oct 2002 nec_cstart.2002.4 Bug(s) fixed: 374 mxm.f failed to compile due to misaligned preprocessing flags 28 Oct 2002 all_eom5.2002.4 all new local eom + changes in calls to cckext,cc3ext_loc etc. in all places, move EOM keys from local to eom registry + corresponding changes in get_inpx correct some recent undefs 29 Oct 2002 cceom_loc.2002.4 Fixes undefined variable n_s_v in cceom_loc, causing overflow of variables. 28 Oct 2002 lsluccsd.2002.4 makes luccsd faster 29 Oct 2002 hayecp_cs.2002.4 Fix a small typo in hayecp.libmol: Ca ECP2 --> Cs ECP2. 29 Oct 2002 gopenmol_ucase.2002.4 Bug(s) fixed: 378 Fixed problem with uppercase letters in directory names being clobbered. 29 Oct 2002 buildpatch_parents.2002.4 changed from using [[ ... ]] to [ ... ] conditional expressions in order to avoid buildpatch failing to create correct parent list with old versions of bash 29 Oct 2002 newmyrinetformat2.2002.4 This patch makes sure that the specified myrinet input format is effective independently from the chosen GA version. This is important for people who frequently switch GA library versions. 30 Oct 2002 warnings_eom2.2002.4 get rid of some compilation warnings, change the first character to blank in outputs 30 Oct 2002 eomloc.2002.4 Adds missing character statement in opdom_eom.f Fixes some formats in cceom.f and eomloc.f 30 Oct 2002 formats.2002.4 Adds missing commas in various format statements 30 Oct 2002 luqcisd.2002.4 fixes bugs for call cc3ext_loc and for lucisd,luqcisd files uccsd_loc.f src/lccsd/ 30 Oct 2002 multi_mxact2.2002.4 Checks if mxact and mxval are exceeded. Fixes wrong print of number of CSFs in certain cases with restrict cards 30 Oct 2002 call_cckext.2002.4 add a missing parameter in calling cckext 31 Oct 2002 ciinp_rotate.2002.4 Makes the default for orbital rotation 90 degrees (was zero) 31 Oct 2002 gmalloc_remove.2002.4 Remove all references to gmalloc. 31 Oct 2002 hpconf_1122a.2002.4 Adapt hp.conf to most recent compiler version for hp-ux11.22 Add HP compiler flags ta mxmadd.f to avoid optimization errors due to overlapping aguments 31 Oct 2002 klen0.2002.4 initialize klen in cceom_loc 1 Nov 2002 follow2.2002.4 activate FOLLOW (root-homing mechanism) for local>1, change the testjob accordingly 1 Nov 2002 mpham.2002.4 Implement transformed hamiltonian H(alpha)=H(0)+alpha*(H-H(0)) where H(0) is sum of Fock operators. 1 Nov 2002 buildpatch_new_patcher2.2002.4 Make warning message about upgrading patcher more truthful 1 Nov 2002 patcher_available2.2002.4 Bug(s) fixed: 383 Proper filling in of available list for already installed patches 1 Nov 2002 syntax.2002.4 Bug(s) fixed: 385 Fix various benign fortran syntax errors 2 Nov 2002 cihtran.2002.4 Fixes htran (orbitals missing) 2 Nov 2002 cckext_loceom.2002.4 Fixes endless loop in mpp case in cckext.f 2 Nov 2002 stacktrace.2002.4 Implemnts a subroutine stack. The stack is updated by calls to either citime, or pricall/pridone (citime calls pricall/pridone internally). pricall and pridone must be called pairwise, as citime. print,cpu is the minimum needed to obtain a reasonably complete stack. If print,test is given, each subroutine call is printed. if print,test=1 also the current stack is printed at each call. print,test implies print,cpu. The subroutine stack is printed when an error occurs (fehler). Most calls to accnt have been replaced by calls to citime. 3 Nov 2002 mp2_fitbasen.2002.4 include ahlrichs mp2 fitting basis svp, tzvpp and HJW's poisson bases. Only for H,C,N,O,F so far. Aux sets include s-functions since otherwise libmol gets confused 3 Nov 2002 mp2_fitbasen2.2002.4 include full set of ahlrichs mp2 fitting bases svp, tzvp and tzvpp. 4 Nov 2002 mpham_more.2002.4 Implement transformed hamiltonian H(alpha)=H(0)+alpha*(H-H(0)) where H(0) is sum of Fock operators. 4 Nov 2002 newmyrinetformat3.2002.4 Bug(s) fixed: 390 Fix for a logical mistake that left MPIGM_FORMAT uninitialized. The bug showed up when building tcgmsg or mpi versions in cases where no CONFIG file existed. 4 Nov 2002 ftconv_width.2002.4 Bug(s) fixed: 391 Maximum line length in ftconv increased from 255 to 1024