changelog

一个很好的分子动力学程序

      (3) Add random number generator to parse.c

March 26, 1998
Version 2.4.35
      (1) Removed change (2) above from version 2.4.34, and
          added 0.5 in input function of RCV and COR.

March 26, 1998
Version 2.4.34
     (1)  Fixed bad bug that happens with FILL command.
          In cdfill.c  called AddParticle()  which called
          ReallocateParticle() in cdalloc.c.  The problem
          was that when ReallcateParticle() allocated new
          storage, it would update the value of a->np.
          But AddParticle() did not anticipate that, and 
          it would increment a->np, resulting in an extra
          particle with coordinate 0.0.0 inserted in array.
     (2)  Found that WRITE RCV and WRITE COR always have
          negative error.  Have added 0.5 floating value
          version of integer coordinate, this rounds it 
          properly. 

March 20, 1998
Version 2.4.33
     (1)  Fixed AddPariticle() routine in cdfill.c.
          It mistakenly incremented a->np before calling
          RealloateParticle(), but that function expected
          a->np to hold the old number of particles.
     (2)  In CalcAngularVelocity() in cditemp.c, was incorrectly
          using if (Select[ipart]) instead of if
          (IS_SELECT2(Select,ipart)).
     (3)  In cdalloc.c, ReallocateParticle(), add particles in
          groups of no less than 1024 (NUM_ALLOC_INC).

March 11, 1998
Version 2.4.32
     (1)  Remove annoying debugin message (done?).
     (2)  Perform particle wrap before WRITE PARTICLE
          (Problem reported by Mario G. Del Popolo)
     (3)  Correct WRITE TEMP command to correct for fixed particles.
          (Problem reported by Mario G. Del Popolo)
     (4)  Fix read_constrain().  Previously assume each call was
          the first call to read_constrain() and it would re-allocate
          the a->cons[] array.  Now it checks to see if array is 
          needed before doing the reallocation.

Feb 4, 1998
Version 2.4.31
     (1)  Added test for PRESSURE CLAMP with no repeating boundaries
          in cdmd.c CheckSimpleErrors().
     (2)  Add Stillinger-Weber Potential (cdstil.c).

Feb 2, 1998
Version 2.4.30
     (1)  WRITE STATE command previously appended blanks to
          end of output file name if present in *.xm file.
          This is now fixed.
     (2)  In cdcsi.c, replace dynamics with static allocation
          for BondZetaD_m, BondVector_m, TotalCutoff_m, TotalCutoffSqr_m
     (3)  In cdeam.c, re-arrange Embedding calculation so that
          "Electron Density Out of Range" error accompanied by
           index of offending atom.

Jan 27, 1998
Version 2.4.29
     (1)  Fixed error in units for PRESSURE EXTPRESSURE

Jan 27, 1998
Version 2.4.28
     (1)  Used new -c command line option to find memory leak,
          was in neigh.c.  NeighRadSqr[] was ALLOCATE()'ed at
          line 281, but was never FREE()'ed.

Jan 26, 1998
Version 2.4.27
     (1)  Add -c command line option to dump memory information

Jan 7, 1998
Version 2.4.26
     (1)  Add pair potential to Tersoff.  Can have up to
          5 atom types. Types 1,2 are C,Si, and can additionally
          include pair potentials.  Types 3,4,5 are pair only.
     (2)  Updated cddoc.c to include missing commands.
     (3)  Renamed VOLUME command PRESSURE (VOLUME still works though).
     (4)  Added external pressure to PRESSURE CLAMP command.

Dec 26, 1997
Versoin 2.4.25
     (1)  Fix compiler  warnings from Borland C++ 4.02
     (2)  Fix error reported by R. Ge (Dec 24, 1997)

Dec 16, 1997
Version 2.4.24
    (1) cwrite.c, neigh.c cdsearch.c
        - replace bondd() in bondsub.c with CalcNeighborPerAtom()
          from neigh.c
        - Add SEL option to WRITE RDF command.
    (2) Add SEL option to ENERGY command (now WRITE ENERGY and
        WRITE EPOT are identical).

Dec 12, 1997
Version 2.4.23
    (1) cdstate.c - correct storage of Volume Clamp info

Dec 11, 1997
Version 2.4.22
    (1) cdstate.c - correct storage of a->UseClamp[] and stuff.

Nov 20, 1997
Version 2.4.21
    (1) cdsubs.c - correct read_clamp().  Was not working without SEL
                   option.
    (2) cdmd.c   - correct ClampVolume() to only scale box directions
                   with repeating boundary conditions

Nov 19, 1997
Version 2.4.20
    (1) Change VOLUME CLAMP command to use B instead of
        C11,C12,C13,C22..

Nov 17, 1997
Version 2.4.18
    (1) cdsubs.c - corrected read_volume() command.  Forgot to
                   read parameters correctly.
Nov 17, 1997
Version 2.4.19
    (1) particile.h - add IsInitializedSurface
    (2) cdalloc.c - initialize surf[] to true,
                    IsInitializedSurface to false
    (3) cdsubs.c -
        read_box() Set surf[] to false if first box call and
    (4) cdsubs.c -
        read_surf() Set IsInitializedSurface to TRUE

Nov 14, 1997
Version 2.4.17
    (1)  cdfill.c - fix Orientation calculation.  Was using
        Orient_m instead of OrientToReal_m and ...


Oct 31, 1997
Version 2.4.16
   (1)  Add FILL command
   (2)  Add multiple temperature CLAMPs
   (3)  Add VOLUME CLAMP  (not done yet)
   (4)  correct cddoc.c, #define INT4 long  ->  #define INT4 int
        The previous version did not work on DEC Alpha.  The
        new one should work on DEC, Linux, AIX (but probably
        not 16bit DOS).
   (5)  plotc.c - initialize PlotInfo structure so that UseBond
        and UseDisplacement are explicitly intialized to FALSE
        (before this seemed to happend automatically on Linux, AIX,
        but not on DEC ALPHA).
   (6)  plotc.c - replace single call to gr_setrgbcolor() to
        gr_setcolor().  This prevents mis-setting colors in
        static variable within gr_setcolor().  When gr_setrbgcolor()
        was called directly, color commands got out of sync.
   (7)  grdevsub.c - change PS header so that Ghostview can
        recognize pages and present on-screen page list.

Oct 7, 1997
Version 2.4.15
   (1)  In cdcsi.c GetEnergy() ASSERT to check that different energy
        sums give same answer failed when energy is zero.  Have
        fixed it by changing  v1 < v2  to v1 <= v2.
   (2)  Add VOLUME QUENCH command to quench volume.

Sep 15, 1997
Version 2.4.14
    (1)  Found similar problems to those fixed in previous version.
         Have fixed them all by having strhed() truncate all
         white-space from end of tail string.
    (2)  In plot.c when error messages were printed program was not
         ending.  Have inserted calls to CleanAfterError() after all
         such error messages.

Sep 12, 1997
Version 2.4.13
    (1)  Uncovered bug, FIX, ECHO and REFSTEP commands.  When reading
         option used (!strcmpi(instr,"ON")) for example.  The problem
         is sometimes there is a trailing blank(s) on instr which
         would miss a match between "ON" and "ON ".  Solution is
         to use   tokstr = strhed(&instr)  which would truncate
         trailing blanks.

Sep 11, 1997
Version 2.4.12
    (1)  Uncovered bug, SURFACE command was not reseting neighbor list.

Sep 8, 1997
Version 2.4.11
    (1)  There was a bug when calling READ from within a REPEAT
         statement - the REPEAT statement would treat the
         EOF of the READ file as an END command.  Fixed by
         having read_read() scan entire file and remove it
         from list before returning.  Previously, read_read()
         simply added file to the list and let subsequent
         code deal with closing the file.
    (2)  Added ERASE file command.
    (3)  Added options AVG, MIN, MAX to WRITE PARTICLE, VELOCITY, FORCE
    (4)  Added WRITE POSITION alias for WRITE PARTICLE
    (5)  In cdmd.c ClampVelocity()
         VelocityFactor calculated using
          VelocityFactor = exp (0.5*log(s->tempc/CurrentTemp)/s->cstep);
         This would return NaN when s->tempc=0.  Changed to
          VelocityFactor = pow( s->tempc/CurrentTemp, 1.0/(2*s->cstep));
    (6)  In cdmd.c ClampVelocity().
            Was using IS_SELECT(ipart) instead of IS_CLAMP(ipart) when
         adjusting atom velocities.

Sep 4, 1997
Version 2.4.10
    (1)  Per requiest of Francios-Xavier and Selena Zajac,
         altered the SELECT NEAR command, adding options
         SET and INDEX, POINT, and when called without options
         using the first selected atom.

August 25, 1997
Version 2.4.9
    (1)  Update cddoc.c to reflect latest changes in commands.
    (2)  Update DUP command to include duplication of mass[] info.
    (3)  EAM POTENTIAL CLEAR command was fixed.  The string
         LeadingTokenPtr was not re-loaded after testing positive
         for string "CLEAR".
    (4)  There was a bug in WRITE EATOM command - it tried to
         use a->v[] to calculate Kinetic Energy even when
         a->v[]==NULL.
    (5)  Added nrand() routine to cdsubs.c to calculate
         normally distributed random numbers.  This is
         now the default velocity generator called by itemp().
         The previous generator - randfloat() - is specified
         the OLD option to the ITEMP command.
    (6)  Update DUP command to include duplication of velocity
         and higher derivatives.
    (7)  Modify format of POTENTIAL SCALE command from
            POTENTIAL SCALE PAIR 1 1  scale
         to
            POTENTIAL SCALE scale PAIR 1 1
    (8)  Update ITEMP to zero higher derivatives when initializing
         velocity.
    (9)  Fix WRITE RDF command to use no types.
   (10)  Add SEL option to WRITE TEMP, EPOT, EKIN

August 18, 1997
Version 2.4.8
   (1)  In cdeam.eam include FREE() Rc_m[] and RcSqr_m[] in
        _free_eam() function.
   (2)  Add ability to use atom coordinates symbolically as
        variables X,Y,Z in CONSTRAIN command (analogous to
        X,Y,Z in DAMP, EXTFORCE, EXTSPRING, and MOVE commands.
        Add EvaluateFormula() to cdsubs.h.
   (3)  Corrected csi_calcforce(), was not initailized AtomBondEnergy.
   (4)  Optimized EvaluateFormula() function.  Previously XMD was
        taking 0.52sec to move 4000 particles with MOVE 0.1 0.1 0.1.
        Now it takes 0.02sec.  Change involved calling
        AssignVariable() function from parse.c to set X=?, Y=?, Z=?.
        Previously was setting three separate strings "X=?", "Y=?",
        etc. and calling dblstrf() three separate times.

August 13, 1997
Version 2.4.7
   (1)  Correct SELECT command that prevented combination of
        previous selections with current selection (the AND, OR and XOR
        options)
   (2)  Correct REMOVE command - it was not saving tag() information.
        This was corrected by modifying CopyParticle() in cdalloc.c.
        Also, tag() was improperly handled in other cdalloc.c functions.
        In particular, whatever is done to a->Selection() should also
        be done to a->tag() (with the exception that a->Selection()
        is allocated by default, a->tag() only when needed).
   (3)  Modified readstate() to correctly update a->IsInitializedMass
        and a->IsInitializedPotential flag.  Also in readarray()
        use FREE() on array before ALLOCATE().  This removed annoying
        message.
   (4)  Added KE to WRITE EATOM command.
   (5)  WRITE EATOM and WRITE EPOT now perform energy calculation
        (instead of just printing most recent values).

August 11, 1997
Version 2.4.6
   (1)  Add auxillary pair potential to cdcsi.c
   (2)  In cdeam.c was freeing a pointer twice, which ultimately
        resulted in a run-time segmentation fault.

10 Jun 1997
Version 2.4.1
   (1)  Add CLAMP SEL command (lost original version, not fully
        restored yet)
Version 2.4.2
   (1)  Add PLOT SIZE command
   (2)  Redirected all messages that were sent to stderr to stdout.
   (3)  Removed need for allocate command.
Version 2.4.3
   (1)  Add select option to ITEMP command.
Version 2.4.4
   (1)  Correct bug with SELECT SET command.
        Caused by misplaced ")" in IS_SET() macro.
Version 2.4.5
   (1)  Correct bug with DUP command.
        read_dup() was not recorder new number of particles in variable
        a->np.

        I have put an error checking statement in cdcsi.c GetEnergy()
        to stop the program when the radius is zero.

20 May 1997
Version 2.3.13
   (1)  Uncovered bug in cdcsi.c.  When two particles had zero radius
        between them, division by zero happened in GetEnergy().
        Strangely enough, S. Wang had an input file that would
        trigger this bug on Rose, but not on Aster.  I assume that
        there was a differnce in round-off error between the two
        machines when the second particle was wrapped in coincidence
        with the first.  On ROSE (a P5-133) the radius was observed to be
        exactly zero.  On ASTER (a P6-200) I assume there was a slight
        round-off error which made the separation other than zero.

23 Apr 1997
Version 2.3.11
   (1)  Corrected bug in PLOT SYMBOL FILL command.  Previously
        "FillFlag" variable was uninitialized.
12 May 1997
Version 2.3.12
   (1)  Correct REPEAT { CORE | RCV } command.  Did not properly
        initialize UseSkip.  This caused routine to implement
        step skipping with bogus value of Step increment.

3 Apr  1997
Version 2.3.10
   (1)  Added REPEAT RCV option.  Previously had  REPEAT COR option.

26 Mar 1997
Version 2.3.9
   (1)  Corrected SELECT PLANE command to read plane normal as float
        instead of integer.
   (2)  Added a color option to PLOT SYMBOL command.   This involved
        modifying plot.c, grdevsub.c and grdevsub.h

4 Feb 1997
Version 2.3.8
   (1)  Added the WAVE command for S. Wang's stacking fault simulations.

5 Feb 1997
Version 2.3.7
   (1) Found that ReadCorFile() was not setting a->boxflag TRUE.  This
       resulting in the diabling of the WrapParticle() function.
       Have added line a->boxflag=TRUE; to  ReadCorFile().

3 Feb 1997
Version 2.3.6
   (1)
   Replaced older neigh.c neighbor routine with simplified routine,
   because Elena's run (see aster://~/cdev/bug/bug2) would have
   an abrupt rise in the total energy at around step 361.  The
   simplification involved removing a mapping from real to integer
   coorrdinates before sorting.  This might have been efficient
   in older processors, but not was just making the routine
   harder to understand and probably hiding a bug.
   (2)  WRITE STRESS command now prints stress*vol/np like
        ssave has always done.
   (3)  In cdmd.c Stress[] had wrong sign in formula for
        BoxForce[].
   (4)  If starting a run by reading particles from a COR file,
        XMD would give an error when you tried to use them, saying
        that they were not initialized.  Have corrected this
        by setting the initialization flag in ReadCorFile() function.

28 Jan 1997
  (1) Repleced cell4.c with older neigh.c neighbor routine because
   (a)  there was a failed assertion when doing a run of Elena's
   (b)  S. Wang's run would take forever when a particle left the
        surface of the material and floated away.  This would
        cause the neighbor routine in cell4.c to work very hard
        in created a large neighbor of cells, most of which
        were empty.
  (2)  Added WRAP routine to read_particle and read_surface that
       also prints out warning when particles have been wrapped.