Data File Specification for Program GPRLSA

     The original version of GPRLSA, called PROLSQ, was written by Wayne 
     Hendrickson (1978).  The current program version, GPRLSA, has been 
     modified by W. Furey (University of Pittsburgh). The writeup is presented
     in cooperation with Star Technologies, Inc.

Input files to GPRLSA are described as follows:

  Unit		    Description

ISYSR = 5	Control Cards

IATMR = 10	Unformatted output file from PROTIN containing atomic
                coordinates, distances and codes, planar groups, etc.
		This file can be used for several cycles, but it should be
                regenerated periodically to update the nonbonded contact list
                or when converting from overall to individual thermal factors.

IREFR = 20	Observed reflection data and scattering factors.  This
		unformatted file was created by HKSCAT.

                The file contains records of the form:

   H, K, L, Fob, sigma F, sin theta/lamda, ( FII( N ), N = 1,8), Afix, Bfix
   -------
      |     				Afix, Bfix normally zero, but fixed
   packed in                            atom contributions to Fcalc (on an
   one word                             absolute scale) can be added here i.e.
                                        hydrogens, anomalous dispersion
                                        correction etc.
					

where:


FII(1) = Scattering factor of Carbon at corresponding sin theta/lamda value.
FII(2) =     "        "    "  Nitrogen        "        "    "     "     "  .
FII(3) =     "        "    "  Oxygen          "        "    "     "     "  .
FII(4) =     "        "    "  Sulphur         "        "    "     "     "  .
FII(5) =     "        "    "  Iron (++)       "        "    "     "     "  .
FII(6) =     "        "    "  Hydrogen        "        "    "     "     "  .
FII(7) =     "        "    "  Zinc (++)       "        "    "     "     "  .
FII(8) =     "        "    "  Calcium (++)    "        "    "     "     "  .  

       This file will not change unless new reflection data is obtained or
       fixed atom contributions are to be included or updated by another
       program.



ISHFTR = 15	Shifts in refined parameters obtained from previous cycles.
		(For each cycle the output file from the previous cycle should
                be used as input here). The file contains accumulated shifts
                from all cycles since the last run of PROTIN. It is an
                unformatted file. Read only if JABN (card 10.1) > 0.
		


        Output files generated by GPRLSA are described as follows.

   Unit                  Description

  ISYSW = 6	Messages and results.

  IDISK = 3	Scratch file.  The program stores the structure factor
                data of the accepted reflections in this file.  The file
                consists of unformatted records, each containing:
		
		H, K, L, 1., Fobs, sigma F, sin theta/lambda, 
                        (FII( I ), I = 1, 8), Afix, Bfix

                H, K, and L are stored in a vector, H(N) where N = 1 to 4.
                H( 4 ) is always equal to 1., as indicated above.
		(Historically the program could accept up to 18 scale
                factors and this value, H( 4 ),  would indicate which one
                to use for the present reflection.  The program now accepts
                only one scale factor SC = SC (1) thus H(4) = 1. )

  JDISK = 4	Scratch file, unformatted and same information as in file
		IDISK but for the NSAMPL reflections selected for RTEST
		only.  Not used if IRTEST = 0 (card 11).

  ISHFTW = 16	Contains parameter shifts from current cycle plus shifts from
                previous cycles, to be used as input (file ISHFTR) in the next
                cycle.

  IFOFC = 31	Optional - generated only if REPORT (card 2) > 0. Contains
                structure factors calculated from current model.  This
                file is formatted with records consisting of:

		H, K, L, Fob, Fcalc, phi          (phi in degrees)
 
                in format (3I4, 2F10.2, F7.2)

  IXYZB = 32	Optional - generated only if REPORT (card 2) > 0. Contains
                atomic parameters for the current model i.e. all shifts from
                previous cycles are first applied. Records contain:

		I, ATOM(I), X(I), Y(I), Z(I), BETA(I), Q(I)
	
                in format (I5, 1X, A8, 3F9.4, 2F6.2)

                x, y, z in Angstroms along cell edges.


 File ISYSR,  unit 5 - Control Information

      The following section describes the input control card data read
      from file ISYSR.  Card numbers are enclosed in parenthesis.

  (1)  Title					FORMAT ( A )

		Title to be printed in output listing (only first 40
		characters are printed).




  (2)  General Parameters			FORMAT ( 16I5 )

     1- 5  NCYCCG   Number of conjugate gradient iterations performed in
		    subroutine CGSOLV.  If less than or equal to 0 the
                    program sets NCYCCG = 50.

     6-10  LISTF      = 0   Do not print calculated structure factors. 
                      = 1   Print the complete list.
		      = N   with 1 < N < number of observations:  Print
                            a list of N structure factors selected at
                            intervals of ( NOBS / N ).

    11-15  LISTA      = 0  Do not print the list of parameter shifts.
                      = 1  Print the complete list of parameter shifts.
		      > 1  Print parameter shifts for atoms having a total
		           positional shift exceeding 0.01*LISTA angstroms.

    16-20  LGX        = 0  Do not constrain origin on corresponding axis.
    21-25  LGY        = 1  Constrain, (for polar space groups).
    26-30  LGZ

    31-35  LQ 	    Associated with scaling Lagrange multiplier elements
		    for constraining origin on polar axes. If LQ =< 0 
                    the program sets LQ = 10.


    36-40  REPORT     = 0  normal refinement run - Do NOT write special
		           output files (coordinates and structure
                           factors).  Printing of structure factors, as
                           determined by LISTF, is independent of this
                           parameter.
                    
                      = 1  refinement run  - Also write current coordinates
                           and calculated structure factors in files 
                           IXYZB = 32 and IFOFC = 31 respectively. (Note
                           coordinates and phases are for model INPUT to
                           this cycle, i.e. apply all previous shifts prior
                           to calculation).
                    
		      = 2  skip refinement.  Calculate structure factors
                           based on current model (i.e. apply all previous
                           shifts prior to calculation), and write coordinate
                           and structure factor files as with REPORT=1.
                    

    
    41-45  IDALIZ     = 0  Idealization plus structure factor refinement.
                      = 1  Idealization only.  WARNING:  Selecting this
                           option causes NOCC and ITEMP (see card (3)) to be
                           set to zero, even though you may not notice it!
	                   PDEL (card 8.4) should be set at about 0.10 to
                           assure good conjugate - gradient behavior.

    46-50  INCFIX     = 0  for normal operation.
                      = 1  to include fixed atom contributions to structure
		           factors for each reflection.

    51-55  NFCYCL     =    Number of refinement cycles (if REPORT=2 this is
                           automatically set to 0, if REPORT=1 it is set to 1),
                           default=1. See notes at end of writeup if NFCYCL > 1.
   
    56-60  ITABLE     = 0  Uses table lookup of trig functions in structure
                           factor calculation (default).
                      = 1  Evaluates trig functions explicitly (job takes
                           twice as much time, but slightly more accurate).    

(3)  Sequence Information			FORMAT ( 16I5 )

  NOTE:  The values required in the first 9 fields (columns 1-45 below)
         were printed by the "PROTIN" run that prepared input file IATMR. 

     1- 5  NA       No. of atoms in asymmetric unit

     6-10  NDIS     No. of distances to be restrained.

    11-15  NPLN     No. of planar groups in input file.

    16-20  NCHR     No. of chiral centers in input file.

    21-25  NVDW     No. of possible contacts in input file.

    26-30  NTOR     No. of conformational torsion angles.
             
    31-35  NSYM1    No. of symmetry equivalencies for type 1 symmetries.

    36-40  NSYM2    No. of symmetry equivalencies for type 2 symmetries.

    41-45  NOCC     No. of variable occupancy factors.

    46-50  ITEMP    = 0   use an overall thermal factor.
		    = 1   use individual thermal factors for each atom. 

(4)  Positional Parameters to FREEZE			FORMAT ( 16I5 )

      NKILL          No. of atoms for which coordinates will be held fixed.
                     (up to 500 values)

     (KIAT(I), I = 1,NKILL)       Atom number, in input list, of the
                                  atom which is to have its coordinates
                    "frozen" (example: metal atom of prosthetic group)
                    Probably its position is well determined and no
                    further refinement is desired.

(5)  Unit Cell Constants				FORMAT ( 10F8.3 )

     a, b, c (in Angstroms), alpha, beta, gamma (in degrees)

(5a)  FORMAT ( 26I3 )

     Col 1-3	KILRES = The number of residues to be omitted from 
                         structure factor calculation. Note that this does
                         not freeze atoms like NKILL does, but it removes
                         structure factor contributions from the least
                         squares equations.  Since restraint contributions
                         are still included, it can be used to "idealize"
                         selected troublesome residues.  It can also be used
                         to generate Fourier coeficients for "residue-deleted"
                         electron density maps. Maximum = 100.

     Col 4-6	KRES( I ), I = 1, KILRES
	 7-9	KRES( I ) = the residue numbers of residues to be omitted.
	 etc.
	

(6)  Reflection File Information	    FORMAT ( I10, 4F10.6, I10 )

      1-10  NOBS  = Maximum number of reflections in input file IREFR
                    (from HKSCAT output).

      1-20  FMIN  = Lower cut-off value for F.  Reflections with 
                    F < FMIN  will be rejected.

     21-30  SMIN  = (sin theta/lambda) min        reflections outside this 
     31-40  SMAX  = (sin theta/lambda) max        range will be ignored.

     41-50  SIGMIN  Such that if  F < SIGMIN * sigma-F  the reflection
                    is rejected.

(6a)  FORMAT ( I5 ) Statistical resolution breakdown information.

     Col 1-5	NI = 0 if d min default values of 5.0, 3.0, 2.5, 2.0, 1.8,
		       1.5 and 1.3 are to be used in resolution breakdown
		       segment of program.

                   = 1 if defaults are to be overridden.


	NOTE!  Include card 7 only if NI = 1.

(7)  Number and Limits of Shells for Statistics   FORMAT ( I5, 15F5.2 )

     1-5   N   = Number of shells in which to subdivide data for
		 statistical analysis.

     DMIN( I ), I = 1, N    Resolution limits (d spacings) of the N shells
                            for statistical analysis.


(8)  Weighting Information

8.1)  FORMAT (I8, 2F8.3, 8X, 6F8.3)

     1-8  KFWGT   Determines the weighting scheme applied to the
                  structure factors during refinement.

		= 1   then SIGAPP = SIGDEL

		= 2   then SIGAPP = max of (sigma-F, SIGDEL)

		= 3   then SIGAPP = sigma-F as given in input file  

		= 4   the SIGAPP = max of (( SIGDEL * sigma-F / < sigma-F >),
                                            sigma-F )

		where:

		1/SIGAPP**2 = structure factor weight used by the program
			      for refinement and statistics.

		SIGDEL = AFSIG + BFSIG * (sin theta/lambda -0.1666667)


 9-16  AFSIG     Independent term and coefficient of (sin theta/lambda)
17-24  BFSIG     term for structure factor weighting scheme as specified
	         by KFWGT.
25-32	         Blank.

33-40  WDSKAL  = Overall weight for distance restraints (usually 1.)
41-48  SIGD1   = Estimated standard deviations for distances in various 
49-56  SIGD2     classes.  Note actual weight used is
57-64  SIGD3      
65-72  SIGD4                  (  WDSKAL/SIGD(i) )**2        
73-80  SIGD5  
                Where SIGD(i) corresponds to SIGD1, SIGD2, . . . SIGD5
		according to the kind of distance that a given atom pair
		determines in the code established in program PROTIN
		(as set by KDWT).

		
		= 1   for bonded distances (1 (1) or 1 (3))
		= 2   for angle distance   (2 (2) or 2 (4))
      (i)	= 3   for planar 1-4 distance (e.g., carbonyl O in
                      residue J-1 to C alpha in residue J).
		= 4   for special input rest. dist., H-bond, etc. (4 (4))
		= 5   not used.


(8.1a)  FORMAT ( 10F8.3 )    NOTE!! Include this card only if KFWGT =4.
 
      Col 1- 8  AVSIGF (1)    =	The mean ESD's for reflections in each
      Col 9-16  AVSIGF (2)	resolution range defined by dmin on card
				6a or 7. (These can be obtained from a
                                previous run of GPRLSA).
	    etc.



(8.2)   FORMAT ( 10F8.3 )		

 1- 8  WPSKAL	Overall weight for planar group restraints (usually 1.)
                                      
 9-16  SIGP	Estimated standard deviation from planarity. Actual weight
                used is
                               ( WPSKAL/SIGP )**2 

17-24  WCSKAL	Overall weight for chiral group restraints (usually 1.)
                                      
25-32  SIGC	Estimated standard deviation for chiral volume. Actual
                weight used is
                               ( WCSKAL/SIGC )**2
  
33-40  WBSKAL	Overall weight for thermal factor restraints (usually 1.),
                used only if ITEMP (card 3) =1.
                
41-48  SIGB1    Estimated standard deviations from equality for thermal
49-56  SIGB2    factors of atom pair related by various distance types.	
57-64  SIGB3    Actual weight used is
65-72  SIGB4                   ( WBSKAL/SIGB(i) )**2
73-80  SIGB5	Where SIGB(i) corresponds to SIGB1, SIGB2 ... SIGB5,
        	according to the second distance kind code (KBWT) set
		in PROTIN (input 4):



|-------------------------------|
|SIGB(i)	Distance Code   |
|with i =	   PROTIN       |
|-------------------------------|---------------------------------------|
|   1              1(1)         |     bonded atoms of backbone only.    |
|-------------------------------|---------------------------------------|
|   2              2(2)         |     non-bonded main chain atoms.      |
|-------------------------------|---------------------------------------|
|   3              1(3)         |     bonded atoms - side chain         |
|-------------------------------|---------------------------------------|
|   4         2(4) or 4(4)      |     non-bonded atoms - side chain     |
|				|     or special input bonds            |
|-------------------------------|---------------------------------------|
|   5                           |     not used                          |
|-------------------------------|---------------------------------------|


(8.3)   FORMAT ( 10F8.3 )


 1- 8	WVSKAL	 Overall weight for VDW contact restraints (usually 1.)

 9-16	SIGV	 Estimated standard deviation from ideal contact distance.
                 Actual weight used is
                           ( WVSKAL / SIGV**2 ) ** 2

17-24	DINC(1)  These parameters give the possibility of modifying the
		 minimum "theoretical" van der Waals contact distance as
25-32   DINC(2)	 computed by the PROTIN program and contained in input
                 file IATMR according to the value of KTYP (in input (8)
33-40   DINC(3)  to PROTIN). Repulsion terms are added only if two
                 atoms are less than d(ideal VDW) + DINC(I) angstroms appart,
                 thus DINC values are usually negative to allow some flexibility
                 in VDW contact distances. Note attractive terms are NEVER
                 included.

		 DINC (1):  For atoms with relative position determined
			    by only one torsion angle:  single-torsion
			    contact.

		 DINC (2):  Two or more torsion angles determine the
			    relative position of the two atoms involved:
			    multiple-torsion contact.

		 DINC (3):  Possible hydrogen bond.  (Contacts between
			    any nitrogen or oxygen atom with another
                            nitrogen or oxygen atom but not
			    N main with N main or O main with O main).


41-48  WTSKAL    Overall weight for torsion angle restraints (usually 1.)

        	 Estimated standard deviations (in degrees) from "ideal"
                 torsion angles for various torsion classes are now supplied.
                 Actual weight used is
                                ( WTSKAL/SIGT(i) )**2

49-56  SIGT1     Sigma associated with a prespecified angle (usually phi
		 and psi of a regular secondary structure).

57-64  SIGT2	 Sigma associated with a planar angle (e.g. omega)

65-72  SIGT3	 Sigma associated with a staggered angle (e.g. chi 1 )

73-80  SIGT4	 Sigma associated with an orthonormal angle 
		 (e.g. chi 2 of aromatics).

(8.4)	FORMAT ( 10F8.3 )

 1- 8  PDEL	 Positional shift magnitude restraint (in angstroms).  This
                 parameter restricts the shift magnitudes and stabilizes
                 ill-conditioned refinement problems.  It should be used
                 if the problem is severely underdetermined.  If zero, no
                 restrictions are imposed. 

 9-16  BDEL	 Shift magnitude restraint on individual thermal factors.
		 (in angstroms**2), used only if ITEMP (card 3) =1. Functions
                 similar to PDEL.

17-24  QDEL	 Shift magnitude restraint on variable occupancy factors.
		 Used only if NOCC (card 3) > 0. Functions similar to PDEL.

25-32  WSSKAL	 Overall weight for non-crystallographic symmetry restraints
                 (usually 1.)
                 
                 Estimated standard deviations from superposability for
                 various classes of atomic parameters and restraint types
                 are now supplied. Actual weight used is
                               ( WSSKAL/SIGSP(i) )**2      or
                               ( WSSKAL/SIGSB(i) )**2

33-40  SIGSP1	 Sigmas associated with positional restraints (in angstroms).
41-48  SIGSP2	 (1,2,3) are for (loose, medium, tight) restraints,
49-56  SIGSP3	 respectively.

57-64  SIGSB1	 Sigmas associated with thermal factor restraints (in 
65-72  SIGSB2	 angstroms**2). (1,2,3) are for (loose, medium, tight
73-80  SIGSB3	 respectively).


(9)  Overall Temperature Factor and Scale.    FORMAT ( F8.3, I8, 8F8.3 )

 1- 8	TO	 Overall temperature factor for present cycle.  To be
		 added to individual B's in input file IATMR or used as
                 is if ITEMP (card 3) = 0.

 9-16	NQ	 Historical, always = 1 (It was used to allow for more
		 than one overall scale factor.  The current version of
                 the program is not able to use them).

17-24	SC(1)	 Overall scale factor.  Used in subroutine GCALC for
                 calculated structure factors.

		 Defined such that Fobs = SC * Fcalc(absolute)




(10)	Shift Damping Factors:

(10.1) Positional Damping Factors               FORMAT ( I5, 15F5.2 )

	JABN	 Number of damping factors to be read from this card.
                 (equal to the number of shift records in file ISHFTR).
                 Usually this is equal to the number of cycles that have been
                 run previously since the last run of PROTIN. The program
                 expects at least JABN * NV parameter shifts in file ISHFTR,
                 where NV is the total number of variables.
                                                               
		 	NV = 3 * NA + NOCC + 2	   if ITEMP = 0

			NV = 4 * NA + NOCC + 1	   if ITEMP = 1

		 NA is the number of atoms and NOCC is the number
		 of variable occupancy factors.  One record of NV values
		 is written per cycle of refinement. Maximum JABN=15.

     DAMP( I ),  I = 1, JABN.  Damping factors to be applied to the
		 coordinate shifts of cycle I to obtain the current 
		 refined coordinates from the starting coordinates
                 in file IATMR. DAMP(I) values of 1.0 apply full shift.

		 When running the first cycle:  JABN = 0 and the parameter
                 shifts (file ISHFTR) are not read in.



(10.2)	Individual Temperature Factor Damping Factors

	FORMAT ( 5X, 15F5.2 )

	Include this card only if ITEMP (card 3) =1

     DAMB(I)	 I = 1, JABN:  Similar to DAMP(I) in card 10.1 but for
		 individual thermal factor shifts. (Note that the program
                 expects the same value of ITEMP for all cycles. Changing
                 ITEMP to 1 requires outputing the current coordinates,
                 rerunning PROTIN and starting a new refinement series).

                 Thermal factors < 2.0 are reset to 2.0



(10.3)	Occupancy Factor Damping Factors    FORMAT ( 5X, 15F5.2 )

	Include this card only if NOCC (card 3) > 0.

	DAMQ(I), I = 1 JABN   	Similar to DAMP(I) in card 10.1, but
			        for variable occupancy factor shifts.

        If DAMQ(I) > 0,		The thermal shifts from the Ith cycle
				are not applied to those atoms which
				have variable occupancy factors, unless
                                the occupancy factor is 1. and the shift
                                indicates it should increase.  In that case
                                the occupancy shift is not applied.

	Permissible occupancy factor range is 0.01 < Q < 1.00



(10.4a)	FORMAT ( I3 )

	NSYMM = Number of equivalent positions in the space group.	
                (For centered cells include equivalent positions associated
                with one lattice point only, i.e. for sp. gp. C2 NSYMM=2)
	
(10.4b)	Free FORMAT

	Include NSYMM symmetry cards with one equivalent position per
	card, EXACTLY as in the International Tables.  First card should
        always be x,y,z.

	Example:  For Space Group P2 sub 1
	
		  CARD 1:  X,Y,Z
		  CARD 2:  -X, 1/2+Y, -Z


        NOTE! Card 11 is needed only if REPORT (card 2) is not = 2
      
(11)	R Test on Sample Set			FORMAT ( 4I5, 10F5.2 )
        
	IRTEST     = 0    Do not carry R test on sample set.
                 
	           > 0    Perform R test on sample set. (on parameter
		          shifts computed in the current cycle)

	NSAMPL		Number of reflections that program should select
			for R test.

  JAPN, JABN, SHFTK(I), I=1, JABN		See below

	The purpose of the R test is to find out the effect that different
	damping factors applied to the newly computed parameter shifts would
	have. The way it is done, is by selecting NSAMPL reflections from the
	set of accepted ones;

	Then, for each of the first JAPN cycles, a damping factor =
        SUM SHFTK(i) is applied to the new coordinate shifts only, and with
	the current temperature factors, the values of 
                            
			       SUM |Fob - Fcalc|	
			R =  -----------------------       ,
				   SUM Fob

	and "scale" = SUM(Fcalc * Fob)/SUM (Fcalc) ** 2, (total and in
        different ranges of sin theta/lambda) are computed. (SHFTK(1)
        applied in the first cycle, SHFTK(1) + SHFTK(2) applied in the
        second, etc. )

JABN	Total number of damping factor cycles (greater or equal to JAPN).
        This parameter indirectly determines the number of damping factor
        cycles applied to the new thermal factor shifts (Number of thermal
        damping cycles = JABN-JAPN). The damped thermal shifts are applied
        to the coordinate shift set that gave the lowest R value earlier.
        Thus

	SHFTK(1-->JAPN)      incremental damping factors for coordinates

	SHFTK(JAPN+1-->JABN) incremental damping factors for thermal parameters.
        
        For each cycle R values and statistics are given as described above.
        Maximum allowed value of JABN = 10.

In Short:

JAPN	Number of cycles for trying different values of positional parameter
	damping factors to explore their effect on the R and scale of a "small"
	NSAMPL set of reflections.

JABN	Total number of cycles to be performed (or different damping factors
	to be tried).  First JAPN cycles use initial temperature factors
	(changes in coordinates only).  JAPN+1 to JABN cycles use
	positional parameter shifts that gave minimum R in first JAPN cycles
	and apply SHFTK (JAPN+1)...etc. to explore effect of damping 
        temperature factor shifts.


         ********** Notes on Multicycle jobs  **********

        This version of the program has been modified to allow multiple
        refinement cycles to be run in the same job (parameter NFCYCL, card 1).
        On fast computers (e.g. CRAY, FPS) this is often desirable, but there
        are certain ramifications described below:

        1) Between cycles the parameters AFSIG and BFSIG (card 8.1) are
        adjusted such that the ratio of MODELED to FITTED values found
        in the first cycle is preserved. Accordingly, one should not use
        the multicycling option without first running a single calculation
        to determine what AFSIG and BFSIG values to input. Good results are
        usually obtained when the input values are 40-50% of the FITTED values
        obtained from the previous output.   
     
        2) To minimize printout, lists of bad distances, torsions etc are
        printed only during the first cycle, but summaries for each restraint
        type, R factors and parameter shifts are printed for each cycle.
        WARNING! For multiple cycles, specifying printout of structure factors
        (LISTF, card 2) and/or complete parameter shifts (LISTA=1, card 2)
        can lead to enormous printout. 

        3) All parameter shift sets created in a multicycle job are combined
        and output on file ISHFTW as if they were computed in a single cycle.
        Thus when the shift file is read back in (now on file ISHFTR), the
        parameter JABN (card 10.1) is no longer equal to the number of
        CYCLES performed since the last run of PROTIN, but is the number
        of refinement JOBS run in that period (Actually it is the number
        of shift sets present on the file).

        4) Full parameter shifts are applied during all cycles run in
        the job.    

        5) The scale factor shift is applied between cycles, and the new
        scale factor is printed out.  To resume where you left off after
        a multicycle job, the value given in the last "NEW SCALE FACTOR = "
        line should be input for the next job (parameter SC(1), card 9). 

        6) If an overall thermal factor is used (parameters ITEMP, card 3
        and TO, card 9) it will be held fixed during all cycles.  Shifts
        are computed and listed on the printout, but they are not applied.
        (In most cases the calculated shift is unreasonable anyway). If
        individual thermal factors are used however, full shifts are applied
        exactly as they would be in normal operation.       

        7) If more than one cycle is computed (NFCYCL (card 2) > 1), then
        the R test (card 11) is not meaningful since the damping factors are
        applied only to the shifts computed in the last cycle OF THE CURRENT
        JOB.  Thus any factor determined in this test SHOULD NOT be applied
        to the shift set written to file ISHFTW when it is used in the next run.