USING THE NSWC LIBRARY ON AN IBM PC

    Some PC users compile the NSWC Library and use a linker to
 fetch needed routines. Others use the batch file,FCH.BAT, to fetch
 needed routines from the source library. The advantages of using
 FCH are that a long compilation is avoided, and the user has the option
 of converting the fetched routines to double precision. For details
 concerning this conversion, and for options not explained in this
 document, see the file FCH.HLP.

    For the purposes of this document, a library(or source file) is a
 file containing one or more uncompiled Fortran routines, where a routine
 is a function, subroutine, main program, or block data subprogram.
 Routines may be written in small letters,capitals, or a combination of
 both. FCH will recognize all ANSI 66 and ANSI 77 statements except SAVE,
 ENTRY, and INQUIRE. Tab characters from Edlin and Personal Editor are
 allowed but only one per line. The tab character must be in columns 1
 thru 6 and the tab column must be 7.

    Typing 'FCH h' will display the argument list of FCH on the terminal.
 Typing 'FCH H' will display the argument list of FCH on the terminal and
 give a description of the arguments and the results they produce. As the
 description is displayed on the screen, it is written to the file
 FCH.HLP(which can be sent to a printer). This is the only instance in
 which small letters and capitals produce different results.

    There are two ways to fetch: by routine name and by source file. In
 either case the fetching is done from one or more libraries. The
 fetching continues until all requested routines(including all supporting
 routines) are obtained or it is determined that the remaining needed
 routines are not in the libraries, in which case it is reported to the
 user. Consequently, FCH acts like a linker or loader except that the
 files and routines are in source code rather than object code.
 
    Assume that L1 and L2 are the libraries from which the fetching is
 done and OUT is the file which will contain the fetched routines.
 Initial data in OUT will be lost.

 1. Fetch the routines BRATIO and GRATIO.
    FCH  f  -  OUT  BRATIO+  GRATIO+  L1  L2

    The minus sign indicates that there is no input source file. The plus
 signs indicate routine names rather than library names. Routine names
 must precede library names. The number of libraries must not exceed 15.

 2. Fetch all routines needed by the source file IN.
    FCH  f  IN  OUT  L1  L2
    Assume that IN contains:  PROGRAM MAIN
                              CALL BRATIO
                              CALL GRATIO
                              END

    Examples 1 and 2 will produce the same output file OUT. OUT will
 contain BRATIO and GRATIO(provided they are in L1 or L2), and all
 the routines from L1 and L2 which are used by BRATIO and GRATIO.

   The two methods may be used together, for example
       FCH  f  IN  OUT  RLOG+  ERF+  L1  L2   is permitted.

    FCH will not fetch more than one copy of a routine having a specified
 name. If 'CALL ABC' is in the input file then FCH will fetch the
 first routine(main program,subroutine,function, or block data) it finds
 which is named 'ABC'. The name of each routine fetched is written to
 the file ERRLIS.

    Error messages are written to the file ERR. When the procedure ends
 ERR is listed on the terminal. If ERR is empty then 'ok' is displayed.

    Libraries may be large enough so that fetching is very time
 consuming. If something is known about the library then fetching will be
 performed more efficiently. Example 3 illustrates the use of an index file.

 3. Do examples 1 and 2 above using the index file IND.
    FCH  f1  -   OUT  BRATIO+  GRATIO+  IND
    FCH  f1  IN  OUT  IND

    The first argument of FCH specifies the function to be performed:
    'f'  indicates to fetch without an index file
    'f1' indicates to fetch with an index file

    Note that using an index file is slightly different from the previous
 version of FCH. This new version of FCH does not require the library
 names in the argument list because the library names are in the index
 file. The library names begin in line 2 of the index file, one name
 per line(including the path) beginning in column 1.

    The index file provided for the NSWC math library is named 'INDEX'
 and contains 'C:\NSWC\NSWCLIBR' in line 2. If the library is stored
 with a different path or name then change line 2 accordingly. Assuming
 that the index and the library are not renamed and are both stored in
 directory C:\NSWC we do example 3 using the NSWC library:
    in directory C:\NSWC
    FCH  f1  -   OUT  BRATIO+  GRATIO+  INDEX
    FCH  f1  IN  OUT  INDEX

    in a directory different from C:\NSWC
    FCH  f1  -   OUT  BRATIO+  GRATIO+  C:\NSWC\INDEX
    FCH  f1  IN  OUT  C:\NSWC\INDEX

    In all the examples either or both of the files IN and OUT may be
 prefixed with a path.

    Many PC users use double precision because of the limited
 accuracy on the PC. For this reason and because of the nature of 
 the routines SPMPAR,EXPARG, and EPSLN:
       when SPMPAR is fetched DPMPAR is also fetched
       when EXPARG is fetched DXPARG is also fetched
       when EPSLN  is fetched DEPSLN is also fetched

    The NSWC Library issued on a floppy disk has the function IPMPAR
 initialized for an IBM compatible PC using any of the four compilers:
 RM/Fortran,Microsoft,Lahey,Professional. For other IBM PC compilers and
 for compilers for other computers, see the appendix of the NSWC Math
 Library manual(NSWCDD TR 92/425).

                                  Russ Gnoffo  K13  703-663-8295
�